Pharmacist-led management of chronic pain in primary care: costs and benefits in a pilot randomised controlled trial

Pharmacist-led management of chronicpain in primary care costs and benefitsin a pilot randomised controlled trial

Aileen R Neilson1 Hanne Bruhn2 Christine M Bond2 Alison M Elliott2

Blair H Smith3 Philip C Hannaford2 Richard Holland4 Amanda J Lee5

Margaret Watson2 David Wright6 Paul McNamee1

To cite Neilson AR Bruhn HBond CM et al Pharmacist-led management of chronicpain in primary care costsand benefits in a pilotrandomised controlled trialBMJ Open 20155e006874doi101136bmjopen-2014-006874

Prepublication history forthis paper is available onlineTo view these files pleasevisit the journal online(httpdxdoiorg101136bmjopen-2014-006874)

Received 20 October 2014Revised 22 January 2015Accepted 9 February 2015

1Health Economics ResearchUnit University of AberdeenAberdeen UK2Centre of Academic PrimaryCare University of AberdeenAberdeen UK3Division of PopulationSciences University ofDundee Dundee UK4Norwich Medical SchoolUniversity of East AngliaNorwich UK5Medical Statistics TeamUniversity of AberdeenAberdeen UK6School of PharmacyUniversity of East AngliaNorwich UK

Correspondence toAileen R Neilsonaileenneilsonabdnacuk

ABSTRACTObjectives To explore differences in mean costs(from a UK National Health Service perspective) andeffects of pharmacist-led management of chronic painin primary care evaluated in a pilot randomisedcontrolled trial (RCT) and to estimate optimal samplesize for a definitive RCTDesign Regression analysis of costs and effectsusing intention-to-treat and expected value of sampleinformation analysis (EVSI)Setting Six general practices Grampian (3) EastAnglia (3)Participants 125 patients with complete resourceuse and short form-six-dimension questionnaire(SF-6D) data at baseline 3 months and 6 monthsInterventions Patients were randomised to eitherpharmacist medication review with face-to-facepharmacist prescribing or pharmacist medicationreview with feedback to general practitioner ortreatment as usual (TAU)Main outcome measures Differences in mean totalcosts and effects measured as quality-adjusted lifeyears (QALYs) at 6 months and EVSI for sample sizecalculationResults Unadjusted total mean costs per patient werepound452 for prescribing (SD pound466) pound570 for review (SDpound527) and pound668 for TAU (SD pound1333) After controllingfor baseline costs the adjusted mean cost differencesper patient relative to TAU were pound77 for prescribing(95 CI minus82 to 237) and pound54 for review (95 CIminus103 to 212) Unadjusted mean QALYs were 03213for prescribing (SD 00659) 03161 for review (SD00684) and 03079 for TAU (SD 00606) Relative toTAU the adjusted mean differences were 00069 forprescribing (95 CI minus00091 to 00229) and 00097for review (95 CI minus00054 to 00248) The EVSIsuggested the optimal future trial size was between460 and 690 and between 540 and 780 patients perarm using a threshold of pound30 000 and pound20 000 perQALY gained respectivelyConclusions Compared with TAU pharmacist-ledinterventions for chronic pain appear more costlyand provide similar QALYs However these estimatesare imprecise due to the small size of the pilot trialThe EVSI indicates that a larger trial is necessaryto obtain more precise estimates of differences inmean effects and costs between treatment groups

Trial registration number ISRCTN06131530

INTRODUCTIONChronic pain is recognised as a common andlong-term condition having a significantimpact on the quality of life of individualsand their families along with a considerablefinancial cost to society Its pathophysiologyand psychosocial impact are largelyindependent of its biological aetiology andtherefore comprehensive approaches toidentifying designing and targeting relevantinterventions have been advocated wherebychronic pain is dealt with as a whole clinicalentity regardless of its underlying cause1

Chronic pain is characterised by pain whichpersists despite adequate time for healingAlthough there is no clear definition it isoften defined as pain that has been presentfor more than 3 months2 Approximately 20of the European adult population havechronic pain3 and the economic burden is

Strengths and limitations of this study

This is the first study to assess the differences incosts and effects (in terms of quality-adjustedlife years (QALYs)) associated with pharmacistprescribing andor pharmacist review comparedwith treatment as usual (TAU) in the manage-ment of chronic pain in primary care

The results are imprecise as they were based ona pilot randomised controlled trial and a largertrial based on these results is now needed

Using a value of information approach to deter-mine optimal sample findings suggest a futuretrial requires between 460 and 690 and between540 and 780 patients per arm using a thresholdof pound30 000 and pound20 000 per QALY gainedrespectively in order to obtain more precise esti-mates of differences in mean QALYs and meancosts between treatment groups

Neilson AR et al BMJ Open 20155e006874 doi101136bmjopen-2014-006874 1

Open Access Research

considerable with national healthcare and socioeconomiccosts running into billions annually representingbetween 3 and 10 of European gross domesticproduct (GDP)4 and from $560 to $635 billion annuallyin the USA5 In the UK an estimated 116 of adultssuffer from severe chronic pain representing around 45million people The average annual incidence risk esti-mated from health surveys of developing chronic pain is83 and for those already with chronic pain (ie preva-lent cases) the average annual recovery rate is 547

The total cost of back pain alone has been estimated tobe pound123 billion annually in the UK or 15 of GDP8

