Page 1
For peer review only
Prescriber Preference for a Particular Tumor Necrosis Factor Antagonist Drug and Treatment Discontinuation:
Population-based Cohort
Journal: BMJ Open
Manuscript ID: bmjopen-2014-005532
Article Type: Research
Date Submitted by the Author: 23-Apr-2014
Complete List of Authors: Fisher, Anat; University of British Columbia, Anesthesiology, Pharmacology & Therapeutics Basset, Ken; University of British Columbia, Anesthesiology, Pharmacology
& Therapeutics; University of British Columbia, Family Practice Wright, James (Jim); University of British Columbia, Anesthesiology, Pharmacology & Therapeutics; University of British Columbia, Medicine Brookhart, M.; University of North Carolina at Chapel Hill, Epidemiology Freeman, Hugh; University of British Columbia, Medicine Dormuth, Colin; University of British Columbia, Department of Anesthesiology, Pharmacology and Therapeutics
<b>Primary Subject Heading</b>:
Epidemiology
Secondary Subject Heading: Epidemiology, Health services research, Patient-centred medicine, Research methods, Rheumatology
Keywords: EPIDEMIOLOGY, RHEUMATOLOGY, Health policy < HEALTH SERVICES
ADMINISTRATION & MANAGEMENT
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open on June 29, 2020 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2014-005532 on 30 S
eptember 2014. D
ownloaded from
Page 2
For peer review only
1
Comparative Persistence of TNF Antagonists in
Rheumatoid Arthritis – A Population-Based Cohort
Study
Anat Fisher, MD, PhD; Ken Bassett, MD, PhD; James M. Wright, MD, PhD; M. Alan
Brookhart, PhD; Hugh Freeman, MD; Colin R. Dormuth, ScD
Corresponding Author: Anat Fisher, MD, PhD; Department of Anesthesiology, Pharmacology
and Therapeutics, University of British Columbia, 2176 Health Sciences Mall, Vancouver, BC,
Canada V6T 1Z3, Telephone: 604-822-0700 Fax: 604-822-0701, Email: [email protected]
Ken Bassett: Department of Anesthesiology, Pharmacology & Therapeutics and Department of
Family Practice, University of British Columbia, Vancouver, Canada
James M. Wright: Department of Anesthesiology, Pharmacology & Therapeutics and
Department of Medicine, University of British Columbia, Vancouver, Canada
M. Alan Brookhart: Department of Epidemiology, Gillings School of Global Public Health,
University of North Carolina, Chapel Hill, North Carolina, USA
Hugh Freeman: Department of Medicine, University of British Columbia, Vancouver, Canada
Colin R. Dormuth: Department of Anesthesiology, Pharmacology & Therapeutics, University of
British Columbia, Vancouver, Canada
Keywords: physician prescribing pattern, rheumatoid arthritis, medication persistence, Tumor
Necrosis Factor-alpha
Number of words (excluding title page, abstract, references, figures and tables, endnotes): 2581
Page 1 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 3
For peer review only
2
ABSTRACT
Objective: To assess the effect of physician preference for a particular tumor necrosis factor
alpha (TNF) antagonist on the risk of treatment discontinuation in rheumatoid arthritis.
Design: Population-based cohort study.
Setting: British Columbia administrative health data (inpatients, outpatients and pharmacy)
Participants: 2,742 British Columbia residents who initiated a first course of a TNF antagonist
between 2001 and December 2008, had been diagnosed with rheumatoid arthritis, and were
treated by one of 58 medium to high-volume prescribers.
Independent variable: A level of physician preference for the drug (higher or lower) was
assigned based on preceding prescribing records of the care-providing physician. Higher
preference was defined as at least 60% of TNF antagonist courses initiated in the preceding year.
Sensitivity analysis was conducted, with different thresholds for higher preference.
Main Outcome Measure: Drug discontinuation was defined as a drug-free interval of 180 days
or switching to another TNF antagonist, anakinra, rituximab or abatacept. The risk of
discontinuation was compared between different levels of physician preference using survival
analysis.
Results: Higher preference for the prescribed TNF antagonist was associated with improved
persistence with the drug (4.28 years [95% confidence interval 3.70-4.90] versus 3.27 [2.84-
3.84], with log rank test p-value of 0.017). The adjusted hazard ratio for discontinuation was
Page 2 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 4
For peer review only
3
significantly lower in courses of drugs with higher preference (0.85 [0.76-0.96]). The results
were robust in a sensitivity analysis.
Conclusions: Higher physician preference was associated with decreased risk of discontinuing
TNF antagonists in patients with rheumatoid arthritis. This finding suggests that physicians who
strongly prefer a specific treatment help their patients to stay on treatment for a longer duration.
Similar research on other treatments is warranted.
Strengths and Limitations
• First study to explore within physician variation in prescribing habits, specifically the
effect of prescriber preference to a drug on the decision to discontinue the drug
• The universal nature of the Canadian health care system and a systematic and
standardized approach to data collection in British Columbia which ensured the
generalizability of our results, as well as the large sample and prolonged follow up
• To conquer the absence of access to clinical data we used multiple proxy variables to
adjust for disease severity
• Physician preference was based on previous prescribing habits and not directly measured
Page 3 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 5
For peer review only
4
INTRODUCTION
The term ‘physician preference’ usually refers to favoring a particular drug or a therapeutic
group among several alternatives, and it has been shown to predict treatment choice (1-4). In
studies of administrative heath (claim) data, this preference is often determined by identifying
dispensing of drug prescribed by the specific physician in a pre-determined period, prior to the
event of interest (a new prescribing). Despite an association with new prescribing desicions, the
role of physician preference in treatment discontinuation has not been studied. Recently, the term
‘preference’ has also been used to describe a second phenomenon – in the context of treatment
discontinuation is was used to describe the baseline risk of discontinuing treatment in patients
treated by a specific physician (the physician ‘preference for discontinuation’) (5). This baseline
risk may differ among physicians because physicians may respond differently to similar clinical
situations such as decreased benefit or harmful events. They could recommend patients to
discontinue treatment (or switch to a second drug) or to persist with the treatment (but to adjust
dose, add-on a second drug or be under frequent watch). In this paper, we use the term
‘preference’ to describe the first phenomenon (physician’s favorite drug) and ‘physician-specific
discontinuation risk’ to describe the second.
Treatment with TNF antagonists in patients with rheumatoid arthritis (RA) was considered
especially sensitive to physician preference for two main reasons. First, during the study period
(2001-2009) there was limited clinical evidence on the comparative effectiveness of the drugs,
due mostly to the absence of head-to-head randomized clinical trials, but also because
participants in placebo-controlled trials were not representative of patients treated in routine clinical
Page 4 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 6
For peer review only
5
settings (6-9). Second, published indications for discontinuation of TNF antagonists were vague
and confusing, and therefore care-providing physicians could reasonably be expected to reach
different clinical decisions given the same clinical situation. Consequently, the decisions about
which TNF antagonist to prescribe first and when to discontinue treatment were likely subject to
physicians’ individual preference.
This study analyzed data of first courses of a TNF antagonist in British Columbia patients with
RA. The prescriber recorded on the first dispensing claim for a TNF antagonist was used as a
proxy of the care-providing physician. The study objective was to estimate the effect of
physician preference on the risk of discontinuation. The null hypothesis tested was that physician
preference for a TNF antagonist when treatment has been initiated does not influence the risk of
discontinuing the treatment in RA patients.
PATIENTS AND METHODS
The study cohort was identified using four British Columbia Ministry of Health administrative
databases: PharmaNet (prescription dispensing data), Medical Service Plan (MSP) registration
information (demographic data), MSP Payment Information (fee-for-service payments to
physicians and alternative providers), and the Discharge Abstract Database (hospital
separations). The databases were linkable using a de-identified patient and physician numbers.
Follow-up data were available from 1995 until December 31, 2009. The study cohort included
British Columbia residents who fulfilled all of three conditions: (1) first exposure to a TNF
antagonist between March 1, 2001 and December 31, 2008; (2) a diagnosis of RA; and (3) TNF
antagonist treatment initiated by a medium or high-volume prescriber, defined as a prescriber
Page 5 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 7
For peer review only
6
who initiated 5 or more courses of TNF antagonists in RA patients during the years 2001-2008.
Exposure to a TNF antagonist was based on one or more recorded dispensing claims for
infliximab, adalimumab or etanercept between March 2001 and December 31, 2008. We
excluded patients who were previously treated with anakinra, rituximab or abatacept to decrease
heterogeneity in the population analyzed. The first dispensing claim for a TNF antagonist was
used to identify the study drug (infliximab, adalimumab or etanercept), index date and prescriber
for each patient in the cohort. We defined a pre-study period of three years preceding the index
date during which patients were required to have continuous provincial medical service
coverage. A gap shorter than 30 days was not considered to be an interruption in coverage. The
pre-study period ensured a standard run-in period of at least three years without TNF antagonist
exposure and a standard period during which diagnosis of RA was identified. RA patients were
selected based on similar criteria to previous studies in British Columbia (10-12): either two
outpatient visits in physician clinics with a diagnosis code of RA (ICD-9 714) at least 60 days
apart, or one hospitalization with recorded discharge diagnosis of RA. Patients were excluded if
sex or date of birth was not available, if they had a concurrent diagnosis of Crohn’s disease
(based on at least one outpatient or inpatient diagnosis code in the pre-study period) or if they
were younger than 18 years at the index date (to remove patients with juvenile RA). We also
excluded patients initiated on TNF antagonist treatment by prescribers who cumulatively
initiated less than five patients in the RA cohort (low-volume prescribers).
Physician preference was determined for each patient at the index date and coded as a Bernoulli
variable. It reflects prescribing patterns of the individual physician (who started this course)
during the preceding year (Figure 1). For each individual patient, we determined the date of
course initiation, the prescriber recorded in the first dispensing event and the drug (infliximab,
Page 6 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 8
For peer review only
7
adalimumab or etanercept). From first dispensing records for each patient in the study cohort, we
identified other patients that received any TNF antagonists from the same prescriber during the
year preceding the index date of the current patient, and the drug they were treated with. Next we
calculated the proportion of these patients who received the same drug as the patient of interest.
Based on this proportion, we categorized the prescriber preference to ‘higher’ when the
calculated proportion for this drug and that prescriber was 60% or higher, and ‘lower’ otherwise.
We expected that this threshold would provide reasonable assurance that a course categorized as
having ‘higher” preference for drug actually reflects that the prescriber favored that particular
TNF antagonist at this time, even if only two TNF antagonists were available. We reassessed
preference for drug based on annual data to allow for changes in preference over time and for
accommodating the more recent availability of adalimumab. Adalimumab was available in
Canada since October 2004, while infliximab and etanercept were on the market since 2001. By
including a year of data, differently from other studies that used only the last prescription in
assigning preference (1-4), we created a more accurate indicator. This way we minimized the
effect of factors not related to preference that might have influence a specific prescribing
decision. Sensitivity analysis was conducted to examine the robustness of main results, using
thresholds of 70% or 80% for the level of physician preference.
Multiple covariates, mainly patient characteristics that may influence drug selection and/or
discontinuation were included in the final models and are listed in Table 1. We were limited by
lack of access to clinical data and inability to adjust for clinical variables not captured in
administrative health data, such as disease severity. Hence, we used multiple proxies, including
disease duration, age, extra-articular involvement, antirheumatic drug use, health services
utilization and consumption of pain medications to adjust for disease severity. In order to control
Page 7 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 9
For peer review only
8
for secular trends and late availability of adalimumab in Canada, we included a categorical
variable for the year of treatment initiation.
Table 1 - List of covariates included in the final models
Variable Description
Demographics
Sex Bernoulli variable
Age at index Four categories: 18-29 years, 30-69, 70-79 and ≥80 (categories were
assigned based on similar discontinuation risk in preliminary
analysis, which was not linear)
The annual deductible for
prescription cost at index
(Fair PharmaCare)
The deductible is based on annual income(13). Six categories:
annual deductible of $0, $1-500, $501-2250, >$2250, other
programs and no coverage
Geographical area at
index
Based on first 3 digit of postal code. Five mutually exclusive
categories: Greater Vancouver, Greater Victoria, Vancouver Island,
urban areas and rural areas
Clinical status
Physicians encounters in
the year preceding the
index date
Continuous variable
Number of inpatients
admissions in the year
preceding the index date
Four categories:0, 1, 2, >2
Comorbidities (presence
and severity) during the
pre-study period
Charlson comorbidity score (14) was determined using Quan’s
algorithm for administrative databases (15), excluding rheumatic
diseases. At least two outpatient or one inpatient encounter with the
diagnosis were required. Four categories: 0, 1, 2, >2
Disease duration Measured from the first diagnosis of RA (inpatient or outpatient) in
Page 8 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 10
For peer review only
9
Variable Description
the data. Continuous variable
The presence of extra-
articular manifestation
during the pre-study
period
Based on recorded at least one diagnoses with ICD-9 codes 3571,
3596, 7141, 71481, 7142 (outpatient) or ICD-10 codes G636, G737,
I39, I418, J990, M050, M051, M052, M053 (inpatient). Bernoulli
variable
Drug therapies
Concomitant MTX
during 200 days
preceding the index date
Based on mean plus two standard deviations of between-dispensing
intervals of MTX in the study cohort. Bernoulli variable
Dispensing claims for
NSAIDs during the year
preceding the index date
Bernoulli variable
Number of different
antirheumatic drugs
dispensed in the pre-
study period
Dispensing claims of 10 drugs were included: MTX,
hydroxychloroquine, sulfasalazine, leflunomide, azathioprine,
minocycline, penicillamine, sodium aurothiomalate, prednisone, and
intra-articular triamanolone or methylprednisolone. Four categories:
no drug, 1-2, 3-6, >6 different drugs
Other
Calendar year at index
date
The variable allowed controlling for secular trends in clinical
practice (16-19) and availability of drugs. Eight yearly categories
were included for the years 2001-2008
ICD-10- International Classification of Diseases, 10th edition; ICD-9-CM - International Classification of Diseases, 9th edition, clinical modification; MTX – methotrexate; NSAID – nonsteroidal anti-inflammatory drug
The outcome variable was a discontinuation defined as either switching to another TNF
antagonist including certolizumab and golimumab, anakinra, rituximab or abatacept (‘biologic’
antirheumatic drugs), or a drug-free interval of 180 days after exhaustion of the dispensed days-
Page 9 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 11
For peer review only
10
supply of the latest refill. The discontinuation date was set to the end of the days-supply of the
last refill before discontinuation or the date of the first dispensing of a second ‘biologic’
antirheumatic drug, whichever was earliest. Unless experienced discontinuation, patients were
followed up until December 31, 2009 (end of follow-up period) or until an interruption of more
than six days in the provincial MSP coverage, at which point data were considered censored. The
most common causes of such interruptions were death and emigration from the province. Change
of dose of the TNF antagonist or adding-on a second drug (not TNF antagonist) was not
considered discontinuation, and patients were continued to be followed.
Discrepancies between recorded days-supply and dispensed quantity were common in the
PharmaNet database; hence, we also calculated days-supply based on the quantity dispensed,
which, if required, was imputed using the recorded total cost. Cost was considered the most
accurate and reliable field, since this field serves for claim and payment processing. We used the
longest duration of days-supply, recorded or calculated, to determine both the length of drug-free
intervals and a discontinuation date.
Sample size calculation
Assuming 1:1 ratio of courses with ‘higher’ preference versus courses with ‘lower’ preference
and in the absence of prior estimates, if the true hazard ratio for discontinuing courses with
‘higher’ preference relative to courses with ‘higher’ preference is 0.8, we required 847 courses
with each preference level to be able to reject the null hypothesis with type I error probability
0.05 and power of 0.80.
Statistical Analysis
Page 10 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 12
For peer review only
11
Summary statistics of baseline characteristics were compared across the three drug groups. The
adjusted risk of discontinuation was compared using two different approaches in a Cox
proportional hazards regression. First, we conducted an analysis of nonclustered data. The
multivariate model also included a series of Bernoulli variables, one for each physician (1=the
particular physician; 0=other physicians), that allowed individualization of the risk of drug
discontinuation by physician (physician-specific discontinuation risk). In the second approach, a
marginal model of clustered data allowed an adjustment for possible correlation between patients
treated by the same physician (20-22). We tested for model assumptions (proportional hazard
and absence of interactions) and found them to be valid. We also checked for the linearity of
continuous variables and categorized non-linear variables. All statistical tests were two-sided.
All calculations were performed using the SAS software package (SAS Institute Inc., Cary, NC).
RESULTS
The study cohort included 2,742 RA patients prescribed by a total of 58 medium-high volume
physicians. Figure 2 presents reasons for excluding patients who used TNF antagonists from the
analysis. Baseline characteristics across the three drug groups are presented in Table 2. Not only
was etanercept was the most frequently prescribed drug (1718 patients, 63%), but also it was the
only drug prescribed by all 58 study physicians. Etanercept was usually initiated by a physician
with high preference for etanercept (70% of etanercept courses). Infliximab or adalimumab, on
the other hand, were usually initiated by physicians with lower preference for these drugs (only
34% or 19% of courses were initiated by a physician with high preference for drug,
respectively). Patients treated with adalimumab were significantly older and had a lower income
Page 11 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 13
For peer review only
12
(reflected in lower annual deductible level for prescribing costs). Patients treated with infliximab
had the highest prevalence of concomitant methotrexate therapy and patients with etanercept the
lowest. This reflects differences in the indications mention in the product monographs; while
infliximab is indicated for use in combination with methotrexate (23), the monograph of
etanercept mentioned that it “can be initiated in combination with methotrexate … or used alone”
(24). Similarly, while the provincial special authority policy requires that infliximab is used in
combination with methotrexate (or other drug) and such requirement does not exist for treatment
with etanercept or adalimumab (25).
