Upload
janet-richardson
View
214
Download
0
Embed Size (px)
Citation preview
The use of randomized control trials incomplementary therapies: exploring theissues
Janet Richardson PhD BSc RN DipDN
Lecturer Practitioner in Research and Development,
The Florence Nightingale Division of Nursing and Midwifery,
King's College, James Clerk Maxwell Building,
57 Waterloo Road, London SE1 8WA, England
Accepted for publication 21 January 2000
RICHARDSONRICHARDSON J. (2000)J. (2000) Journal of Advanced Nursing 32(2), 398±406
The use of randomized control trials in complementary therapies: exploring
the issues
The current popularity of complementary therapies presents an interesting
challenge to nurses and midwives. If they are to deliver such therapies
themselves, or support patients in choosing appropriate therapies they will
need to consider the professional and legal issues, in particular those regarding
safety. Evidence for the effectiveness for complementary therapies is also a
requirement in order that their integration into nursing practice can be justi®ed.
Purchasers are currently hampered by the lack of credible evidence for
effectiveness and until that evidence is provided, access to such therapies
through the National Health Service (NHS) will remain limited. The form that
evidence should take has led to a lively debate about possible methodological
approaches. There appears to be a clash between the medical profession and
those working in the ®eld of complementary therapy research, with the medical
establishment advocating randomized control trials (RCTs). This contrasts with
the view held by some advocates of complementary therapies that the RCT
approach is reductionist and not applicable to such approaches. The pivot
of the debate on the methodological approaches for evaluating complementary
therapies is the contrast of two apparently different and diverse world-views,
and the assertion that methods developed in one world-view are not transferable
to the other. There is also some confusion within the ®eld of complementary
therapy over the applicability of RCTs to therapies such as acupuncture, and the
mistaken assumption that trials which include a control group, are also required
to be double-blind. This paper is based on the need for good quality evidence of
effectiveness in complementary therapy. It will set out the concerns associated
with the use of RCTs within complementary therapy, together with the bene®ts
and limitations of this approach. The paper will go on to review research
options and propose some suggestions for future methodological approaches.
Keywords: complementary therapies, randomized control trials, pragmatic,
explanatory, methodological, double-blind, health services research, nursing
Journal of Advanced Nursing, 2000, 32(2), 398±406 Integrative literature reviews and meta-analyses
398 Ó 2000 Blackwell Science Ltd
INTRODUCTION
There is little doubt that interest in complementary
therapies has increased in recent years, and that accom-
panying this increase is large-scale expenditure on
complementary medicines and practitioners. In the
United Kingdom (UK), estimates of the percentage of the
population that uses complementary therapies range from
14%±30% of the general population, with acupuncture,
osteopathy, chiropractic, herbal medicine and homeo-
pathy indicated as the most popular therapies (Sharma
1992). Consumer surveys indicate positive public atti-
tudes in other European countries with 20%±50% of the
European population consulting complementary practi-
tioners, and identifying acupuncture as one of the most
popular forms of complementary treatment (Fisher &
Ward 1994).
For nurses and midwives the increase in public interest
in complementary therapies is proving to be an interesting
challenge and practitioners are responding in a number of
different ways. Some advocate the use of therapies such as
massage and aromatherapy in hospital settings (Styles
1997). Stevensen (1997), for example, claims that comple-
mentary therapies can bene®t patients and suggests how
nurses can work with such therapies in the context of
United Kingdom Central Council for Nursing, Midwifery
and Health Visiting (UKCC) Professional Development
Categories. Whilst massage, re¯exology and aromatherapy
perhaps understandably are popular amongst nurses,
more invasive treatments such as acupuncture are also
receiving attention. In particular midwives are beginning
to offer acupuncture to pregnant and postnatal women
(West 1997), and nurses sometimes provide acupuncture
in chronic pain clinics.
For most nurses, however, rather than providing the
therapy, the primary issue will be related to giving advice
and support to their clients who are keen to integrate
complementary therapies into their healthcare. Dimond
(1995) suggests that midwives have a duty to ensure that
their clients are aware of the advantages and disadvan-
tages and the dangers of particular therapies. In doing this,
however, they will need to be aware of the legal and
ethical implications for their practice (Norton 1995,
Geddes & Henry 1997).
Arguably, if nurses are to advise or support their clients
in the choice to use a complementary therapy, they should
also be aware of the evidence for the effectiveness of the
treatment. The need for research evidence demonstrating
the effectiveness of complementary therapy, and the
methodological issues, have received some attention in
the nursing press (Ersser 1995, Botting 1997). Many
complementary therapies have not been adequately inves-
tigated (Hamilton & Bechtel 1996). So whilst nurses might
facilitate or incorporate such therapies in to their practice,
there is a corresponding need to develop a knowledge base
grounded in effectiveness and outcome measurement
(Norton 1995). Further, as long as purchasers of healthcare
are hampered by the lack of credible evidence for effect-
iveness, access to such therapies through the National
Health Service (NHS) will remain limited.
Five types of evidence have been posited ranging from
systematic reviews of well-designed randomized control
trials (RCTs) to opinions based on clinical evidence or
descriptive studies. The strongest types of evidence
ranked according to Bandolier (1995) are thought to be
provided by systematic reviews of well-designed RCTs
(type i), and strong evidence from at least one properly
designed RCT of appropriate size (type ii). Bandolier rates
the evidence from well-designed trials without random-
ization and time-series or matched case-control studies as
type iii. The evidence provided by well-designed non-
experimental studies or the opinions of respected author-
ities have been ranked as type iv and type v, respectively.
