Abstract

PIP:

Dr Shapiro and his colleagues (July 28, p. 194) continue to insist that the results of their study on oral contraceptives and myocardial infarction are "discordant" with ours. However, when data on a group of their otherwise healthy subjects (provided for the first time in their letter) similar in age to ours are examined, the estimate of risk is entirely compatible with that in our study. The maximum likelihood estimate of relative risk from the entire data is 6.5 and the 95% confidence limits are wide (3.3 and 13). Our estimate of relative risk was 15 with a lower confidence bound of 6.2. The p value testing the discordance between these two estimates is 0.2. Dr Shapiro and associates continue to seriously distort the methodology of our study by giving the impression that the refusal rate connected with knowledge of drug use may have been as high as 85%. This impression is entirely false. Less than 3% of the subjects were excluded because of refusal by the attending physician or the patient to participate. The majority of exclusions (66%) came from hospital administrators who were unwilling to cooperate with the study from the start. An additional 15% were excluded because the woman or her physician had either died or could not be traced. As we have repeatedly pointed out, very few of the exclusions in our study could have introduced bias related to knowledge of drug use. The data of Shapiro et al. fully support our finding of a strong association between oral contraceptives and myocardial infarction in otherwise healthy women despite the fact that their paper contains (1) the inclusion of subjects for whom no inference can be drawn; (2) the failure to control important confounding variables; (3) the presentation of data incorrectly identified; (4) the claim of a discordance for results whose differences may be readily explainable by sampling variability; and (5) the distortion of relevant literature.