The dangers of inferring treatment effects from observational data: a case study in HIV infection

Several recent articles have implicitly questioned the need for randomized controlled trials, based on a comparison of treatment effects from observational studies and related randomized controlled trials. We present here a counterexample of a comparison of two antiretroviral drugs used in the treatment of HIV infection, in which the observational analysis gave a potentially misleading result. Examples such as this emphasize the need to regard randomized controlled trials as the primary mechanism for assessing therapeutic efficacy.