Page 12 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 14
For peer review only
13
Table 2 - Baseline characteristics
Variable Infliximab Adalimumab Etanercept P-value
(between
groups
differences)
Number of patients (% from
cohort)
571 (21) 453 (16) 1718 (63)
Number of prescribers (based on
first dispensing event for each
patient) (total 58 prescribers)
49 46 58
Patients treated with a drug with
higher preference, No (%)
193 (34) 84 (19) 1198 (70) <0.0001
Demographics
Females, No (%) 403 (71) 326 (72) 1239 (72) 0.77
Age at index, median (range) y 56 (18-87) 58 (22-91) 56 (18-92) 0.003
Annual deductible for prescription
cost, No (%)
Very low ($0) 47 (8.3) 80 (18) 199 (12) <0.0001
Low ($1-500) 33 (5.8) 53 (12) 152 (8.9) 0.004
Medium ($501-2250) 92 (16) 109 (24) 315 (18) 0.004
High (>$2250) 35 (6.1) 38 (8.4) 143 (8.3) 0.22
Residence in Greater
Vancouver/Victoria, No (%)
341 (60) 224 (50) 782 (46) <0.0001
Clinical status
Number of physician visits median
(range)
33 (3-158) 31 (2-112) 32 (3-136) 0.25
At least one admission to hospital
No (%)
104 (18) 63 (14) 340 (20) 0.01
Extra-articular manifestations, No 28 (4.9) 14 (3.1) 60 (3.5) 0.23
Page 13 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 15
For peer review only
14
(%)
Presence of comorbidity (score>0),
No (%)
113 (20) 95 (21) 383 (22) 0.41
RA disease duration median,
(range) y
9.2 (0.1-
17.9)
7.7 (0.3-17.9) 8.0 (0-17.8) 0.21
RA drugs
Concomitant MTX, No (%) 388 (68) 264 (58) 856 (50) <0.0001
Dispensing claims for NSAIDs,
No (%)
307 (54) 214 (47) 923 (54) 0.04
Number of different antirheumatic
drugs, median (range)
4 (0-8) 4 (0-8) 4 (0-9) 0.46
%- percent, unless otherwise specified – percentage from patients treated with this drug; $-Canadian dollars; MTX-methotrexate; No-number of patients; NSAID- nonsteroidal anti-inflammatory drug; NSS – not statistically significant; y-years
For detailed descriptions of the variables refer to Table 1
Persistence with the three TNF antagonists was similar (log rank test p-value=0.15). The product
limit median persistence estimates were 3.9 years (95% confidence interval 3.0-5.3), 3.3 (2.6-4.0)
and 3.9 years (3.4-4.4) for infliximab, adalimumab and etanercept, respectively. Higher physician
preference was associated with improved persistence compared to lower preference (Figure 3),
with median time to discontinuation 4.28 years (confidence interval 3.70-4.90) in 1475 courses
with ‘high’ preference, compared with 3.27 (2.85-3.84) in 1267 courses with ‘low’ preference .
In both groups about 50% of the patients discontinued and the other 50% were censored.
Higher physician preference was associated with a significant decrease of 14-15% in the adjusted
hazard for discontinuation (Table 3). No significant interaction between drug and preference was
observed. The results of sensitivity analysis were similar, with overall adjusted hazard ratio
Page 14 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 16
For peer review only
15
between 0.85-0.88 in all models. The results of clustered data analyses with thresholds of 60%,
70% and 80% for physician preference were robust (Table 3). The results of nonclustered data
analyses for thresholds of 70% and 80%, on the other hand, did not reach the significance level.
That may be a result of decreasing numbers of patients treated with a drug of higher physician
preference and therefore decreased power to detect significant difference.
Table 3 - Hazard ratios for drug discontinuation, higher preference versus lower
preference
Approach
Hazard ratio (95% confidence interval)
Preference threshold
60%
Preference threshold
70%
Preference threshold
80%
Patients with higher
preference N (%)
1426 (52) 1234 (45) 987 (36)
Nonclustered
data analysis
Crude 0.88 (0.79-0.98) 0.89 (0.80-0.99) 0.83 (0.81-0.996)
Adjusted* 0.86 (0.76-0.98) 0.88 (0.77-1.01) 0.87 (0.75-1.004)
Marginal
modeling of
clustered data
Crude 0.88 (0.77-1.001) 0.89 (0.78-1.01) 0.89 (0.79-1.01)
Adjusted# 0.85 (0.76-0.96) 0.87 (0.78-0.97) 0.87 (0.78-0.98)
% - percent from overall patients (2742); N – Number of patients; *Adjusted by drug, calendar year, patient’s demographics and clinical status, and prescriber #Adjusted by drug, calendar year and patient’s demographics and clinical status
Page 15 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 17
For peer review only
16
DISCUSSION
We demonstrated two types of variations of physician decisions to discontinue treatment with TNF
antagonists: between-physician and within-physician. Between-physician variability in response
to similar clinical situations, such as decreased benefit or a harmful event is expressed as a
different baseline risk of discontinuation, depending on the treating physician. We used the term
physician-specific discontinuation risk for this phenomenon. It could be a result of differences in
education or experience, adherence to different guidelines and possible differences in patient
case mix. Within-physician variability, on the other hand, is a different response to a similar
clinical situation (similar patients, similar drug effects) by the same physician. In this study we
showed that this variability correlated to physician preference for drug – and specifically, higher
physician preference for a specific TNF antagonist drug was associated with a decreased risk of
discontinuing TNF antagonists in RA patients. This preference most probably reflected the
physician’s beliefs regarding the relative effectiveness and safety of the specific drug compared
to the alternatives, where a preferred drug is thought to be superior. When a physician treated
with a drug believed to be superior compared to comparators, patients were encouraged to stay
longer on the drug and the risk of discontinuation, measured in this study, decreased. We believe
that a different response (level of encouragement) to a similar clinical situation applied in
indefinite clinical situations, such as mild harmful effect or questionable benefit.
Strengths and weaknesses of the study
The main advantages of our study are the universal nature of the Canadian health care system
and a systematic and standardized approach to data collection in British Columbia which ensured
the generalizability of our results, as well as the large sample and prolonged follow up that
Page 16 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 18
For peer review only
17
increased the power of the study to detect differences in discontinuation risk. The main
disadvantage is the absence of access to clinical data; hence adjustment to patient mix was
difficult. To conquer this problem, we used multiple proxy variables to adjust for disease
severity, such as extra-articular involvement, antirheumatic drug use, health services utilization
and consumption of pain medications.
There are several limitations to ascertaining physician preference based on previous prescribing
habits. First, preference was not defined for all patients. For example, it was not defined for the
first patient treated by the physician, or if the physician did not prescribe this class of drugs in
the preceding year. Second, we assumed that increased loyalty for a particular TNF antagonist
reflects preference, but we did not measure preference directly. Nevertheless, stated preference
and prescribing habits were shown to correlate in previous research (26). Third, we measured
preference at the time of treatment initiation and not at discontinuation, and preference might
have changed over time. Finally, we assumed strong preference for one drug only and regarded
situations in which the physician preferred two of the TNF antagonists to be a situation of
treatment with a drug of lower preference.
Overview of previous studies
Treatment persistence was hypothesized to reflect therapeutic benefit and harm in chronic non-
curable diseases, such as rheumatoid arthritis (27). In support of this hypothesis, previous studies
in RA patients using TNF antagonists demonstrated that the main reasons for discontinuing or
switching were decreased benefit (36-67% of the discontinuations) or perceived harm (30-58%)
(28-33). As such, the risk of discontinuing treatment is assumed to be influenced solitarily by
drug properties and patient characteristics. The demonstrated effect of care-provider
Page 17 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 19
For peer review only
18
characteristics, specifically prescribing habits, on the risk of discontinuation calls into question
this hypothesis.
Between-physician differences in response to the same clinical situation (physician-specific
discontinuation risk) have previously been studies in RA patients treated by TNF antagonists.
Differences in response to harm were reported by Cush et al 2005 based on an on-line survey of
rheumatologists (1) and in response to decreased benefit – by Zhang et al 2011 (5). Using mixed
effect model with a random intercept to cluster patients at physician level, Zhang demonstrated
the importance of clustering by physicians, even after adjustment for both baseline disease
activity and improvement in disease activity.
Within-physician variation in prescribing decisions and the effect of physician preference have
infrequently been studied. While the effect of physician preference on discontinuation has not
been previously studied, in this therapeutic class or in other conditions, physician preference was
found to be an important predictor of drug selection in treatment initiation. Physician preference
for therapeutic class was the most important determinant in initiating treatment with TNF
antagonists compared to the alternative: prescribing synthetic antirheumatic drugs (1) and in
other treatment situations (3,4,34). Physician preference for an individual TNF antagonist was
studied by Kamal et al 2006 (35). In response to mailed questionnaires, most American
rheumatologists said they preferred etanercept over adalimumab or infliximab and considered
etanercept the most efficacious of the three drugs with less harm. In our study, etanercept was the
most prescribed TNF drug, and it was prescribed by all 58 medium to high-volume prescribers.
Explanations and interpretation
Page 18 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 20
For peer review only
19
We suggest several possible explanations for the finding of decreased risk of discontinuation
with increased physician preference for the prescribed TNF antagonist. First, the results could be
explained in light of the theory of cognitive dissonance (36). Dissonance is an uncomfortable
feeling caused by holding conflicting ideas simultaneously. In line with this theory, treating with
a drug believed to be inferior (lower preference) induces dissonance. In indefinite clinical
situations, such as mild harmful effect or questionable benefit, early drug discontinuation
supports the physician’s belief that the selected drug was inferior compared to the alternatives.
On the other hand, in a similar situation, a drug with a higher preference might be continued, to
support a belief in its superiority. The interpretation of our results using the theory of cognitive
dissonance is restricted by the lack of direct measurement of physician dissonance at the time of
treatment discontinuation. The indirect measure of dissonance we used, physician preference,
could also be influenced by factors unrelated to prescriber beliefs, such as limited availability of
a drug or higher relative cost. Direct measures of cognitive dissonance are complicated and
involve activities beyond the scope of the current study (37). Second, the observed association
between higher preference and decreased risk of discontinuation might be confounded by
experience with the drug, namely the total number of patients a physician has previously treated
with the same drug. Increased experience with a specific drug may be associated with higher
preference as a result of the algorithm we used to assign the level of preference, but also because
of a tendency to continue doing what is familiar. Theoretically, increased experience with a
specific drug would improve patient selection, and therefore is associated with improved benefit,
decreased harm and decreased risk of discontinuation in these patients. Lastly, a physician’s
stated preference for a particular drug has been shown to correlate with patient preference. The
selection of a specific TNF antagonist mostly depends on physician and/or patient preference
Page 19 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 21
For peer review only
20
because the benefit and harm profiles of the three drugs are considered to be similar (38-40)
despite limited relative effectiveness data. Physician and patient preferences for an individual
TNF antagonist are likely to be correlated, as was showed in studies of NSAID treatment in RA
of RA patients (41,42), although this was not studied in patients treated with TNF antagonists. In
addition, 67% of 77 Canadian rheumatologists surveyed reported concordance with the patient
preference more than 80% of the time (43). If physician and patient preference for TNF
antagonist do correlate, then improved persistence may have been a result of higher patient rather
than physician preference.
Our findings suggest that administrative restrictions alone, without a change in physicians’
preference and beliefs may not achieve the desired effect on practice. A change in drug coverage
policy, for example, may achieve the desired effect in short-term, since physicians would follow
the new policy and prescribe mainly the reimbursed drug. In the longer term, however, unless a
change in physician preference is achieved, the recommended drug would be discontinued early,
and patients would be switched to a second-line drug. As a result, the overall effect of a change
in policy alone would be less than desired.
Unanswered questions and future research
We theorize that in other conditions, when treating with a therapeutic class in which drugs are
considered to have similar benefits and harms, a similar association exists. Further study of these
conditions is required. In addition, we suggest exploring the explanation we suggested for the
association observed, such as measuring cognitive dissonance during prescribing decision (37) or
patient preference. Awareness of the possible role of cognitive dissonance in clinical decision-
making, based on beliefs not necessarily supported by clinical evidence, can contribute to the
Page 20 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 22
For peer review only
21
development of education programs for physicians. Further research is warranted to identifying
effective policy implementation approaches.
CONCLUSIONS
Higher physician preference, estimated using their prescribing habits, was associated with
decreased discontinuation risk in RA patients treated with TNF antagonists. This finding
highlights the limitation of introducing a new drug coverage policy without encouraging change
in physicians’ drug preference. Similar research on other treatments is warranted.
Page 21 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 23
For peer review only
22
Contributors: All authors contributed to the study design and interpreted the data, drafted the
article or revised it critically for important intellectual content, and gave final approval of the
version to be published. Analysis was performed as part of research conducted under the
supervision of CD as part of the requirements for AF’s PhD project.
Funding: The authors declare (1) The study was supported by the University of British Columbia
Graduate Fellowship and a grant to the University of British Columbia from the British
Columbia Ministry of Health. The publication of study results was not contingent on their
approval; (2) no financial relationships with commercial entities that might have an interest in
the submitted work; (3) no spouses, partners, or children with relationships with commercial
entities that might have an interest in the submitted work; and (4) no non-financial interests that
may be relevant to the submitted work.
Ethical approval: The study protocol was approved by the Clinical Research Ethics Board of the
University of British Columbia (number H11-00303).
Data sharing: patient level data available from Population Data BC (https://www.popdata.bc.ca/).
Patient consent was not obtained but the presented data are anonymised and risk of identification
is low. Statistical code available from the corresponding author.
Copyrights: The Corresponding Author has the right to grant on behalf of all authors and does
grant on behalf of all authors, a worldwide licence to the Publishers and its licensees in
perpetuity, in all forms, formats and media (whether known now or created in the future), to i)
publish, reproduce, distribute, display and store the Contribution, ii) translate the Contribution
into other languages, create adaptations, reprints, include within collections and create
Page 22 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 24
For peer review only
23
summaries, extracts and/or, abstracts of the Contribution, iii) create any other derivative work(s)
based on the Contribution, iv) to exploit all subsidiary rights in the Contribution, v) the inclusion
of electronic links from the Contribution to third party material where-ever it may be located;
and, vi) licence any third party to do any or all of the above.
Transparency: the lead author affirms that the manuscript is an honest, accurate, and transparent
account of the study being reported; that no important aspects of the study have been omitted;
and that any discrepancies from the study as planned have been explained.
Page 23 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 25
For peer review only
24
REFERENCES
(1) Cush JJ. Biological drug use: US perspectives on indications and monitoring. Ann Rheum
Dis 2005 Nov;64(Suppl 4):iv18-iv23.
(2) Curtis JR, Chen L, Harrold LR, Narongroeknawin P, Reed G, Solomon DH. Physician
preference motivates the use of anti-tumor necrosis factor therapy independent of clinical disease
activity. Arthritis Care Res 2010 Jan 15;62(1):101-107.
(3) Schneeweiss S, Glynn RJ, Avorn J, Solomon DH. A Medicare database review found that
physician preferences increasingly outweighed patient characteristics as determinants of first-
time prescriptions for COX-2 inhibitors. J Clin Epidemiol 2005 Jan;58(1):98-102.
(4) Solomon DH, Schneeweiss S, Glynn RJ, Levin R, Avorn J. Determinants of selective
cyclooxygenase-2 inhibitor prescribing: Are patient or physician characteristics more important?
Am J Med 2003 Dec 15;115(9):715-720.
(5) Zhang J, Shan Y, Reed G, Kremer J, Greenberg JD, Baumgartner S, et al. Thresholds in
disease activity for switching biologics in rheumatoid arthritis patients: Experience from a large
US cohort. Arthritis Care Res 2011 Dec;63(12):1672-1679.
(6) Sokka T, Pincus T. Eligibility of patients in routine care for major clinical trials of anti-tumor
necrosis factor alpha agents in rheumatoid arthritis. Arthritis Rheum 2003 Feb;48(2):313-318.
(7) Zink A, Strangfeld A, Schneider M, Herzer P, Hierse F, Stoyanova-Scholz M, et al.
Effectiveness of tumor necrosis factor inhibitors in rheumatoid arthritis in an observational
Page 24 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 26
For peer review only
25
cohort study: Comparison of patients according to their eligibility for major randomized clinical
trials. Arthritis Rheum 2006 Nov;54(11):3399-3407.
(8) Kievit W, Fransen J, Oerlemans AJ, Kuper HH, van der Laar MA, de Rooij DJ, et al. The
efficacy of anti-TNF in rheumatoid arthritis, a comparison between randomised controlled trials
and clinical practice. Ann Rheum Dis 2007 Nov;66(11):1473-1478.
(9) Berthelot JM, Benoist-Gerard S, le Goff B, Muller-Chevalet F, Maugars Y. Outcome and
safety of TNFalpha antagonist therapy in 475 consecutive outpatients (with rheumatoid arthritis
or spondyloarthropathies) treated by a single physician according to their eligibility for clinical
trials. Joint Bone Spine 2010 Dec;77(6):564-569.
(10) Lacaille D, Anis AH, Guh DP, Esdaile JM. Gaps in care for rheumatoid arthritis: A
population study. Arthritis Rheum 2005 Apr 15;53(2):241-248.