The majority of patients with chronic pain aremanaged outside the specialist setting mainly in primarycare however patients will make a variety of choices forself-management of their pain3 Pharmacological andnon-pharmacological management options exist withprovision of the former involving mostly primary care orpurchase of over-the-counter medicines from the com-munity pharmacy or supermarket the latter optionsinclude cognitive behavioural therapy physiotherapy oralternative therapies and are often provided privatelyThe recent Scottish Intercollegiate Guidelines Network(SIGN) guideline on chronic pain1 reviewed the evi-dence for assessment and management of chronic non-malignant pain in the adult population in the non-specialist setting and reported on the current lack ofgood quality evidence to guide management The guide-line indicated a range of recommendations relating topharmacological therapies including for example theprovision of at least an annual assessment of patients onpharmacotherapy for chronic painWe recently reported findings from the lsquoPharmacist-

led management of chronic pain in primary carersquo(PIPPC) pilot trial in which pharmacist medicationreview with face-to-face pharmacist prescribing was com-pared with pharmacist medication review with feedbackto the general practitioner (GP) or usual GP-led care9

This was the first pilot randomised controlled trial(RCT) to compare key clinical outcomes (rather thanprocess end points) between pharmacist-led interven-tions for adults with chronic pain with the usual GPcare Although it was a pilot trial to inform the designfor a subsequent definitive trial the findings suggestedthat compared to usual care participantsrsquo pain levels asmeasured by the chronic pain grade (CPG) improvedin both pharmacist-led intervention groups particularlyin the group managed by pharmacist prescribing Thepilot trial also collected data on resource use andhealth-related quality of life This provided an opportun-ity to assess the impact of pharmacist-led interventionson health economic outcomes and to offer some insightinto the potential value for money of the study interven-tions The objective of this paper therefore is to calcu-late the differences in mean costs and effects of theinterventions and the precision of those estimates so asto inform sample size and other design considerationsfor a future definitive trial

METHODSOverview of the PIPPC trialThe PIPPC study9 was a UK-based 6-month open explora-tory pilot RCT primarily designed to investigate theeffectiveness of either pharmacist medication review andface-to-face pharmacist prescribing or pharmacist medi-cation review and feedback to GP relative to treatment asusual (usual GP-led care) The study was performed insix general practices with an attached independentpharmacist prescriber (3 in Grampian and 3 in EastAnglia) Full details have been reported previously9

Briefly a total of 196 patients were recruited into the trialbetween March and June 2010 Included patients wereover 18 years of age living in their own home and receiv-ing regular prescribed medication for pain (defined asreceiving within the previous 120 days either two or moreacute prescriptions andor one repeat prescription foran analgesic and or a non-steroidal anti-inflammatorydrug (NSAID)) Patients were randomised in a 111 ratioto receive either pharmacist medication review withpharmacist prescribing (n=70) or pharmacist review only(n=63) or TAU (n=63) Patients and their healthcare pro-viders were unblinded due to the nature of the interven-tion Trial outcomes assessed by self-report in postalquestionnaires at baseline and at 3 and 6 monthsrsquofollow-up included the short form mental (SF-12 MCS)and physical (SF-12 PCS) component summary scoresthe generic preference-based health state utilitiesmeasure short form-six-dimension questionnaire(SF-6D) the CPG and the Hospital Anxiety andDepression Scale (HADS) Preference weights (utilities)for the SF-6D health status measure were derived usingthe recommended tariffs derived from a UK generalpopulation survey10 Data on health service resource usewere also collected at three time points baseline (tomeasure service use 3 months prerandomisation)3 months (to measure resource use 3 months postrando-misation) and 6 months (to measure resource use3ndash6 months postrandomisation) as described below9

Resource use and unit costsThe economic analysis included the direct costs asso-ciated with the intervention (pharmacist trainingpharmacist and GP time involved in delivering the inter-vention along with related follow-up) pain-related hospi-talisation (number of hospital inpatient days day casesand outpatient visits) primary care visits for chronic pain(GP nurse healthcare assistant appointments) primarycare telephone contacts for chronic pain prescribed andnon-prescribed OTC pain-related medicationsFrequency of the above primary and secondary care

visits and names and quantity of medicine prescribedwere measured and valued using information frompatientsrsquo medical records and by patient self-reportedquestionnaire Standard unit costs from published UKsources were attached to these items of resource useResource use quantities and the assigned unit costs usedin this analysis are summarised in table 1 For study

2 Neilson AR et al BMJ Open 20155e006874 doi101136bmjopen-2014-006874

Open Access

medications unit costs were obtained from the BritishNational Formulary11 Hospital inpatient and outpatientunit costs were based on average costs from the ScottishHealth Service Costs Book12 Intervention-relatedpharmacist and GP time were based on costs from thePersonal Social Services Research Unit (PSSRU)13

Intervention-related costs (pharmacist training etc)was based on information obtained from follow-up semi-structured interviews with pharmacists and GPs in par-ticipating practicesOther costs borne by patients (eg OTC medicines

travel) or their carers and productivity losses to society

Table 1 Unit costs applied to NHS resource use in the PIPPC study

Resource use item Value (pound) Unit Sourcecomments

Intervention costs

Pharmacist training Prescribing

pound1085

Review pound1067

Per patient PSSRU 2009201013 2 days (ie16 h) pound40h

(includes travel)=pound40times16=pound640 this cost being

apportioned over the number of patients in each

intervention group

Intervention delivery Prescribing

pound667

Review pound44

Per patient PSSRU 2009201013 pound40h (includes travel)