In complementary therapy the form the evidence should
take has led to a lively debate about possible methodolo-
gical approaches. The pivot of this debate is the proposed
contrast of two apparently different and diverse world-
views often referred to as paradigms (Patel 1987, Mercer
et al. 1995, Vickers 1996), and the assertion that methods
developed in one world-view (reductionist/biomedical)
are not transferable to the other (holistic). The arguments
put forward in this paper are based on the need for good
quality evidence of effectiveness in complementary ther-
apy if such therapies are to be integrated into nursing
practice. This paper does not attempt to present an
in-depth monologue of the arguments for and against
RCTs, and the author accepts that different views exist
regarding the use of RCTs within the nursing and medical
professions as well as in complementary therapy. It does,
however, set out the concerns associated with the use of
RCTs within complementary therapy, together with the
bene®ts of this approach. Ultimately, the position adopted
is one which advocates that the best `scienti®c' approach
to the inquiry is to adopt a methodology which is most
appropriate to the question and the context.
RANDOMIZED CONTROL TRIALS
The randomized control trial (RCT) is rooted in a `posit-
ivist' science. Positivism has its origins in the natural
sciences. It is concerned with events which can be
observed, the research requires a stable environment, is
`value free and objective', should be quanti®able, and aims
to establish causal relationships (Proctor 1998). The need
to eliminate error and sources of bias is paramount,
therefore tight control is required in order to provide
results which support or reject hypotheses and create
generalizable laws (Proctor 1998). The RCT has widely
Integrative literature reviews and meta-analyses Complementary therapies
Ó 2000 Blackwell Science Ltd, Journal of Advanced Nursing, 32(2), 398±406 399
become regarded as the principle method for obtaining a
reliable assessment of treatment effect and involves four
phases. The ®rst phase is primarily concerned with drug
safety, toxicity and side-effects. Phase 2 involves small-
scale investigations into effectiveness and safety. Phase 3
is a full-scale evaluation of treatment, comparing the new
drug or interaction with current standard (or placebo)
treatment. In this phase each patient is randomly assigned
to the treatment or placebo group. Phase 4 is the post-
marketing surveillance (Pocock 1983).
Before the Second World War there were no formal
requirements for clinical trials before a drug could be
freely marketed. The ®rst RCTs were performed in agri-
culture where various crops and fertilisers were assigned
in random arrangements. The ®rst properly randomized
control trial in medicine was conducted by the Medical
Research Council (MRC) in 1948. In this trial patients with
pulmonary tuberculosis were randomized to streptomycin
and bed rest or bed rest only; no `placebo' drug was used.
The thalidomide tragedy in the 1960s caused a tightening
of government regulations in the United States of America
(USA) and UK. Regulatory requirements for the USA Food
and Drug Administration require evidence from two
placebo controlled trials. This will generally provide good
evidence that a drug works under speci®c conditions, but
it is less clear how the drug will work under different
conditions and in patients who do not resemble those in
the trial. Other interventions such as surgical procedures
and diagnostic imaging, however, are often adopted on the
basis of little or uncertain evidence (Henry & Hill 1995).
Randomization and the use of a control group remain an
important component of trials attempting to assess the
effectiveness of an intervention. Uncontrolled trials have
the potential to provide a very distorted view of therapy. A
therapy may gain widespread acceptance on the basis of
uncontrolled trials prior to RCTs that may prove to be less
optimistic. Further, wide variations in response rates may
be due to patient selection and disease progression. In
nursing and complementary therapy the practitioner is an
integral part of the intervention. This is also true in much
orthodox medicine. For example, the skill of a surgeon is
likely to impact on the patient's prognosis, thus general-
izing results from highly skilled surgeons to routine
clinical practice may be problematic (Pocock 1983).
One of the major objections to the use of RCTs in the
evaluation of complementary therapies is the blinding
procedure (Anthony 1993). The purpose of blinding is to
exclude non-speci®c factors (placebo effects) which may
produce a desirable outcome, but which are not due
directly to the active intervention. Double-blind trials are
the `gold standard' of clinical trial research, yet double-
blind studies for interventions other than pharmaceutical
ones are very rare, and for some treatments it is totally
impossible to arrange a double-blind trial, for example in
cancer trials where complex drug dose calculations are
required (Pocock 1983). Anthony (1993) believes that
blind designs are impossible in complementary therapy as
the therapist is an integral part of the intervention.
There are, however, clear distinctions between blinding
and randomization, and good quality RCTs can be
conducted in the absence of a blinding procedure. Ulti-
mately the aim of randomization is to provide equivalent
groups for comparison. Where comparisons are made
between groups which are not equivalent, research can be
fundamentally ¯awed and lead to insupportable conclu-
sions (Bagenal et al. 1990, Hayes et al. 1990, James & Reed
1990, Monro & Payne 1990, Sheard 1990a, b). However
even careful randomization can lead to non-equivalent
group comparisons (Cassidy et al. 19901 ), which highlights
the importance of assessing group equivalence, not just at
entry to the study to establish an equivalent base-line, but
at each stage of the analysis.
There are a number of practical dif®culties in the
interpretation of clinical trials. For example, an overem-
phasis on signi®cance testing and inadequate sample sizes
increase the probability of a type 2 error. Repeated
signi®cance testing on accumulating data or the use of
multiple endpoints increases the probability of a signi®-
cant result due to chance. Further, care is required when
carrying out subgroup analysis, in particular the sample
size and power of the test should be suf®cient to detect a
signi®cant result. In general it is recommended that both
con®dence intervals (for the treatment difference) and
P-values are reported, as the use of con®dence intervals
re¯ects the true magnitude of the effect (Pocock 1983,
Pocock et al. 1987, Altman 1991). Problems may arise if
the practitioners participating are unrepresentative, if
patients or treatments are atypical, or if there is poor
recruitment. Furthermore it has been suggested that the
RCT will be out of context, and patient/practitioner beliefs
and wishes will be excluded (Black 1996). The importance
of asking patients' their views is now widely acknow-
ledged and such approaches can be incorporated into
research into complementary therapy (Kacperek 1997).
