-
The Core Components of Cardiac Rehabilitation for Health Related
Quality of Life in Coronary Heart Disease
Patients: A Systematic Review and Meta-Analysis of Randomized
Controlled Trials
by
Troy Anthony Francis
A thesis submitted in conformity with the requirements for the
degree of Master of Science
Graduate Department of Pharmaceutical Sciences University of
Toronto
© Copyright by Troy A. Francis, 2016
-
ii
The Core Components of Cardiac Rehabilitation for Health
Related Quality of Life in Coronary Heart Disease Patients:
A
Systematic Review and Meta-Analysis of Randomized Controlled
Trials
Troy Anthony Francis
Master of Science
Graduate Department of Pharmaceutical Sciences
University of Toronto
2016
Abstract
Background: Cardiac rehabilitation (CR) is a comprehensive
program offered to patients with coronary
heart disease (CHD). The aim of this study was to assess the
effectiveness of providing any core
component of CR on health related quality of life (HRQOL) in
adult patients with CHD.
Methods: We performed a systematic review, meta-analysis and
meta-regression of randomized
controlled trials examining the core components of CR.
Identified sources were published between
database inception and July 16th, 2014. Outcomes included
overall, physical, emotional and social
HRQOL. Outcomes were reported as standardized mean change (SMC)
with 95% confidence intervals.
Results: Summary effect sizes were (SMC 0.14; 95% CI 0.03 to
0.25), (SMC 0.23; 95% CI 0.08 to 0.38),
(SMC 0.11; 95% CI -0.03 to 0.24) and (SMC 0.03; 95% CI -0.07 to
0.13) for overall, physical, emotional
and social HRQOL respectively.
Conclusion: Receiving any CR intervention was shown to improve
overall and physical HRQOL.
-
iii
Acknowledgments
I would like to acknowledge several people for guiding and
encouraging me during my Masters
work. Foremost, I would like to thank my supervisor, Dr. Murray
Krahn, and my advisor, Dr.
Valeria Rac, for providing me with this opportunity, and
continuous support and motivation.
They both frequently inspired me and allowed me to develop
independent thinking and research
skills, which has greatly assisted me throughout this whole
process and will continue to do so in
the future.
Besides my advisors, I would like to express my gratitude to the
rest of my thesis committee:
Petros Pechlivanoglou for his time, patience and mentorship;
Nicholas Mitsakakis for his
encouragement and knowledge; and David Alter for his valued
criticism during each step of my
research.
I wish to also thank my mentor, Nader Kabboul, who continually
strives to make me a better
person. Without his direction, support and dedication none of
this would have been possible.
Additionally, I would like to recognize Joanna Bieleki and the
Toronto Health Economics
Technology Assessment Collaborative for their backing and
commitment to this study.
Last but not the least, I would be remiss to not express my
appreciation to my family and friends,
who have supported me on this journey and who constantly inspire
me to accomplish more.
-
iv
Table of Contents
Acknowledgments..........................................................................................................................
iii
Table of Contents
...........................................................................................................................
iv
List of Tables
................................................................................................................................
vii
List of Figures
..............................................................................................................................
viii
List of Appendices
.........................................................................................................................
ix
Chapter 1 Introduction
.....................................................................................................................1
1 Introduction
.................................................................................................................................1
1.1 Cardiac Rehabilitation and Secondary Prevention Programs
..............................................3
1.1.1 Core Components of Cardiac Rehabilitation
...........................................................4
1.1.2 Cardiac Rehabilitation’s Proposed Mechanism of Action
.......................................6
1.1.3 Complexity of Cardiac Rehabilitation
.....................................................................6
1.1.4 Psychosocial Outcomes of Cardiac Rehabilitation
..................................................7
1.1.5 Physical Outcomes of Cardiac Rehabilitation
.........................................................8
1.2 Patient Reported Outcomes and Cardiac Rehabilitation
......................................................8
1.2.1 Health Related Quality of Life Instruments
...........................................................10
1.2.2 Health Related Quality of Life and Cardiac Rehabilitation
...................................12
1.2.3 Long-Term Sustainability of HRQOL after Cardiac
Rehabilitation ......................13
1.3 Summary
............................................................................................................................14
1.4 Aim
....................................................................................................................................15
1.5
Hypothesis..........................................................................................................................15
Chapter 2 Methods
.........................................................................................................................16
2 Methods
.....................................................................................................................................16
2.1 Eligibility Criteria
..............................................................................................................16
2.1.1 Study Design
..........................................................................................................16
-
v
2.1.2 Information Sources
...............................................................................................17
2.1.3 Search Strategy
......................................................................................................18
2.1.4 Study Selection and Screening Process
.................................................................18
2.1.5 Data Collection and Extraction
..............................................................................19
2.1.6 Risk of Bias and Quality Assessment
....................................................................19
2.2 Conceptualization of HRQOL Domains
............................................................................19
2.2.1 Challenges in Pooling Heterogeneous HRQOL Data
............................................20
2.3 Statistical Analysis
.............................................................................................................21
2.3.1 Methods for Pooling Heterogeneous Data
.............................................................21
2.3.2 Investigating Sources of Heterogeneity
.................................................................22
2.3.3 Outcome
Measurements.........................................................................................23
2.3.4 Health Related Quality of Life
...............................................................................24
2.3.5 Meta
Regression.....................................................................................................24
2.3.6 Demographics
........................................................................................................26
2.3.7 Data Synthesis
........................................................................................................26
3 Results
.......................................................................................................................................28
3.1 Study Demographics
..........................................................................................................28
3.1.1 Risk of Bias Assessment
........................................................................................29
3.2 Health Related Quality of Life
...........................................................................................30
3.2.1 Overall Health Related Quality of Life
..................................................................30
3.2.2 Physical Health Related Quality of Life
................................................................31
3.2.3 Emotional Health Related Quality of Life
.............................................................32
3.2.4 Social Health Related Quality of Life
....................................................................32
3.3 Meta-Regression
................................................................................................................33
3.3.1 Overall Health Related Quality of Life
..................................................................33
3.3.2 Physical Health Related Quality of Life
................................................................33
-
vi
3.3.3 Emotional Health Related Quality of Life
.............................................................34
3.3.4 Social Health Related Quality of Life
....................................................................34
4 Discussion
.................................................................................................................................34
4.1 Health Related Quality of Life
...........................................................................................35
4.2 Meta
Regression.................................................................................................................36
4.3 Strengths of Study
..............................................................................................................37
4.4 Study Limitations
...............................................................................................................37
4.5 Implications for Practice
....................................................................................................39
4.6 Future Directions
...............................................................................................................40
4.7 Conclusion
.........................................................................................................................41
Tables
.............................................................................................................................................42
Figures............................................................................................................................................56
Appendices
.....................................................................................................................................79
-
vii
List of Tables
Table 1: Baseline Demographics
..................................................................................................
42
Table 2: HRQOL Instruments Used by Investigators
...................................................................
43
Table 3: Overall Health Related Quality of Life (Meta-Analysis)
............................................... 48
Table 4: Physical Health Related Quality of Life (Meta-Analysis)
.............................................. 49
Table 5: Emotional Health Related Quality of Life
(Meta-Analysis) ........................................... 50
Table 6: Social Health Related Quality of Life (Meta-Analysis)
................................................. 51
Table 7: Overall Health Related Quality of Life Regression
Output............................................ 52
Table 8: Physical Health Related Quality of Life Regression
Output .......................................... 53
Table 9: Emotional Health Related Quality of Life Regression
Output ....................................... 54
Table 10: Social Health Related Quality of Life Regression
Output............................................ 55
-
viii
List of Figures
Figure 1: PRISMA Flow Diagram Preferred Reporting Items for
Systematic Reviews and Meta-
Analyses (PRISMA)
.....................................................................................................................
56
Figure 2: Risk of Bias Graph: review authors’ judgments about
each risk of bias item presented
as percentages across all included studies
....................................................................................
57
Figure 3: Risk of Bias Summary: review authors’ judgments about
each risk of bias item for each
included study
...............................................................................................................................
58
Figure 4: Overall Health Related Quality of Life Forest Plot
....................................................... 61
Figure 5: Overall Health Related Quality of Life Funnel Plot
...................................................... 62
Figure 6: Physical Health Related Quality of Life Forest Plot
..................................................... 63
Figure 7: Physical Health Related Quality of Life Funnel Plot
.................................................... 64
Figure 8: Emotional Health Related Quality of Life Forest Plot
.................................................. 65
Figure 9: Emotional Health Related Quality of Life Funnel Plot
................................................. 66
Figure 10: Social Health Related Quality of Life Forest Plot
....................................................... 67
Figure 11: Social Health Related Quality of Life Funnel Plot
...................................................... 68
-
ix
List of Appendices
Appendix 1 Literature Search Strategies
..................................................................................