Prescribing mean duration 100 min Review mean

duration 66 min

Pharmacist follow-up

appointments

Prescribing

pound618

Review pound0

Per patient PSSRU 2009201013 pound40h

Prescribing 4553 (849) patients received

follow-up appointments 0 (151) 1 (623)

mean duration 11 min 2 (151) mean duration

10 min 3 (75) mean duration 12 min Most were

conducted by phone Weighted average cost=

pound618

Review no follow-up appointments

GP review pound459 Per patient PSSRU 2009201013 pound106h of GMS activity

Mean duration 26 min

Primary care costs

GP-home pound108 Per consultation PSSRU 2009201013

GP-surgery pound32 Per consultation PSSRU 2009201013

GP-telephone pound20 Per consultation PSSRU 2009201013

Nurse-home pound20 Per consultation PSSRU 2009201013

Nurse-surgery pound12 Per consultation PSSRU 2009201013

Nurse-telephone pound750 Per consultation PSSRU 2009201013 (based on an assumed

multiplier for telephone consultations for nurse and

healthcare assistant of 0625 that is 2032

Other-home pound9 Per consultation PSSRU 2009201013 (clinical support worker

nursingmdashcommunity)

Healthcare

assistant-surgery

consultation

pound750 Per consultation Based on an assumed multiplier for telephone

consultations for nurse and healthcare assistant of

0625 that is 2032

Hospital costs

Inpatient stay Orthopaedics=

pound873

General

medicine=pound395

Per bed day ISD Scotland Scottish Health Service Costs NHS

Grampian year ended 31 March 201012

R040 (ampLS)mdashSpecialty Group CostsmdashInpatients

Day case Pain relief=pound541 (Net) per case ISD Scotland Scottish Health Service Costs NHS

Grampian year ended 31 March 201012 R042mdash

Specialty CostsmdashDay Cases

Outpatient clinicmdash

consultant led

Pain relief=pound122 Per attendance ISD Scotland Scottish Health Service Costs NHS

Grampian year ended 31 March 201012 R044mdash

Specialty CostsmdashConsultant Outpatient Clinics

Outpatient clinicmdashother

(eg physiotherapist)

Pain relief=pound37 Per attendance ISD Scotland Scottish Health Service Costs NHS

Grampian year ended 31 March 201012 R046mdash

Specialty CostsmdashAHP Outpatient Clinics

Medications Varied by

medication

Dose duration pack

size number supplied

etc

BNF 61 March 201111

PIPPC study Pharmacist-led management of chronic pain in primary care PSSRU Personal Social Services Research Unit NHS NationalHealth Service

Neilson AR et al BMJ Open 20155e006874 doi101136bmjopen-2014-006874 3

Open Access

were considered outside the scope of the NationalHealth Service (NHS) perspective of the present ana-lysis All costs were in UK pounds sterling (pound) undis-counted (as the duration of the trial was less than1 year) The price year used for all costs was 20092011

EffectsThe effects of the pharmacist-led interventions were esti-mated as gain in quality-adjusted life years (QALYs)Based on the SF-6D data collected at baseline 3-monthand 6-month follow-up the number of QALYs over6 months follow-up associated with each study arm wereapproximated by applying the area under the curve(AUC) method (implemented by summing the areas ofthe geometrical shapes obtained by linearly interpolat-ing between utility scores over the study period)10

QALYs were undiscounted due to the short duration ofthe trial

Methods used to estimate differences in mean costs andeffectsSimilar to the main analysis of clinical outcomes9 theeconomic analysis was conducted on an intention-to-treat basis for participants with complete SF-6D dataat baseline 3-month and 6-month follow-up Whileresource use information was available for 178 patientsonly 125 patients had complete SF-6D data for all threestudy assessment time-points These 125 patients formedthe sample for the main analysis Using the datadescribed above estimates of the total mean costs andtotal mean QALYs per group were calculated Regressionanalyses calculated the differences (and associated 95CIs) in mean total costs and differences in mean totalQALYs per patient while controlling for differences intotal costs at baseline baseline SF-6D14 and other base-line patient characteristics (age sex marital status workstatus education income baseline CPGmdashintensity)

Sensitivity analysesWe performed a number of sensitivity analyses toexplore areas of uncertainty in the main analysis Firstas noted above approximately one-third of patients wereexcluded due to missing some SF-6D data at follow-up(no missing cost data was apparent and so imputationwas not necessary) To assess whether this introducedbias we performed a sensitivity analysis with imputedvalues for missing SF-6D data and re-ran the regressionanalysis using data from all 178 patients The missingSF-6D data were imputed by multiple imputation viachained equations with the STATA 13 lsquomi imputersquocommand programme to account for patients withmissing data at either 3 or 6 months assuming that datawere missing at random (and creating five imputed datasets)15 16 Second to investigate the impact on differen-tial costs from patients who experienced very high costsfrom the analysis we conducted an analysis excludinghospital inpatient costs on the basis that some of thesecosts may not be directly related to chronic pain Finally

we explored whether the main results were affected bycontrolling only for baseline differences between groupsin total costs and SF-6D (ie excluding sociodemographicand economic factors and other health status measuresie the CPG)