(11) Lacaille D, Guh DP, Abrahamowicz M, Anis AH, Esdaile JM. Use of nonbiologic disease-
modifying antirheumatic drugs and risk of infection in patients with rheumatoid arthritis.
Arthritis Rheum 2008 Aug;59(8):1074-1081.
(12) Li X, Guh D, Lacaille D, Esdaile J, Anis AH. The impact of cost sharing of prescription
drug expenditures on health care utilization by the elderly: Own- and cross-price elasticities.
Health Policy 2007 Aug;82(3):340-347.
(13) British Columbia Ministry of Health. Your level of Fair PharmaCare coverage in Fair
PharmaCare. 2011; Available at:
http://www.health.gov.bc.ca/pharmacare/plani/planiindex.html#7. Accessed March, 2011.
Page 25 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 27
For peer review only
26
(14) Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic
comorbidity in longitudinal studies: Development and validation. J Chronic Dis 1987
May;40(5):373-383.
(15) Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC, et al. Coding algorithms
for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005
Nov;43(11):1130-1139.
(16) Bensen B, Atkins C, Sheriff M, Kelsall J, Khalil H. Profile of patients with rheumatoid
arthritis treated with infliximab in Canada-trends toward less DMARD use prior to a biologic,
earlier use of infliximab and differences in baseline disease [abstract]. J Rheumatol 2011
Jun;38(6):1155.
(17) Buchanan J, Rahman MU, Doyle MK, Hsia EC, Gathany T, Parasuraman S, et al. Historical
trends of patient characteristics in anti-TNF clinical trials for rheumatoid arthritis: An analysis of
the literature over the past 16 years [abstract]. Arthritis Rheum 2009 (first published online: Mar
2010);60(Supple 10):1012.
(18) Hyrich KL, Watson KD, Lunt M, Symmons DP, on behalf of the British Society for
Rheumatology Biologics Register (BSRBR). Changes in disease characteristics and response
rates among patients in the United Kingdom starting anti-tumour necrosis factor therapy for
rheumatoid arthritis between 2001 and 2008. Rheumatology (Oxford) 2011 Jul;50(1):117-123.
(19) Lee SJ, Chang H, Yazici Y, Greenberg JD, Kremer JM, Kavanaugh A. Utilization trends of
tumor necrosis factor inhibitors among patients with rheumatoid arthritis in a United States
observational cohort study. J Rheumatol 2009 Aug;36(8):1611-1617.
Page 26 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 28
For peer review only
27
(20) Lee EW, Wei LJ, Amato DA. Cox-type tegression analysis for large numbers of small
groups of correlated failure time observations. In: Klein JP, Goel PK, editors. Survival analysis:
State of the art. NATO ASI series. Series E, Applied sciences ed. Boston, Massachusetts, U.S.:
Kluwer Academic Publishers; 1992. p. 237-247.
(21) Gui-shuang Y, Chengcheng L. Statistical analysis of clustered data using SAS® system
(NESUG 2006). 2006; Available at: http://www.nesug.org/proceedings/nesug06/an/da01.pdf.
Accessed November, 2010.
(22) Gharibvand L, Liu L. Analysis of Survival Data with Clustered Events (paper 237-2009).
2009; Available at: http://support.sas.com/resources/papers/proceedings09/237-2009.pdf;.
Accessed November, 2010.
(23) eCPS (online Compendium of Pharmaceuticals and Specialties). Remicade® (infliximab).
2008. Accessed October, 2008.
(24) eCPS (online Compendium of Pharmaceuticals and Specialties). Enbrel® (Etanercept).
2009. Accessed June, 2009.
(25) British Columbia Ministry of Health. Special Authority request; Abatacept/ Adalimumab/
Certolozumab/ Etanercept/ Golimumab/ Infliximab Rheumatoid Arthritis (Initial or Switch).
2011; Available at: https://www.health.gov.bc.ca/exforms/pharmacare/5345fil.pdf. Accessed
May, 2012.
(26) Mark TL, Swait J. Using stated preference and revealed preference modeling to evaluate
prescribing decisions. Health Econ 2004 Jun;13(6):563-573.
Page 27 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 29
For peer review only
28
(27) Wolfe F. The epidemiology of drug treatment failure in rheumatoid arthritis. Baillieres Clin
Rheumatol 1995 Nov;9(4):619-632.
(28) Hyrich KL, Lunt M, Watson KD, Symmons DPM, Silman AJ. Outcomes after switching
from one anti-tumor necrosis factor alpha agent to a second anti-tumor necrosis factor alpha
agent in patients with rheumatoid arthritis: Results from a Large UK National Cohort Study.
Arthritis Rheum 2007 Jan;56(1):13-20.
(29) Carmona L, Gomez-Reino JJ, on behalf of the BIOBADASER Group. Survival of TNF
antagonists in spondylarthritis is better than in rheumatoid arthritis. Data from the Spanish
registry BIOBADASER. Arthritis Res Ther 2006 Apr;8(3):R72.
(30) Hetland ML, Christensen IJ, Tarp U, Dreyer L, Hansen A, Hansen IT, et al. Direct
comparison of treatment responses, remission rates, and drug adherence in patients with
rheumatoid arthritis treated with adalimumab, etanercept, or infliximab: Results from eight years
of surveillance of clinical practice in the nationwide Danish DANBIO registry. Arthritis Rheum
2010 Jan;62(1):22-32.
(31) Fernandez-Nebro A, Irigoyen MV, Urena I, Belmonte-Lopez MA, Coret V, Jimenez-Nunez
FG, et al. Effectiveness, predictive response factors, and safety of anti-tumor necrosis factor
(TNF) therapies in anti-TNF-naive rheumatoid arthritis. J Rheumatol 2007 Dec;34(12):2334-
2342.
(32) Kievit W, Fransen J, Adang EMM, den Broeder AA, Bernelot Moens HJ, Visser H, et al.
Long-term effectiveness and safety of TNF-blocking agents in daily clinical practice: Results
Page 28 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 30
For peer review only
29
from the Dutch Rheumatoid Arthritis Monitoring Register. Rheumatology (Oxford) 2011
Jan;50(1):196-203.
(33) Filippini M, Bazzani C, Favalli EG, Marchesoni A, Atzeni F, Sarzi-Puttini P, et al. Efficacy
and safety of anti-tumour necrosis factor in elderly patients with rheumatoid arthritis: An
observational study. Clin Rev Allergy Immunol 2010 Apr;38(2-3):90-96.
(34) Baser O, Wang L, Xie L, Dysinger A, Gust C, Yuce H, et al. Deriving doctors' prescribing
patterns from health care claims: An instrumental variable analysis [abstract]. Value Health 2010
November;13(7):A425.
(35) Kamal KM, Madhavan SS, Hornsby JA, Miller L, Kavookjian J, Scott V. Use of tumor
necrosis factor inhibitors in rheumatoid arthritis: A national survey of practicing United States
rheumatologists. Joint Bone Spine 2006 Dec;73(6):718-724.
(36) Festinger L. A theory of cognitive dissonance. Evanston, Ill.: Row, Peterson; 1957.
(37) Visser PS, Cooper J. Dissonance and attitude change: A matter of measurement, in Chapter
10: Attitude change. In: Hogg MA, Cooper J, editors. The Sage handbook of social psychology
Thousand Oaks, California, U.S.: Sage; 2003. p. 220.
(38) Yazici Y, Simsek I. Treatment options for rheumatoid arthritis beyond TNF-alpha
inhibitors. Expert Rev Clin Pharmacol 2010 Sep;3(5):663-666.
(39) Dudler J, Moller B, Michel BA, Villiger PM. Biologics in rheumatoid arthritis (RA) -
Recommendations for Swiss practice. Swiss Med Wkly 2011(w13189).
Page 29 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 31
For peer review only
30
(40) Lu TY, Hill C. Managing patients taking tumour necrosis factor inhibitors. Australian
Prescriber 2006 Jun;29(3):67-70.
(41) Gall EP, Caperton EM, McComb JE, Messner R, Multz CV, O'Hanlan M, et al. Clinical
comparison of ibuprofen, fenoprofen calcium, naproxen and tolmetin sodium in rheumatoid
arthritis. J Rheumatol 1982 May-Jun;9(3):402-407.
(42) Wasner C, Britton MC, Kraines RG, Kaye RL, Bobrove AM, Fries JF. Nonsteroidal anti-
inflammatory agents in rheumatoid arthritis and ankylosing spondylitis. JAMA 1981 Nov
13;246(19):2168-2172.
(43) Choquette D, Shaikh SA, Abu-Hakima M. Rheumatoid Arthritis Patient Preference Survey
(RAPPS) (Presentation number: 965). 2006; Available at:
http://acr.confex.com/acr/2006/webprogram/Paper5515.html;. Accessed December, 2011.
Page 30 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 32
For peer review only
31
Figure 1: Calculating the value of the independent variable physician preference
Figure 2: Study patients’ flaw
Figure 3: Persistence with TNF antagonists by physician preference levels
Page 31 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 33
For peer review only
Calculating the value of the independent variable physician preference 136x174mm (300 x 300 DPI)
Page 32 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 34
For peer review only
Study patients’ flaw
170x98mm (300 x 300 DPI)
Page 33 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 35
For peer review only
Persistence with TNF antagonists by physician preference levels 40x30mm (300 x 300 DPI)
Page 34 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 36
For peer review only
1
Prescriber Preference for a Particular Tumor Necrosis Factor Antagonist Drug and
Treatment Discontinuation: Population-based Cohort
STROBE Statement—Checklist of items that should be included in reports of cohort studies
Item
No Recommendation
Page
number
Title and abstract 1 (a) Indicate the study’s design with a commonly used term in the title or
the abstract
2
(b) Provide in the abstract an informative and balanced summary of
what was done and what was found
2-3
Introduction
Background/rationale 2 Explain the scientific background and rationale for the investigation
being reported
4-5
Objectives 3 State specific objectives, including any prespecified hypotheses 5
Methods
Study design 4 Present key elements of study design early in the paper 5
Setting 5 Describe the setting, locations, and relevant dates, including periods of
recruitment, exposure, follow-up, and data collection
5-6
Participants 6 (a) Give the eligibility criteria, and the sources and methods of selection
of participants. Describe methods of follow-up
5-6
(b) For matched studies, give matching criteria and number of exposed
and unexposed
n/a
Variables 7 Clearly define all outcomes, exposures, predictors, potential
confounders, and effect modifiers. Give diagnostic criteria, if applicable
5-11
Data sources/
measurement
8* For each variable of interest, give sources of data and details of
methods of assessment (measurement). Describe comparability of
assessment methods if there is more than one group
5-11
Bias 9 Describe any efforts to address potential sources of bias 8-11
Study size 10 Explain how the study size was arrived at 11
Quantitative variables 11 Explain how quantitative variables were handled in the analyses. If
applicable, describe which groupings were chosen and why
5-11
Statistical methods 12 (a) Describe all statistical methods, including those used to control for
confounding
5-11
(b) Describe any methods used to examine subgroups and interactions 11
(c) Explain how missing data were addressed n/a
(d) If applicable, explain how loss to follow-up was addressed 10
(e) Describe any sensitivity analyses 7,11
Results
Participants 13* (a) Report numbers of individuals at each stage of study—eg numbers
potentially eligible, examined for eligibility, confirmed eligible,
included in the study, completing follow-up, and analysed
11, figure
2
(b) Give reasons for non-participation at each stage Figure 2
(c) Consider use of a flow diagram Figure 2
Descriptive data 14* (a) Give characteristics of study participants (eg demographic, clinical,
social) and information on exposures and potential confounders
11-12,
table 2
(b) Indicate number of participants with missing data for each variable
of interest
n/a
Page 35 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 37
For peer review only
2
(c) Summarise follow-up time (eg, average and total amount) 14, figure
3
Outcome data 15* Report numbers of outcome events or summary measures over time Figure 3
Main results 16 (a) Give unadjusted estimates and, if applicable, confounder-adjusted
estimates and their precision (eg, 95% confidence interval). Make clear
which confounders were adjusted for and why they were included
Table 3
(b) Report category boundaries when continuous variables were
categorized
Table 1
(c) If relevant, consider translating estimates of relative risk into
absolute risk for a meaningful time period
n/a
Other analyses 17 Report other analyses done—eg analyses of subgroups and interactions,
and sensitivity analyses
14-15,
table 3
Discussion
Key results 18 Summarise key results with reference to study objectives 16
Limitations 19 Discuss limitations of the study, taking into account sources of potential
bias or imprecision. Discuss both direction and magnitude of any
potential bias
16-17
Interpretation 20 Give a cautious overall interpretation of results considering objectives,
limitations, multiplicity of analyses, results from similar studies, and
other relevant evidence
18-20
Generalisability 21 Discuss the generalisability (external validity) of the study results 16-17, 20
Other information
Funding 22 Give the source of funding and the role of the funders for the present
study and, if applicable, for the original study on which the present
article is based
22
*Give information separately for exposed and unexposed groups.
Note: An Explanation and Elaboration article discusses each checklist item and gives methodological background and
published examples of transparent reporting. The STROBE checklist is best used in conjunction with this article (freely
available on the Web sites of PLoS Medicine at http://www.plosmedicine.org/, Annals of Internal Medicine at
http://www.annals.org/, and Epidemiology at http://www.epidem.com/). Information on the STROBE Initiative is
available at http://www.strobe-statement.org.
Page 36 of 36
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 38
For peer review only
Prescriber Preference for a Particular Tumor Necrosis Factor Antagonist Drug and Treatment Discontinuation:
Population-based Cohort
Journal: BMJ Open
Manuscript ID: bmjopen-2014-005532.R1
Article Type: Research
Date Submitted by the Author: 10-Sep-2014
Complete List of Authors: Fisher, Anat; University of British Columbia, Anesthesiology, Pharmacology & Therapeutics Basset, Ken; University of British Columbia, Anesthesiology, Pharmacology
& Therapeutics; University of British Columbia, Family Practice Wright, James (Jim); University of British Columbia, Anesthesiology, Pharmacology & Therapeutics; University of British Columbia, Medicine Brookhart, M.; University of North Carolina at Chapel Hill, Epidemiology Freeman, Hugh; University of British Columbia, Medicine Dormuth, Colin; University of British Columbia, Department of Anesthesiology, Pharmacology and Therapeutics
<b>Primary Subject Heading</b>:
Epidemiology
Secondary Subject Heading: Epidemiology, Health services research, Patient-centred medicine, Research methods, Rheumatology
Keywords: EPIDEMIOLOGY, RHEUMATOLOGY, Health policy < HEALTH SERVICES
ADMINISTRATION & MANAGEMENT
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open on June 29, 2020 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2014-005532 on 30 S
eptember 2014. D
ownloaded from
Page 39
For peer review only
1
Prescriber Preference for a Particular Tumor Necrosis
Factor Antagonist Drug and Treatment Discontinuation:
Population-based Cohort
Anat Fisher, MD, PhD; Ken Bassett, MD, PhD; James M. Wright, MD, PhD; M. Alan
Brookhart, PhD; Hugh Freeman, MD; Colin R. Dormuth, ScD
Corresponding Author: Anat Fisher, MD, PhD; Department of Anesthesiology, Pharmacology
and Therapeutics, University of British Columbia, 2176 Health Sciences Mall, Vancouver, BC,
Canada V6T 1Z3, Telephone: 604-822-0700 Fax: 604-822-0701, Email: [email protected]
Ken Bassett: Department of Anesthesiology, Pharmacology & Therapeutics and Department of
Family Practice, University of British Columbia, Vancouver, Canada
James M. Wright: Department of Anesthesiology, Pharmacology & Therapeutics and
Department of Medicine, University of British Columbia, Vancouver, Canada
M. Alan Brookhart: Department of Epidemiology, Gillings School of Global Public Health,
University of North Carolina, Chapel Hill, North Carolina, USA
Hugh Freeman: Department of Medicine, University of British Columbia, Vancouver, Canada
Colin R. Dormuth: Department of Anesthesiology, Pharmacology & Therapeutics, University of
British Columbia, Vancouver, Canada
Keywords: physician prescribing pattern, rheumatoid arthritis, medication persistence, Tumor
Necrosis Factor-alpha
Number of words (excluding title page, abstract, references, figures and tables, endnotes): 2581
Page 1 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 40
For peer review only
2
ABSTRACT
Objective: To assess the effect of physician preference for a particular tumor necrosis factor
alpha (TNF) antagonist on the risk of treatment discontinuation in rheumatoid arthritis.
Design: Population-based cohort study.
Setting: British Columbia administrative health data (inpatients, outpatients and pharmacy)
Participants: 2,742 British Columbia residents who initiated a first course of a TNF antagonist
between 2001 and December 2008, had been diagnosed with rheumatoid arthritis, and were
treated by one of 58 medium to high-volume prescribers.
Independent variable: A level of physician preference for the drug (higher or lower) was
assigned based on preceding prescribing records of the care-providing physician. Higher
preference was defined as at least 60% of TNF antagonist courses initiated in the preceding year.
Sensitivity analysis was conducted with different thresholds for higher preference.
Main Outcome Measure: Drug discontinuation was defined as a drug-free interval of 180 days
or switching to another TNF antagonist, anakinra, rituximab or abatacept. The risk of
discontinuation was compared between different levels of physician preference using survival
analysis.
Results: Higher preference for the prescribed TNF antagonist was associated with improved
persistence with the drug (4.28 years [95% confidence interval 3.70-4.90] versus 3.27 [2.84-
3.84], with log rank test p-value of 0.017). The adjusted hazard ratio for discontinuation was
Page 2 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 41
For peer review only
3
significantly lower in courses of drugs with higher preference (0.85 [0.76-0.96]). The results
were robust in a sensitivity analysis.
