Placebo response in the treatment of CFS: systematic review & meta-analysis (Cho et al., 2005)

[Dolphin]

Free full text: http://simonwessely.com/Downloads/Publications/CFS/176.pdf

Psychosom Med. 2005 Mar-Apr;67(2):301-13.
The placebo response in the treatment of chronic fatigue syndrome: a systematic review and meta-analysis.
Cho HJ1, Hotopf M, Wessely S.
Author information

Abstract
OBJECTIVE:
The placebo response is conventionally asserted to be high in chronic fatigue syndrome (CFS) because of the latter's subjective nature and obscure pathogenesis, but no systematic review of placebo responses has been undertaken.

We report such a study.

Patient expectation is known to be important in the placebo response.

It is also known that CFS patients attending specialist clinics often have strong physical attributions regarding causation and hence skepticism about psychological or psychiatric interventions.

If so, the placebo response in CFS may be influenced by the type of intervention according to its perceived rationale.

We aimed to estimate the summary placebo response in clinical trials of CFS and to determine whether intervention type influences the placebo response in CFS.

METHODS:

We searched Medline, Embase, Cochrane Library, PsychInfo, and the references of the identified articles, and contacted experts for controlled trials (randomized or nonrandomized) of any intervention on CFS patients reporting the placebo response as a clinical improvement in physical or general outcomes.

Data were extracted from the articles and validity assessment conducted by one reviewer and checked by a second. Meta-analysis and metaregression were performed.

RESULTS:

The pooled placebo response was 19.6% (95% confidence interval, 15.4-23.7), lower than predicted and lower than in some other medical conditions.

The meta-regression revealed that intervention type significantly contributed to the heterogeneity of placebo response (p = .03).

CONCLUSION:

In contrast with the conventional wisdom, the placebo response in CFS is low.

Psychological-psychiatric interventions were shown to have a lower placebo response, perhaps linked to patient expectations.

PMID: 15784798

DOI: 10.1097/01.psy.0000156969.76986.e0

 

[Dolphin]

The comparison made between the placebo response in CFS and other conditions may not be like-for-like i.e. the threshold for being a responder might be different in other conditions.

This is the information we are given on the other conditions. I haven't looked into how a responder was defined or whether these are representative figures or not.

The placebo response has been classically considered as the rough proportion of one third in many illnesses after the publication of a review article by Henry K. Beecher in 1955 (35.2%) (3). The response rates computed in recent metaanalyses seem to be in a reasonable accordance with this classic one third. A systematic review of the clinical trials for major depression has estimated the placebo response as 29.7% (4). A similar approach to the treatment of duodenal ulcer suggests a healing rate of 44.2% in trials with a frequency of placebo administration four times a day and 36.2% in trials with administration twice a day (5). Other examples are 29.0% in the acute treatment of migraine (6) and 26.8% in the treatment of reflux esophagitis (7).

In contrast with the initial hypothesis, the pooled placebo response was substantially lower than the usually reported one third response in other medical conditions.

 

[worldbackwards]

In contrast with the conventional wisdom, the placebo response in CFS is low.

Psychological-psychiatric interventions were shown to have a lower placebo response, perhaps linked to patient expectations.

Is this why PDW wants to shut out the world from patients doing his treatments? He could do with a good placebo response.

 

[Dolphin]

Pooled Placebo Response

Low (N=8): 14.0%
Medium (N=5): 16.5%
High (N=16): 24.0%

We classified the type of intervention according to the study hypothesis: high, medium, and low. Interventions based on infectious or immunological assumptions were hypothesized to have a high placebo response and those based on psychological or psychiatric assumptions a low response. Alternative therapies are also popular in CFS patients, and we therefore hypothesized they also would elicit a positive expectancy among CFS patients, and a high placebo response. Finally, other interventions either with an obscure or neutral theoretical background were hypothesized to have a medium placebo response. The former included galanthamine, sulbutiamine, and oral nicotinamide adenine dinucleotide. The latter included hormones, so-called neuroendocrinological agents acting locally on the central nervous system and systemically on the whole body.