EVSI analysisThe expected value of sample information (EVSI)17 18

was calculated to assess whether conducting a largerRCT of pharmacist prescribing versus TAU or pharma-cist review versus TAU would be worthwhile Thisapproach determines the optimal sample size that maxi-mises the difference between the expected total cost(Total Cost) of a future trial and the monetary value ofthe information that it provides (EVSI) The EVSI is theanticipated value of the health gain (or the expectednet benefit of sample information (lsquoExpected Net GainrsquoENG)) over the period of t years the chosen timehorizon used for the analysis This time horizon repre-sents how long we expect the lsquonew intervention(s)rsquo tobecome established as the lsquostandard of care in the UKrsquoshould it be shown to be cost effective Total Cost is cate-gorised into three main elements a fixed component torepresent the human (ie labourstaff resource use) andother resources associated with the setting up andrunning of the trial a variable component to reflect perpatient accrual (ie per additional patient recruited orincreases in each participant taking part and who mightpotentially benefit from the intervention) follow-up anddata collection costs and an opportunity costmdashhealthloss to participants randomised to the control group(assuming the intervention is effective) To generateEVSI estimates the difference in effectiveness costs andassociated variances and covariances observed within thePIPPC trial were used In addition assumptions weremade regarding trial fixed costs variable costs and thenumber of future patients with chronic pain expected tobenefit from the proposed pharmacist prescribing andreview intervention For our analysis we made the fol-lowing assumptions time horizon=30 years annual inci-dence=540 000 (this value was calculated based on theaverage annual incidence of chronic pain is 83 inadults aged 25 years and older7 and the number ofadult population is around 42 million (last UK census)that gives 35 million This implies that approximately12 million (13times35 million) people develop chronicpain that is ongoinglong-standing (ie lastinggt3months) On the basis of the PIPPC trial9 only a pro-portion of these people were receiving medicationmdashwecalculate this to be 1492 people from a total of 3281 or45 So the assumed incidence used 540 000) a fixedtrial cost=pound14 million and a variable trial cost=pound220 Wegenerated EVSI estimates using a willingness to pay perQALY gained of pound20 000 and pound30 000 respectivelyAdditional sensitivity analyses checked whether the

EVSI results were robust to variations in the main ana-lysis assumptions including the underlying incidencerate the number of patients who would benefit and the

4 Neilson AR et al BMJ Open 20155e006874 doi101136bmjopen-2014-006874

Open Access

recruitment cost per patient to a trial Specifically wehalved and doubled the values used in the main analysisin each alternative case with the exception of timehorizon (1020 years)

RESULTSResource use and costsTable 2 presents the number of patients using each typeof resource (and percentages) and per patient meanquantities used mean unadjusted cost per type ofresource unadjusted total mean costs and the adjustedcost differences by trial arm from baseline to 6-monthfollow-up As the table shows there was some variationin the resources used and associated costs betweenpatients in each study arm For example the resultssuggest that GP resource use is higher in the reviewgroup relative to either prescribing or TAU groups andalso that nurse practice consultations by phone is higherin the prescribing group relative to review and TAUThe bar charts in figure 1 show per patient the differ-

ent cost components as a proportion of the total(unadjusted) costs in each study arm At both baselineand follow-up medications accounted for the largestpercentage of the total cost in all study arms (prescrib-ing 37 review 31 TAU 55) outpatient hospitalisa-tions for TAU (19) intervention-related costs forpharmacist prescribing (18) and primary care costsexcluding pharmacist visit costs for pharmacist reviewonly (20) Both pharmacist-led intervention arms wereless costly than TAU based on the raw unadjusted meantotal costs The TAU group however was also the mostcostly treatment group observed prerandomisation andthis was largely driven by medication costs this suggestspotential imbalances between study arms at baseline(figure 1) Following adjustment for differences in base-line costs and controlling for other baseline patientcharacteristics (age sex marital status work status edu-cation income baseline CPGmdashintensity and baselineSF-6D) yielded positive incremental mean costs differ-ences for both pharmacist-led interventions relative toTAU prescribing pound775 (95 CI minuspound817 to pound2367) andreview pound544 (95 CI minuspound1033 to pound2121) In otherwords the review and prescribing groups relative to TAUwere now more expensive rather than cheaper relativeto usual care Adjusting for baseline costs were largelyresponsible for this resulting change which was statistic-ally significant (with a regression coefficient p=00000)No other variables reached significance

Quality-adjusted life yearsTable 3 shows the SF-6D health utility scores by study armThe values were broadly similar across all arms at eachtime point although pharmacist prescribing showedslightly higher SF-6D scores at all three time pointsBoth pharmacist-led intervention arms generated

slightly more QALYs than TAU based on the rawunadjusted mean total QALYs but the magnitude of

QALY gains relative to TAU were small (approximately001 extra QALYs) in both casesAfter adjusting for baseline SF-6D score baseline costs

and controlling for other baseline patient characteristics(age sex marital status work status education incomebaseline CPGmdashintensity) QALY gains relative to TAUwere largely unchanged for pharmacist prescribing00069 (minus00091 to 00229) and pharmacist review00097 (minus00054 to 00248)

Sensitivity analysesThe sensitivity analysis with missing SF-6D data imputed(ie 125 out of 178 complete 53 incomplete so imputed)by using multiple imputations (MI) produced similarvalues to the main analyses for estimated difference inmean total QALYs however differences in QALYs wereslightly larger for pharmacist prescribing versus TAU00065 (minus00075 to 00205) than pharmacist review versusTAU 00047 (minus00086 to 00181) Using the full resourceuse data set and rerunning the cost regression analysisshowed the two pharmacist-led interventions still to bemore costly than TAU though the magnitude of theadjusted difference in cost was reduced in the prescribinggroup and increased in the review group relative to TAUprescribing pound21 (minus124 to pound167) and review pound75 (minus72 topound221) Excluding from the total costs the few patients ineach arm having hospital inpatient care (hospital stayand day cases) reduced raw unadjusted mean total costsper patient by around pound90 pound130 and pound40 in the pharma-cist prescribing pharmacist review and TAU armsrespectively The adjusted differences in mean total costs(after controlling for baseline costs SF-6D age sexmarital status work status education income and CPGmdashintensity) were increased for both pharmacist-led inter-vention groups relative to TAU prescribing pound112 (pound24 topound200) and review pound88 (minus8 to pound185) Controlling only forbaseline costs and baseline SF-6D (ie excluding age sexmarital status work status education income baselineCPGmdashintensity) in regression analyses increased incre-mental costs for both pharmacist-led interventions rela-tive to TAU prescribing pound125 (pound82 to pound242) and reviewpound76 (minuspound40 to pound192) this however had little impact onQALY gains relative to TAU prescribing 00017 (minus00127to 00160) and review 00040 (minus00099 to 00179)