Conclusions: Higher physician preference was associated with decreased risk of discontinuing
TNF antagonists in patients with rheumatoid arthritis. This finding suggests that physicians who
strongly prefer a specific treatment help their patients to stay on treatment for a longer duration.
Similar research on other treatments is warranted.
Strengths and Limitations
• First study to explore within-physician variation in prescribing habits, specifically the
effect of prescriber preference to a drug on the decision to discontinue the drug
• The universal nature of the Canadian health care system and a systematic and
standardized approach to data collection in British Columbia, which ensured the
generalizability of our results, as well as the large sample and prolonged follow up
• To conquer the absence of access to clinical data, we used multiple proxy variables to
adjust for disease severity
• Physician preference was not directly measured but instead based on previous prescribing
habits.
Page 3 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 42
For peer review only
4
INTRODUCTION
The term ‘physician preference’ usually refers to favoring a particular drug or a therapeutic
group among several alternatives, and it has been shown to predict treatment choice.1-4 In studies
of administrative health (claim) data, this preference is often determined by identifying
dispensing of drug prescribed by the specific physician in a pre-determined period, prior to the
event of interest (a new prescribing). Despite an association with new prescribing decisions, the
role of physician preference in treatment discontinuation has not been studied. Recently, the term
‘preference’ has also been used to describe a second phenomenon – in the context of treatment
discontinuation, it was used to describe the baseline risk of discontinuing treatment in patients
treated by a specific physician (the physician ‘preference for discontinuation’). 5 This baseline
risk may differ among physicians because physicians may respond differently to similar clinical
situations such as decreased benefit or harmful events. They could recommend patients to
discontinue treatment (with or without switching to a second drug) or to persist with the
treatment (but to adjust dose, add-on a second drug or be under frequent watch). In this paper, we
use the term ‘preference’ to describe the first phenomenon (physician’s favorite drug) and
‘physician-specific discontinuation risk’ to describe the second.
Treatment with tumor necrosis factor alpha (TNF) antagonists in patients with rheumatoid
arthritis (RA) was considered especially sensitive to physician preference for two main reasons.
First, during the study period (2001-2009) there was limited clinical evidence on the comparative
effectiveness of the drugs, mainly due to the absence of head-to-head randomized clinical trials,
but also because participants in placebo-controlled trials were not representative of patients treated
Page 4 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 43
For peer review only
5
in routine clinical settings.6-9 Second, published indications for discontinuation of TNF
antagonists were vague and confusing, and therefore care-providing physicians could reasonably
be expected to reach different clinical decisions given the same clinical situation. Consequently,
the decisions about which TNF antagonist to prescribe first and when to discontinue treatment
were likely subject to physicians’ individual preference.
This study analyzed data of first courses of a TNF antagonist in British Columbia patients with
RA. The prescriber recorded on the first dispensing claim for a TNF antagonist was used as a
proxy of the care-providing physician. The study objective was to estimate the effect of
physician preference on the risk of discontinuation. The null hypothesis tested was that physician
preference for a TNF antagonist when treatment has been initiated does not influence the risk of
discontinuing the treatment in RA patients.
PATIENTS AND METHODS
The study cohort was identified using four British Columbia Ministry of Health administrative
databases: PharmaNet (prescription dispensing data), Medical Service Plan (MSP) registration
information (demographic data), MSP Payment Information (fee-for-service payments to
physicians and alternative providers), and the Discharge Abstract Database (hospital
separations). The databases were linkable using a de-identified patient and physician numbers.
Follow-up data were available from 1995 until December 31, 2009. The study cohort included
British Columbia residents who fulfilled all of three conditions: (1) first exposure to a TNF
antagonist between March 1, 2001 and December 31, 2008; (2) a diagnosis of RA; and (3) TNF
antagonist treatment initiated by a medium or high-volume prescriber, defined as a prescriber
Page 5 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 44
For peer review only
6
who initiated 5 or more courses of TNF antagonists in RA patients during the years 2001-2008.
Exposure to a TNF antagonist was based on one or more recorded dispensing claims for
infliximab, adalimumab or etanercept between March 2001 and December 31, 2008. We
excluded patients who were previously treated with anakinra, rituximab or abatacept to decrease
heterogeneity in the population analyzed. The first dispensing claim for a TNF antagonist was
used to identify the study drug (infliximab, adalimumab or etanercept), index date and prescriber
for each patient in the cohort. We defined a pre-study period of three years preceding the index
date during which patients were required to have continuous provincial medical service
coverage. A gap shorter than 30 days was not considered to be an interruption in coverage. The
pre-study period ensured a standard run-in period of at least three years without TNF antagonist
exposure and a standard period during which diagnosis of RA was identified. RA patients were
selected based on similar criteria to previous studies in British Columbia.10-12 either two
outpatient visits in physician clinics with a diagnosis code of RA (ICD-9 714) at least 60 days
apart, or one hospitalization with recorded discharge diagnosis of RA. Patients were excluded if
sex or date of birth was not available, if they had a concurrent diagnosis of Crohn’s disease
(based on at least one outpatient or inpatient diagnosis code in the pre-study period) or if they
were younger than 18 years at the index date (to remove patients with juvenile RA). We also
excluded patients initiated on TNF antagonist treatment by prescribers who cumulatively
initiated less than five patients in the RA cohort (low-volume prescribers).
Physician preference was determined for each patient at the index date and coded as a Bernoulli
variable. It reflects prescribing patterns of the individual physician (who started this course)
during the preceding year (Figure 1). For each individual patient, we determined the date of
course initiation, the prescriber recorded in the first dispensing event and the drug (infliximab,
Page 6 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 45
For peer review only
7
adalimumab or etanercept). From first dispensing records for each patient in the study cohort, we
identified other patients that received any TNF antagonists from the same prescriber during the
year preceding the index date of the current patient, and the drug they were treated with. Next we
calculated the proportion of these patients who received the same drug as the patient of interest.
Based on this proportion, we categorized the prescriber preference to ‘higher’ when the
calculated proportion for this drug and that prescriber was 60% or higher, and ‘lower’ otherwise.
We expected that this threshold would provide reasonable assurance that a course categorized as
having ‘higher” preference for drug actually reflects that the prescriber favored that particular
TNF antagonist at this time, even if only two TNF antagonists were available. We reassessed
preference for drug based on annual data to allow for changes in preference over time and for
accommodating the more recent availability of adalimumab. Adalimumab was available in
Canada since October 2004, while infliximab and etanercept were on the market since 2001. By
including a year of data, differently from other studies that used only the last prescription in
assigning preference 1-4, we created a more accurate indicator. This way we minimized the effect
of factors not related to preference that might have influenced a specific prescribing decision.
Sensitivity analysis was conducted to examine the robustness of main results, using thresholds of
70% or 80% for the level of physician preference.
Multiple covariates, mainly patient characteristics that may influence drug selection and/or
discontinuation were included in the final models and are listed in Table 1. We were limited by
lack of access to clinical data and inability to adjust for clinical variables not captured in
administrative health data, such as disease severity. Hence, we used multiple proxies, including
disease duration, age, extra-articular involvement, antirheumatic drug use, health services
utilization and consumption of pain medications to adjust for disease severity. In order to control
Page 7 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 46
For peer review only
8
for secular trends and late availability of adalimumab in Canada, we included a categorical
variable for the year of treatment initiation.
Table 1 - List of covariates included in the final models
Variable Description
Demographics
Sex Bernoulli variable
Age at index Four categories: 18-29 years, 30-69, 70-79 and ≥80 (categories were
assigned based on similar discontinuation risk in preliminary
analysis, which was not linear)
The annual deductible for
prescription cost at index
(Fair PharmaCare)
The deductible is based on annual income.13 Six categories: annual
deductible of $0, $1-500, $501-2250, >$2250, other programs and
no coverage
Geographical area at
index
Based on first 3 digit of postal code. Five mutually exclusive
categories: Greater Vancouver, Greater Victoria, Vancouver Island,
urban areas and rural areas
Clinical status
Physicians encounters in
the year preceding the
index date
Continuous variable
Number of inpatients
admissions in the year
preceding the index date
Four categories: 0, 1, 2, >2
Comorbidities (presence
and severity) during the
pre-study period
Charlson comorbidity score14 was determined using Quan’s
algorithm for administrative databases,15 excluding rheumatic
diseases. At least two outpatient or one inpatient encounter with the
diagnosis were required. Four categories: 0, 1, 2, >2
Disease duration Measured from the first diagnosis of RA (inpatient or outpatient) in
Page 8 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 47
For peer review only
9
Variable Description
the data. RA disease duration captured in Canadian administrative
data has been found to agree with reported duration by physicians.16
Continuous variable
The presence of extra-
articular manifestation
during the pre-study
period
Based on recorded at least one diagnoses with ICD-9 codes 3571,
3596, 7141, 71481, 7142 (outpatient) or ICD-10 codes G636, G737,
I39, I418, J990, M050, M051, M052, M053 (inpatient). Bernoulli
variable
Drug therapies
Concomitant MTX
during 200 days
preceding the index date
Based on mean plus two standard deviations of between-dispensing
intervals of MTX in the study cohort. Bernoulli variable
Dispensing claims for
NSAIDs during the year
preceding the index date
Bernoulli variable
Number of different
antirheumatic drugs
dispensed in the pre-
study period
Dispensing claims of 10 drugs were included: MTX,
hydroxychloroquine, sulfasalazine, leflunomide, azathioprine,
minocycline, penicillamine, sodium aurothiomalate, prednisone, and
intra-articular triamanolone or methylprednisolone. Four categories:
no drug, 1-2, 3-6, >6 different drugs
Other
Calendar year at index
date
The variable allowed controlling for secular trends in clinical
practice17-20 and availability of drugs. Eight yearly categories were
included for the years 2001-2008
ICD-10- International Classification of Diseases, 10th edition; ICD-9-CM - International Classification of Diseases, 9th edition, clinical modification; MTX – methotrexate; NSAID – nonsteroidal anti-inflammatory drug
The outcome variable was discontinuation defined as either switching to another TNF antagonist
including certolizumab and golimumab, anakinra, rituximab or abatacept (‘biologic’
Page 9 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 48
For peer review only
10
antirheumatic drugs), or a drug-free interval of 180 days after exhaustion of the dispensed days-
supply of the latest refill. Discontinuation date was set to the end of the days-supply of the last
refill before the 180-days drug-free interval or the date of the first dispensing of a second
‘biologic’ antirheumatic drug, whichever was earliest. Unless discontinuation occurred, patients
were followed up until either December 31, 2009 (end of follow-up period) or an interruption of
more than six days in the provincial MSP coverage, at which point data were considered
censored. The most common causes of coverage interruptions were death and emigration from
the province. Change of dose of the TNF antagonist or addition of a second drug (not TNF
antagonist) was not considered discontinuation.
Discrepancies between recorded days-supply and dispensed quantity were common in the
PharmaNet database; hence, we also calculated days-supply based on the quantity dispensed,
which, if required, was imputed using the recorded total cost. Cost was considered the most
accurate and reliable field, since this field serves for claim and payment processing. We used the
longest duration of days-supply, recorded or calculated, to determine both the length of drug-free
intervals and a discontinuation date.
Sample size calculation
Assuming 1:1 ratio of courses with ‘higher’ preference versus courses with ‘lower’ preference
and in the absence of prior estimates, if the true hazard ratio for discontinuing courses with
‘higher’ preference relative to courses with ‘lower’ preference is 0.8, we required 847 courses at
each preference level to be able to reject the null hypothesis with type I error probability of 0.05
and power of 0.80.
Page 10 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 49
For peer review only
11
Statistical Analysis
Patients’ characteristics were compared across the three drugs. The adjusted risk of
discontinuation was compared using two different approaches in a Cox proportional hazards
regression. First, we conducted an analysis of nonclustered data. The multivariate model also
included a series of Bernoulli variables, one for each physician (1=the particular physician;
0=other physicians), which allowed individualization of the risk of drug discontinuation by
physician (physician-specific discontinuation risk). In the second approach, a marginal model of
clustered data allowed an adjustment for possible correlation between patients treated by the
same physician.21-23 We tested for model assumptions (proportional hazard and absence of
interactions) and found them to be valid. We also checked for the linearity of continuous
variables and categorized non-linear variables. All statistical tests were two-sided. All
calculations were performed using the SAS software package (SAS Institute Inc., Cary, NC).
RESULTS
The study cohort included 2,742 RA patients prescribed by a total of 58 medium-high volume
physicians. Figure 2 presents reasons for excluding patients who used TNF antagonists from the
analysis. Baseline characteristics across users of the three drugs are presented in Table 2. Not
only was etanercept was the most frequently prescribed drug (1718 patients, 63%), but it was
also the only drug prescribed by all 58 study physicians. Etanercept was usually initiated by a
physician with high preference for etanercept (70% of etanercept courses). Infliximab or
adalimumab, on the other hand, were usually initiated by physicians with lower preference for
these drugs (only 34% or 19% of courses were initiated by a physician with high preference for
Page 11 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 50
For peer review only
12
drug, respectively). Patients treated with adalimumab were significantly older and had a lower
income (reflected in lower annual deductible level for prescribing costs). Patients treated with
infliximab had the highest prevalence of concomitant methotrexate therapy and patients with
etanercept the lowest. This reflects differences in the indications mentioned in the product
monographs; while infliximab is indicated for use in combination with methotrexate,24 the
monograph of etanercept mentioned that it “can be initiated in combination with methotrexate …
or used alone”.25 Similarly, while the provincial special authority policy requires that infliximab
is used in combination with methotrexate (or other drug), such requirement does not exist for
treatment with etanercept or adalimumab.26
Page 12 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 51
For peer review only
13
Table 2 - Baseline characteristics
Variable Infliximab Adalimumab Etanercept P-value
(between
groups
differences)
Number of patients (% from
cohort)
571 (21) 453 (16) 1718 (63)
Number of prescribers (based on
first dispensing event for each
patient) (total 58 prescribers)
49 46 58
Patients treated with a drug with
higher preference, No (%)
193 (34) 84 (19) 1198 (70) <0.0001
Demographics
Females, No (%) 403 (71) 326 (72) 1239 (72) 0.77
Age at index, median y (range) 56 (18-87) 58 (22-91) 56 (18-92) 0.003
Annual deductible for prescription
cost, No (%)
Very low ($0) 47 (8.3) 80 (18) 199 (12) <0.0001
Low ($1-500) 33 (5.8) 53 (12) 152 (8.9) 0.004
Medium ($501-2250) 92 (16) 109 (24) 315 (18) 0.004
High (>$2250) 35 (6.1) 38 (8.4) 143 (8.3) 0.22
Residence in Greater
Vancouver/Victoria, No (%)
341 (60) 224 (50) 782 (46) <0.0001
Clinical status
Number of physician visits median
(range)
33 (3-158) 31 (2-112) 32 (3-136) 0.25
At least one admission to hospital
No (%)
104 (18) 63 (14) 340 (20) 0.01
Extra-articular manifestations, No 28 (4.9) 14 (3.1) 60 (3.5) 0.23
Page 13 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 52
For peer review only
14
(%)
Presence of comorbidity (score>0),
No (%)
113 (20) 95 (21) 383 (22) 0.41
RA disease duration median, y
(range)
9.2 (0.1-
17.9)
7.7 (0.3-17.9) 8.0 (0-17.8) 0.21
RA drugs
Concomitant MTX, No (%) 388 (68) 264 (58) 856 (50) <0.0001
Dispensing claims for NSAIDs,
No (%)
307 (54) 214 (47) 923 (54) 0.04
Number of different antirheumatic
drugs, median (range)
4 (0-8) 4 (0-8) 4 (0-9) 0.46
%- percent, unless otherwise specified – percentage from patients treated with this drug; $-Canadian dollars; MTX-methotrexate; No-number of patients; NSAID- nonsteroidal anti-inflammatory drug; NSS – not statistically significant; y-years
For detailed descriptions of the variables refer to Table 1
Persistence with the three TNF antagonists was similar (log rank test p-value=0.15). The product
limit median persistence estimates were 3.9 years (95% confidence interval 3.0-5.3), 3.3 (2.6-4.0)
and 3.9 years (3.4-4.4) for infliximab, adalimumab and etanercept, respectively. Higher physician
preference was associated with improved persistence compared to lower preference (Figure 3),
with median time to discontinuation of 4.28 years (confidence interval 3.70-4.90) in 1475
courses with ‘high’ preference, compared with 3.27 (2.85-3.84) in 1267 courses with ‘low’
preference. In both groups, about 50% of the patients discontinued and the other 50% were
censored.
Higher physician preference was associated with a significant decrease of 14-15% in the adjusted
hazard for discontinuation (Table 3). No significant interaction between drug and preference was
Page 14 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 53
For peer review only
15
observed. The results of sensitivity analysis were similar, with overall adjusted hazard ratio
between 0.85-0.88 in all models. The results of clustered data analyses with thresholds of 60%,
70% and 80% for physician preference were robust (Table 3). The results of nonclustered data
analyses for thresholds of 70% and 80%, on the other hand, did not reach the significance level.
That may be a result of decreased numbers of patients treated with a drug of higher physician
preference and therefore decreased power to detect significant difference.
