The need for surrogate measures emerges from an important pragmatic limitation of most trials in chronic diseases: The observation period is relatively short compared with required, long-term (often lifetime) treatment. For example, a meta-analysis of 66 clinical trials reported in 1990 indicated the efficacy of methotrexate in RA was indistinguishable from injectable gold salts, penicillamine and azathioprine.18
In contrast, a study of 1,077 disease-modifying anti-rheumatic drug (DMARD) courses over five years at seven rheumatology care settings indicated that more than 50% of methotrexate courses were continued over five years vs. fewer than 20% of courses of other DMARDs.19 However, over one year, results for all DMARDs in the same patients were indistinguishable, as in clinical trials.19 Therefore, five-year effectiveness of methotrexate vs. other DMARDs was considerably higher; whereas, one-year efficacy was similar for all DMARDs.
Another example of important differences between short-term and long-term results involves the National Institutes of Health clinical trial in systemic lupus erythematosus (SLE) nephritis, which established cyclophosphamide as the standard of care for several decades.20 Significantly greater renal survival was seen over 10 years in patients treated with cyclophosphamide rather than prednisone. But no differences were seen between treatment groups after three years, and differences were marginal even after five years.20 The data raise consideration that some recently reported negative trials in SLE, vasculitis, systemic sclerosis and other diseases over one year may document significant benefit if longer periods were studied.
Even if all pragmatic limitations could be overcome, placebo-controlled clinical trials include intrinsic limitations not widely recognized.12,13 One is that results are reported in groups, but some individual participants do not fit the pattern of the group. In most trials of pain control, some individuals report greater efficacy of placebo than an active treatment reported as, statistically, significantly superior to placebo in patient groups. Therefore, when patients in routine clinical care report that a certain medication is of no value and/or even makes them feel worse, they should not be told the treatment should be efficacious on the basis of reported evidence. The evidence is reported for patient groups, while individual patients often differ widely in responses to treatments or placebos.
Another intrinsic limitation of placebo-controlled clinical trials is that placebo and nocebo effects may be affected by the methodology itself.13 A patient who is told they are receiving the best available treatment may be more likely to respond than a patient who is told they are receiving a treatment in a research study intended to analyze responses to a drug compared with a placebo. Although the placebo control should adjust for differences, stronger treatments are recognized to elicit stronger placebo responses.3