---

I wonder how much of this could be explained by placebo type (behavioral, oral, or injected)

Independent Variable
Placebo type

N of Studies With Data
29

Unit of Increase
1 category (out of 3)

Coefficient in %
5.3

95% CI in %
1.3 to 11.9

p-Value
0.12

They do say:

Finally, caution is needed to interpret the findings, because a relatively large number of regression parameters were estimated against a small number of observations (N 29). This means that the parameters we estimated are imprecise. Had the power of this study been higher, the independent variables with a marginal effect size such as strictness of response criteria and placebo type could have been significant predictors.

The Coefficient in % for Intervention type was 5.0

I wouldn't be surprised if similar patterns are not found in other conditions (they don't mention this).

 

[BruceInOz]

So the percentage of responders to placebo among CFS patients is 19.6%, in other conditions it's around 30%. The "recovery" rates reported in the PACE recovery paper (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3776285/) are APT: 8%, CBT: 21%, GET: 21%, SMC: 7%. Go figure!

 

[Dolphin]

The authors seem to believe in placebos having real effects

e.g.

A similar approach to the treatment of duodenal ulcer suggests a healing rate of 44.2% in trials with a frequency of placebo administration four times a day and 36.2% in trials with administration twice a day (5). Other examples are 29.0% in the acute treatment of migraine (6) and 26.8% in the treatment of reflux esophagitis (7).

rather than the findings being artefacts of one sort or another e.g. spontaneous remission.

 

[Dolphin]

Buried in the paper are these important points. It would be interesting to know whether the authors had them in the initial draft or whether they were suggested by a reviewer

Finally, the low placebo response could relate to the natural history of CFS. By definition, it is a chronic condition with duration of at least 6 months. Many of the sufferers entered into trials have illnesses that have lasted many years, and the disorder has a poor prognosis. Researchers have suggested that the response rate in the placebo arm of a clinical trial—placebo response as operationally defined—may include not only the pure placebo effect but also the other components such as spontaneous improvement, regression to the mean, measurement bias, and unidentified parallel interventions (106,107). A controversial meta-analysis of the trials comparing placebo with no treatment— an attempt to distinguish the placebo effect from the other components—has found little evidence that placebos had powerful clinical effects, and this seems to accord with the thesis (108). Given this context, our finding may be partly explained by the low rate of spontaneous remission in CFS.

Notice how they use the word "controversial" suggesting the authors do not necessarily agree with the findings.

 

[Dolphin]

First, for each study, we calculated the placebo response by dividing the number of placebo responders by the number of participants assigned to the placebo arm (rather than the number of study completers). If outcome data were provided only on study completers, we assumed that noncompleters had not responded.

If studies in other conditions did not deal with missing data in this way, this could give an artificially low percentage of responders in CFS studies compared to other conditions.

 

[Dolphin]

They refer to "strict" and "loose". However it doesn't read necessarily mean that it would be harder to achieve a response with a strict criterion versus a loose criterion; see how they define them here:

Strictness of criteria for placebo response, a binary variable (strict or loose), was added to this list after the data extraction because of the perceived heterogeneity of measurement systems (Table 1), indicating more caution in the categorization procedure and interpretation of the results. When a trial had set a more elaborate criterion to designate the response, eg, an increase of 10 points or more in the Karnofsky scale rather than improved or much improved, it was categorized as strict.

 

[Dolphin]

Five had used behavioral placebos such as relaxation or standardized medical care and hence were not double-blind (35,36,72,77,93), 16 had oral placebo (34,38,49,56,61,62,64,67,68,70,75,83,96,100,102,103), and 8 had injected placebo (28,65,66,71,73,80,90,97).

Is relaxation therapy really a good placebo?

The placebo has been defined as “any therapeutic procedure which has an effect on a patient, symptom, syndrome or disease, but which is objectively without specific activity for the condition being treated” (1).

Also "standardised medical care" sounds more like it might involve no therapy. I wonder does it include "waiting-list controls" who, it has been said, sometimes report poor results due to dissatisfaction with not getting a therapy.

 

[Dolphin]

The median of follow-up duration was 13 weeks (range 3–61).

This is not that long. The longer the follow-up period, more chance there is for spontaneous improvement. I wonder how long the follow-up was in the non-CFS studies.