79
Appendix 2 Standardized Mean Change Formula
....................................................................
85
Appendix 3 List of Validated HRQOL Instruments for use in CHD
patients .......................... 86
Appendix 4 Characteristics of Included Studies
.......................................................................
91
Appendix 5 List of Excluded Studies (Full-Text Review Subset
Only) .................................. 93
Appendix 6 R Source Code
....................................................................................................
164
-
1
Chapter 1 Introduction
1 Introduction
Cardiovascular disease (CVD) is the leading cause of mortality
and a major cause of disability
across the globe in both adult men and women. CVD refers to a
myriad of diseases involving the heart,
blood vessels and poor blood flow due to the hardening and
narrowing of vascular walls leading to a
diseased cardiovascular system (1, 2). CVD accounts for an
estimated 17.5 million deaths per year
globally and is expected to exceed 23.6 million by 2030 (3, 4).
In Canada CVD claims more than 66
000 lives per year, which equates to one death every 7 minutes
(5). The burden of CVD does not just
affect low and middle income countries but is increasingly a
global health issue. With the advent of
globalization diets have changed and more people are consuming
refined processed foods and as a
consequence are adopting a fast food culture. Increases in risk
factors such as hypertension, obesity,
dyslipidemia and a sedentary lifestyle all contribute to the
increase in CVD morbidity seen around the
world (6).
The global cost of CVD, including direct and indirect costs, is
estimated to be US$863 billion,
and is projected to rise to US$1044 billion by 2030 (1).
Currently the annual direct and indirect costs
of CVD per year in Canada are estimated to be $20.9 billion.
(7). By 2020, it is estimated that the total
costs will reach $28.3 billion per year (8).
CVD represents more than one condition. CVD can be broken down
into coronary heart disease
(CHD), stroke and congestive heart failure (CHF). A stroke
occurs due to cerebral ischemia which
occurs when blood stops flowing to any part of the brain. Most
strokes are ischemic and are caused by
blockages or a clot in blood vessels due to a buildup of plaque
which causes damage to brain cells
which cannot be repaired. Stroke has been estimated to have
caused 5.7 million deaths a year in 2005
-
2
and the number of global deaths is projected to be 6.5 million
in 2015 and 7.8 million in 2030 (9). CHF
can manifest itself in a variety of different ways but ischemic
heart disease is thought to be a major risk
factor for CHF (10). CHF is known to have two subcategories;
left ventricular dysfunction and
preserved ejection fraction, with preserved ejection fraction
CHF having a better prognosis (11). CHF
is estimated to have a prevalence of 23 million worldwide, with
a lifetime risk of developing CHF of
one in five (10).
CHD is the most common form of heart disease and reflects the
narrowing of the blood vessels
supplying the heart muscle due to artheroscelorosis and presents
symptoms as angina, acute coronary
insufficiency, and myocardial infarction (MI)(3, 12). The World
Health Organization estimated that
CHD accounted for 7.4 million deaths a year globally according
to 2012 statistics (3). In Canada CHD
is one of the leading causes of death which claims more than 33
600 lives per year (13). CHD
morbidity continues to rise globally through an increased number
of post MI patients living longer
with CHD symptoms due to improvements in cardiac care (2, 12).
The economic burden of CHD when
accounting for direct and indirect costs worldwide was estimated
to be $108.9 billion and is predicted
to reach $218.7 billion by 2030 (4). In the current economic
state the costs of CHD in Canada are
estimated to be $11 billion when measuring the economic burden
of illness and are forecasted to reach
$17 billion by 2020 (7).
While there is no direct cure for CHD there are many possible
treatments directed toward
slowing disease progression and preventing secondary events.
Treatments for CHD include
pharmaceutical interventions, surgical procedures, lifestyle
modification through secondary prevention
services and cardiac rehabilitation (CR). For those with
established CHD pharmaceutical treatments of
statins, beta-blockers, angiotensin-converting enzyme inhibitors
and anti-platelet medicine are
commonly prescribed to reduce the risk of MI (3). Patients with
severe CHD can receive surgical
operations such as coronary artery bypass graft surgery (CABG)
or percutaneous transluminal
-
3
coronary angioplasty (PCI) to treat the artherosclerosis and
improve blood flow but the progression of
CHD will not change without lifesyle modification and drug
therapy (14).
1.1 Cardiac Rehabilitation and Secondary Prevention Programs
CR is a secondary prevention program that aims to prolong
survival from acute disease
manifestations through an improvement in day to day
functionality (2). Secondary prevention services
work to reduce cardiac mortality and morbidity, through
pharmacological therapy, surgical
interventions, and risk factor modification (15). CR programs
have been created and promoted as a
way to recovery following acute coronary events and is defined
as “…the coordinated sum of
interventions required to ensure the best physical,
psychological and social conditions so that patients
with chronic or post-acute cardiovascular disease may, by their
own efforts, preserve or resume
optimal functioning in society and, through improved health
behaviours, slow or reverse progression
of disease” (16, 17). The use of CR can decrease the burden of
CVD through a reduction in risk factors
and an improvement in functionality, while being found to be
effective in patients with diagnosed
acute coronary syndrome, heart failure and those who have
undergone coronary revascularization (18,
19).
The American Heart Association (AHA) and the American
Association of
Cardiovascular and Pulmonary Rehabilitation (AACVPR) conclude
that CR programs should
offer a multidisciplinary approach to CHD risk reduction and
that programs should consist of
more than just exercise training alone (20, 21). While exercise
training is an important factor in
CR, it is one element of many different therapies. The AHA and
the AACVPR recommend that
all CR and secondary prevention services should contain specific
core components that utilize
baseline patient assessment, nutritional counselling, risk
factor management (lipids, blood
-
4
pressure, weight, diabetes mellitus, and smoking cessation),
psychosocial interventions, physical
activity counseling and individualized exercise training (20,
21).
1.1.1 Core Components of Cardiac Rehabilitation
The core components of CR are an essential part of the
contemporary care for patients
with CHD (22). Defining the core components of CR provides a
foundation for programs
around the world to be build on, which can then be tailored for
specific settings and populations.
The first core component is baseline patient assessment. This
includes a physical examination to
determine the extent of the patients comorbidities and assesses
their perceived health status.
This information can then be used to create individualistic
treatment plans for each patient based
on their medical history and direct the implementation of
further core components of CR (21).
The next core component is nutritional counselling performed by
a registered dietitian or trained
health professional. This involves educating and prescribing
specific dietary modifications to
closely match the therapeutic lifestyle change diet. The
therapeutic lifestyle change diet
recommends the reduction of saturated fats, trans fats,
cholesterol and sodium while increasing
the intake of fruits, vegetables, whole grains and fish into the
diet (23).
The following core component is risk factor management which
targets the modifiable
CVD risks such as hypertension, dyslipidemia, obesity, and a
sedentary lifestyle (6). Weight
management regimens create achievable goals to reduce body
weight in patients with a BMI >
25kg/m2
and/or waist circumference > 40 inches in men and > 35
inches in women. Blood
pressure management involves drug therapy and lifestyle
modification in hypertensive patient’s
( ≥ 140 mm Hg systolic or ≥ 90 mm Hg diastolic). Lipid
manangement is aimed at discovering
and treating those patients with dyslipidemia by obtaining
fasting cholesterol, lipoprotein and
triglyceride levels (21, 24). Management of dyslipidemia
involves drug therapy and dietary
-
5
changes (increasing plant sterol, fibre, and omega-3 fatty acid
intake) (21). Diabetes
management screens for the presence of diabetes in all
participants and if present educates the
patient on treatment options stressing the importance of
complaince to diet, drug therapy and
blood sugar monitoring. Smoking cessation is offered to patients
who are current smokers and
past smokers who have quit in the preceding 12 months and are
likely to relapse. Patients are
given one on one or family counselling by a trained health
professional to assist the smoker in
quitting and preventing relapse. Patients are offered
pharmaceutical support in the form of
nicotine replacement therapy, bupropion or varenicline (24).
The subsequent core component is psychosocial management which
is delivered by
registered psychologists or trained healthcare workers.