Jones (1995) suggests that the establishment of evi-
dence-based practice depends on research from both
qualitative and quantitative traditions. Qualitative
methods are well established in nursing (see for example
Morse 1989, Morse & Field 1996); case study approaches
have also been used to assess the effectiveness of comple-
mentary therapy (Ballard 1996, Brooker et al. 1997). It is
interesting to note that qualitative research methodologies
are becoming increasingly popular in orthodox medicine
(Fitzpatrick & Boulton 1994, Britten 1995, Pope & Mays
1995, Greenhalgh & Taylor 1997).
Sackett et al. (1991) emphasized the importance of
`clinical' as well as statistical signi®cance. This can mean
two things, the clinical signi®cance to the patient, or the
extent to which the results of a trial lead to changes in
clinical behaviour. Many RCTs compare treatments that
J. Richardson
400 Ó 2000 Blackwell Science Ltd, Journal of Advanced Nursing, 32(2), 398±406
will produce only moderate differences in outcome, but
these differences may be clinically important, hence the
need to combine trial data in systematic reviews and meta-
analysis (Sackett et al. 1991, Clarke & Stewart 1995).
Clinical research is ultimately about making a difference
to patient care, quality of life or health outcome.
In RCTs external validity is often poor, as the outcome is
in¯uenced by the characteristics of the provider, thus
Fitter & Thomas (1997) emphasize the importance of
evaluating the therapies under `normal' service condi-
tions, leaving practitioners free to give individualized
patient treatments. External validity requires attention
when conducting or evaluating clinical trials (Vickers
1995b). However, issues of external validity receive little
attention in the instruments developed to judge methodo-
logical quality of trials (Moher et al. 1995).
The fundamental underpinning of the RCT is to assess
and to isolate potential sources of bias. This includes
observation bias, sampling bias, non-speci®c factors
(placebo effects), natural disease progression (large
number of diseases are self-limiting), and cointerventions
(Vickers 1995c). A number of important methodological
innovations (such as quasi-experimental designs), encour-
age the application of controlled trials in clinical practice
for assessing the outcome of an intervention, by providing
a means by which potential biases can be identi®ed and
dealt with (see for example Cook & Campbell 1979).
EVALUATING COMPLEMENTARY THERAPIES:THE METHODOLOGICAL DEBATE
In spite of the current methodological debate in comple-
mentary therapy, RCTs have been conducted and a
number of systematic reviews exist (see for example
Richardson & Vincent 1986, ter Riet et al. 1990, 1991b,
Kleijnen et al. 1991a). However, Mercer et al. (1995)
suggest a number of potential barriers to the use of RCTs
in complementary therapy, which include con¯icting
concepts and theories of health and disease, lack of
agreement about diagnostic criteria, contrasting views
about the therapeutic process and different theories of
causation. Patel (1987) identi®es `paradigmatic' problems
and suggests that a difference of world-views presents
more of a problem for the integration of complementary
with orthodox medicine than lack of evidence. Vickers
(1996) explains that `paradigm' is a term borrowed from
the history and philosophy of science, and a shift from one
paradigm to another is a transition to different ways of
viewing the world.
It might be argued that paradigms in complementary
and orthodox medicine are widely different, therefore
research methods developed in one paradigm are not
transferable to the other and those working within a
paradigm are unable to look outside it. Black (1996)
suggests that the arti®ciality of the RCT may reduce the
placebo element of any intervention, failing to capitalize
on the non-speci®c treatment effects, therefore the trial
will inevitably re¯ect the minimum level of bene®t that
can be expected. However, it is possible to apply RCTs
without adopting a reductionist approach, for example a
number of RCTs have been applied to the evaluation of
aromatherapy and massage, psychotherapeutic tech-
niques, healing and prayer (Dunn et al. 1995, Vickers
1996).
In attempting to disentangle the issues relating to the
use of RCTs in the context of complementary (and to some
extent orthodox) therapies the research question becomes
paramount, as different questions require different
approaches (Canter & Nanke 1993, Vickers 1995a). A
report prepared by the National Institutes for Health (NIH)
concludes that the RCT is the gold standard and most
reliable method yet developed, and should always be used
where it is practical and ethical to do so. Complementary
therapies are often very individualistic in approach and
cannot always be standardized as a treatment for large
groups of individuals in the context of an RCT (Mercer
et al. 1995). Some interventions are impossible to blind,
and due to the nature of different diagnostic systems,
patient selection may be problematic. Further, one meth-
odological assumption of the RCT is that the majority of
patients have no preference for one treatment or another
(Fitter & Thomas 1997).
A number of ways of dealing with these problems have
been proposed. Wiegant et al. (1991) and Brewin &
Bradley (1989) set out procedures for patient selection.
Others have suggested alternative methodologies for evalu-
ating effectiveness such as single case designs, qualit-
ative approaches, outcome assessment and clinical audit,
and observational studies (Aldridge 1988, Reilly & Taylor
1993, St George 1994, Mercer et al. 1995, Black 1996,
Melchart et al. 1997). Heron & Reason (1984) have gone so
far as to categorize the RCT as a source of alienation Ð
because the individual is separated from what is going on
in their body and from decisions about treatment. As the
RCT remains the `gold standard' within conventional
medical research and has in some cases been applied to
the evaluation of complementary therapies, it is important
that this form of evidence is used, where possible, to
provide evidence of effectiveness.