Expected value of sample information analysisThe EVSI investigated the cost-effectiveness of carryingout a larger RCT of pharmacist prescribing and pharma-cist review versus TAU The parameters used in the EVSIcalculation and the resulting expected EVSI is given infigure 2 which also shows the expected costs of the RCTand the resultant ENG The expected costs of runningthe proposed trial rise linearly as the sample size (n) isincreased however the expected benefits diminish as thesample size increases after a certain point giving theoptimal number of patients in each arm of the RCT Thelsquooptimal trial sizersquo (SS) is estimated as 780 per arm (pre-scribing vs TAU comparison) and 540 per arm (review vs

Neilson AR et al BMJ Open 20155e006874 doi101136bmjopen-2014-006874 5

Open Access

Table

2MeanNHSresourcesusedandassociatedcosts

perpatientfrom

baselineto

6monthsfollow-upin

thePIPPC

study(complete

caseanalysis

n=125)

Prescribing(n=39)

Review

(n=44)

TAU

(n=42)

Resourceuseitem

n(

)Quantity

Costpound

n(

)Quantity

Costpound

n(

)Quantity

Costpound

Intervention-related

Pharm

acisttraining(hours)

39(100)

16

109

44(100)

16

107

NA

ndashndash

Pharm

acistactivitiesrelatedto

interventiondelivery

(mins)

39(100)

100dagger

667

44(100)

66Dagger

44

NA

ndashndash

Pharm

acistfollow-up(m

ins)

39(100)

93sect

62

44(100)

0ndash

NA

ndashndash

GPreview

ofpharm

acistrecommendations(m

ins)

NA

`ndash

44(100)

26

459

NA

ndashndash

Primary

care

GPconsultations-surgery

25(64)

26

845

32(73)

35

113

30(71)

32

1024

GPconsultations-phone

4(10)

13

25

13(30)

18

353

8(19)

24

475

GPconsultations-home

0(0)

ndashndash

1(2)

10

108

1(2)

60

648

Practicenurseconsultations-surgery

9(23)

24

293

10(23)

27

324

9(21)

20

24

Practicenurseconsultations-home

0(0)

ndash18

0(0)

ndashndash

1(2)

10

20

Practicenurseconsultations-phone

13(33)

16

1211

3(7)

17

125

3(7)

10

75

HCA-home

1(3)

20

18

2(5)

20

18

3(7)

13

12

HCA-surgery

9(23)

46

342

13(30)

37

277

11(26)

18

1364

Secondary

care

Hospitalinpatientstay(days)

2(5)

20

790

2(5)

35

1621

1(2)

20

790

Hospitalinpatientdaycase

1(3)

10

541

1(2)

10

541

4(10)

10

541

Hospitaloutpatient-consultantled

7(18)

29

3486

14(32)

20

244

15(36)

26

3172

Hospitaloutpatient-other

6(15)

18

678

7(16)

27

1004

4(10)

33

1203

Medications

39(100)

Variouspara

1652

43(98)

Variouspara

1814

42(100)

Variouspara

3649

Unadjustedtotalcostsmean(SD)daggerdagger

4522

(4660)dagger

5697

(5266)

6682

(13334)

Adjusteddifferencein

totalcosts

versusTAUmean(95

CI)DaggerDagger

775

(minus817

to2367)sectsect

544

(minus1033

to2121)sectsect

Pharm

acisttrainingtimeapportionedacrossallpatients

ineachstudygroup

daggerPharm

acisttime(prescribinggroup)spentontasksincludingface-to-faceconsultationsrecord-basedmedicationreviewcareplanpreparationmeetingGP

DaggerPharm

acisttime(review

group)spentontasksincludingrecord-basedmedicationreviewcareplanpreparationandmeetingGP

sectBasedonPIPPC

studydata

around85

patients

had(between1and3)follow-upvisitsin

theprescribinggroupwithaweightedaveragedurationacrossallpatients

of93

min

paraOwingto

thelargenumberofdifferenttypesofmedicationsused(totaln=48)onlysummary

costdetails

are

presentedhere

Raw

unadjustedmeantotalcosts

overthetotalpatients

perstudygroup

daggerdaggerForcomparisontheraw

unadjusted(3

monthsperiod)prerandomisationtotalmeancosts

bystudygroupwereprescribing=pound3648review=pound4366TAU=pound6247

DaggerDaggerEstimatesfrom

regressionanalyseswithadjustm

entfordifferencesin

baselinecostsbaselineSF-6D

healthutilityscore

andotherpatientcharacteristics(agesexmaritalstatuswork

status

educationincomebaselineCPGmdashintensity)Thenumberofpatients

withdata

onallthesebaselinevariablesprescribing(n=35)review

(n=39)TAU

(n=34)

sectsectCostdifferencesare

calculatedasthedifferencein

costbetweentheinterventiongroup(prescribingorreview)andtheTAU

group

CPGchronic

pain

gradeGPgeneralpractitionerHCAhealthcare

assistantNHSNationalHealthServicePIPPC

studyPharm

acist-ledmanagementofchronic

pain

inprimary

careTAU

treatm

entasusual

6 Neilson AR et al BMJ Open 20155e006874 doi101136bmjopen-2014-006874

Open Access

TAU comparison) using a cost per QALY threshold ofpound20 000 and 690 per arm (prescribing vs TAU compari-son) and 460 per arm (review vs TAU comparison) usinga cost per QALY threshold of pound30 000The results of the sensitivity analyses around the main