Psychosocial management is designed to
identify psychological distress due to CHD using standardized
instruments (22). In cases of
suspected depression or anxiety individual or group counselling
was offered to patient’s and
family. Psychological interventions that are available include
stress management which uses
cognitive behavioural strategies to help patient’s reduce or
manage stress, as well as relaxation
and self-instruction training (25, 26). Physical activity has
long been known to have a positive
effect on improving ones overall health and well being (6). The
ensuing core component is
physical activity counselling. This includes the assessment of
current physical activity levels
through questionnaires and pedometers and addresses readiness to
change barriers to physical
activity. This information is then used by counsellors who
encourage patients to gain 30-60
minutes of moderate intensity physical activity atleast 5 days a
week and warns patient’s of
spontaneous vigourous physical activity risks (27).
The final core component is indivdualized exercise training.
This component of CR is
based on the original baseline patient assessment from the
physical examination. The
-
6
individualized exercise regimen is modeled on a rough aerobic
exercise prescription of 3-5 days
a week at 50 - 80% of the patients exercise capacity for 20 – 60
minutes a session using any
continous modality such as walking or cycling. For resistance
exercise patients are advised to
workout 2-3 days a week performing 1 – 3 sets of 10 – 15
repetitions performing calistenics,
free weights, and band exercises (21, 27).
1.1.2 Cardiac Rehabilitation’s Proposed Mechanism of Action
The main mechanistic evidence for how CR would work to improve
one’s health
revolves around the exercise component. For patients with CHD
exercise training has direct
benefits on the heart and coronary vasculature. Aerobic exercise
has been shown to improve
myocardial oxygen demand, endothelial function, and autonomic
tone, while reducing
inflammatory markers and clotting factors (28). The hypothesis
is that the other core
components work to help reduce mortality and improve day to day
functioning through an
indirect improvement in risk factors (lipids, smoking and blood
pressure)(6, 29).
1.1.3 Complexity of Cardiac Rehabilitation
CR is a comprehensive program and shares many characteristics of
a complex
intervention as defined by the Medical Research Councils 2000
Guidelines for developing and
evaluating complex interventions. In order for an intervention
to be “complex” it needs to have a
number of interacting components, requires a number and
difficulty of behaviors by those
delivering or receiving the intervention, there has to be a
variability of outcomes and a degree of
flexibility or tailoring of the intervention is permitted
(30).
Each secondary prevention service used in CR is distinct but
when used together creates
the whole of the intervention. The ‘whole’ intervention refers
to CR as a single entity and relates
to its ability to influence important health behaviors
associated with CVD greater than the
-
7
individual use of the core components (31). The concept of
“complex” interventions is relatively
new and as such there is a debate on how these interventions
should be described and evaluated
(31). It is important to note that complex interventions are
formed from many different parts
which could be material, human, theoretical, social or
procedural in nature but are synergistic
when brought together (30, 31). CR should be thought of as the
sum of its parts and each
component should be individually researched and evaluated for
their overall benefit.
1.1.4 Psychosocial Outcomes of Cardiac Rehabilitation
The psychosocial and behavioral changes associated with CR are
complicated.
Psychosocial dysfunction which is characterized as depression,
anxiety and or social isolation is
normally seen in patients receiving CR (22). In order to
determine if there was an association
between psychosocial disorders and cardiovascular events a large
randomized multicenter trial,
Enhanced Recovery in Coronary Heart Disease Patients (ENRICHD),
was performed (32). The
ENRICHD trial was conducted using 2481 MI patients (1084 women,
1397 men) with
diagnosed major or minor depression and low social support.
Patients were randomized to a
cognitive behavoural psychosocial intervention or usual medical
care and treated with selective
serotonin reuptake inhibitors, when indicated. The objective of
this landmark study was to
determine whether treating depression and increasing social
support in patients who recently
suffered an MI would reduce the risk of recurrent non fatal
infarctions and sudden death (33).
The ENRICHD intervention did not improve event-free survival in
comparison to patients
receiving usual medical care. However, depression and social
isolation improved in both groups
but psychosocial interventions were unable to modify CHD (22,
33). In practice, it is generally
accepted that both men and women with varying degrees of CHD
benefit from CR in terms of
quality of life and well-being.
-
8
1.1.5 Physical Outcomes of Cardiac Rehabilitation
Exercise based CR has been shown to improve physical outcomes in
most groups of
patients with established CHD through an improvement in
cardiovascular function leading to
improved strength and fitness (14). Exercise based CR has also
been shown to significantly
reduce all-cause (13-25%) and cardiovascular mortality (26-37%)
based on several systematic
reviews and meta-analysis (6, 8, 29). While, recent studies West
et al, 2012 and Anderson et al,
2016 have stated that there is no significant difference in
patients referred to CR in terms of
mortality these studies were underpowered and/or using outdated
study designs (34, 35). More
recent unpublished research using indirect treatment comparisons
has reinforced CRs ability to
reduce all cause and cardiovascular mortality (36).
CHD patients receiving psychological and education based
interventions alone with no
associated exercise program have shown little or no influence on
mortality or hospitalization
(25, 37). However, contemporary CR has transitioned from
exercise only interventions to more
comprehensive secondary prevention programs that utilize all of
the core components and has
been shown to provide the same overall mortality reduction as
exercise based CR (22).
1.2 Patient Reported Outcomes and Cardiac Rehabilitation
The importance of highlighting patient centered care in
designing and implementing a
comprehensive CR program allows for greater attention to sudden
changes in overall health.
Working to improve a patients health status is an aspects of CR
which is extremely important. A
patients health status is composed of their burden of symptoms,
functional limitations and health
related quality of life (HRQOL) (38). The burden of symptoms a
patient has to deal with refers
to the type and frequency of symptoms that has manifested in
relation to their disease or the
-
9
medical treatment, while a patients functional status includes
their physical, mental and social
limitations (39).
HRQOL is a multidimensional concept that represents a patient’s
perception of the
discrepancy between actual and desired functional status and the
overall impact of disease on
their own well-being (39, 40). An individual’s HRQOL is affected
by factors such as
impairments, functional stress, perceptions and social
opportunities and influenced by disease,
injury and treatment (41). Each patient has a varying degree to
which symptoms, functional
limitations and medical interventions influence their well-being
causing HRQOL to only
accurately be quantified by patient self-report (39).
Health status as related to quality of life consists of four
domains that are important
measures for cardiac survivorship and provide prognostic
information which reflect the aims of
CR (40). These four domains reflect subjective assessments of
physical, emotional, and social
functioning as well as global perceptions on health (42). These
four domains are conceptually
different but there is an overlap between outcomes because it is
rare for an illness or disease to
affect only one domain.
Patient-reported outcomes (PROs) are any reports coming directly
from patients about
how they feel in relation to a health condition and its therapy,
without interpretation of their
responses by a clinician, or aid (43). PROs are important
because they provide the patients
perspective on treatment benefit and provide another opportunity
to measure treatment benefit
beyond survival, disease and physiological markers. They can
also be used to report on
treatment satisfaction, HRQOL and compliance to treatments (44).
PROs are sometimes used as
primary outcomes in clinical trials, predominantly when no other
substitute measure of direct
benefit such as survival or death is available. Although
clinical trials are incorporating more
-
10
PROs they are currently underused, and are usually used as
secondary add on measures (39).
The lack of PROs being used as primary outcomes in clinical
trials may be due to a perception
that these outcomes are “soft” and may not be useful in clinical
practice or interpretable (38).
However, in previous prospective studies a patients health
status has been shown to be a strong
independent predictor for health outcomes such as mortality,
cardiovascular events and
hospitalization (45).
1.2.1 Health Related Quality of Life Instruments
PROs such as HRQOL can be collected using instruments that are
disease and condition
specific or generic in nature. There are two varieties of
measures that are currently used to
collect HRQOL scores. The first collection of instruments are
health profiles which measure
HRQOL using individual scores of dimensions or domains and
psychometric profiles which use
one or multiple scales to measure patient chracteristics or
attributes. The second type of
instrument which can be used to conseptualize HRQOL are
preference/utilty based measures
which can estimate HRQOL scores using either direct or indirect
methods (46). Direct measures
of utility can be achieved through asking respondents to trade
health states for different risks of
death or remaining years of life. While, indirect utilities can
be achieved through HRQOL
questionnaires using weights or tariffs. Each item on the
instruments questionnaire measures an
aspect of HRQOL, but in order for an instrument to detect
significant effects of a treatment it
must be valid, reliable, responsive and interpretable (47).