The historical purpose of medical research has been to
understand the mechanisms of disease and produce new
treatments, not to question effectiveness or the implemen-
tation of those treatments in routine clinical practice
(Smith 1995). Many conventional treatments have not
been subjected to randomized control trials, and as a
result new interventions have been introduced on a wide-
scale on the basis of opinion rather than evidence from
RCTs (Smith & Rennie 1995). Treatments have been
introduced and used widely before being shown to be
harmful and of no bene®t, for example the use of insulin
Integrative literature reviews and meta-analyses Complementary therapies
Ó 2000 Blackwell Science Ltd, Journal of Advanced Nursing, 32(2), 398±406 401
coma therapy, leucotomy, and regressive electro-convul-
sive therapy (ECT) for the treatment of schizophrenia
(Andrews 1989).
Lack of evidence can prohibit the effective use of
resources. For example, the prevalence of pressure sores
in the UK is 5±9% of patients receiving hospital and
community care; treatment costs are estimated as high
as £755 million a year, but there is little evidence
regarding the effectiveness of available treatments
(Bandolier 1994).
In biomedicine the overall approach places an emphasis
on generating objective `hard data' to establish particular
cause±effect relationships for disease. The experimental
methodology ®ts closely with the theoretical assumptions
contained in the biomedical model, and has been adopted
as the `gold standard'. Strenuous efforts are made to
isolate independent and dependent variables within the
experimental research design and eliminate `non-speci®c'
variables. In assessing the impact of health care, however,
`cure' is not the only factor to be considered; increasingly
lay people place importance on how far treatment
enhances or sustains emotional well-being, a healthier
life-style and more satisfying relationships (Mercer et al.
1995). So arguably, clinical trials should include some
measure of quality of life and patients' perceptions
(Delamothe 1994, Sackett 1994).
CONTROLLING FOR PLACEBO EFFECTS
The terms `placebo' and `placebo effect' are often used
interchangeably; however, it is important to make a clear
distinction between the two terms. A placebo is an inert
substance used to persuade an individual that she/he has
received an active, usually therapeutic, intervention. In
contrast a `nocebo' is an inert substance which produces a
negative or harmful effect. The placebo-effect is an effect
rather than a substance. It is mediated by unde®ned
mechanisms which facilitate a therapeutic change, but
which are usually viewed as less valid than pharmaco-
logical or other `active' interventions (Watkins 1994).
Lewith (1993) suggests that the placebo response is
consistent at 35á2% (� 2á2%), a ®gure which is generally
thought to come from Beecher (1955) who quotes the
average of his own 11 studies, all of which varied widely.
In fact the placebo response in double-blind studies is not
constant and has been shown to vary from 0% to 100%
(Ernst & Resch 1995, Wall 1996).
There is no evidence that the placebo-effect differenti-
ates between organic and mental illness, or that speci®c
groups of the population are `placebo responders' with a
particular type of mentality (Wall 1996). In studies of pain
(a multidimensional experience) all dimensions appear to
be equally involved in the placebo response, either
separately or together. The non-speci®c effects which
bring about a placebo response can be extremely powerful
and are complex in nature. Watkins (1994) suggests that
the power of the placebo-effect has been recognized and
used to enhance the well-being and state of mind of
patients since the time of Aristotle. It is thought to be an
essential component of all forms of medicine and healing
(Helman 1984, Lewith 1993).
The power of the placebo-effect has been demonstrated
in a number of studies (for example Dimond et al. 1958,
Cobb et al. 1959, Moscucci et al. 1987). Though the nature
and cause of such effects are poorly understood, they have
been attributed to an interaction of the self-healing
properties of the body, changes induced by therapist and
environment, the cultural context, the power of expect-
ancy and belief and physical/pharmacological interven-
tions (Helman 1984).
Studies suggest patients are searching for subtle clues
about what to expect and that these expectations affect
responses to active or inactive interventions (Bootzin
1985). This raises the question of the comparable nature
of the active intervention and placebo intervention.
Active interventions, for example morphine, may
produce immediate side-effects. It is often possible for
patients who participate in double-blind RCTs to guess
accurately which experimental group they have been
assigned to (for example on the basis of their experience
of side-effects), hence the importance of ensuring blind-
edness at outcome (Moscucci et al. 1987). So a thorough
placebo comparison might involve the use of `active
placebos', that is the placebo substance should mimic
the active substance without containing the agent
designed to bring about the desired outcome (Finkel
1985, Wall 1996). In complementary therapy, nursing
and many forms of orthodox medicine identifying a
comparable manoeuvre against which to test an inter-
vention (for example acupuncture) is fraught with dif®-
culty (Wall 1996).
Placebos used in acupuncture trials include the use of
needles that are either rubbed against the skin, or glued to
it. However, it seems unlikely that patients would accept
these procedures as credible forms of treatment (Vincent
1993). Alternatively needles may be inserted at incorrect
(theoretically irrelevant) sites as in `sham acupuncture'.
However, there are some doubts as to whether acupunc-
ture and non-acupuncture sites are distinguishable, and
some studies have shown no difference between sham and
real acupuncture sites (Vincent 1993). Vincent suggests
the use of mock TENS treatment or minimal acupuncture
as a possible placebo solution; or, alternatively two types
of control trials, a no-treatment waiting list control, or
acupuncture as an addition to traditional treatment.
Lewith & Machin (1983) argue that sham acupuncture
cannot be considered a control in pain relief studies, as
there is some evidence that chronic pain is attenuated
through mechanisms such as diffuse noxious inhibitory
control.
J. Richardson
402 Ó 2000 Blackwell Science Ltd, Journal of Advanced Nursing, 32(2), 398±406
PRAGMATIC AND EXPLANATORY STUDIES
One useful approach is to make a distinction between
pragmatic trials, where the intention is to study the whole
policy of a particular approach to therapy, and explan-
atory trials which try to subdivide the policy, separating it
out into the constituent parts (Schwartz & Lellouch 1967).