Table 3 - Hazard ratios for drug discontinuation, higher preference versus lower
preference
Approach
Hazard ratio (95% confidence interval)
Preference threshold
60%
Preference threshold
70%
Preference threshold
80%
Patients with higher
preference N (%)
1426 (52) 1234 (45) 987 (36)
Nonclustered
data analysis
Crude 0.88 (0.79-0.98) 0.89 (0.80-0.99) 0.83 (0.81-0.996)
Adjusted* 0.86 (0.76-0.98) 0.88 (0.77-1.01) 0.87 (0.75-1.004)
Marginal
modeling of
clustered data
Crude 0.88 (0.77-1.001) 0.89 (0.78-1.01) 0.89 (0.79-1.01)
Adjusted# 0.85 (0.76-0.96) 0.87 (0.78-0.97) 0.87 (0.78-0.98)
% - percent from overall patients (2742); N – Number of patients; *Adjusted by drug, calendar year, patient’s demographics and clinical status, and prescriber #Adjusted by drug, calendar year and patient’s demographics and clinical status
Page 15 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 54
For peer review only
16
DISCUSSION
We demonstrated two types of variations of physician decisions to discontinue treatment with
TNF antagonists: between-physician and within-physician. Between-physician variability is
expressed as a different baseline risk of discontinuation depending on the treating physician in
response to similar clinical situations, such as decreased benefit or a harmful event. We used the
term physician-specific discontinuation risk for this phenomenon. It could be a result of
differences in education or experience, adherence to different guidelines and possible differences
in patient case mix. Within-physician variability, on the other hand, is a different response to a
similar clinical situation (similar patients, similar drug effects) by the same physician. In this
study we showed that this variability correlated to physician preference for drug – and
specifically, higher physician preference for a specific TNF antagonist drug was associated with
a decreased risk of discontinuing TNF antagonists in RA patients. This preference most probably
reflected the physician’s beliefs regarding the relative effectiveness and safety of the specific
drug compared to the alternatives, where a preferred drug is thought to be superior. When a
physician prescribed a drug believed to be superior compared to comparators, patients were
encouraged to stay longer on the drug and the risk of discontinuation, as measured in this study,
decreased. We believe that a different response (level of encouragement) to a similar clinical
situation applied in indefinite clinical situations, such as mild harmful effect or questionable
benefit.
Strengths and weaknesses of the study
The main advantages of our study are the universal nature of the Canadian health care system
and a systematic and standardized approach to data collection in British Columbia, which
Page 16 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 55
For peer review only
17
ensured the generalizability of our results, as well as the large sample and prolonged follow up
that increased the power of the study to detect differences in discontinuation risk. The main
disadvantage is the absence of access to clinical data; hence adjustment to patient mix was
difficult. To conquer this problem, we used multiple proxy variables to adjust for disease
severity, such as extra-articular involvement, antirheumatic drug use, health services utilization
and consumption of pain medications.
There are several limitations to ascertaining physician preference based on previous prescribing
habits. First, preference was not defined for all patients. For example, it was not defined for the
first patient treated by the physician, or if the physician did not prescribe this class of drugs in
the preceding year. Second, we assumed that increased loyalty for a particular TNF antagonist
reflects preference, but we did not measure preference directly. Nevertheless, stated preference
and prescribing habits were shown to correlate in previous research.27 Third, we measured
preference at the time of treatment initiation and not at discontinuation, and preference might
have changed over time. Finally, we assumed strong preference for one drug only and regarded
situations in which the physician preferred two of the TNF antagonists to be a situation of
treatment with a drug of lower preference.
Overview of previous studies
Treatment persistence was hypothesized to reflect therapeutic benefit and harm in chronic non-
curable diseases, such as RA.28 In support of this hypothesis, previous studies in RA patients
using TNF antagonists demonstrated that the main reasons for discontinuing or switching were
decreased benefit (36-67% of the discontinuations) or perceived harm (30-58%).29-34 As such, the
risk of discontinuing treatment is assumed to be influenced solitarily by drug properties and
Page 17 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 56
For peer review only
18
patient characteristics. The demonstrated effect of care-provider characteristics, specifically
prescribing habits, on the risk of discontinuation calls into question this hypothesis.
Between-physician differences in response to the same clinical situation (physician-specific
discontinuation risk) have previously been studied in RA patients treated by TNF antagonists.
Differences in response to harm were reported by Cush et al 2005 based on an online survey of
rheumatologists 1 and in response to decreased benefit – by Zhang et al 2011 5. Using mixed
effect model with a random intercept to cluster patients at physician level, Zhang demonstrated
the importance of clustering by physicians, even after adjustment for both baseline disease
activity and improvement in disease activity.
Within-physician variation in prescribing decisions and the effect of physician preference have
infrequently been studied. While the effect of physician preference on discontinuation has not
been previously studied in this therapeutic class or in other conditions, physician preference was
found to be an important predictor of drug selection in treatment initiation. Physician preference
for therapeutic class was the most important determinant in initiating treatment with TNF
antagonists compared to the alternative: prescribing synthetic antirheumatic drugs 1 and in other
treatment situations.3, 4, 35 Physician preference for an individual TNF antagonist was studied by
Kamal et al 2006.36 In response to mailed questionnaires, most American rheumatologists said
they preferred etanercept over adalimumab or infliximab and considered etanercept the most
efficacious of the three drugs with less harm. In our study, etanercept was the most prescribed
TNF drug, and it was prescribed by all 58 medium to high-volume prescribers.
Explanations and interpretation
Page 18 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 57
For peer review only
19
We suggest several possible explanations for the finding of decreased risk of discontinuation
with increased physician preference for the prescribed TNF antagonist. First, the results could be
explained in light of the theory of cognitive dissonance.37 Dissonance is an uncomfortable
feeling caused by holding conflicting ideas simultaneously. In line with this theory, treating with
a drug believed to be inferior (lower preference) induces dissonance. In indefinite clinical
situations, such as mild harmful effect or questionable benefit, early drug discontinuation
supports the physician’s belief that the selected drug was inferior compared to the alternatives.
On the other hand, in a similar situation, a drug with a higher preference might be continued, to
support a belief in its superiority. The interpretation of our results using the theory of cognitive
dissonance is restricted by the lack of direct measurement of physician dissonance at the time of
treatment discontinuation. The indirect measure of dissonance we used, physician preference,
could also be influenced by factors unrelated to prescriber beliefs, such as limited availability of
a drug or higher relative cost (although during the study period, availability or cost were not
factors). Direct measures of cognitive dissonance are complicated and involve activities beyond
the scope of the current study.38 Second, the observed association between higher preference and
decreased risk of discontinuation might be confounded by experience with the drug, namely the
total number of patients a physician has previously treated with the same drug. Increased
experience with a specific drug may be associated with higher preference as a result of the
algorithm we used to assign the level of preference, but also because of a tendency to continue
doing what is familiar. Theoretically, increased experience with a specific drug would improve
patient selection, and therefore is associated with improved benefit, decreased harm and
decreased risk of discontinuation in these patients. Lastly, a physician’s stated preference for a
particular drug has been shown to correlate with patient preference. The selection of a specific
Page 19 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 58
For peer review only
20
TNF antagonist mostly depends on physician and/or patient preference because the benefit and
harm profiles of the three drugs are considered to be similar39-41 despite limited relative
effectiveness data. Physician and patient preferences for an individual TNF antagonist are likely
to be correlated, as was shown in studies of nonsteroidal anti-inflammatory drug (NSAID)
treatment in RA patients.42, 43 In regard to TNF antagonists, studies have showed that patients
preferred shared treatment decisions or responsibility of the health professional when choosing a
TNF antagonist.44 In addition, 67% of 77 Canadian rheumatologists surveyed reported
concordance with the patient preference more than 80% of the time.44 In addition, 67% of 77
Canadian rheumatologists surveyed reported concordance with the patient preference more than
80% of the time.45 If physician and patient preference for TNF antagonist do correlate, then
improved persistence may have been a result of higher patient rather than physician preference.
Our findings suggest that administrative restrictions alone, without a change in physicians’
preference and beliefs may not achieve the desired effect on practice. A change in drug coverage
policy, for example, may achieve the desired effect in short-term, since physicians would follow
the new policy and prescribe mainly the reimbursed drug. In the longer term, however, unless a
change in physician preference is achieved, the recommended drug would be discontinued early,
and patients would be switched to a second-line drug. As a result, the overall effect of a change
in policy alone would be less than desired. In a similar manner, the study indicates that the
pharmaceutical industry may have a strong interest in influencing physicians despite
administrative restrictions.46
Unanswered questions and future research
Page 20 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 59
For peer review only
21
We theorize that in other conditions, when treating with a therapeutic class in which drugs are
considered to have similar benefits and harms, a similar association exists. Further study of these
conditions is required. In addition, we suggest exploring the explanation we suggested for the
association observed, such as measuring cognitive dissonance during prescribing decision38 or
patient preference. Awareness of the possible role of cognitive dissonance in clinical decision-
making, based on beliefs not necessarily supported by clinical evidence, can contribute to the
development of education programs for physicians. We also suggest exploring whether patients
who had been initiated on a non-preferred TNF inhibitor were more likely to be switched to a
preferred TNF antagonist. Further research is warranted to identify effective approaches to
policy implementations.
CONCLUSIONS
Higher physician preference, estimated using their prescribing habits, was associated with
decreased discontinuation risk in RA patients treated with TNF antagonists. This finding
highlights the limitation of introducing a new drug coverage policy without encouraging change
in the physicians’ drug preference. Similar research on other treatments is warranted.
Contributors: All authors contributed to the study conception and design and interpretation of the
data, drafting and revising the article critically for important intellectual content, and gave final
approval of the version to be published. Analysis was performed as part of research conducted
under the supervision of CD as part of the requirements for AF’s PhD project.
Page 21 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 60
For peer review only
22
Funding: The authors declare that (1) the study was supported by the University of British
Columbia Graduate Fellowship and a grant to the University of British Columbia from the
British Columbia Ministry of Health. The publication of study results was not contingent on their
approval; (2) no financial relationships with commercial entities that might have an interest in
the submitted work; (3) no spouses, partners, or children with relationships with commercial
entities that might have an interest in the submitted work; and (4) no non-financial interests that
may be relevant to the submitted work.
Ethical approval: The study protocol was approved by the Clinical Research Ethics Board of the
University of British Columbia (number H11-00303).
Data sharing: patient level data available from Population Data BC (https://www.popdata.bc.ca/).
Patient consent was not obtained but the presented data are anonymised and risk of identification
is low. Statistical code is available from the corresponding author.
Page 22 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 61
For peer review only
23
REFERENCES
1. Cush JJ. Biological drug use: US perspectives on indications and monitoring. Ann Rheum Dis
2005; Nov;64(Suppl 4):iv18-23.
2. Curtis JR, Chen L, Harrold LR, Narongroeknawin P, Reed G, Solomon DH. Physician
preference motivates the use of anti-tumor necrosis factor therapy independent of clinical disease
activity. Arthritis Care Res 2010; Jan 15;62(1):101-7.
3. Schneeweiss S, Glynn RJ, Avorn J, Solomon DH. A Medicare database review found that
physician preferences increasingly outweighed patient characteristics as determinants of first-
time prescriptions for COX-2 inhibitors. J Clin Epidemiol 2005; Jan;58(1):98-102.
4. Solomon DH, Schneeweiss S, Glynn RJ, Levin R, Avorn J. Determinants of selective
cyclooxygenase-2 inhibitor prescribing: Are patient or physician characteristics more important?.
Am J Med 2003; Dec 15;115(9):715-20.
5. Zhang J, Shan Y, Reed G, Kremer J, Greenberg JD, Baumgartner S, et al. Thresholds in
disease activity for switching biologics in rheumatoid arthritis patients: Experience from a large
US cohort. Arthritis Care Res 2011; Dec;63(12):1672-9.
6. Sokka T, Pincus T. Eligibility of patients in routine care for major clinical trials of anti-tumor
necrosis factor alpha agents in rheumatoid arthritis. Arthritis Rheum 2003; Feb;48(2):313-8.
7. Zink A, Strangfeld A, Schneider M, Herzer P, Hierse F, Stoyanova-Scholz M, et al.
Effectiveness of tumor necrosis factor inhibitors in rheumatoid arthritis in an observational
Page 23 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 62
For peer review only
24
cohort study: Comparison of patients according to their eligibility for major randomized clinical
trials. Arthritis Rheum 2006; Nov;54(11):3399-407.
8. Kievit W, Fransen J, Oerlemans AJ, Kuper HH, van der Laar MA, de Rooij DJ, et al. The
efficacy of anti-TNF in rheumatoid arthritis, a comparison between randomised controlled trials
and clinical practice. Ann Rheum Dis 2007; Nov;66(11):1473-8.
9. Berthelot JM, Benoist-Gerard S, le Goff B, Muller-Chevalet F, Maugars Y. Outcome and
safety of TNFalpha antagonist therapy in 475 consecutive outpatients (with rheumatoid arthritis
or spondyloarthropathies) treated by a single physician according to their eligibility for clinical
trials. Joint Bone Spine 2010; Dec;77(6):564-9.
10. Lacaille D, Anis AH, Guh DP, Esdaile JM. Gaps in care for rheumatoid arthritis: A
population study. Arthritis Rheum 2005; Apr 15;53(2):241-8.
11. Lacaille D, Guh DP, Abrahamowicz M, Anis AH, Esdaile JM. Use of nonbiologic disease-
modifying antirheumatic drugs and risk of infection in patients with rheumatoid arthritis.
Arthritis Rheum 2008; Aug;59(8):1074-81.
12. Li X, Guh D, Lacaille D, Esdaile J, Anis AH. The impact of cost sharing of prescription drug
expenditures on health care utilization by the elderly: Own- and cross-price elasticities. Health
Policy 2007; Aug;82(3):340-7.
13. Your level of Fair PharmaCare coverage in Fair PharmaCare. Available at:
http://www.health.gov.bc.ca/pharmacare/plani/planiindex.html#7. Accessed March 05, 2011.
Page 24 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 63
For peer review only
25
14. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic
comorbidity in longitudinal studies: Development and validation. J Chronic Dis 1987;
May;40(5):373-83.
15. Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC, et al. Coding algorithms
for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005;
Nov;43(11):1130-9.
16. Widdifield J, Bernatsky S, Paterson JM, Tu K, Ng R, Thorne JC, et al. Accuracy of Canadian
health administrative databases in identifying patients with rheumatoid arthritis: a validation
study using the medical records of rheumatologists. Arthritis Care Res (Hoboken) 2013;
Oct;65(10):1582-91.
17. Bensen B, Atkins C, Sheriff M, Kelsall J, Khalil H. Profile of patients with rheumatoid
arthritis treated with infliximab in Canada-trends toward less DMARD use prior to a biologic,
earlier use of infliximab and differences in baseline disease [abstract]. J Rheumatol 2011;
Jun;38(6):1155.
18. Buchanan J, Rahman MU, Doyle MK, Hsia EC, Gathany T, Parasuraman S, et al. Historical
trends of patient characteristics in anti-TNF clinical trials for rheumatoid arthritis: An analysis of
the literature over the past 16 years [abstract]. Arthritis Rheum 2009; (first published online: Mar
2010);60(Supple 10):1012.
19. Hyrich KL, Watson KD, Lunt M, Symmons DP, on behalf of the British Society for
Rheumatology Biologics Register (BSRBR). Changes in disease characteristics and response
Page 25 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 64
For peer review only
26
rates among patients in the United Kingdom starting anti-tumour necrosis factor therapy for
rheumatoid arthritis between 2001 and 2008. Rheumatology (Oxford) 2011; Jul;50(1):117-23.
20. Lee SJ, Chang H, Yazici Y, Greenberg JD, Kremer JM, Kavanaugh A. Utilization trends of
tumor necrosis factor inhibitors among patients with rheumatoid arthritis in a United States
observational cohort study. J Rheumatol 2009; Aug;36(8):1611-7.
21. Lee EW, Wei LJ, Amato DA. Cox-type tegression analysis for large numbers of small groups
of correlated failure time observations. In: Klein JP, Goel PK, editors. Survival analysis: State of
the art. NATO ASI series. Series E, Applied sciences ed. Boston, Massachusetts, U.S.: Kluwer
Academic Publishers; 1992. p. 237-47.
22. Statistical analysis of clustered data using SAS® system (NESUG 2006). Available at:
http://www.nesug.org/proceedings/nesug06/an/da01.pdf. Accessed November 12, 2010.
23. Analysis of Survival Data with Clustered Events (paper 237-2009). Available at:
http://support.sas.com/resources/papers/proceedings09/237-2009.pdf;. Accessed November 12,
2010.
24. Remicade® (infliximab) in eCPS. Accessed October 24, 2008.
25. Enbrel® (Etanercept) in eCPS. Accessed June 03, 2009.
26. Special Authority request; Abatacept/ Adalimumab/ Certolozumab/ Etanercept/ Golimumab/
Infliximab Rheumatoid Arthritis (Initial or Switch). Available at:
https://www.health.gov.bc.ca/exforms/pharmacare/5345fil.pdf. Accessed May 10, 2012.
Page 26 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 65
For peer review only
27
27. Mark TL, Swait J. Using stated preference and revealed preference modeling to evaluate
prescribing decisions. Health Econ 2004; Jun;13(6):563-73.
28. Wolfe F. The epidemiology of drug treatment failure in rheumatoid arthritis. Baillieres Clin
Rheumatol 1995; Nov;9(4):619-32.
29. Hyrich KL, Lunt M, Watson KD, Symmons DPM, Silman AJ. Outcomes after switching
from one anti-tumor necrosis factor alpha agent to a second anti-tumor necrosis factor alpha
agent in patients with rheumatoid arthritis: Results from a Large UK National Cohort Study.