The reliability of a tool refers to its capacity to produce
dependable results over time
rated by Cronbachs’s alpha statistics. Test-retest reliability
is an aspect of an instrument which
is a critical factor in making it consistent. Test-retest
reliabilty measures if the repeated
administration of a measure to patients at different time points
yields similar results (48).
-
11
Validity of an instrument seeks to determine the extent to which
the tool measures what it is
intented to measure. Validity is assessed using criterion, face
and construct validity. Criterion
validity seeks to measure the accuracy of the instrument in
comparision to a gold standard. An
instrument has face validity if it contains items that reflect
the specific disease or condition
being examined. While an instrument has construct validity if it
is consistant with the concepts
being measured and relates to other tools (46, 48). The
responsiveness and sensitivity of an
instrument is measured by its ability to detect clinically
important changes in health status even
when they are small (47).
There is an overabundance of different instruments which could
be used to measure
HRQOL in CHD patients, but not all measurement scales are equal
and useful. The MacNew
HRQOL instrument is a disease specific measure for use in
post-MI patients with CHD and has
previously been shown to be valid, reliable and responsive in
both interviewer and self
administered modes. (49). Disease specific health surveys
measure symptom burden, functional
limitations and HRQOL related to a specific disease state.
Condition specific measures describe
patient symptoms or experiences related to a specific condition
or a particular intervention or
treatment (39). The 36-Item Short Form Health Survey (SF-36) is
one of the most readily used
and recognizable generic surveys. Generic measures are designed
for use with any population
sample and summarize overall functional well being but do not
give any information about
symptoms and functional limitations related to a specific
desease. Generic HRQOL instruments
allow for comparisons between impact of treatments across
diseases or conditions (44).
Preference based measures can also be used to measure health
outcomes as a supplement to both
health profiles. The Euroqol quality of life scale (EQ-5D) is an
indirect standardized and
validated measure which can be converted into a utility score
and is applicable to many
conditions and treatments (50). Utility refers to the
desirability or preference that individuals
-
12
exhibit for a condition (51). Patient health utilties are
measured for various possible health states
and range between 0-1 (0 death, 1 perfect health) and represents
ones overall health state (52).
When assessing HRQOL in patients with CHD both a disease
specific and generic measure
should be used allowing for a more comprehensive assessment of
health status in patients
receiving CR.
1.2.2 Health Related Quality of Life and Cardiac
Rehabilitation
Over the years CR has been said to improve HRQOL when adherence
rates to the
program are high by decreasing disease specific symptoms and
increasing functional capacity
(53). Some improvement in HRQOL are attributed to the natural
recovery process after cardiac
events, but CR has been shown to assist patients in reaching
HRQOL scores similar to the
population norm (53). However, for all of the positive results
following CRs ability to reduce
morbidity and mortality there are no meta-analyses which report
on HRQOL in CHD patients
because of the heterogeneity in outcome measures and
inconsistency in the reporting of
findings. A low HRQOL measured at baseline prior to CR has been
shown to be one of the
strongest independent predictors of an improved response to CR
(41).
There has been a lack of consistency in the reporting of CRs
effect on HRQOL domains.
While it is anecdotally thought that receiving any core
component of CR would improve
HRQOL in patients with CHD, it has not been systematically
proven. Prior research into
physical dimension outcomes has shown that CR may improve
physical functioning and
physical well-being when compared to controls who were not
receiving structured exercise
therapy (54, 55). When comparing hospital based CR to home CR
interventions no significant
difference in physical HRQOL domains were observed (54, 56). The
age of the patients
engaging in CR may have an effect on their sensitivity to CR
interventions. Older patients are
-
13
usually underrepresented in CR programs and may have an
improvement in physical HRQOL
(57, 58). Lie et al, 2009 demonstrated that even interventions
with no exercise component given
to CABG patients could improve physical domain HRQOL at 6 month
follow-up, but no
between group differences were observed in comparison to the
control.
Exploration into previous emotional or psychological domains has
not revealed
consistent outcomes. On average CR was shown to improve state
anxiety scores in patients but
did not have a significant effect on depression scores (56, 58).
The psychological effects of CR
are difficult to quantify, while there are increases in
particularly the SF-36 mental component
score there are no between group differences (59, 60). It is
possible that psychological domain
effects may be represented in the physical domain (40). In terms
of the social domain relatively
few trials employed instruments which aimed to measure social
constructs. Dalal et al, 2007,
Robinson et al, 2011 and Roncella et al, 2013 reported social
subscales scores of the MacNew
instrument and found there was no significant difference between
groups at follow-up, however
improvements were reported. The same phenomenon was seen in
other studies using the SF-36
social functioning subscale (34, 59, 61). The lack of consistent
findings in the psychological and
social HRQOL domains could possibly be due to the sensitivity of
the instrument or a lack of
adherence to the intervention by patients. Generic measures have
primarily been used in all
trials evaluating HRQOL in CHD patients, most notably the SF-36
because of its ability to be
administered quickly. Generic measures may lack the sensitivity
to change with cardiac
treatment in comparison to disease specific measures (29).
1.2.3 Long-Term Sustainability of HRQOL after Cardiac
Rehabilitation
Even though there is a vast amount of evidence showing the
benefit of CR, research has
not yet considered the effectiveness of CR programs in terms of
long term sustainability of
-
14
health status. There are very few randomized controlled trials
that observe the effects of CR
with a follow-up greater than one year which could possibly show
its effects long term. There
have been many reports of improvements in health status achieved
in the first year following the
intervention that were reduced as time went on. This was
demonstrated by Murchie et al, 2004
and Cupples et al, 1999 who found that at 4-5 years follow-up
improvements in health status
and HRQOL were reduced and no longer statistically significant
between groups (62, 63). The
mode of exercise has also been associated with longer term
sustainability of HRQOL benefits
when concurrent aerobic-strength training is performed in
relation to aerobic training only (64).
1.3 Summary
The use of CR is a key tool in decreasing the burden of CHD
through a reduction in
disease specific symptoms and increasing functional capacity.
Unfortunately, no systematic
review has examined the effects of the core components of CR on
overall HRQOL and HRQOL
domains among adult patients with CHD. The clinical
effectiveness of CR on long-term
outcomes such as HRQOL is an area which has not been regularly
explored. An overview of
cochrane reviews examining CR in CHD patients found that
comparing HRQOL findings in CR
studies is difficult because of the complexity of the
intervention, heterogeneity in the HRQOL
instruments, patient populations, and a lack of studies
reporting patient reported health status
(12, 29). The AHA advocates for the inclusion of patient
reported health status as a clinical
measure with an emphasis on using validated CHD specific
instruments (39). Improvements in
patients HRQOL following CR and secondary prevention programs
has not been consistently
reported. Accordingy, research is needed to explore the core
components of CR effect on
HRQOL domains.
-
15
1.4 Aim
The aims of the present study were to evaluate the effectiveness
of providing any core
component of CR delivered in the context of CR on overall,
physical, emotional, and social
HRQOL domains in adult patients with CHD. Additionally, this
study aimed to explore the
potential effect of the study-level predictors of CR and
secondary prevention programs on
HRQOL in patients with CHD using meta-regression.
1.5 Hypothesis
The current study was premised on the hypothesis that receiving
any core component of
CR would show an improvement on HRQOL domains based on exercise
based CR effect on
reducing CVD morbidity and improving functional capacity.
However, literature in this area is
limited but it is expected that whilst some HRQOL domains will
change based on patients
receiving some non-pharmacological secondary prevention services
it is theorized that exercise
will drive most of the improvement in HRQOL domains.
-
16
Chapter 2 Methods
2 Methods
A systematic review and meta-analysis of randomized controlled
trials (RCTs) and
cluster RCTs examining CR interventions for CHD patients was
performed (65).
2.1 Eligibility Criteria
2.1.1 Study Design
This systematic review included RCTs and cluster RCTs which
utilized a pretest-
posttest-control (PPC) design. In the PPC design participants
were randomized to the treatment
or control groups and each participant was measured before
(baseline) and after (follow-up) the
treatment has been administered (66). The use of repeated
measurements in the PPC design
allows each individual to be used as their own control, which
typically increases the power and
precision of statistical tests (67). The PPC design compares the
pre-post change in the treatment
group to the amount of change in the control.