The explanatory trial is aimed at understanding which
speci®c component of an intervention produces the
outcome; the pragmatic trial is aimed at decision, that is
it seeks to answer the question `Do patients bene®t from
this treatment?' or `Which of two treatments are prefer-
able?' (Schwartz & Lellouch 1967).
One of the differences between pragmatic studies and
explanatory studies is the extent to which protocol viol-
ations and major deviations (such as patient non-compli-
ance) are considered in the interpretation of the results
(Pocock 1983). Pocock suggests that all eligible patients,
regardless of compliance with the protocol, should be
included in the analysis where possible. This pragmatic
approach is sometimes called `intention to treat analysis'
and provides a more valid assessment of treatment effect-
iveness in actual clinical practice. The alternative `explan-
atory' approach con®nes the analysis to only those
patients who received therapy according to the protocol.
In most full-scale trials of drug therapy, and for the
majority of health care evaluation, the aim is to evaluate
treatment policies as they relate to actual clinical practice
rather than purely scienti®c evaluations of drug effect. In
such studies pragmatic trials are more important than
explanatory trials (Pocock 1983).
In pragmatic trials `analysis by intention to treat' is
recommended, thus withdrawals and deviations are
accounted for in the analysis (Altman 1991). `Analysis of
compliers only' can distort treatment comparisons. This is
particularly important if patients are withdrawn because
of side-effects (as these side-effects and treatment failures
will not then be reported in the analysis). In some trials it
is not easy to include withdrawals, as lack of response
data once a patient is withdrawn leaves no alternative but
to exclude them from the corresponding analysis. It is
therefore important to develop alternatives to the standard
intention-to-treat analysis that can compensate in some
way for non-compliance, for example sensitivity analysis
(Pocock & Abdalla 19982 ).
In the debate regarding the appropriate methodology for
assessing the effectiveness of complementary therapies,
the term pragmatic has a different meaning. It refers to a
research design that allows practitioners to treat patients
individually, in a way more in line with clinical practice,
and capitalizes on rather than restricts the non-speci®c
`placebo' effects (Thomas & Fitter 1997). Knipschild
(1993) suggests the use of trials comparing the best
orthodox treatment and complementary therapy for
patients who seem suitable for both. For example Meade
& Frank (1993) compared chiropractic and outpatient
management for the treatment of low back pain.
Prior to commencing a clinical research trial the nature
of the research questions needs to be clearly de®ned, and
the type of trial (pragmatic or explanatory) stated. Further,
the selected outcome measures should be appropriate,
valid and reliable. Reliability refers to the extent to which
a measure provides consistent results; validity is the
extent to which the results of a test pertain directly to
the characteristic being measured (see Bowling 1991). If
the study utilizes a qualitative approach it will need to
demonstrate rigour in processing and interpreting the data
(Burnard 1991, Forchuk & Roberts 1993, Greenhalgh &
Taylor 1997).
PATIENT SELECTION
Two major problems exist for patient selection to comple-
mentary therapy research trials. The ®rst problem also
relates to trials in orthodox treatments and originates from
the premise that patients play an active part in the
outcome of all treatments (Brewin & Bradley 1989). The
patient's motivation to follow treatment regimens is likely
to be in¯uenced by preferences before treatment is started.
The greater the need for participation (for example
following a special diet) the greater the motivation will
in¯uence outcome. In pragmatic trials the objective is to
discover the effect and outcome of `packages' of manage-
ment, even though the individual components of the
package cannot necessarily be reduced. Most treatments
are complex, consisting of a mix of supposedly active
components with contextual factors such as the way
treatment is given. In explanatory trials the contextual
factors are `equalised' by randomization and arti®cial
constraints. In pragmatic trials contextual factors are
optimized so approximating actual practice, and the role
expectations play in treatment outcome can be evaluated.
Brewin and Bradley suggest an alternative to the conven-
tional RCT is to optimize motivation by ascertaining
patients' preferences prior to randomization and allo-
cating patients to their preferred treatment choice.
Patients who express no preference are randomized to
either treatment or control group.
The second problem relates to patient selection that can
either be based on orthodox or complementary system
diagnosis. Wiegant (1991) proposes four possible ways of
dealing with this, from selection based on orthodox
diagnosis, to selection criteria de®ned by complementary
therapy diagnosis alone.
A pragmatic trial which accounts for patient preference
and which allows practitioners to treat patients on an
individual basis, goes some way to addressing the criti-
cisms of the application of RCTs to the evaluation of
Integrative literature reviews and meta-analyses Complementary therapies
Ó 2000 Blackwell Science Ltd, Journal of Advanced Nursing, 32(2), 398±406 403
complementary therapies. Blinding should be included in
the design where possible, and every effort made to
identify and control for potential bias in order to ensure
that the outcome is due to the intervention.
SUMMARY AND CONCLUSIONS
Many of the challenges facing research in complementary
therapies are consistent with those facing research in
nursing, indeed a number of nursing interventions (for
example massage) are also de®ned as complementary
therapy. There is clearly a need for good quality studies
evaluating the effectiveness of such therapies, but there is
also disagreement about the use of a `reductionist'
approach. Randomized-controlled trials are thought to be
the `gold standard' of clinical research. They do, however,
present a number of problems, in particular the approach
attempts to `control out' the non-speci®c (placebo) effects
which appear to play a signi®cant role in producing a
positive outcome; issues of external validity and general-
izibility to clinical practice are not straightforward. There
are also dif®culties in identifying suitable `placebo' inter-
ventions, and some evidence that patients are able to
identify whether they are in a placebo or active treatment
group.
The methodological debate appears to be quite complex,
with proponents of different approaches, and often even
different world-views. Many of the arguments focus on the
`gold standard' double-blind placebo control trial and the
possibility of this method of scienti®c inquiry being
transferable to complementary therapies (see Mercer et al.