Arthritis Rheum 2007; Jan;56(1):13-20.
30. Carmona L, Gomez-Reino JJ, on behalf of the BIOBADASER Group. Survival of TNF
antagonists in spondylarthritis is better than in rheumatoid arthritis. Data from the Spanish
registry BIOBADASER. Arthritis Res Ther 2006; Apr;8(3):R72.
31. Hetland ML, Christensen IJ, Tarp U, Dreyer L, Hansen A, Hansen IT, et al. Direct
comparison of treatment responses, remission rates, and drug adherence in patients with
rheumatoid arthritis treated with adalimumab, etanercept, or infliximab: Results from eight years
of surveillance of clinical practice in the nationwide Danish DANBIO registry. Arthritis Rheum
2010; Jan;62(1):22-32.
32. Fernandez-Nebro A, Irigoyen MV, Urena I, Belmonte-Lopez MA, Coret V, Jimenez-Nunez
FG, et al. Effectiveness, predictive response factors, and safety of anti-tumor necrosis factor
(TNF) therapies in anti-TNF-naive rheumatoid arthritis. J Rheumatol 2007; Dec;34(12):2334-42.
Page 27 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 66
For peer review only
28
33. Kievit W, Fransen J, Adang EMM, den Broeder AA, Bernelot Moens HJ, Visser H, et al.
Long-term effectiveness and safety of TNF-blocking agents in daily clinical practice: Results
from the Dutch Rheumatoid Arthritis Monitoring Register. Rheumatology (Oxford) 2011;
Jan;50(1):196-203.
34. Filippini M, Bazzani C, Favalli EG, Marchesoni A, Atzeni F, Sarzi-Puttini P, et al. Efficacy
and safety of anti-tumour necrosis factor in elderly patients with rheumatoid arthritis: An
observational study. Clin Rev Allergy Immunol 2010; Apr;38(2-3):90-6.
35. Baser O, Wang L, Xie L, Dysinger A, Gust C, Yuce H, et al. Deriving doctors' prescribing
patterns from health care claims: An instrumental variable analysis [abstract]. Value Health
2010; November;13(7):A425.
36. Kamal KM, Madhavan SS, Hornsby JA, Miller L, Kavookjian J, Scott V. Use of tumor
necrosis factor inhibitors in rheumatoid arthritis: A national survey of practicing United States
rheumatologists. Joint Bone Spine 2006; Dec;73(6):718-24.
37. Festinger L. A theory of cognitive dissonance. Evanston, Ill.: Row, Peterson; 1957.
38. Visser PS, Cooper J. Dissonance and attitude change: A matter of measurement, in Chapter
10: Attitude change. In: Hogg MA, Cooper J, editors. The Sage handbook of social psychology
Thousand Oaks, California, U.S.: Sage; 2003. p. 220.
39. Yazici Y, Simsek I. Treatment options for rheumatoid arthritis beyond TNF-alpha inhibitors.
Expert Rev Clin Pharmacol 2010; Sep;3(5):663-6.
Page 28 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 67
For peer review only
29
40. Dudler J, Moller B, Michel BA, Villiger PM. Biologics in rheumatoid arthritis (RA) -
Recommendations for Swiss practice. Swiss Med Wkly 2011;(w13189).
41. Lu TY, Hill C. Managing patients taking tumour necrosis factor inhibitors. Australian
Prescriber 2006; Jun;29(3):67-70.
42. Gall EP, Caperton EM, McComb JE, Messner R, Multz CV, O'Hanlan M, et al. Clinical
comparison of ibuprofen, fenoprofen calcium, naproxen and tolmetin sodium in rheumatoid
arthritis. J Rheumatol 1982; May-Jun;9(3):402-7.
43. Wasner C, Britton MC, Kraines RG, Kaye RL, Bobrove AM, Fries JF. Nonsteroidal anti-
inflammatory agents in rheumatoid arthritis and ankylosing spondylitis. Jama 1981; Nov
13;246(19):2168-72.
44. Chilton F, Collett RA. Treatment choices, preferences and decision-making by patients with
rheumatoid arthritis. Muscoskel Care 2008; Mar;6(1):1-14.
45. Rheumatoid Arthritis Patient Preference Survey (RAPPS) (Presentation number: 965).
Available at: http://acr.confex.com/acr/2006/webprogram/Paper5515.html;. Accessed December
05, 2011.
46. Spurling GK, Mansfield PR, Montgomery BD, Lexchin J, Doust J, Othman N, et al.
Information from pharmaceutical companies and the quality, quantity, and cost of physicians'
prescribing: a systematic review. PLoS Med 2010; Oct 19;7(10):e1000352.
Page 29 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 68
For peer review only
30
Page 30 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 69
For peer review only
31
Figure 1: Calculating the value of the independent variable physician preference
Figure 2: Participants’ flow
Figure 3: Persistence with TNF antagonists by physician preference levels
Page 31 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 70
For peer review only
Calculating the value of the independent variable physician preference 215x224mm (300 x 300 DPI)
Page 32 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 71
For peer review only
Participants’ flow
215x137mm (300 x 300 DPI)
Page 33 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 72
For peer review only
Persistence with TNF antagonists by physician preference levels 54x40mm (300 x 300 DPI)
Page 34 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 73
For peer review only
1
Prescriber Preference for a Particular Tumor Necrosis Factor Antagonist Drug and
Treatment Discontinuation: Population-based Cohort
STROBE Statement—Checklist of items that should be included in reports of cohort studies
Item
No Recommendation
Page
number
Title and abstract 1 (a) Indicate the study’s design with a commonly used term in the title or
the abstract
2
(b) Provide in the abstract an informative and balanced summary of
what was done and what was found
2-3
Introduction
Background/rationale 2 Explain the scientific background and rationale for the investigation
being reported
4-5
Objectives 3 State specific objectives, including any prespecified hypotheses 5
Methods
Study design 4 Present key elements of study design early in the paper 5
Setting 5 Describe the setting, locations, and relevant dates, including periods of
recruitment, exposure, follow-up, and data collection
5-6
Participants 6 (a) Give the eligibility criteria, and the sources and methods of selection
of participants. Describe methods of follow-up
5-6
(b) For matched studies, give matching criteria and number of exposed
and unexposed
n/a
Variables 7 Clearly define all outcomes, exposures, predictors, potential
confounders, and effect modifiers. Give diagnostic criteria, if applicable
5-11
Data sources/
measurement
8* For each variable of interest, give sources of data and details of
methods of assessment (measurement). Describe comparability of
assessment methods if there is more than one group
5-11
Bias 9 Describe any efforts to address potential sources of bias 8-11
Study size 10 Explain how the study size was arrived at 11
Quantitative variables 11 Explain how quantitative variables were handled in the analyses. If
applicable, describe which groupings were chosen and why
5-11
Statistical methods 12 (a) Describe all statistical methods, including those used to control for
confounding
5-11
(b) Describe any methods used to examine subgroups and interactions 11
(c) Explain how missing data were addressed n/a
(d) If applicable, explain how loss to follow-up was addressed 10
(e) Describe any sensitivity analyses 7,11
Results
Participants 13* (a) Report numbers of individuals at each stage of study—eg numbers
potentially eligible, examined for eligibility, confirmed eligible,
included in the study, completing follow-up, and analysed
11, figure
2
(b) Give reasons for non-participation at each stage Figure 2
(c) Consider use of a flow diagram Figure 2
Descriptive data 14* (a) Give characteristics of study participants (eg demographic, clinical,
social) and information on exposures and potential confounders
11-12,
table 2
(b) Indicate number of participants with missing data for each variable
of interest
n/a
Page 35 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 74
For peer review only
2
(c) Summarise follow-up time (eg, average and total amount) 14, figure
3
Outcome data 15* Report numbers of outcome events or summary measures over time Figure 3
Main results 16 (a) Give unadjusted estimates and, if applicable, confounder-adjusted
estimates and their precision (eg, 95% confidence interval). Make clear
which confounders were adjusted for and why they were included
Table 3
(b) Report category boundaries when continuous variables were
categorized
Table 1
(c) If relevant, consider translating estimates of relative risk into
absolute risk for a meaningful time period
n/a
Other analyses 17 Report other analyses done—eg analyses of subgroups and interactions,
and sensitivity analyses
14-15,
table 3
Discussion
Key results 18 Summarise key results with reference to study objectives 16
Limitations 19 Discuss limitations of the study, taking into account sources of potential
bias or imprecision. Discuss both direction and magnitude of any
potential bias
16-17
Interpretation 20 Give a cautious overall interpretation of results considering objectives,
limitations, multiplicity of analyses, results from similar studies, and
other relevant evidence
18-20
Generalisability 21 Discuss the generalisability (external validity) of the study results 16-17, 20
Other information
Funding 22 Give the source of funding and the role of the funders for the present
study and, if applicable, for the original study on which the present
article is based
22
*Give information separately for exposed and unexposed groups.
Note: An Explanation and Elaboration article discusses each checklist item and gives methodological background and
published examples of transparent reporting. The STROBE checklist is best used in conjunction with this article (freely
available on the Web sites of PLoS Medicine at http://www.plosmedicine.org/, Annals of Internal Medicine at
http://www.annals.org/, and Epidemiology at http://www.epidem.com/). Information on the STROBE Initiative is
available at http://www.strobe-statement.org.
Page 36 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 75
For peer review only
1
Prescriber Preference for a Particular Tumor Necrosis
Factor Antagonist Drug and Treatment Discontinuation:
Population-based Cohort
Anat Fisher, MD, PhD; Ken Bassett, MD, PhD; James M. Wright, MD, PhD; M. Alan
Brookhart, PhD; Hugh Freeman, MD; Colin R. Dormuth, ScD
Corresponding Author: Anat Fisher, MD, PhD; Department of Anesthesiology, Pharmacology
and Therapeutics, University of British Columbia, 2176 Health Sciences Mall, Vancouver, BC,
Canada V6T 1Z3, Telephone: 604-822-0700 Fax: 604-822-0701, Email: [email protected]
Ken Bassett: Department of Anesthesiology, Pharmacology & Therapeutics and Department of
Family Practice, University of British Columbia, Vancouver, Canada
James M. Wright: Department of Anesthesiology, Pharmacology & Therapeutics and
Department of Medicine, University of British Columbia, Vancouver, Canada
M. Alan Brookhart: Department of Epidemiology, Gillings School of Global Public Health,
University of North Carolina, Chapel Hill, North Carolina, USA
Hugh Freeman: Department of Medicine, University of British Columbia, Vancouver, Canada
Colin R. Dormuth: Department of Anesthesiology, Pharmacology & Therapeutics, University of
British Columbia, Vancouver, Canada
Keywords: physician prescribing pattern, rheumatoid arthritis, medication persistence, Tumor
Necrosis Factor-alpha
Number of words (excluding title page, abstract, references, figures and tables, endnotes): 2581
Page 37 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 76
For peer review only
2
ABSTRACT
Objective: To assess the effect of physician preference for a particular tumor necrosis factor
alpha (TNF) antagonist on the risk of treatment discontinuation in rheumatoid arthritis.
Design: Population-based cohort study.
Setting: British Columbia administrative health data (inpatients, outpatients and pharmacy)
Participants: 2,742 British Columbia residents who initiated a first course of a TNF antagonist
between 2001 and December 2008, had been diagnosed with rheumatoid arthritis, and were
treated by one of 58 medium to high-volume prescribers.
Independent variable: A level of physician preference for the drug (higher or lower) was
assigned based on preceding prescribing records of the care-providing physician. Higher
preference was defined as at least 60% of TNF antagonist courses initiated in the preceding year.
Sensitivity analysis was conducted with different thresholds for higher preference.
Main Outcome Measure: Drug discontinuation was defined as a drug-free interval of 180 days
or switching to another TNF antagonist, anakinra, rituximab or abatacept. The risk of
discontinuation was compared between different levels of physician preference using survival
analysis.
Results: Higher preference for the prescribed TNF antagonist was associated with improved
persistence with the drug (4.28 years [95% confidence interval 3.70-4.90] versus 3.27 [2.84-
3.84], with log rank test p-value of 0.017). The adjusted hazard ratio for discontinuation was
Page 38 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 77
For peer review only
3
significantly lower in courses of drugs with higher preference (0.85 [0.76-0.96]). The results
were robust in a sensitivity analysis.
Conclusions: Higher physician preference was associated with decreased risk of discontinuing
TNF antagonists in patients with rheumatoid arthritis. This finding suggests that physicians who
strongly prefer a specific treatment help their patients to stay on treatment for a longer duration.
Similar research on other treatments is warranted.
Strengths and Limitations
• First study to explore within- physician variation in prescribing habits, specifically the
effect of prescriber preference to a drug on the decision to discontinue the drug
• The universal nature of the Canadian health care system and a systematic and
standardized approach to data collection in British Columbia, which ensured the
generalizability of our results, as well as the large sample and prolonged follow up
• To conquer the absence of access to clinical data, we used multiple proxy variables to
adjust for disease severity
• Physician preference was not directly measured but instead based on previous prescribing
habits. and not directly measured
Page 39 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 78
For peer review only
4
INTRODUCTION
The term ‘physician preference’ usually refers to favoring a particular drug or a therapeutic
group among several alternatives, and it has been shown to predict treatment choice.1-4 In studies
of administrative heath health (claim) data, this preference is often determined by identifying
dispensing of drug prescribed by the specific physician in a pre-determined period, prior to the
event of interest (a new prescribing). Despite an association with new prescribing
decisionsdesicions, the role of physician preference in treatment discontinuation has not been
studied. Recently, the term ‘preference’ has also been used to describe a second phenomenon –
in the context of treatment discontinuation, it is was used to describe the baseline risk of
discontinuing treatment in patients treated by a specific physician (the physician ‘preference for
discontinuation’). 5 This baseline risk may differ among physicians because physicians may
respond differently to similar clinical situations such as decreased benefit or harmful events.
They could recommend patients to discontinue treatment (with or without switchingswitch to a
second drug) or to persist with the treatment (but to adjust dose, add-on a second drug or be
under frequent watch). In this paper, we use the term ‘preference’ to describe the first
phenomenon (physician’s favorite drug) and ‘physician-specific discontinuation risk’ to describe
the second.
Treatment with tumor necrosis factor alpha (TNF) antagonists in patients with rheumatoid
arthritis (RA) was considered especially sensitive to physician preference for two main reasons.
First, during the study period (2001-2009) there was limited clinical evidence on the comparative
effectiveness of the drugs, mainly due mostly to the absence of head-to-head randomized clinical
Page 40 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 79
For peer review only
5
trials, but also because participants in placebo-controlled trials were not representative of patients
treated in routine clinical settings.6-9 Second, published indications for discontinuation of TNF
antagonists were vague and confusing, and therefore care-providing physicians could reasonably
be expected to reach different clinical decisions given the same clinical situation. Consequently,
the decisions about which TNF antagonist to prescribe first and when to discontinue treatment
were likely subject to physicians’ individual preference.
This study analyzed data of first courses of a TNF antagonist in British Columbia patients with
RA. The prescriber recorded on the first dispensing claim for a TNF antagonist was used as a
proxy of the care-providing physician. The study objective was to estimate the effect of
physician preference on the risk of discontinuation. The null hypothesis tested was that physician
preference for a TNF antagonist when treatment has been initiated does not influence the risk of
discontinuing the treatment in RA patients.
PATIENTS AND METHODS
The study cohort was identified using four British Columbia Ministry of Health administrative
databases: PharmaNet (prescription dispensing data), Medical Service Plan (MSP) registration
information (demographic data), MSP Payment Information (fee-for-service payments to
physicians and alternative providers), and the Discharge Abstract Database (hospital
separations). The databases were linkable using a de-identified patient and physician numbers.
Follow-up data were available from 1995 until December 31, 2009. The study cohort included
British Columbia residents who fulfilled all of three conditions: (1) first exposure to a TNF
antagonist between March 1, 2001 and December 31, 2008; (2) a diagnosis of RA; and (3) TNF
Page 41 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 80
For peer review only
6
antagonist treatment initiated by a medium or high-volume prescriber, defined as a prescriber
who initiated 5 or more courses of TNF antagonists in RA patients during the years 2001-2008.
Exposure to a TNF antagonist was based on one or more recorded dispensing claims for
infliximab, adalimumab or etanercept between March 2001 and December 31, 2008. We
excluded patients who were previously treated with anakinra, rituximab or abatacept to decrease
heterogeneity in the population analyzed. The first dispensing claim for a TNF antagonist was
used to identify the study drug (infliximab, adalimumab or etanercept), index date and prescriber
for each patient in the cohort. We defined a pre-study period of three years preceding the index
date during which patients were required to have continuous provincial medical service
coverage. A gap shorter than 30 days was not considered to be an interruption in coverage. The
pre-study period ensured a standard run-in period of at least three years without TNF antagonist
exposure and a standard period during which diagnosis of RA was identified. RA patients were
selected based on similar criteria to previous studies in British Columbia.10-12 either two
outpatient visits in physician clinics with a diagnosis code of RA (ICD-9 714) at least 60 days
apart, or one hospitalization with recorded discharge diagnosis of RA. Patients were excluded if
sex or date of birth was not available, if they had a concurrent diagnosis of Crohn’s disease
(based on at least one outpatient or inpatient diagnosis code in the pre-study period) or if they
were younger than 18 years at the index date (to remove patients with juvenile RA). We also
excluded patients initiated on TNF antagonist treatment by prescribers who cumulatively
initiated less than five patients in the RA cohort (low-volume prescribers).