Saying that, follow-up duration length was not associated with placebo response in this study.

 

[Dolphin]

No objective outcome measures were used:

Outcomes: physical (eg, fatigue, energy, pain, sleep, and functional status) or general (eg, quality of life, well-being, clinical improvement, and overall symptom measure) outcomes measuring placebo response as a binary variable, eg, “improved or not” and “responded or not.” More stringent criteria on outcomes would have compromised the generalizability of the review results.

 

[Dolphin]

Fair points:

The major limitation of the review was the heterogeneity of the outcome measurement systems across the trials. Different scales and instruments were used to define and measure the endpoint, clinical improvement.

Five trials were not double-blind, which may have led to bias in the assessment of response, but the metaregression showed that this did not contribute to heterogeneity (p = .31).

 

[Dolphin]

The final paragraph contains the most psychobabble it seems to me::

At the clinical practice level, the finding of the overall low placebo response emphasizes the need to enhance the nonspecific effects in the current treatment of CFS. Contextual factors such as a collaborative therapeutic relationship should be maximized in the management of CFS, hence increasing the overall effect of an active treatment, which consists of an active component and a nonspecific component—the placebo effect. The role of contextual factors may be even more critical for CBT, graded exercise therapy (GET), and antidepressants, because at least CBT and GET are validated treatments for CFS (27) and antidepressants effective for comorbid depression in both physical and psychological disorders (109). It is of course both intriguing and paradoxical to note the disconnection between expectations of improvement and the actual effectiveness of interventions such as CBT and GET—an area worthy of closer study, perhaps using observation methods. Whatever explanation is favored, the clinical implication is the need to provide existing evidence supportive of CBT and GET in a language accessible to patients, and if antidepressants are to be used, to make it clear that this is a treatment for depression rather than CFS itself. These strategies may assist in eliciting positive expectations in patients and hence improving outcomes.

Technically speaking this study gives us no information about the specific expectations regarding CBT and GET. Placebo effects could be higher with them for various reasons perhaps e.g. claims made about efficacy, energy participants need to invest, reframing of symptoms, etc. and how this might lead to response biases.

I'm not convinced one should hype the efficacy of therapies in order to increase the "placebo response". I'm not sure that a "placebo response" has much effect on objective measures. Getting more people to say they feel a bit better without necessarily being any better does not seem to me to be a worthy aim in most contexts.

 

[Simon]

Finally, caution is needed to interpret the findings, because a relatively large number of regression parameters were estimated against a small number of observations (N 29).

I think that might render the whole analysis invalid: you generally need a lot more observations than paramaters to avoid violating assumptions needed to apply the tecnique. Any experts here?

 

[Keith Geraghty]

Whatever explanation is favored, the clinical implication is the need to provide existing evidence supportive of CBT and GET in a language accessible to patients, and if antidepressants are to be used, to make it clear that this is a treatment for depression rather than CFS itself.

...I read this as "we must convince patients that theres a benefit and use lanaguage to do this" - perhaps the bias of this paper is the authors seeking to find a lower placebo expectancy so they can say in future "placebo is lower in CBT for CFS than in other conditions " (but on a limited number of studies with no real blinding - we cant sorry)

This paper struggles to convince me on many levels. 1. the studies reviewed have such different outcomes and approaches, making cross comparison difficult 2. Many of the studies didnt really try to capture expectancy so hard to do a post-hoc analysis of expectancy. 3. I dont understand the search limits of 2000-2002, 4. there is a lot of design bias in this study - the authors basically draw up their own parameters of assessment to pool data, 5. I think some distinction needs to be made between pre-CBT expectancy and post-CBT expectancy ie CFS patients may have low expectancy going in, but may have higher expectancy after spending time with the therapy and we need to extend the term out from just expectancy of benefit to "return on investment of time and effort" ie the patients who stick with CBT may feel they have invested in it, thus there must be a benefit, that goes beyond simple expectancy 5. No blinding, no real controls, sorry, weak here.

*in drug trials of placebo, the sugar pill is comapred against the drug - two white pills given in the same way - why do the authors not ask for research on sham CBT given in the same way as real CBT to do the same as drug trials? - thats the big question, why is CBT protected from placebo research