Inclusion criteria: RCTs evaluating any core component of CR
delivered in the context
of CR which measured patients HRQOL at baseline and follow-up in
both the active and control
arms were included. Instruments could be generic or disease
specific but needed to be validated
for us in CVD patients, and encompass one or all of the relevant
HRQOL domains. Additionally
studies were required to have a minimum of six months follow up
in order to be included in the
analysis.
Exclusion criteria: Studies using non-validated instruments for
CVD patients were not
included. Studies which do not report both intervention and
control group arm HRQOL scores at
-
17
baseline and follow-up, as well as studies of participants who
completed cardiac rehabilitation
programs prior to randomization were also excluded.
Intervention: Based on the AHA and AACVPR scientific statement
for CR and
secondary prevention programs (21). The focus was on the core
components of CR: nutritional
counselling, risk factor management, psychosocial interventions,
patient education, physical
activity counseling and individualized exercise training (20,
21).
Participants: The population consisted of adult men and women,
who have had a
myocardial infarction (MI), have undergone revascularization
(coronary artery bypass graft
(CABG) or percutaneous coronary interventions (PCI)), and who
have angina pectoris or
coronary artery disease defined by angiography was included.
Patients with heart failure, heart
valve surgery, heart transplants or implanted with either
cardiac resynchronization therapy or
implantable defibrillators were excluded.
Comparators: The comparator (usual care / standard of care)
could include standard
medical care, such as drug therapy, but patients were not
randomized to receive any of the core
components of CR.
Setting: Hospital, community and home based settings.
Language: Only English language publications were included in
the review.
2.1.2 Information Sources
Eligible studies were identified through a systematic search of
the following databases:
Cochrane Central Register of Controlled Trials (CENTRAL), Health
Technology Assessment
(HTA), and Database of Abstracts of Reviews of Effects (DARE) in
The Cochrane Library,
-
18
MEDLINE, EMBASE, CINAHL, SCI-EXPANDED, PsychINFO and Web of
Science (WOS),
all from their inception to July 16th
, 2014.
Reference lists of all identified systematic reviews and
meta-analyses published since
inception of any of the above databases to July 16th
, 2014 were fully screened, and relevant titles
were imported for evaluation of their eligibility for this
systematic review.
2.1.3 Search Strategy
The search strategy was designed with reference to those of the
previous systematic
reviews evaluating the core components of CR (6, 8, 25, 29) and
was conducted by an
information specialist experienced in systematic reviews (68).
MEDLINE, EMBASE and
CINAHL were searched using a strategy combining selected MeSH
terms and free text terms
relating to the core components of CR and coronary heart disease
with RCT filters. The
MEDLINE search strategy was translated into the other databases
using the appropriate
controlled vocabulary as applicable. Consideration was given to
variations in terms used and
spellings of terms in different countries so that studies will
not missed by the search strategy
because of such variations. The detailed search strategy used is
this study is provided in
Appendix 1.
2.1.4 Study Selection and Screening Process
The titles and abstracts of all citations identified by the
electronic searches were
examined for possible inclusion by two reviewers (NNK and TAF)
working independently. Full
publications of potentially relevant studies were retrieved and
reviewed by two reviewers (NNK
and TAF) who then independently determined study eligibility
using a standardized inclusion
form. Any disagreements about study eligibility were resolved by
discussion and, if necessary, a
third reviewer (MDK) was asked to arbitrate. Studies were
excluded if they were a commentary,
-
19
editorial, or a letter to the editor. Masking was complete when
outcome assessors were
concealed. Patient or performing physician masking was not
deemed pertinent because of the
procedural nature of CR as an intervention.
2.1.5 Data Collection and Extraction
Data from included studies was extracted independently by two
reviewers (NNK and
TAF) using a standardized data extraction tool. For each trial
characteristics of the study, trial
population, intervention and outcome data were extracted and
cross-checked. If data were
presented numerically (in tables or text) and graphically (in
figures), the numeric data was used
because of possible measurement error when estimating from
graphs. Any discrepancies were
resolved by the third reviewer (MDK).
2.1.6 Risk of Bias and Quality Assessment
In order to assess the quality of the included studies two
reviewers (NNK and TAF)
independently assessed the risk of bias in included studies
using the Cochrane Collaboration’s
recommended tool. The Cochrane Collaboration’s tool is a
domain-based critical evaluation of
the following domains: sequence generation; allocation
concealment; blinding of outcome
assessment; incomplete outcome data; and selective outcome
reporting (69). Any disagreements
were resolved by a third reviewer (MDK). Assessments of risk of
bias were provided in the risk
of bias table and summary for each study.
2.2 Conceptualization of HRQOL Domains
A subsequent literature review was performed to determine which
HRQOL instruments
extracted from the retrieved studies were validated for use in
CVD patients. Each instrument
was assessed for the specific patient population in which
validation occurred as well as which
domains and subscales each instrument evaluated. The Quality of
Life after Myocardial
-
20
Infarction (QLMI), MacNew Heart Disease Questionnaire (MacNew),
Leiden Quality of Life
Questionnaire, Angina Pectoris Quality of Life Questionnaire
(AP-QLQ), Seattle Angina
Questionnaire (SAQ), The Myocardial Infarction Dimensional
Assessment Scale (MIDAS),
Quality of Life Index–Cardiac Version III (QLI), Short Form-36
(SF-36), Short Form -12 (SF-
12), Nottingham Health Profile (NHP), Dartmouth COOP Quality of
Life instrument, Duke
Activity Status Index (DASI) , Cantril Ladder of Life, Short
Form – 6D (SF-6D), Euroqol -5
Dimension (EQ-5D), and Time Trade Off (TTO) were deemed to be
valid.
Based on CRs statement to improve a patient’s physical,
psychological and social
conditions HRQOL outcomes were created to reflect these
purported changes. HRQOL
outcomes were stratified into overall, physical, emotional and
social HRQOL domains. Overall
HRQOL included perspectives on one’s life as a whole which also
encompasses the physical,
emotional and social domains. Physical HRQOL included
performance of self-care activities,
mobility and physical activities. Emotional HRQOL functions
included mental health and
emotional reactions, while social HRQOL included social
interactions, behaviours and isolation
(70).
2.2.1 Challenges in Pooling Heterogeneous HRQOL Data
CR is a complex intervention with many interacting components
(31). The heterogeneity
in cardiac rehabilitation programs (patient population, and core
components) and the multitude
of PRO instruments that can be used to measure HRQOL all create
problems when attempting
to pool the data. When performing a meta-analysis of HRQOL
outcomes challenges in
interpretation occur because of the different instruments used
to measure similar constructs (71).
Additionally, because of various measures used interpreting the
magnitude of the effect
becomes an issue from a clinical and decision making standpoint
(72).
-
21
The underlying difficulty when attempting to understand HRQOL
scores is providing a
meaningful interpretation of what those scores actually
represent. While there is consensus that
HRQOL is an important endpoint in clinical trials there is still
a glaring gap in how to use these
results in practice. In an attempt to make HRQOL results more
meaningful researchers have
begun to look at the minimum important difference (MID) which is
“the smallest difference in a
score in the domain of interest that patients perceive as
important, either beneficial or harmful,
and that would lead to a change in the patients management”(47).
While the concept of the MID
is a worthwhile attempt at helping to bring meaningfulness to
interpreting HRQOL results, we
need to be mindful that these are estimates prone to sampling
variation and influenced by the
patient population and should only be used as rough guidelines
(73).
In order to pool HRQOL scores two important requirements must be
met. First,
instruments scores must correlate with one another by measuring
similar constructs in order to
be combined. Second, each measure must have similar
responsiveness to change, even if small.
If instruments are less responsive than their counter parts
treatment effects will be
underestimated and heterogeneity may be incorrectly attributed
to differences in patients or
intervention (71, 72, 74).
2.3 Statistical Analysis
2.3.1 Methods for Pooling Heterogeneous Data
Researchers deal with the challenge of studies using multiple
HRQOL measures to
measure the same construct by using an effect size summary
statistic to standardize all scales to
a common metric (71). This research uses a repeated measure PPC
design of HRQOL scores
and as such the metric of standardized mean change (SMC) was
used. SMC is the mean pre-post
change in the treatment group minus the mean pre-post change in
the control group, divided by
-
22
the pretest standard deviation (66). This approach will provide
a single unit free estimate of
treatment effect in standard deviation units. The formulas used
to calculate sampling variances
can be found in Appendix 2.