Physician preference was determined for each patient at the index date and coded as a Bernoulli
variable. It reflects prescribing patterns of the individual physician (who started this course)
during the preceding year (Figure 1). For each individual patient, we determined the date of
Page 42 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 81
For peer review only
7
course initiation, the prescriber recorded in the first dispensing event and the drug (infliximab,
adalimumab or etanercept). From first dispensing records for each patient in the study cohort, we
identified other patients that received any TNF antagonists from the same prescriber during the
year preceding the index date of the current patient, and the drug they were treated with. Next we
calculated the proportion of these patients who received the same drug as the patient of interest.
Based on this proportion, we categorized the prescriber preference to ‘higher’ when the
calculated proportion for this drug and that prescriber was 60% or higher, and ‘lower’ otherwise.
We expected that this threshold would provide reasonable assurance that a course categorized as
having ‘higher” preference for drug actually reflects that the prescriber favored that particular
TNF antagonist at this time, even if only two TNF antagonists were available. We reassessed
preference for drug based on annual data to allow for changes in preference over time and for
accommodating the more recent availability of adalimumab. Adalimumab was available in
Canada since October 2004, while infliximab and etanercept were on the market since 2001. By
including a year of data, differently from other studies that used only the last prescription in
assigning preference 1-4, we created a more accurate indicator. This way we minimized the effect
of factors not related to preference that might have influencedinfluence a specific prescribing
decision. Sensitivity analysis was conducted to examine the robustness of main results, using
thresholds of 70% or 80% for the level of physician preference.
Multiple covariates, mainly patient characteristics that may influence drug selection and/or
discontinuation were included in the final models and are listed in Table 1. We were limited by
lack of access to clinical data and inability to adjust for clinical variables not captured in
administrative health data, such as disease severity. Hence, we used multiple proxies, including
disease duration, age, extra-articular involvement, antirheumatic drug use, health services
Page 43 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 82
For peer review only
8
utilization and consumption of pain medications to adjust for disease severity. In order to control
for secular trends and late availability of adalimumab in Canada, we included a categorical
variable for the year of treatment initiation.
Table 1 - List of covariates included in the final models
Variable Description
Demographics
Sex Bernoulli variable
Age at index Four categories: 18-29 years, 30-69, 70-79 and ≥80 (categories were
assigned based on similar discontinuation risk in preliminary
analysis, which was not linear)
The annual deductible for
prescription cost at index
(Fair PharmaCare)
The deductible is based on annual income.13 Six categories: annual
deductible of $0, $1-500, $501-2250, >$2250, other programs and
no coverage
Geographical area at
index
Based on first 3 digit of postal code. Five mutually exclusive
categories: Greater Vancouver, Greater Victoria, Vancouver Island,
urban areas and rural areas
Clinical status
Physicians encounters in
the year preceding the
index date
Continuous variable
Number of inpatients
admissions in the year
preceding the index date
Four categories: 0, 1, 2, >2
Comorbidities (presence
and severity) during the
pre-study period
Charlson comorbidity score14 was determined using Quan’s
algorithm for administrative databases,15 excluding rheumatic
diseases. At least two outpatient or one inpatient encounter with the
Page 44 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 83
For peer review only
9
Variable Description
diagnosis were required. Four categories: 0, 1, 2, >2
Disease duration Measured from the first diagnosis of RA (inpatient or outpatient) in
the data. RA disease duration captured in Canadian administrative
data has been found to agree with reported duration by physicians.16
Continuous variable
The presence of extra-
articular manifestation
during the pre-study
period
Based on recorded at least one diagnoses with ICD-9 codes 3571,
3596, 7141, 71481, 7142 (outpatient) or ICD-10 codes G636, G737,
I39, I418, J990, M050, M051, M052, M053 (inpatient). Bernoulli
variable
Drug therapies
Concomitant MTX
during 200 days
preceding the index date
Based on mean plus two standard deviations of between-dispensing
intervals of MTX in the study cohort. Bernoulli variable
Dispensing claims for
NSAIDs during the year
preceding the index date
Bernoulli variable
Number of different
antirheumatic drugs
dispensed in the pre-
study period
Dispensing claims of 10 drugs were included: MTX,
hydroxychloroquine, sulfasalazine, leflunomide, azathioprine,
minocycline, penicillamine, sodium aurothiomalate, prednisone, and
intra-articular triamanolone or methylprednisolone. Four categories:
no drug, 1-2, 3-6, >6 different drugs
Other
Calendar year at index
date
The variable allowed controlling for secular trends in clinical
practice17-20 and availability of drugs. Eight yearly categories were
included for the years 2001-2008
ICD-10- International Classification of Diseases, 10th edition; ICD-9-CM - International Classification of Diseases, 9th edition, clinical modification; MTX – methotrexate; NSAID – nonsteroidal anti-inflammatory drug
Page 45 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 84
For peer review only
10
The outcome variable was a discontinuation defined as either switching to another TNF
antagonist including certolizumab and golimumab, anakinra, rituximab or abatacept (‘biologic’
antirheumatic drugs), or a drug-free interval of 180 days after exhaustion of the dispensed days-
supply of the latest refill. DiscontinuationThe discontinuation date was set to the end of the days-
supply of the last refill before the 180-days drug-free intervaldiscontinuation or the date of the
first dispensing of a second ‘biologic’ antirheumatic drug, whichever was earliest. Unless
experienced discontinuation occurred, patients were followed up until either December 31, 2009
(end of follow-up period) or until an interruption of more than six days in the provincial MSP
coverage, at which point data were considered censored. The most common causes of
coveragesuch interruptions were death and emigration from the province. Change of dose of the
TNF antagonist or addition ofadding-on a second drug (not TNF antagonist) was not considered
discontinuation, and patients were continued to be followed.
Discrepancies between recorded days-supply and dispensed quantity were common in the
PharmaNet database; hence, we also calculated days-supply based on the quantity dispensed,
which, if required, was imputed using the recorded total cost. Cost was considered the most
accurate and reliable field, since this field serves for claim and payment processing. We used the
longest duration of days-supply, recorded or calculated, to determine both the length of drug-free
intervals and a discontinuation date.
Sample size calculation
Assuming 1:1 ratio of courses with ‘higher’ preference versus courses with ‘lower’ preference
and in the absence of prior estimates, if the true hazard ratio for discontinuing courses with
‘higher’ preference relative to courses with ‘lower’higher’ preference is 0.8, we required 847
Page 46 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 85
For peer review only
11
courses atwith each preference level to be able to reject the null hypothesis with type I error
probability of 0.05 and power of 0.80.
Statistical Analysis
Patients’Summary statistics of baseline characteristics were compared across the three drugsdrug
groups. The adjusted risk of discontinuation was compared using two different approaches in a
Cox proportional hazards regression. First, we conducted an analysis of nonclustered data. The
multivariate model also included a series of Bernoulli variables, one for each physician (1=the
particular physician; 0=other physicians), whichthat allowed individualization of the risk of drug
discontinuation by physician (physician-specific discontinuation risk). In the second approach, a
marginal model of clustered data allowed an adjustment for possible correlation between patients
treated by the same physician.21-23 We tested for model assumptions (proportional hazard and
absence of interactions) and found them to be valid. We also checked for the linearity of
continuous variables and categorized non-linear variables. All statistical tests were two-sided.
All calculations were performed using the SAS software package (SAS Institute Inc., Cary, NC).
RESULTS
The study cohort included 2,742 RA patients prescribed by a total of 58 medium-high volume
physicians. Figure 2 presents reasons for excluding patients who used TNF antagonists from the
analysis. Baseline characteristics across users of the three drugsdrug groups are presented in
Table 2. Not only was etanercept was the most frequently prescribed drug (1718 patients, 63%),
but also it was also the only drug prescribed by all 58 study physicians. Etanercept was usually
Page 47 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 86
For peer review only
12
initiated by a physician with high preference for etanercept (70% of etanercept courses).
Infliximab or adalimumab, on the other hand, were usually initiated by physicians with lower
preference for these drugs (only 34% or 19% of courses were initiated by a physician with high
preference for drug, respectively). Patients treated with adalimumab were significantly older and
had a lower income (reflected in lower annual deductible level for prescribing costs). Patients
treated with infliximab had the highest prevalence of concomitant methotrexate therapy and
patients with etanercept the lowest. This reflects differences in the indications mentionedmention
in the product monographs; while infliximab is indicated for use in combination with
methotrexate,24 the monograph of etanercept mentioned that it “can be initiated in combination
with methotrexate … or used alone”.25 Similarly, while the provincial special authority policy
requires that infliximab is used in combination with methotrexate (or other drug),) and such
requirement does not exist for treatment with etanercept or adalimumab.26
Page 48 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 87
For peer review only
13
Table 2 - Baseline characteristics
Variable Infliximab Adalimumab Etanercept P-value
(between
groups
differences)
Number of patients (% from
cohort)
571 (21) 453 (16) 1718 (63)
Number of prescribers (based on
first dispensing event for each
patient) (total 58 prescribers)
49 46 58
Patients treated with a drug with
higher preference, No (%)
193 (34) 84 (19) 1198 (70) <0.0001
Demographics
Females, No (%) 403 (71) 326 (72) 1239 (72) 0.77
Age at index, median y (range) y 56 (18-87) 58 (22-91) 56 (18-92) 0.003
Annual deductible for prescription
cost, No (%)
Very low ($0) 47 (8.3) 80 (18) 199 (12) <0.0001
Low ($1-500) 33 (5.8) 53 (12) 152 (8.9) 0.004
Medium ($501-2250) 92 (16) 109 (24) 315 (18) 0.004
High (>$2250) 35 (6.1) 38 (8.4) 143 (8.3) 0.22
Residence in Greater
Vancouver/Victoria, No (%)
341 (60) 224 (50) 782 (46) <0.0001
Clinical status
Number of physician visits median
(range)
33 (3-158) 31 (2-112) 32 (3-136) 0.25
At least one admission to hospital
No (%)
104 (18) 63 (14) 340 (20) 0.01
Extra-articular manifestations, No 28 (4.9) 14 (3.1) 60 (3.5) 0.23
Page 49 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 88
For peer review only
14
(%)
Presence of comorbidity (score>0),
No (%)
113 (20) 95 (21) 383 (22) 0.41
RA disease duration median, y
(range) y
9.2 (0.1-
17.9)
7.7 (0.3-17.9) 8.0 (0-17.8) 0.21
RA drugs
Concomitant MTX, No (%) 388 (68) 264 (58) 856 (50) <0.0001
Dispensing claims for NSAIDs,
No (%)
307 (54) 214 (47) 923 (54) 0.04
Number of different antirheumatic
drugs, median (range)
4 (0-8) 4 (0-8) 4 (0-9) 0.46
%- percent, unless otherwise specified – percentage from patients treated with this drug; $-Canadian dollars; MTX-methotrexate; No-number of patients; NSAID- nonsteroidal anti-inflammatory drug; NSS – not statistically significant; y-years
For detailed descriptions of the variables refer to Table 1
Persistence with the three TNF antagonists was similar (log rank test p-value=0.15). The product
limit median persistence estimates were 3.9 years (95% confidence interval 3.0-5.3), 3.3 (2.6-4.0)
and 3.9 years (3.4-4.4) for infliximab, adalimumab and etanercept, respectively. Higher physician
preference was associated with improved persistence compared to lower preference (Figure 3),
with median time to discontinuation of 4.28 years (confidence interval 3.70-4.90) in 1475
courses with ‘high’ preference, compared with 3.27 (2.85-3.84) in 1267 courses with ‘low’
preference . In both groups, about 50% of the patients discontinued and the other 50% were
censored.
Higher physician preference was associated with a significant decrease of 14-15% in the adjusted
hazard for discontinuation (Table 3). No significant interaction between drug and preference was
Page 50 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 89
For peer review only
15
observed. The results of sensitivity analysis were similar, with overall adjusted hazard ratio
between 0.85-0.88 in all models. The results of clustered data analyses with thresholds of 60%,
70% and 80% for physician preference were robust (Table 3). The results of nonclustered data
analyses for thresholds of 70% and 80%, on the other hand, did not reach the significance level.
That may be a result of decreaseddecreasing numbers of patients treated with a drug of higher
physician preference and therefore decreased power to detect significant difference.
Table 3 - Hazard ratios for drug discontinuation, higher preference versus lower
preference
Approach
Hazard ratio (95% confidence interval)
Preference threshold
60%
Preference threshold
70%
Preference threshold
80%
Patients with higher
preference N (%)
1426 (52) 1234 (45) 987 (36)
Nonclustered
data analysis
Crude 0.88 (0.79-0.98) 0.89 (0.80-0.99) 0.83 (0.81-0.996)
Adjusted* 0.86 (0.76-0.98) 0.88 (0.77-1.01) 0.87 (0.75-1.004)
Marginal
modeling of
clustered data
Crude 0.88 (0.77-1.001) 0.89 (0.78-1.01) 0.89 (0.79-1.01)
Adjusted# 0.85 (0.76-0.96) 0.87 (0.78-0.97) 0.87 (0.78-0.98)
% - percent from overall patients (2742); N – Number of patients; *Adjusted by drug, calendar year, patient’s demographics and clinical status, and prescriber #Adjusted by drug, calendar year and patient’s demographics and clinical status
Page 51 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 90
For peer review only
16
DISCUSSION
We demonstrated two types of variations of physician decisions to discontinue treatment with
TNF antagonists: between-physician and within-physician. Between-physician variability in
response to similar clinical situations, such as decreased benefit or a harmful event is expressed
as a different baseline risk of discontinuation, depending on the treating physician in response to
similar clinical situations, such as decreased benefit or a harmful event. We used the term
physician-specific discontinuation risk for this phenomenon. It could be a result of differences in
education or experience, adherence to different guidelines and possible differences in patient
case mix. Within-physician variability, on the other hand, is a different response to a similar
clinical situation (similar patients, similar drug effects) by the same physician. In this study we
showed that this variability correlated to physician preference for drug – and specifically, higher
physician preference for a specific TNF antagonist drug was associated with a decreased risk of
discontinuing TNF antagonists in RA patients. This preference most probably reflected the
physician’s beliefs regarding the relative effectiveness and safety of the specific drug compared
to the alternatives, where a preferred drug is thought to be superior. When a physician
prescribedtreated with a drug believed to be superior compared to comparators, patients were
encouraged to stay longer on the drug and the risk of discontinuation, as measured in this study,
decreased. We believe that a different response (level of encouragement) to a similar clinical
situation applied in indefinite clinical situations, such as mild harmful effect or questionable
benefit.
Strengths and weaknesses of the study
Page 52 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 91
For peer review only
17
The main advantages of our study are the universal nature of the Canadian health care system
and a systematic and standardized approach to data collection in British Columbia, which
ensured the generalizability of our results, as well as the large sample and prolonged follow up
that increased the power of the study to detect differences in discontinuation risk. The main
disadvantage is the absence of access to clinical data; hence adjustment to patient mix was
difficult. To conquer this problem, we used multiple proxy variables to adjust for disease
severity, such as extra-articular involvement, antirheumatic drug use, health services utilization
and consumption of pain medications.
There are several limitations to ascertaining physician preference based on previous prescribing
habits. First, preference was not defined for all patients. For example, it was not defined for the
first patient treated by the physician, or if the physician did not prescribe this class of drugs in
the preceding year. Second, we assumed that increased loyalty for a particular TNF antagonist
reflects preference, but we did not measure preference directly. Nevertheless, stated preference
and prescribing habits were shown to correlate in previous research.27 Third, we measured
preference at the time of treatment initiation and not at discontinuation, and preference might
have changed over time. Finally, we assumed strong preference for one drug only and regarded
situations in which the physician preferred two of the TNF antagonists to be a situation of
treatment with a drug of lower preference.
Overview of previous studies
Treatment persistence was hypothesized to reflect therapeutic benefit and harm in chronic non-
curable diseases, such as RA.rheumatoid arthritis28 In support of this hypothesis, previous studies
in RA patients using TNF antagonists demonstrated that the main reasons for discontinuing or
Page 53 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 92
For peer review only
18
switching were decreased benefit (36-67% of the discontinuations) or perceived harm (30-
58%).29-34 As such, the risk of discontinuing treatment is assumed to be influenced solitarily by
drug properties and patient characteristics. The demonstrated effect of care-provider
characteristics, specifically prescribing habits, on the risk of discontinuation calls into question
this hypothesis.
Between-physician differences in response to the same clinical situation (physician-specific
discontinuation risk) have previously been studiedstudies in RA patients treated by TNF
antagonists. Differences in response to harm were reported by Cush et al 2005 based on an
onlineon-line survey of rheumatologists 1 and in response to decreased benefit – by Zhang et al
2011 5. Using mixed effect model with a random intercept to cluster patients at physician level,
Zhang demonstrated the importance of clustering by physicians, even after adjustment for both
baseline disease activity and improvement in disease activity.
Within-physician variation in prescribing decisions and the effect of physician preference have
infrequently been studied. While the effect of physician preference on discontinuation has not
been previously studied, in this therapeutic class or in other conditions, physician preference was
found to be an important predictor of drug selection in treatment initiation. Physician preference
for therapeutic class was the most important determinant in initiating treatment with TNF
antagonists compared to the alternative: prescribing synthetic antirheumatic drugs 1 and in other
treatment situations.3, 4, 35 Physician preference for an individual TNF antagonist was studied by
Kamal et al 2006.36 In response to mailed questionnaires, most American rheumatologists said
they preferred etanercept over adalimumab or infliximab and considered etanercept the most
Page 54 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 93
For peer review only
19
efficacious of the three drugs with less harm. In our study, etanercept was the most prescribed
TNF drug, and it was prescribed by all 58 medium to high-volume prescribers.
