Interactions between specific and non-specific treatment effects

Weatherley-Jones et al [ ¹ ] postulate that placebo-controlled trials of individualized homoeopathy may not be an adequate test of effectiveness/efficacy mainly for two reasons: (1) specific and non-specific effects may not simply be additive but could interact in a more complex fashion, (2) homoeopaths need to know whether, during a trial, a patient did receive the prescribed medicine (rather than a placebo) in order to make correct decisions about further individualized prescribing.

Both assumptions require empirical evidence. Weatherley Jones et al argue that the first assumption cannot be tested with a two-by-two design because ‘it is not possible…to prescribe individualized homoeopathic without an appropriate consultation’. This is, of course, true. It is, however, conceivable to prescribe homoeopathy once with a full consultation and once with a minimal consultation. The two-by-two design does not require one of the levels to be zero. This should provide valuable information about the postulated interactions—if they exist.

The second assumption could be tested by a variation of a typical randomized clinical trial (RCT) ([Figure 1]). In this design, patients could be randomized first into Groups I and II. Group I is submitted to a standard placebo-controlled RCT with two parallel groups A (individualized homoeopathy) and B (placebo). Homoeopaths are not informed that Group II patients will also be randomized into similar Groups A and B. Instead they are instructed to treat all patients with individualized homoeopathy. In this part of the study, the treating homoeopaths would be as much the research subjects as the patients, and obviously one would have to debrief them after the trial. With this device, the uncertainty introduced into the mind of the homeopath by knowing that some patients are receiving placebo exists in Group I but not in Group II. By comparing results at the end of the treatment period (several, different comparisons are possible, eg AI vs BI, AII vs BII and AI vs AII), one would be able to test whether this uncertainty does have the effect Weatherley-Jones et al assume it has. Alternatively one could conduct a trial only with sub-group II and inform therapists that all patients will receive active treatment with a view to differentiating responders from non-responders.

Figure 1 Schematic outline of proposed trial design.

A further, even simpler test could be conducted. In a standard placebo-controlled RCT of individualized homoeopathy, prescribers felt (according to Weatherley-Jones et al) ‘confused as to how to decide on the next prescription…when well-indicated remedies failed to act’. If one documented the exact nature and frequency of this confusion in both treatment arms (comparing homoeopathy vs placebo), one would generate valuable information about whether it is a meaningful phenomenon. If, for instance, it occurs with the same frequency in both treatment arms it is likely to be an imagined rather than a real problem. This approach is similar to testing the success of blinding in a standard RCT. If control subjects really do fail to respond, as expected, to well-indicated remedies, then the homeopaths should, on this basis, be able to discriminate experimental patients from placebo patients at a rate above that predicted by chance.

Our overall point is that the assumptions of Weatherley-Jones et al are testable and must be tested before they can be accepted.

• References

• 1 Weatherley-Jones E, Thompson E.A, Thomas K.J. The placebo-controlled trial as a test of complementary and alternative medicine. observations from research experience of individualised homeopathic treatment. Homp 2004; 4: 186-189.