EVSI results (see table 4) showed that this finding wasgenerally robust to assumptions about annual incidenceand recruitment cost but sensitive to annual accrualand the time horizon used

DISCUSSIONThis current study estimated the differences in meancosts and mean effectiveness (in terms of QALYs) of

pharmacist medication review with or without prescrib-ing as compared with usual GP care for the treatment ofchronic pain in primary care based on data from thePIPPC pilot trial9 We also assessed the EVSI for a defini-tive trial The results suggested that the pharmacist inter-ventions were likely to be more costly than TAU whendifferences in costs at baseline were taken into accountbut there was a large degree of uncertainty surroundingthe estimates of differences in mean costs and meaneffects (ie QALYs) Given this uncertainty the economiccase for pharmacist prescribing with or without reviewover TAU can only be established with a larger RCTThe potential gain from a future trial was estimated bythe EVSI this showed that the gains in monetary termsfrom a future trial exceed the expected costsIn the main clinical paper9 the CPG was shown to

have potential to be able to discriminate betweenpatients who improved postintervention and suggestedmaximum benefit for those in the pharmacist prescrib-ing arm (effect size of 045)9 However the difference ineffect observed with the CPG was not reflected in a(large) effect difference in terms of QALYS We believea larger trial is now warranted to determine more pre-cisely the true effects of pharmacist medication reviewwith or without prescribing measured in terms of CPGand QALYsTo the best of our knowledge no other published

studies have assessed the costs and effects (in terms ofQALYs) of pharmacist interventions for chronic pain orpharmacist independent prescribing for any conditionA small feasibility study19 examining the impact of intro-duction of a combined nursepharmacist-led clinic formanaging chronic pain in primary care was reported tolead to improvements in management of pain as well asa reduction in use of secondary care resources and highrates of satisfaction Stewart et al20 have argued for moreresearch into the clinical economic and humanisticeffects of different forms of lsquonon-medical prescribingrsquo(NMP) NMP involves the prescribing of medicines byhealthcare professionals (who are not doctors egnurses pharmacists) but who have undergone appropri-ate training (including pharmacovigilance) and aretherefore qualified to prescribe either in a supplemen-tary or independent role within their areas of compe-tence20 Our study results provide new informationabout the costs and QALY effects of pharmacist medica-tion review with or without prescribing for chronic painOne important aspect of the design of the cost effect-

iveness component of RCTsmdashincluding those evaluatingpharmacist interventions for painmdashrelates to the selec-tion of the most appropriate health utility measuresThe SF-6D was used in the current trial An alternative isthe EQ-5D and this has become the instrument ofchoice for many agencies including the NationalInstitute for Health and Care (formerly lsquoClinicalrsquo)Excellence (NICE)21 The SF-6D cannot easily be com-pared with the EQ-5D because of differences in theirdescriptive systems values applied to health states and

Figure 1 Proportion of unadjusted total mean costs per

patient prerandomisation (ie baseline) and at 6 months

follow-up by each main cost component and study arm

Neilson AR et al BMJ Open 20155e006874 doi101136bmjopen-2014-006874 7

Open Access

contextual basis (ie recall period) Moreover a numberof studies have reported that these two measures cannotbe used interchangeably in patients with chronicpain22ndash27 In a study of 389 patients with chronic kneepain Barton et al28 reported that in contrast to theEQ-5D the SF-6D was unable to discriminate betweenpatients who improved postintervention and those whodid not but in another study of patients with chroniclow back pain and degenerative disc disease25 the SF-6Dhad the best ability to detect changes and correctly iden-tify patients as improved or non-improved In a UKgeneral population study29 both measures were shownto discriminate between those self-reporting chronic

pain or no chronic pain with and without neuropathiccharacteristics However both measures generatedwidely different health utility scores for the same patientgroups for example the mean utilities for severe pain(EQ-5D 033 vs SF-6D 058) In another study30 theEQ-5D was found to be more responsive to deteriorationin health whereas the SF-6D was more responsive toimprovement in health in patients with knee pain dueto inflammatory arthritis The choice of instrumentmight have a considerable impact on the conclusionsreached about the cost-effectiveness of the interventionsbeing evaluated For instance a Norwegian cost-effectiveness study of total disc replacement versus

Table 3 SF-6D health utility scores and QALYs over 6 months follow-up in the PIPPC study (complete case analysis n=125)

Prescribing (n=39) Review (n=44) TAU (n=42)

SF-6D at baseline mean (SD) 06349 (01336) 06173 (01431) 06077 (01140)

SF-6D at 3 months mean (SD) 06428 (01396) 06411 (01469) 06226 (01405)

SF-6D at 6 months mean (SD) 06500 (01462) 06291 (01471) 06105 (01336)

Unadjusted total QALYs mean (SD) 03213 (00659) 03161 (00684) 03079 (00606)

Adjusted difference in total QALYs versus TAU

mean (95 CI)

00069 (minus00091 to 00229) 00097 (minus00054 to 00248)

The number of patients with data on all these baseline variables prescribing (n=35) review (n=39) TAU (n=34)Estimates from regression analyses with adjustment for differences in baseline costs baseline SF-6D health utility score and other patientcharacteristics (age sex marital status work status education income baseline CPGmdashintensity)QALYs quality-adjusted life years SF-6D short form six-dimension TAU treatment as usual

Figure 2 Value of expected value of sample information (EVSI) analysis The expected net benefit of sampling at specified

randomised controlled trial (RCT) sizes using a threshold of pound20 000 per quality-adjusted life years (QALY) gained for

