On comparing groups in studies of pain treatment.

Studies of pain treatment involve comparing groups. In observational studies scores for outcome variables in the one group are compared before and after treatment. In controlled trials, scores are compared between groups undergoing different treatments. Statistics based on group scores might show that a statistically significant change has occurred but they do not reveal how well a treatment works. That is revealed by categorical data that show the proportion of patients that benefitted by how much after treatment. Those proportions are qualified by the 95% confidence interval of the proportion. In observational studies the magnitude of the success rate, and its 95% confidence interval, is enough to indicate how well the treatment worked. In controlled trials, success rates can be compared to determine how more often is one treatment successful than another. Statistical significance of the difference is established if the confidence intervals of the respective success rates do not overlap, or if the 95% confidence interval of the difference between success rates does not include zero. Pragmatic trials compare one treatment with another, but the comparison does not show if either treatment works well. Each arm of a pragmatic trial constitutes an observational study and the data in each ach show how well each treatment works. Explanatory trials investigate the extent to which the observed success rate is provided by responses to non-specific effects of treatment. The extent to which active treatment is more often effective than is sham treatment can be demonstrated by comparing the success rates of the two treatments, by comparing their survival curves, and by calculating the attributable effect and the number needed to treat of the active treatment.