𝑔 = 𝑔𝑇 − 𝑔
𝐶= 𝐶(𝑛𝑇 − 1)
�̅�𝑝𝑜𝑠𝑡,𝑇 − �̅�𝑝𝑟𝑒,𝑇
𝑆𝐷𝑝𝑟𝑒,𝑇
− 𝐶(𝑛𝐶 − 1)�̅�𝑝𝑜𝑠𝑡,𝐶 − �̅�𝑝𝑟𝑒,𝐶
𝑆𝐷𝑝𝑟𝑒,𝐶
Where x̅ pre, T and x̅ post, T are the treatment group pretest
and posttest means, SDpre, T is the
standard deviations of the pretest scores, C () is the
bias-correction factor to account for the
overestimation of the effect sizes in small samples, nT is the
size of the treatment group, x̅ pre C,
x̅ post C, SD pre C, and nC are the equivalent values for the
control group and g represents sample
standardized mean change effect size (66).
A random effects multilevel meta-analysis was performed in order
to account for some
of the heterogeneity that would be present when pooling these
results to find the summary effect
of the intervention. The random-effects model estimates the mean
of a distribution of effects.
Each study provides information about a different effect size,
and the random effects model
incorporates each effect size into a summary estimate. In order
to obtain the most precise
estimate of the summary effect in a random effects model both
the within-study and between
study variances, tau2 (T
2), need to be known (75).
2.3.2 Investigating Sources of Heterogeneity
To investigate sources of variability in meta-analyses one of
the most commonly utilised
methods to examine heterogeneity is meta-regression (76).
Meta-regression merges meta-
analytic techniques with linear regression principles to
determine whether a linear association
exists between explanatory variables and a comparative treatment
effect (77-79). In meta-
regression, the outcome variable is the effect estimate (SMC)
and the explanatory variables are
-
23
characteristics of the studies that might influence the size of
intervention effect (80). A random
effects meta-regression was used in this study to allow for
residual heterogeneity by assuming
the underlying effects follow a normal distribution (N) around
the predictive covariates (76).
The effect size (gi) was estimated by the treatment effects (yi)
in study i (i =,…,k). The
estimated variance (vi) of the treatment effects was assumed to
be known. In the random effects
model, X represents the matrix of study level covariates and
intercept, while β represents a
vector of the coefficients. The heterogeneity variance
parameter, T2, represents the between
study variance (76).
Fixed effect meta-regression
yi ~ N(gi ,vi)
gi = Xiβ
Random effect meta-regression
gi ~ N(Xiβ,T2)
Meta-regression is a useful tool for investigating sources of
heterogeneity in meta-
analysis when there are a large number of trials. However,
meta-regression is an observational
meta-analytic technique and thus cannot be used to draw
inferences about causal relationships
(76, 78).
2.3.3 Outcome Measurements
Study outcomes were measured at baseline (entry to CR) and at
follow-up (minimum 6
months). The collected variables included overall, physical,
emotional and social health related
quality of life, meta-regression proportion of variance
explained and demographic information.
-
24
2.3.4 Health Related Quality of Life
The instruments used to measure overall HRQOL were the Cantril
Ladder of Life,
MacNew, Leiden Quality of Life Questionnaire, AP-QLQ, QLI,
Dartmouth COOP Quality of
Life instrument, QLMI, SF-6D, EQ-5D, and TTO. To determine
physical HRQOL the MacNew,
Leiden Quality of Life Questionnaire, QLMI, DASI, AP-QLQ, SAQ,
MIDAS, QLI, SF-36, SF-
12, and the NHP were used. Each of these subscales measured an
aspect of physical functioning,
mobility or physical limitations. To evaluate emotional HRQOL
the SF-36, SF-12, MacNew
Heart Disease Questionnaire, Leiden Quality of Life
Questionnaire, QLMI, MIDAS, NHP, QLI,
and the AP-QLQ were used. These instruments all provided
subscales on emotional limitations
or mental health. While, the social domain was created using
social functioning subscales of the
MacNew, Leiden Quality of Life Questionnaire, QLMI, SF-36, NHP,
and the QLI. A complete
breakdown of each validated instrument can be found in Appendix
3.
2.3.5 Meta Regression
The regression coefficient (β) obtained from a meta-regression
analysis describes how
the treatment effect changes with a unit increase in the
explanatory variables. In our analysis,
positive effect sizes indicate that the intervention had a
better outcome than the control group.
The proportion of variance explained in the meta-regression
analysis is calculated by comparing
the estimate of T2 with the covariate to T
2 when no covariate is used in the model (79).
𝑅2 = 1 − 𝑇2 𝑢𝑛𝑒𝑥𝑝𝑙𝑎𝑖𝑛𝑒𝑑
𝑇2 𝑡𝑜𝑡𝑎𝑙
A univariable meta-regression was undertaken to explore
potential treatment effect
modifiers. This study attempted to shed light on the complexity
of CR through identifying
heterogeneity in the intervention, patient population and HRQOL
instrument. Variables which
-
25
were shown to be important in the literature but were not
previously assessed in a univariable
model were evaluated (12, 25, 29). Five a priori covariates were
explored: year of publication,
duration of follow-up, the proportion of MI patients, type of CR
intervention, and type of
instrument.
Year of publication was included as a continuous covariate in
order to explore the
change in the standard of usual care over time in patients with
established CHD to reflect the
improvement in pharmacological interventions. The duration of
follow-up was explored as a
continuous potential treatment effect modifier in order to
determine if the length of the follow-
up was associated with HRQOL scores. It was hypothesized that
HRQOL scores would be
lower in studies that had longer follow-up because they assessed
HRQOL farther from the
intervention. The proportion of MI patients was used as a
continuous covariate in order to
explore if having more post-MI patients in the program was
associated with HRQOL scores.
The type of CR intervention received (exercise vs non exercise)
and (psychosocial management
only vs other core components) was explored to determine if type
of intervention had a
differential effect on HRQOL scores. In order to evaluate type
of intervention in the meta-
regression model two analyses were performed. Intervention was
represented as a categorical
covariate with two levels, exercise and non-exercise. Exercise
was used as the reference level in
the model. The second intervention analysis was represented as a
categorical covariate with two
levels, psychosocial management only and other core components.
Other core components were
used as the reference level in the model. The type of Instrument
used was also included as a
categorical covariate in order to determine if there was a
significant difference in the type of
measure used to conceptualize HRQOL scores. Three levels were
used in the meta-regression
model generic, disease specific and preference based measures.
Disease specific measures were
used as the reference level in each analysis. In cases were
preference based measures were not
-
26
used meta-regression models only had two levels, generic and
disease specific measures.
Disease specific instruments are said to be in general more
responsive than generic instruments
which could possibly lead to an underestimation of treatment
effect and as such had to be
investigated (81).
2.3.6 Demographics
Each study had demographic characteristics assessed at baseline
extracted. Demographic
variables included: study location, publication date, age,
gender, and comorbidities.
2.3.7 Data Synthesis
Data synthesis and analyses were performed using Microsoft
Excel, and R software
using the package “metafor” (82, 83). A direct head-to-head
pair-wise frequentist analysis was
used to compare receiving any core component of CR to usual
care. Trials reported data on the
continuous outcome of HRQOL and/or HRQOL domains. Continuous
outcomes are expressed
using the metric of standardized mean change (SMC) to combine
data from different
instruments measuring the same constructs. A random-effects
model was used to account for
residual heterogeneity. For all outcomes, data was collected
from intention-to-treat (ITT)
analyses, but in cases where ITT results were not provided per
protocol results were used. With
regards to the inconsistency in the reporting of outcomes in the
absence of mean scores medians
were used as a replacement in order to include as many studies
as possible. Additionally, in
cases where no standard deviations were given for associated
means standard deviation were
estimated by transforming given standard errors or confidence
intervals (Appendix 2).
A multilevel meta-analytic model using independent pairwise
crossed random effects
were used in order to account for the correlation between HRQOL
instruments and studies with
multiple reports of HRQOL outcomes.
-
27
𝑔𝑖𝑗 = 𝜷𝟎 + 𝜷𝟏𝑋1𝑖 + ⋯ + 𝜷𝒑𝑋𝑝𝑖 + 𝜂𝑖 + 𝜗𝑗 + 𝜖𝑖𝑗
Where gij is the observed effect sizes with i representing the
study and j representing the
instrument, β0 is the intercept of the dependent variable; X
represents the matrix of study level
covariates and intercept, while β represents a vector of the
coefficients, Xki (k = 1, …p) p
covariates for study i, εij is the random error term accounting
for the within study variance and
the variation between instruments, with ηi representing the
study specific random effects and ϑj
representing the HRQOL instrument specific random effects
(84).
