VARIATION IN THE PLACEBO-EFFECT IN RANDOMIZED CONTROLLED TRIALS OF ANALGESICS - ALL IS AS BLIND AS IT SEEMS

The objective was to investigate the relationship between pain relief scores produced by placebo and by active interventions in randomised c ontrolled trials (RCTs). Individual patient categorical pain relief sc ores from 5 placebo-controlled single-dose parallel-group RCTs in acut e postoperative pain were used to calculate the percentage of the maxi mum possible pain relief score (%maxTOTPAR) for the different treatmen ts. One hundred and thirty of the 525 patients in the 5 trials had a p lacebo. Individual patients' scores with placebo varied from 0 to 100% of the maximum possible pain relief. The proportion who obtained more than 50% of the maximum possible pain relief with placebo varied from 7% to 37% across the trials; with the active drugs the variation was from 5 to 63%. Mean placebo scores were related to the mean score for the active treatments in each study; the higher the mean active score, the higher the mean placebo score. This relationship disappeared when median values were used. Medical folklore has it that the amount of r elief obtained with placebo is one-third of the maximum possible (and does not vary), and that one-third of patients respond to placebo. The results show that the amount of relief obtained with placebo varies c onsiderably between patients, that 38% of patients obtained more than 10% of the maximum possible relief, and 16% obtained greater than 50%. In double-blind, randomised parallel-group studies of high quality pl acebo scores should not vary. Despite these conditions being met the p lacebo scores did vary. The previous explanation, of a relationship be tween the mean placebo scores and the mean scores for the active treat ments was not supported.