(A) Prescribing versus treatment as usual (TAU) (C) Review versus TAU and using a threshold of pound30 000 per QALY gained for

(B) Prescribing versus TAU and (D) Review versus TAU From regression analysis with adjustment for baseline costs baseline

short form six-dimension (SF-6D) and other patient characteristics (including age sex material status education work status

income chronic pain grade (CPG)-intensity) The number of participants with data on all these baseline variables prescribing

(n=35) review (n=39) TAU (n=34) ENG expected net gain EVSI expected value of sample information n number of patients

in each arm SS sample size

8 Neilson AR et al BMJ Open 20155e006874 doi101136bmjopen-2014-006874

Open Access

multidisciplinary rehabilitation in patients with chroniclow back pain found that total disc replacement was cost-effective when EQ-5D was used but not when SF-6D wasused26 Additionally the practical issue of instrumentcompletion is important in future trial design In ouranalysis we found around one-third of SF-6D measure-ments were incomplete Similar results have been foundin other studies in patients with pain where rates ofcompletion were significantly better for the EQ-5D overthe course of an RCT in patients with non-specific neckpain27 and in patients with low back pain and degenera-tive disease25 Finally it should be noted that a new5-level version of the EQ-5D has recently beenlaunched31 and its application in patients with variouschronic conditions (eg osteoarthritis) suggests it mighthave improved discriminative capacity and sensitivity tochange than the EQ-5D 3-level version32 The use ofboth the EQ-5D and the SF-6D might be the bestapproach in future trials in patients with chronic painIn conclusion the present study suggests that

pharmacist-led medication review with or without pre-scribing in patients with chronic pain in primary carehad similar effects in terms of QALYs compared withtreatment as usual but was more expensive These resultshowever are highly uncertain due to the small samplesize of the pilot trial and do not reflect the previouslyreported significantly improved chronic pain grade TheEVSI analysis indicated that this current evidence is insuf-ficient for decision-making and that a future larger trial isworthwhile Such a trial would require between 460 and690 patients and between 540 and 780 patients per armusing a threshold of pound30 000 and pound20 000 per QALYgained respectively in order to obtain more precise esti-mates of differences in mean effects (QALYs) and meancosts between treatment groups

Twitter Follow Margaret Watson at MagsWatson1

Acknowledgements The authors thank the participating patients practicesand pharmacists The authors would also like to thank Annie Blyth who wasthe lead researcher in East Anglia Kirsten Harrild (Medical Statistics UOA) forstatistical support Rick Adams (School of Pharmacy UEA) helped design anddeliver the pharmacist training and Lesley Thomson (NHS Grampian) helped

design the pharmacist data collection forms The patient postal questionnairewas based on work by Nicola Cooper and the Norfolk Arthritis Register(NOAR) research team The Pharmacy-Led Management of Chronic PainStudy Team acknowledges the financial support of NHS Research Scotland(NRS) through Scottish Primary Care Research Network Northeast The workwas conducted as part of the Aberdeen Pain Research Collaboration Allauthors had access to all of the study data and can take responsibility for theintegrity of the data and the accuracy of the data analysis

Contributors CMB conceived the study and led the writing of proposal forfunding provided day to day supervision of study conduct analysis andinterpretation of data and co-led the writing of the paper including revisionsand final draft AME PCH RH AJL PM BHS DW and MW werecoinvestigators contributed to writing of the proposal oversight of the studyconduct data analysis and interpretation drafting of the paper and approvedthe final version PM had particular responsibility for the health economiccomponents and AJL for the statistical analyses BHS and AME providedspecific advice on chronic pain and its measurement and DW provided theexpertise on the design and delivery of the pharmacist training HBcoordinated the two study sites and was lead researcher for Grampian Shewas responsible for daily study conduct and coordination acquisition of thedata analysis producing of tables and figures and interpretation of data Sheco-led the drafting of the clinical outcomes paper with CMB and approved thefinal version ARN undertook the analysis and interpretation of the healtheconomic data and lead the drafting of the economic research article Allauthors participated in a critical revision of the manuscript and have approvedthe final version

Funding The project was funded by the Medical Research Council (grant ID85356) The Health Economics Research Unit is partly funded by the ChiefScientist Office of the Scottish Government Health and Social Care Directorates

Competing interests None

Patient consent Obtained

Ethics approval North of Scotland Research Ethics Service (ref 09S0801107) NHS Research and Development approval was granted by NHSGrampian and East Norfolk amp Waveney Research Governance CommitteesPatients gave informed consent before taking part

Provenance and peer review Not commissioned externally peer reviewed

Data sharing statement Consent was not obtained from participants for datasharing and so cannot be released to third parties We welcome howeveropportunities to work collaboratively with other investigators to use the datafor further analyses

Open Access This is an Open Access article distributed in accordance withthe terms of the Creative Commons Attribution (CC BY 40) license whichpermits others to distribute remix adapt and build upon this work forcommercial use provided the original work is properly cited See httpcreativecommonsorglicensesby40

Table 4 EVSI sensitivity analyses

WTP of pound20 000 per QALY gained WTP of pound30 000 per QALY gained

Prescribing vs TAU Review vs TAU Prescribing vs TAU Review vs TAU

Base case 780 540 690 460

Annual incidence halved 780 540 690 460

Annual incidence doubled 780 540 690 460

Recruitment cost halved 780 540 690 460

Recruitment cost doubled 780 540 690 460

Annual accrual rate halved 550 370 320 320

Annual accrual rate doubled 1110 770 670 670

Time horizon 10 years 300 300 250 250

Time horizon 20 years 440 430 370 370

Estimates rounded to the nearest 10EVSI expected value of sample information TAU treatment as usual WTP willingness to pay