Subgroup analysis by means of stratified meta-analysis using
random and fixed effects
was performed according to the HRQOL instrument in order to
observe individual subscale
scores. Random effects subgroup analysis was performed when
there was a large amount of
studies; otherwise fixed effects models were used. Additionally,
95% confidence intervals (CI)
were calculated for each effect estimate. HRQOL meta-analysis
results are represented using
forest plots. Post-test-pretest correlation coefficients
estimates for each HRQOL instrument
needed for the SMC analysis were extracted from the literature
or imputed based on expert
opinion. The correlation coefficients represent the relationship
between an instrument’s baseline
and follow-up scores in relation to its reliability. In order to
interpret the meta-analysis results
the criterion created by Cohen, 1988 which states that effect
size changes of 0.2 SD units
reflects a small difference, 0.5 SD units a moderate difference
and 0.8 SD units a large
difference were used (85).
Heterogeneity amongst included studies was quantitatively
assessed using the I2 statistic,
tau2 (T
2) and qualitatively by comparing characteristics of included
studies and visually
inspecting forest plots. Given that using the I2 statistic is
not precise an uncertainty interval was
-
28
also given (86, 87). In order to interpret the inconsistency
seen in I2 the guide of 0% to 40%:
minimal heterogeneity; 30% to 60%: moderate heterogeneity; 50%
to 90%: substantial
heterogeneity; 75% to 100%: considerable heterogeneity (77).
A P-value of 0.10, rather than the conventional level of 0.05
were used to determine
statistical significance to account for domains with small
sample sizes or low power. A
sensitivity analysis was undertaken to assess the various
differences in imputed correlation
coefficients used in the SMC analysis. In order to examine small
study and publication bias a
funnel plot and a rank correlation test were performed. A rank
correlation test using Kendall’s
tau was performed to investigate for correlation between the
effect size estimate and sampling
variance to identify possible publication bias (88).
3 Results
3.1 Study Demographics
Figure 1 outlines the selection of potentially eligible studies.
A total of 1,270 potential
studies were identified, 1,205 were excluded because they were
not randomized (n=142),
included the wrong patient population (n=138), did not evaluate
an eligible intervention (n=55),
had a study duration of less than 6 months (n=66), did not
report outcome of interest (n=750),
included patients who were randomized after CR (n=6), were not
published in English (n=16),
did not report full outcome data (n=27) or were randomized
before cardiac surgery (n=5). Sixty-
five reports of 52 RCTs with 13,360 participants were included
in the multilevel meta-analysis.
The study and patient demographics are outlined in Table 1.
Studies were conducted in
North America (25%), Australia (9.6%), Asia (7.7%) and Europe
(58%). In terms of publication
dates studies ranged from 1990 – 1999 (8%), 2000 – 2009 (57%)
and 2010 – 2014 (35%). The
-
29
mean age of participants was 62 years and 66% of the
participants were male. Eleven percent of
patients were diagnosed with diabetes and 24% were previous
smokers. In terms of studies
reporting the primary objectives 19 RCTs with 3,892 patients
reported overall HRQOL, 46
RCTs with 12,523 patients reported physical HRQOL, 39 RCTs with
11,539 patients reported
emotional HRQOL, and 27 RCTs with 8,209 patients reported social
HRQOL. A full list of
included and excluded studies can be found in Appendix 5 and 6
respectively.
3.1.1 Risk of Bias Assessment
All included trials were assessed using the Cochrane risk of
bias assessment tool (69).
For each trial the risk of bias was presented using ‘risk of
bias’ graph (Figure 2). In addition an
overall summary of risk of bias is given in Figure 3. Included
RCTs ranged from the year 1995
to 2014.
Thirty-two trials (62%) were at low risk of selection bias due
to the satisfactory
generation of the randomization sequence. One trial (2%) had a
high risk of selection bias
because of a non-random method used to generate their sequence.
Nineteen studies (37%) were
judged to have an unclear risk of selection bias because the
method used to generate the random
sequence was not described in the paper. Thirty-one studies
(60%) were at a low risk of
selection bias owing to proper concealment allocation of the
intervention to participants and
investigators. One trial (2%) had a high risk of selection bias
as the participants or investigators
could foresee assignment. Twenty studies (38%) were judged to
have an unclear risk of
selection bias because the method of concealment was not
described in detail allowing for a
definite judgement.
Twenty trials (38%) were at a low risk for performance bias
because investigators and
key personal were blinded to allocation. Three studies (6%) were
at a high risk of performance
-
30
bias owing to investigators and participants not being blinded
to allocation. Twenty-seven
studies (52%) were judged to have unclear risk of performance
bias. Seventeen trials (33%)
were at a low risk for detection bias because investigators were
unaware of the allocation of
patients. Four studies (8%) were at a high risk of detection
bias and thirty-one studies (60%)
were at a low risk of detection bias.
Thirty-eight studies (73%) were judged to be at low risk of
attrition bias due to the
nature of handling of incomplete outcome data and four trials
were measured to have an unclear
risk for attrition bias. Ten trials (19%) were at a high risk
for attrition bias. Thirty-eight studies
(73%) were judged to be at low risk of reporting bias because
based on information provided by
the authors regarding primary and secondary outcomes. Five
studies (10%) were at a high risk
of selective reporting and nine studies (17%) were judged to
have an unclear risk of selective
reporting.
3.2 Health Related Quality of Life
The QLMI, MacNew, Leiden Quality of Life Questionnaire, AP-QLQ,
SAQ, MIDAS,
QLI, SF-36, SF-12, NHP, Dartmouth COOP Quality of Life
instrument, DASI, Cantril Ladder
of Life, SF-6D, EQ-5D, and TTO were used to evaluate the changes
in each HRQOL domain.
Table 2 outlines the specific studies, subscales and or domains
which were used to
conceptualize each HRQOL domain.
3.2.1 Overall Health Related Quality of Life
Receiving any core component of CR improved overall HRQOL when
compared to
usual care. Using 21 reports of 19 trials and 3,892 participants
the SMC, with respect to overall
HRQOL was 0.14 (95% CI 0.03 to 0.25) (Figure 4). Using Cohen’s
criteria, this would be
categorized as a small treatment effect. There was a substantial
amount of clinical and statistical
-
31
heterogeneity when combining all the studies in the SMC model
(I2
= 68%; 95% UI 44-87%).
From the I2
of 69% about 52% of the inconsistency was attributed to between
study variance
with 16% of the inconsistency due to within study variance.
Fixed effect subgroup analysis
results based on the MacNew, EQ-5D and TTO overall HRQOL scores
were inconsistent in
showing an improvement in overall HRQOL. The SMC, with respect
to the MacNew’s overall
HRQOL subscale was -0.04 (95% CI -0.12 to 0.04), while, the
EQ-5D’s SMC was 0.04 (95%
CI -0.06 to 0.14). Based on Cohen’s classification of effect
sizes no treatment effect was
observed. The SMC, with regard to the TTO was 0.31 (95% CI 0.12
to 0.49). In relation to
Cohen’s criteria this effect would be categorized as a small
moderate effect. Table 3 provides an
outline of the overall summary effect and subgroup analysis of
all the HRQOL instruments used
to measure overall HRQOL. The rank correlation test (tau =
0.3810) provided evidence of
publication bias (p = 0.02) in terms of funnel plot asymmetry
(Figure 5).
3.2.2 Physical Health Related Quality of Life
Receiving any core component of CR improved physical HRQOL when
compared to
usual care. Using 53 reports of 46 trials and 12,523
participants the SMC, with regard to
physical HRQOL was 0.23 (95% CI 0.08 to 0.38) (Figure 6). Using
Cohen’s criteria, a small
treatment effect was observed. However, there was a considerable
amount of clinical and
statistical heterogeneity when combining all the studies in the
SMC model (I2
= 92%; 95% UI
89-95%). In relation to the I2 about 59% of the inconsistency
observed was due to between
study variance with 33% due to within study variance. Random and
fixed effect subgroup
analysis results of the SF-36/12 PCS, SF-36 physical functioning
subscale, MacNew physical
well-being subscale and SAQ physical limitations subscale were
inconsistent in showing an
improvement in physical HRQOL. The SF-36 physical functioning
subscale and SAQ physical
limitation subscale had a SMC of 0.26 (95% CI 0.07 to 0.46), and
0.19 (95% CI 0.08 to 0.29)
-
32
respectively. All of which reported a small treatment effect
size. The SF-36 PCS, MacNew and
SF-12 provided a SMC of 0.08 (95% CI - 0.02 to 0.19), 0.04 (95%
CI -0.07 to 0.15) and 0.05
(0.00 to 0.11) respectively and provided no treatment effect.