19953 ). There is, however, a misconception that all RCTs
need to be blinded and that treatments within an RCT
cannot be individualized (Pocock 1993).
The purpose of randomization is to ensure that the
groups (treatment, placebo and/or control) are equivalent,
and that any differences between groups are due to the
interventions and not due to bias or chance. The use of
credible placebos in much of conventional and comple-
mentary therapies is impractical. However, it is practical
to evaluate the treatment on offer compared with either
the best available treatment or a no treatment control
group (Ernst & Resch 1995).
The decision regarding methodological approach and
the use of a placebo-controlled trial will depend on the
research question. Ultimately the methodology should be
appropriate for the question. The question could ask
`What is the patient's experience of therapeutic massage?'
or `Is osteopathy more effective than physiotherapy for the
treatment of back pain?' Clearly the two questions require
different approaches.
The provision of an evidence-based health service
requires not only explanatory trials to identify the
precise nature of what causes a therapeutic effect, but
also pragmatic trials, which capitalize on the therapeutic
intervention as a `complete package', involving individ-
ualized treatments and an interaction with a health care
provider. Where possible, such pragmatic trials should
be randomized and/or the experimental groups assessed
for equivalence in order to identify potential sources of
bias.
Avis & Robinson (1996) suggest that methodological
positions proposing quantitative or qualitative research
should be put aside in favour of an `interdisciplinary'
approach to research. This might, for example, involve the
use of pragmatic clinical trials, which incorporate a
control group, but allow treatment to be carried out in
line with everyday clinical practice. Such trials could
involve group comparisons, as with the conventional RCT,
or adopt a single-case control approach. Greater emphasis
should be placed on the use of con®dence intervals, the
size of the treatment effect and its clinical signi®cance.
Patients who are withdrawn from the study should be
accounted for in the analysis. Where randomization is not
possible, procedures are available for assessing group
equivalence and dealing with biases (Cook & Campbell
1979). Qualitative approaches can provide a rich source of
data that allows the experience of the subject to be
reported and can be used independently or alongside
quantitative approaches. The use of objective measures
can be combined with a subjective patient assessment.
However, such measures will need to be appropriate,
valid, reliable and sensitive to changes over time (Guyatt
et al. 1987).
Nurses need to be aware of the research issues and how
to judge the quality of research evidence in complement-
ary therapy, in order that they can provide appropriate
advice to patients, and move towards integrating effective
therapies into their practice.
References
Aldridge D. (1988) Single-case research designs. Complementary
Medical Research 3, 37±46.
Altman D.G. (1991) Practical Statistics for Medical Research.
Chapman & Hall, London.
Andrews G. (1989) Evaluating treatment effectiveness. Australian
and New Zealand Journal of Psychiatry 23, 181±186.
Anthony H.M. (1993) Clinical research: questions to ask and the
bene®ts of asking them. In Clinical Research Methodology for
Complementary Therapies (Lewith G.T. & Aldridge D. eds),
Hodder & Stoughton, London4 .
Avis M. & Robinson J. (1996) Continuing dilemma in health-care
research. Nursing Times Research 1, 9±11.
Bagenal F.S., Easton D.F., Harris E., Chilvers C.E.D. & McElwain
T.J. (1990) Survival of patients with breast cancer attending
Bristol Cancer Help Centre. Lancet 336, 606±610.
Ballard A.E. (1996) Traditional and complementary therapies
used together in the treatment, relief and control of Chron's
disease and polyarthritis. Complementary Therapies in Nursing
and Midwifery 2, 52±54.
J. Richardson
404 Ó 2000 Blackwell Science Ltd, Journal of Advanced Nursing, 32(2), 398±406
Bandolier 65 (1994) Pressure Sores. Five Reviews and an RCT.
Anglia and Oxford Regional Health Authority, Oxford.6
Bandolier 12 (1995) Evidence-Base Everything. Anglia and Oxford
Regional Health Authority, Oxford.7
Beecher H.K. (1955) The powerful placebo. Journal of the Amer-
ican Medical Association 159, 1602±1606.
Bowling A. (1991) Measuring Health: A Review of Quality of Life
Measurement Scales. Open University Press, Buckingham.
Black N. (1996) Why we need observational studies to evaluate
the effectiveness of health care. British Medical Journal 312,
1215±1218.
Bootzin R.R. (1985) The role of expectancy in behavior change. In
Placebo: Theory, Research and Mechanisms (White L., Tursky
B. & Schwartz G. eds), Guilford Press, New York, pp. 198±2038 .
Botting D. (1997) Review of literature on the effectiveness of
re¯exology. Complementary Therapies in Nursing and
Midwifery 3, 123±130.
Brewin C.R. & Bradley C. (1989) Patient preferences and rando-
mised clinical trials. British Medical Journal 299, 313±315.
Britten N. (1995) Qualitative interviews in medical research.
British Medical Journal 311, 251±253.
Brooker D.J., Snape M., Johnson E., Ward D. & Payne M. (1997)
Single case evaluation of the effects of aromatherapy and
massage on disturbed behaviour in severe dementia. British
Journal of Clinical Psychology 36, 287±296.
Burnard P. (1991) A method of analysing interview transcripts in
qualitative research. Nurse Education Today 11, 461±466.
Canter C. & Nanke L. (1993) Emerging priorities in complement-
ary medical research. In Clinical Research Methodology for
Complementary Therapies (Lewith G.T. & Aldridge D. eds),
Hodder & Stoughton, London.
Cassidy J.D., Lopes A.A. & Yong-Hing K. (1990) The immediate
effect of manipulative vs. mobilization on pain and range of
motivation in the cervical spine: a randomised controlled trial.