Neilson AR et al BMJ Open 20155e006874 doi101136bmjopen-2014-006874 9

Open Access

REFERENCES1 Scottish Intercollegiate Guidelines Network (SIGN) Management of

chronic pain Edinburgh SIGN 2013 (SIGN publication no 136)httpwwwsignacuk

2 Classification of chronic pain Descriptions of chronic painsyndromes and definitions of pain terms Prepared by theInternational Association for the Study of Pain Subcommittee onTaxonomy Pain Suppl 19863S1ndash226

3 Breivik H Collett B Ventafridda V et al Survey of chronic pain inEurope prevalence impact on daily life and treatment Eur J Pain200610287ndash333

4 Breivik H Eisenberg E OrsquoBrien T on behalf of OPENMinds Theindividual and societal burden of chronic pain in Europe the case forstrategic prioritisation and action to improve knowledge andavailability of appropriate care BMC Public Health 2013131229

5 Gaskin DJ Richard P The economic costs of pain in the UnitedStates J Pain 201213715ndash24

6 The National Pain Audit Report 2012 httpwwwnationalpainauditorgmediafilesNationalPainAudit-2012pdf

7 Elliott AM Smith BH Hannaford PC et al The course of chronicpain in the community results of a 4-year follow-up study Pain200299299ndash307

8 Maniadkis N Gray A The economic burden of back pain in the UKPain 20008495ndash103

9 Bruhn H Bond CM Elliott AM et al Pharmacist-led management ofchronic pain in primary care results from a randomised controlledexploratory trial BMJ Open 20133e002361

10 Brazier J Roberts J Deverill M The estimation of a preference-basedmeasure of health from the SF-36 J Health Econ 200221271ndash92

11 Royal Pharmaceutical Society of Great Britain British NationalFormulary 61 London BMJ Group 2011 httpwwwbnforgbnf

12 Scottish Health Services Costs 2010 (financial year 200910) httpwwwisdscotlandorghealth-topicsfinancepublications

13 Curtis L Unit costs of health and social care 2010 Personal SocialServices Research Unit Cantebury University of Kent 2010 httppssruacuk

14 Glick HA Jalpa A Doshi JA et al Economic evaluation in clinicaltrials (Handbooks in Health Economic Evaluation) Oxford OxfordUniversity Press 2007

15 Rubin DB Inference and missing data Biometrika 197663581ndash9216 Rubin DB Multiple imputation for non-response in surveys

New York John Wiley Sons 198717 Willan AR Clinical decision-making and the expected value of

information Clinical Trials 20074279ndash8518 Willan AR Goeree R Boutis K Value of information methods for

planning and analysing clinical studies optimize decision making andresearch planning J Clin Epidemiol 201265870ndash6

19 Briggs M Closs SJ Marczewski K et al A feasibility study of acombined nursepharmacist-led chronic pain clinic in primary careQual Prim Care 20081691ndash4

20 Stewart D MacLure K George J Educating nonmedical prescribersBr J Clin Pharmacol 201274662ndash7

21 National Institute for Health and Clinical Excellence Guide to themethods of technology appraisal London National Institute forHealth and Clinical Excellence 2008

22 Grieve R Grishchenko M Cairns J SF-6D versus EQ-5D reasonsfor differences in utility scores and impact on reported cost-utilityEur J Health Econ 20091015ndash23

23 Obradovic N Lal A Liedgens H Validity and responsiveness ofEuroQol-5 dimensions (EQ-5D) versus Short Form-6 dimension(SF-6D) questionnaire in chronic pain Health Qual Life Outcomes201311110

24 Soslashgaard R Christensen FB Videbaek TS et al Interchangeabilityof the EQ-5D and the SF-6D in long-standing low back pain ValueHealth 200912606ndash12

25 Johnsen LG Hellum C Nygaard OP et al Comparison of theSF6D the EQ5D and the Oswestry disability index in patients withchronic low back pain and degenerative disc disease BMCMusculoskelet Disord 201314148

26 Johnsen LG Hellum C Storheim K et al Norwegian Spine StudyGroup Cost-effectiveness of total disc replacement versusmultidisciplinary rehabilitation in patients with chronic low back paina Norwegian multicentre RCT Spine 20143923ndash32

27 Whitehurst DG Bryan S Another study showing that twopreference-based measures of health-related quality of life (EQ-5Dand SF-6D) are not interchangeable But why should we expectthem to be Value Health 201114531ndash8

28 Barton GR Sach TH Avery AJ et al Comparing the performance ofthe EQ-5D and SF-6D when measuring the benefits of alleviatingknee pain Cost Effectiveness Resource Allocation 2009712

29 Torrance N Lawson KD Afolabi E et al Estimating the burden ofdisease in chronic pain with and without neuropathic characteristicsdoes the choice between the EQ-5D and SF-6D matter Pain20141551996ndash2004

30 Harrison MJ Davies LM Bansback NJ The comparativeresponsiveness of the EQ-5D and SF-6D to change in patients withinflammatory arthritis Qual Life Res 2009181195ndash205

31 Herdman M Gudex C Lloyd A et al Development and preliminarytesting of the new five-level version of EQ-5D (EQ-5D-5L) Qual LifeRes 2011201727ndash36

32 Janssen MF Simon Pickard A Golicki D et al Measurementproperties of the EQ-5D-5L compared to the EQ-5D-3L across eightpatient groups a multi-country study Qual Life Res2013221717ndash27

10 Neilson AR et al BMJ Open 20155e006874 doi101136bmjopen-2014-006874

Open Access