The Impact of Bias in the Tri-State Toxic-Shock Syndrome Study-Reply

We shall respond separately to the several points made by Drs Osterholm, Davis, Gibson, and Mandel. First, as we noted previously and in the article discussing temporal precedence, the five case-control studies of tampons and TSS had differences in the procedures used for assembling patients and ascertaining data. The number of cases ranged from 29 to 80, and there were one to four controls per case; the controls were selected from such diverse sources as neighborhoods where the cases lived, designated friends of the cases, and women attending gynecologic or adolescent medicine clinics. These differences in research tactics, however, are overshadowed by the basically similar methods that made each of the studies vulnerable to bias in diagnosis, data collection, and temporal precedence. Effects of Temporal Precedence on Calculation of the Odds Ratio 20% Error Rate 30% Error Rate Cases Controls Cases Controls Tampons used before illness onset 72 123