Journal of Manipulative and Physiological Therapeutics 16,
279.9
Clarke M.J. & Stewart L.A. (1995) Obtaining data from randomised
controlled trials: how much do we need for reliable and
informative meta-analyses? In Systematic Reviews (Chalmers I.
& Altman D.G. eds), BMJ Publishing, London, pp. 37±4710 .
Cobb L.A., Thomas G.I., Dillard D.H., Merendino K.A. & Bruce
R.A. (1959) An evaluation of internal mammary artery ligation
by a double blind technique. New England Journal of Medicine
20, 1115±1118.
Cook T.D. & Campbell D.T. (1979) Quasi-Experimentation Design
and Analysis Issues for Field Settings. Houghton Mif¯in,
Boston.
Delamothe T. (1994) Using outcomes research in clinical practice.
British Medical Journal 308, 1583±1584.
Dimond B. (1995) Complementary therapy and the mother's
wishes. Modern Midwife 5, 34±35.
Dimond E.G., Kittle C.F. & Crockett J.E. (1958) Evaluation of
internal mammary ligation and sham procedure in angina
pectoris. Circulation 18, 712±713.
Dunn C., Sleep J. & Collett D. (1995) Sensing an improvement: an
experimental study to evaluate the use of aromatherapy,
massage and periods of rest in and intensive care unit. Journal
of Advanced Nursing 21, 34±40.
Ernst E. & Resch K.L. (1995) Concept of true and perceived
placebo effects. British Medical Journal 311, 551±553.
Ersser S.J. (1995) Complementary therapies and nursing research:
issues and practicalities. Complementary Therapies in Nursing
and Midwifery 1, 44±50.
Finkel M.J. (1985) Placebo controls are not always necessary. In
Placebo Theory, Research and Mechanisms (White L., Tursky
B. & Schwartz G.), The Guilford Press, New York11 .
Fisher P. & Ward A. (1994) Complementary medicine in Europe.
British Medical Journal 309, 107±111.
Fitter M.J. & Thomas K.J. (1997) Evaluating complementary
therapies for use in the National Health Service: `Horses for
Courses'. Part 1: The design challenge. Complementary Ther-
apies in Medicine 5, 90±93.
Fitzpatrick R. & Boulton M. (1994) Qualitative methods for
assessing health care. Quality in Health Care 3, 107±113.
Forchuk C. & Roberts J. (1993) How to critique qualitative research
articles. Canadian Journal of Nursing Research 25, 47±56.
Geddes N. & Henry J.K. (1997) Nursing and alternative medicine.
Legal and practice issues. Journal of Holistic Nursing 15,
271±281.
Greenhalgh T. & Taylor R. (1997) Papers that go beyond numbers
(qualitative research). British Medical Journal 315, 740±743.
Guyatt G.H., Walter S. & Norman G. (1987) Measuring change over
time: assessing the usefulness of evaluation instruments.
Journal of Chronic Diseases 40, 171±178.
Hamilton D. & Bechtel G.A. (1996) Research implications for
alternative therapies. Nursing Forum 31, 6±10.
Hayes R.J., Smith P.G. & Carpenter L. (1990) Bristol Cancer Help
Centre. Lancet 336, 1185.
Henry D. & Hill S. (1995) Comparing treatments (Editorial). British
Medical Journal 310, 1279.
Helman C. (1984) Culture, Health and Illness. Butterworth
Heinemann, Oxford.
Heron J. & Reason P. (1984) New paradigm research and holistic
medicine. British Journal of Holistic Medicine 1, 86±91.
James N. & Reed A. (1990) Bristol Cancer Help Centre. Lancet
336, 744.
Jones R. (1995) Why do qualitative research? (Editorial). British
Medical Journal 311, 2.
Kacperek L. (1997) Patients' views on the factors which in¯uence
the use of an aromatherapy massage outpatient service. Comple-
mentary Therapies in Nursing and Midwifery 3, 51±57.
Kleijnen J., ter Riet G. & Knipschild P. (1991a) Acupuncture and
asthma: a review of controlled trials. Thorax 46, 799±802.
Kleijnen J., Knipschild P. & ter Riet G. (1991b) Clinical trials of
homoeopathy. British Medical Journal 302, 316±323.
Knipschild P. (1993) Trials and errors. British Medical Journal
309, 1706±1707.
Lewith G.T. (1993) Every doctor a walking placebo. In Clinical
Research Methodology for Complementary Therapies (Lewith
G.T. & Aldridge D. eds), Hodder & Stoughton, London12 .
Lewith G.T. & Machin D. (1983) On the evaluation of the clinical
effects of acupuncture. Pain 16, 111±127.
Meade T.W. & Frank A.O. (1993) Manipulation and low back pain:
an example of principles and practice. In Clinical Research
Methodology for Complementary Therapies (Lewith G.T. &
Aldridge D. eds), Hodder and Stoughton, London, pp. 411±42013 .
Integrative literature reviews and meta-analyses Complementary therapies
Ó 2000 Blackwell Science Ltd, Journal of Advanced Nursing, 32(2), 398±406 405
Melchart D., Linde K. & Weidenhammer W. (1997) Systematic
clinical auditing in complementary medicine: rationale,
concept, and a pilot study. Alternative Therapies in Health
and Medicine 3, 33±39.
Mercer G., Long A.F. & Smith I.J. (1995) Researching and
Evaluating Complementary Therapies: The State of the Debate.
Collaborating Centre for Health Service Research, Nuf®eld
Institute for Health, Leeds.
Moher D., Jadad A.R., Nichol G. et al. (1995) Assessing the quality
of randomised controlled trials: an annotated bibliography of
scales and checklists. Controlled Clinical Trials 16, 62±73.