Table 4 provides an outline of the
overall summary effect and subgroup analysis of all the HRQOL
instruments used to measure
physical HRQOL. The rank correlation test (tau = 0.2685)
provided evidence of publication bias
(p = 0.004) for funnel plot asymmetry (Figure 7).
3.2.3 Emotional Health Related Quality of Life
Receiving any core component of CR did not improve emotional
HRQOL when
compared to usual care. Using 42 reports of 39 trials and 11,539
participants the SMC, with
reference to emotional HRQOL was 0.11 (95% CI -0.03 to 0.24)
(Figure 8). There was a
substantial amount of clinical and statistical heterogeneity
when combining all the studies in the
SMC model (I2
= 86%; 95% UI 71-90%). Forty-one percent of the inconsistency
seen in I2 was
due to between study variance, with 45% being due to within
study variance. Random and fixed
effect subgroup analysis results of the SF-36/12 MCS, SF-36
emotional limitation subscale, and
the MacNew emotional well-being subscale were consistent in not
showing an improvement in
emotional HRQOL. Table 5 provides an outline of the overall
summary effect and subgroup
analysis of all the HRQOL instruments used to measure emotional
HRQOL. The rank
correlation test (tau = 0.0395) provided no evidence of
publication bias (p = 0.71) due to funnel
plot asymmetry (Figure 9).
3.2.4 Social Health Related Quality of Life
Receiving any core component of CR did not improve social HRQOL
when compared to
usual care. Using 29 reports of 27 trials and 8,209 participants
the SMC relating to social
HRQOL was 0.03 (95% CI -0.07 to 0.13) (Figure 10). There was a
substantial amount of
-
33
clinical and statistical heterogeneity when combining all the
studies in the SMC model (I2
=
75%; 95% UI 60-90%). With respect to the inconsistency seen in
I2, 72% of the heterogeneity
was due to between study variance and only 3% due to within
study variance. Random and fixed
effect subgroup analysis results of the SF-36 social functional
scale and MacNew social well-
being subscale were shown to provide no treatment effect. Table
6 provides an outline of the
overall summary effect and subgroup analysis of all the HRQOL
instruments used to measure
social HRQOL. The rank correlation test (tau = 0.0640) provided
no evidence of publication
bias (p = 0.64) due to funnel plot asymmetry (Figure 11).
3.3 Meta-Regression
To explore the substantial heterogeneity seen in each HRQOL
domain a meta-regression
was performed to examine the five a priori covariates:
instrument type (generic, disease specific,
and preference), type of intervention (exercise, non-exercise,
psychosocial only), year, duration
of follow-up and proportion of MI patients. Potential effect
modifiers were entered into
univariable models to determine the percentage of heterogeneity
explained by the covariates.
3.3.1 Overall Health Related Quality of Life
There was a substantial amount of clinical and statistical
heterogeneity when combining
all the studies in the overall SMC model (I2
= 68%; 95% UI 44-87%). Only one of the
previously described variables explained any of the
heterogeneity, which was duration of
follow-up explaining about 7% of the variance (Table 7). None of
the covariates were shown to
be significant predictors of study effect sizes.
3.3.2 Physical Health Related Quality of Life
There was a considerable amount of heterogeneity when combining
all the studies in the
physical SMC model (I2
= 92%; 95% UI 89-95%). Table 8 presents the estimates of the
-
34
synthesized univariable meta-regression where the influence of
instrument type, type of
intervention, year, duration of follow-up and proportion of MI
patients was observed in regards
to HRQOL effect sizes. Type of intervention, type of instrument,
year, duration of follow-up
and proportion of MI patients were not significant predictors of
HRQOL effect sizes. None of
the examined covariates helped explain a significant amount of
the between study heterogeneity
seen in the model.
3.3.3 Emotional Health Related Quality of Life
There was a considerable amount of clinical and statistical
heterogeneity when
combining all the studies in the SMC model (I2
= 86%; 95% UI 74-91%). No covariates were
shown to be significant predictors of study effect sizes.
Additionally, none of the examined
potential treatment effect modifiers explained a significant
amount of between study
heterogeneity in the model. Only proportion of MI patients was
shown to explain about 5% of
the unexplained variance.
3.3.4 Social Health Related Quality of Life
There was a substantial amount of clinical and statistical
heterogeneity when combining
all the studies in the SMC model (I2
= 75%; 95% UI 60-90%). Only year explained any variance
(~6%). No potential treatment effect modifiers were shown to be
significant predictors of study
effect sizes or shown to explain any significant amount of
between study heterogeneity (Table
10).
4 Discussion
This systematic review and meta-analysis of RCTs examining CR
interventions for CHD
patients was designed to determine if receiving any core
component of CR was in general able
-
35
to improve HRQOL. While CR has been previously shown to improve
mortality in CHD
patients little research has been done on CRs ability to
influence HRQOL domains because of
the heterogeneity in instruments and their varying psychometric
properties. To our knowledge
this is one of the first studies to attempt to attain an overall
summary effect of receiving any core
component of CR/secondary prevention program on HRQOL domains.
The use of random
effects multilevel meta-analysis and univariable meta-regression
helped clarify in general if
receiving any core component of CR was effective in improving
overall, physical, emotional
and social HRQOL domains. Several important findings were
discovered during this exploratory
analysis.
4.1 Health Related Quality of Life
Fifty-two trials with 13,360 adult CHD patients were included in
order to determine
improvements in patients HRQOL following receiving any core
component of CR. Receiving
any core component of CR resulted in a summary effect size of
(SMC 0.14; 95% CI 0.03 to
0.25) and (SMC 0.23; 95% CI 0.08 to 0.38) for overall and
physical HRQOL respectively.
While these effect sizes may be considered small treatment
effects, they none the less represent
and incremental benefit in comparison to the usual standard of
care (85). This improvement in
HRQOL domains was seen even though there is a considerable
amount of variability in each
study because of differences in duration, delivery format,
setting, population and intervention.
There was no difference seen between receiving any core
component of CR and usual care
shown by the estimated effect sizes of the emotional (SMC 0.11;
95% CI - 0.03 to 0.24) and
social (SMC 0.03; 95% CI - 0.07 to 0.13) HRQOL domains. An
effect size of zero demonstrates
that the treatment was not any different from the control and
that there was no improvement in
HRQOL.
-
36
When performing subgroup analysis on the different instruments
used to assess each
HRQOL domain, no consistent treatment effects were seen. This
was in part due to the
variability in patient types, interventions received, duration
of follow-up and difference in
session lengths. Although some significant effects in each
subgroup were observed they tended
to be in subgroups with small patient populations and did not
involve many studies undermining
our confidence in the observed treatment effects. It is
important to note that whilst some
domains were not statistically significant that does not reflect
an absence of benefit altogether,
these HRQOL domains are not mutually exclusive and do influence
one another. A non-
significant HRQOL outcome could reflect a lack of statistical
power in the model to detect any
relevant change.
4.2 Meta Regression
Through the use of univariable random effects meta-regression we
attempted to
investigate heterogeneity within the overall, physical,
emotional and social HRQOL domains.
Each model had significant heterogeneity after pooling and the
covariates defined a priori were
used to assess the proportion of variance explained. No
statistically significant associations were
seen in any HRQOL domains using the potential effect modifiers
type of type of intervention,
instrument type, year, duration of follow-up and proportion of
MI patients. Additionally, the a
priori covariates did not explain much if any of the between
study variance. Caution is needed
when interpreting our findings, especially in domains with low
power. The lack of associations
discovered in our meta-regression should not take away from the
possible importance of our
selected covariates in the use of explaining heterogeneity
between CR studies in the future.
-
37
4.3 Strengths of Study
One of the major strengths of this study was that we were able
to examine and organize
an extensive amount of studies and instruments looking at the
core components of CR to
determine whether their use provided an overall treatment effect
on HRQOL. This was achieved
using a comprehensive search strategy which included nine
electronic databases and the review
of reference lists