Explanations and interpretation
We suggest several possible explanations for the finding of decreased risk of discontinuation
with increased physician preference for the prescribed TNF antagonist. First, the results could be
explained in light of the theory of cognitive dissonance.37 Dissonance is an uncomfortable
feeling caused by holding conflicting ideas simultaneously. In line with this theory, treating with
a drug believed to be inferior (lower preference) induces dissonance. In indefinite clinical
situations, such as mild harmful effect or questionable benefit, early drug discontinuation
supports the physician’s belief that the selected drug was inferior compared to the alternatives.
On the other hand, in a similar situation, a drug with a higher preference might be continued, to
support a belief in its superiority. The interpretation of our results using the theory of cognitive
dissonance is restricted by the lack of direct measurement of physician dissonance at the time of
treatment discontinuation. The indirect measure of dissonance we used, physician preference,
could also be influenced by factors unrelated to prescriber beliefs, such as limited availability of
a drug or higher relative cost (although during the study period, availability or cost were not
factors).. Direct measures of cognitive dissonance are complicated and involve activities beyond
the scope of the current study.38 Second, the observed association between higher preference and
decreased risk of discontinuation might be confounded by experience with the drug, namely the
total number of patients a physician has previously treated with the same drug. Increased
experience with a specific drug may be associated with higher preference as a result of the
algorithm we used to assign the level of preference, but also because of a tendency to continue
Page 55 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 94
For peer review only
20
doing what is familiar. Theoretically, increased experience with a specific drug would improve
patient selection, and therefore is associated with improved benefit, decreased harm and
decreased risk of discontinuation in these patients. Lastly, a physician’s stated preference for a
particular drug has been shown to correlate with patient preference. The selection of a specific
TNF antagonist mostly depends on physician and/or patient preference because the benefit and
harm profiles of the three drugs are considered to be similar39-41 despite limited relative
effectiveness data. Physician and patient preferences for an individual TNF antagonist are likely
to be correlated, as was shownshowed in studies of nonsteroidal anti-inflammatory drug
(NSAID) treatment in RA of RA patients.42, 43 In regard to, although this was not studied in
patients treated with TNF antagonists, studies have showed that patients preferred shared
treatment decisions or responsibility of the health professional when choosing a TNF
antagonist.44. In addition, 67% of 77 Canadian rheumatologists surveyed reported concordance
with the patient preference more than 80% of the time.44 In addition, 67% of 77 Canadian
rheumatologists surveyed reported concordance with the patient preference more than 80% of the
time.45 If physician and patient preference for TNF antagonist do correlate, then improved
persistence may have been a result of higher patient rather than physician preference.
Our findings suggest that administrative restrictions alone, without a change in physicians’
preference and beliefs may not achieve the desired effect on practice. A change in drug coverage
policy, for example, may achieve the desired effect in short-term, since physicians would follow
the new policy and prescribe mainly the reimbursed drug. In the longer term, however, unless a
change in physician preference is achieved, the recommended drug would be discontinued early,
and patients would be switched to a second-line drug. As a result, the overall effect of a change
in policy alone would be less than desired. In a similar manner, the study indicates that the
Page 56 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 95
For peer review only
21
pharmaceutical industry may have a strong interest in influencing physicians despite
administrative restrictions.46
Unanswered questions and future research
We theorize that in other conditions, when treating with a therapeutic class in which drugs are
considered to have similar benefits and harms, a similar association exists. Further study of these
conditions is required. In addition, we suggest exploring the explanation we suggested for the
association observed, such as measuring cognitive dissonance during prescribing decision38 or
patient preference. Awareness of the possible role of cognitive dissonance in clinical decision-
making, based on beliefs not necessarily supported by clinical evidence, can contribute to the
development of education programs for physicians. We also suggest exploring whether patients
who had been initiated on a non-preferred TNF inhibitor were more likely to be switched to a
preferred TNF antagonist. Further research is warranted to identifyidentifying effective policy
implementation approaches to policy implementations.
CONCLUSIONS
Higher physician preference, estimated using their prescribing habits, was associated with
decreased discontinuation risk in RA patients treated with TNF antagonists. This finding
highlights the limitation of introducing a new drug coverage policy without encouraging change
in the physicians’ drug preference. Similar research on other treatments is warranted.
Page 57 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 96
For peer review only
22
Contributors: All authors contributed to the study conception and design and interpretation of the
data, drafting and revising the article critically for important intellectual content, and gave final
approval of the version to be published. Analysis was performed as part of research conducted
under the supervision of CD as part of the requirements for AF’s PhD project.
Funding: : The authors declare that (1) the study was supported by the University of British
Columbia Graduate Fellowship and a grant to the University of British Columbia from the
British Columbia Ministry of Health. The publication of study results was not contingent on their
approval; (2) no financial relationships with commercial entities that might have an interest in
the submitted work; (3) no spouses, partners, or children with relationships with commercial
entities that might have an interest in the submitted work; and (4) no non-financial interests that
may be relevant to the submitted work.
Ethical approval: The study protocol was approved by the Clinical Research Ethics Board of the
University of British Columbia (number H11-00303).
Data sharing: patient level data available from Population Data BC (https://www.popdata.bc.ca/).
Patient consent was not obtained but the presented data are anonymised and risk of identification
is low. Statistical code is available from the corresponding author.
Page 58 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 97
For peer review only
23
REFERENCES
1. Cush JJ. Biological drug use: US perspectives on indications and monitoring. Ann Rheum Dis
2005; Nov;64(Suppl 4):iv18-23.
2. Curtis JR, Chen L, Harrold LR, Narongroeknawin P, Reed G, Solomon DH. Physician
preference motivates the use of anti-tumor necrosis factor therapy independent of clinical disease
activity. Arthritis Care Res 2010; Jan 15;62(1):101-7.
3. Schneeweiss S, Glynn RJ, Avorn J, Solomon DH. A Medicare database review found that
physician preferences increasingly outweighed patient characteristics as determinants of first-
time prescriptions for COX-2 inhibitors. J Clin Epidemiol 2005; Jan;58(1):98-102.
4. Solomon DH, Schneeweiss S, Glynn RJ, Levin R, Avorn J. Determinants of selective
cyclooxygenase-2 inhibitor prescribing: Are patient or physician characteristics more important?.
Am J Med 2003; Dec 15;115(9):715-20.
5. Zhang J, Shan Y, Reed G, Kremer J, Greenberg JD, Baumgartner S, et al. Thresholds in
disease activity for switching biologics in rheumatoid arthritis patients: Experience from a large
US cohort. Arthritis Care Res 2011; Dec;63(12):1672-9.
6. Sokka T, Pincus T. Eligibility of patients in routine care for major clinical trials of anti-tumor
necrosis factor alpha agents in rheumatoid arthritis. Arthritis Rheum 2003; Feb;48(2):313-8.
7. Zink A, Strangfeld A, Schneider M, Herzer P, Hierse F, Stoyanova-Scholz M, et al.
Effectiveness of tumor necrosis factor inhibitors in rheumatoid arthritis in an observational
Page 59 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 98
For peer review only
24
cohort study: Comparison of patients according to their eligibility for major randomized clinical
trials. Arthritis Rheum 2006; Nov;54(11):3399-407.
8. Kievit W, Fransen J, Oerlemans AJ, Kuper HH, van der Laar MA, de Rooij DJ, et al. The
efficacy of anti-TNF in rheumatoid arthritis, a comparison between randomised controlled trials
and clinical practice. Ann Rheum Dis 2007; Nov;66(11):1473-8.
9. Berthelot JM, Benoist-Gerard S, le Goff B, Muller-Chevalet F, Maugars Y. Outcome and
safety of TNFalpha antagonist therapy in 475 consecutive outpatients (with rheumatoid arthritis
or spondyloarthropathies) treated by a single physician according to their eligibility for clinical
trials. Joint Bone Spine 2010; Dec;77(6):564-9.
10. Lacaille D, Anis AH, Guh DP, Esdaile JM. Gaps in care for rheumatoid arthritis: A
population study. Arthritis Rheum 2005; Apr 15;53(2):241-8.
11. Lacaille D, Guh DP, Abrahamowicz M, Anis AH, Esdaile JM. Use of nonbiologic disease-
modifying antirheumatic drugs and risk of infection in patients with rheumatoid arthritis.
Arthritis Rheum 2008; Aug;59(8):1074-81.
12. Li X, Guh D, Lacaille D, Esdaile J, Anis AH. The impact of cost sharing of prescription drug
expenditures on health care utilization by the elderly: Own- and cross-price elasticities. Health
Policy 2007; Aug;82(3):340-7.
13. Your level of Fair PharmaCare coverage in Fair PharmaCare. Available at:
http://www.health.gov.bc.ca/pharmacare/plani/planiindex.html#7. Accessed March 05, 2011.
Page 60 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 99
For peer review only
25
14. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic
comorbidity in longitudinal studies: Development and validation. J Chronic Dis 1987;
May;40(5):373-83.
15. Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC, et al. Coding algorithms
for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005;
Nov;43(11):1130-9.
16. Widdifield J, Bernatsky S, Paterson JM, Tu K, Ng R, Thorne JC, et al. Accuracy of Canadian
health administrative databases in identifying patients with rheumatoid arthritis: a validation
study using the medical records of rheumatologists. Arthritis Care Res (Hoboken) 2013;
Oct;65(10):1582-91.
17. Bensen B, Atkins C, Sheriff M, Kelsall J, Khalil H. Profile of patients with rheumatoid
arthritis treated with infliximab in Canada-trends toward less DMARD use prior to a biologic,
earlier use of infliximab and differences in baseline disease [abstract]. J Rheumatol 2011;
Jun;38(6):1155.
18. Buchanan J, Rahman MU, Doyle MK, Hsia EC, Gathany T, Parasuraman S, et al. Historical
trends of patient characteristics in anti-TNF clinical trials for rheumatoid arthritis: An analysis of
the literature over the past 16 years [abstract]. Arthritis Rheum 2009; (first published online: Mar
2010);60(Supple 10):1012.
19. Hyrich KL, Watson KD, Lunt M, Symmons DP, on behalf of the British Society for
Rheumatology Biologics Register (BSRBR). Changes in disease characteristics and response
Page 61 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 100
For peer review only
26
rates among patients in the United Kingdom starting anti-tumour necrosis factor therapy for
rheumatoid arthritis between 2001 and 2008. Rheumatology (Oxford) 2011; Jul;50(1):117-23.
20. Lee SJ, Chang H, Yazici Y, Greenberg JD, Kremer JM, Kavanaugh A. Utilization trends of
tumor necrosis factor inhibitors among patients with rheumatoid arthritis in a United States
observational cohort study. J Rheumatol 2009; Aug;36(8):1611-7.
21. Lee EW, Wei LJ, Amato DA. Cox-type tegression analysis for large numbers of small groups
of correlated failure time observations. In: Klein JP, Goel PK, editors. Survival analysis: State of
the art. NATO ASI series. Series E, Applied sciences ed. Boston, Massachusetts, U.S.: Kluwer
Academic Publishers; 1992. p. 237-47.
22. Statistical analysis of clustered data using SAS® system (NESUG 2006). Available at:
http://www.nesug.org/proceedings/nesug06/an/da01.pdf. Accessed November 12, 2010.
23. Analysis of Survival Data with Clustered Events (paper 237-2009). Available at:
http://support.sas.com/resources/papers/proceedings09/237-2009.pdf;. Accessed November 12,
2010.
24. Remicade® (infliximab) in eCPS. Accessed October 24, 2008.
25. Enbrel® (Etanercept) in eCPS. Accessed June 03, 2009.
26. Special Authority request; Abatacept/ Adalimumab/ Certolozumab/ Etanercept/ Golimumab/
Infliximab Rheumatoid Arthritis (Initial or Switch). Available at:
https://www.health.gov.bc.ca/exforms/pharmacare/5345fil.pdf. Accessed May 10, 2012.
Page 62 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 101
For peer review only
27
27. Mark TL, Swait J. Using stated preference and revealed preference modeling to evaluate
prescribing decisions. Health Econ 2004; Jun;13(6):563-73.
28. Wolfe F. The epidemiology of drug treatment failure in rheumatoid arthritis. Baillieres Clin
Rheumatol 1995; Nov;9(4):619-32.
29. Hyrich KL, Lunt M, Watson KD, Symmons DPM, Silman AJ. Outcomes after switching
from one anti-tumor necrosis factor alpha agent to a second anti-tumor necrosis factor alpha
agent in patients with rheumatoid arthritis: Results from a Large UK National Cohort Study.
Arthritis Rheum 2007; Jan;56(1):13-20.
30. Carmona L, Gomez-Reino JJ, on behalf of the BIOBADASER Group. Survival of TNF
antagonists in spondylarthritis is better than in rheumatoid arthritis. Data from the Spanish
registry BIOBADASER. Arthritis Res Ther 2006; Apr;8(3):R72.
31. Hetland ML, Christensen IJ, Tarp U, Dreyer L, Hansen A, Hansen IT, et al. Direct
comparison of treatment responses, remission rates, and drug adherence in patients with
rheumatoid arthritis treated with adalimumab, etanercept, or infliximab: Results from eight years
of surveillance of clinical practice in the nationwide Danish DANBIO registry. Arthritis Rheum
2010; Jan;62(1):22-32.
32. Fernandez-Nebro A, Irigoyen MV, Urena I, Belmonte-Lopez MA, Coret V, Jimenez-Nunez
FG, et al. Effectiveness, predictive response factors, and safety of anti-tumor necrosis factor
(TNF) therapies in anti-TNF-naive rheumatoid arthritis. J Rheumatol 2007; Dec;34(12):2334-42.
Page 63 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 102
For peer review only
28
33. Kievit W, Fransen J, Adang EMM, den Broeder AA, Bernelot Moens HJ, Visser H, et al.
Long-term effectiveness and safety of TNF-blocking agents in daily clinical practice: Results
from the Dutch Rheumatoid Arthritis Monitoring Register. Rheumatology (Oxford) 2011;
Jan;50(1):196-203.
34. Filippini M, Bazzani C, Favalli EG, Marchesoni A, Atzeni F, Sarzi-Puttini P, et al. Efficacy
and safety of anti-tumour necrosis factor in elderly patients with rheumatoid arthritis: An
observational study. Clin Rev Allergy Immunol 2010; Apr;38(2-3):90-6.
35. Baser O, Wang L, Xie L, Dysinger A, Gust C, Yuce H, et al. Deriving doctors' prescribing
patterns from health care claims: An instrumental variable analysis [abstract]. Value Health
2010; November;13(7):A425.
36. Kamal KM, Madhavan SS, Hornsby JA, Miller L, Kavookjian J, Scott V. Use of tumor
necrosis factor inhibitors in rheumatoid arthritis: A national survey of practicing United States
rheumatologists. Joint Bone Spine 2006; Dec;73(6):718-24.
37. Festinger L. A theory of cognitive dissonance. Evanston, Ill.: Row, Peterson; 1957.
38. Visser PS, Cooper J. Dissonance and attitude change: A matter of measurement, in Chapter
10: Attitude change. In: Hogg MA, Cooper J, editors. The Sage handbook of social psychology
Thousand Oaks, California, U.S.: Sage; 2003. p. 220.
39. Yazici Y, Simsek I. Treatment options for rheumatoid arthritis beyond TNF-alpha inhibitors.
Expert Rev Clin Pharmacol 2010; Sep;3(5):663-6.
Page 64 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 103
For peer review only
29
40. Dudler J, Moller B, Michel BA, Villiger PM. Biologics in rheumatoid arthritis (RA) -
Recommendations for Swiss practice. Swiss Med Wkly 2011;(w13189).
41. Lu TY, Hill C. Managing patients taking tumour necrosis factor inhibitors. Australian
Prescriber 2006; Jun;29(3):67-70.
42. Gall EP, Caperton EM, McComb JE, Messner R, Multz CV, O'Hanlan M, et al. Clinical
comparison of ibuprofen, fenoprofen calcium, naproxen and tolmetin sodium in rheumatoid
arthritis. J Rheumatol 1982; May-Jun;9(3):402-7.
43. Wasner C, Britton MC, Kraines RG, Kaye RL, Bobrove AM, Fries JF. Nonsteroidal anti-
inflammatory agents in rheumatoid arthritis and ankylosing spondylitis. Jama 1981; Nov
13;246(19):2168-72.
44. Chilton F, Collett RA. Treatment choices, preferences and decision-making by patients with
rheumatoid arthritis. Muscoskel Care 2008; Mar;6(1):1-14.
45. Rheumatoid Arthritis Patient Preference Survey (RAPPS) (Presentation number: 965).
Available at: http://acr.confex.com/acr/2006/webprogram/Paper5515.html;. Accessed December
05, 2011.
46. Spurling GK, Mansfield PR, Montgomery BD, Lexchin J, Doust J, Othman N, et al.
Information from pharmaceutical companies and the quality, quantity, and cost of physicians'
prescribing: a systematic review. PLoS Med 2010; Oct 19;7(10):e1000352.
Page 65 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 104
For peer review only
30
Page 66 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from
Page 105
For peer review only
31
Figure 1: Calculating the value of the independent variable physician preference
Figure 2: Participants’ flow
Figure 3: Persistence with TNF antagonists by physician preference levels
Page 67 of 67
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on June 29, 2020 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2014-005532 on 30 Septem
ber 2014. Dow
nloaded from