Monro J. & Payne M. (1990) Bristol Cancer Help Centre. Lancet
336, 743±744.
Morse J.M. (1989) Qualitative Nursing Research: A Contemporary
Dialogue. Sage, California.14
Morse J.M. & Field P.A. (1996) Nursing Research. The Application
of Qualitative Approaches. Stanley Thornes, Cheltenham.
Moscucci M., Byrne L., Weintraub M. & Cox C. (1987) Blinding,
unblinding, and the placebo effect: an analysis of patients'
guesses of treatment assignment in a double-blind clinical trial.
Clinical Pharmacology & Therapeutics 41, 259±265.
Norton L. (1995) Complementary therapies in practice: the ethical
issues. Journal of Clinical Nursing 4, 343±348.
Patel M.S. (1987) Problems in the evaluation of alternative
medicine. Social Science Medicine 25, 669±678.
Pocock S.J. (1983) Clinical Trials: A Practical Approach. John
Wiley & Sons, Chichester.
Pocock S.J. (1993) Error and bias in single group and controlled
data trials. In Clinical Research Methodology for Complement-
ary Therapies (Lewith G.T. & Aldridge D. eds), Hodder &
Stoughton, London, pp. 35±3715 .
Pocock S.J., Hughes M.D. & Robert J.L. (1987) Statistical problems
in the reporting of clinical trials. New England Journal of
Medicine 317, 426±432.
Pocock S.J. & Abdalla M.16 (1995) The hope and the hazard of using
compliance data in randomized controlled trials. Statistics in
Medicine 17, 303±317.
Pope C. & Mays N. (1995) Reaching the parts other methodologies
cannot reach: an introduction to qualitative methods in health
and health services research. British Medical Journal 311,
42±45.
Proctor S. (1998) Linking philosophy and method in the research
process: the case for realism. Nurse Researcher 5, 73±90.
Reilly D. & Taylor M. (1993) Developing integrated medicine:
Report of the RCCM Research Fellowship in Complementary
Medicine, The University of Glasgow 1987±90. Complementary
Therapies in Medicine 1(Suppl. 1), 1±50.
Richardson P.H. & Vincent C.A. (1986) Acupuncture for the
treatment of pain: a review of evaluation research. Pain 24,
15±40.
Sackett D.L. (1994) Understanding clinical trials (Editorial).
British Medical Journal 309, 755±756.
Sackett D.L., Haynes R.B., Guyatt G.H. & Tugwell P. (1991)
Clinical Epidemiology: A Basic Science for Clinical Medicine.
Little, Brown, Boston.
Schwartz D. & Lellouch J. (1967) Explanatory and pragmatic
attitudes in therapeutic trials. Journal of Chronic Disease 20,
637±648.
Sharma U. (1992) Complementary Medicine Today: Practitioners
and Patients. Tavistock/Routledge, London.
Sheard T.A.B. (1990a) Bristol Cancer Help Centre. Lancet 336,
683.
Sheard T.A.B. (1990b) Bristol Cancer Help Centre. Lancet 336,
1185±1186.
Smith R. (1995) The scienti®c basis of health services. British
Medical Journal 311, 961±962.
Smith R. & Rennie D. (1995) And now, evidence based editing.
British Medical Journal 311, 826.
Stevensen C. (1997) Complementary therapies and their role in
nursing care. Nursing Standard 11, 49±53.
St George D. (1994) Towards a research & development strategy
for complementary medicine. Homoeopath 54, 254±257.
Styles J.L. (1997) The use of aromatherapy in hospitalised
children with HIV disease. Complementary Therapies in
Nursing and Midwifery 3, 16±20.
ter Riet G., Kleijnen J. & Knipschild P. (1990) Acupuncture and
chronic pain: a criteria-based meta-analysis. Journal of Clinical
Epidemiology 43, 1191±1199.
Thomas K.J. & Fitter M.J. (1997) Evaluating complementary
therapy for use in the National Health Service: `Horses for
courses'. Part 2: alternative research strategies. Complementary
Therapies in Medicine 5, 94±98.
Vickers A.J. (1995a) A basic introduction to medical research.
Part ii: an overview of different research methods. Comple-
mentary Therapies in Nursing and Midwifery 1, 113±117.
Vickers A.J. (1995b) What conclusions should we draw from the
data? British Homoeopathic Journal 84, 95±101.
Vickers A.J. (1995c) A basic introduction to medical research.
Part i: what is research and why do it? Complementary
Therapies in Nursing and Midwifery 1, 85±88.
Vickers A.J. (1996) Research paradigms in mainstream and
complementary medicine. In Complementary Medicine: An
Objective17 Appraisal (Ernst E. ed.), Butterworth Heinemann,
Oxford, pp. 3±11.
Vincent C. (1993) Acupuncture as a treatment for chronic pain.
In Clinical Research Methodology for Complementary Ther-
apies (Lewith G.T. & Aldridge D. eds), Hodder & Stoughton,
London, pp. 293±29418 .
Wall P.D. (1996) The placebo effect. In The Science of Conscious-
ness. Psychological, Neuropsychological and Clinical Reviews
(Velmans M. ed.), Routledge, London19 .
Watkins A.D. (1994) The role of alternative therapies in the
treatment of allergic disease. Clinical and Experimental Allergy
24, 813±825.
West Z. (1997) Acupuncture within the National Health Service:
a personal perspective. Complementary Therapies in Nursing
and Midwifery 3, 83±86.
Wiegant F.A.C., Kramers C.W. & van Wijik R. (1991) Clinical
research in complementary medicine: the importance of patient
selection. Complementary Medical Research 5, 110±115.
J. Richardson
406 Ó 2000 Blackwell Science Ltd, Journal of Advanced Nursing, 32(2), 398±406