Regarding the paper by Tordoff et al.1 in the September 2019 issue of Epidemiology, the authors have identified an error in the interpretation of the minimum and maximum selection bias odd ratio and the risk ratio due to confounding.

The Results section on page 675 states, in error, that: “The closer the minima and maxima are to one, the more robust the estimates are to bias. Conversely, the larger the range in which the resulting PR can be either above or below one, the more vulnerable an estimate was to selection bias and confounding.” The footnote in Table 5 similarly states that: “The closer the minima and maxima are to one, the more robust the estimates are to bias.” In both cases, the statements should be: “The closer the minima or maxima are to 1.0, the more sensitive the results are to selection bias or confounding (i.e., a smaller degree of bias is needed to change the direction of the association).”

This extends to several interpretive statements in the Results and Discussion. Estimates for hysterectomy and lifetime PSA testing among gender nonconforming individuals as well as Pap testing and pregnancy were all robust to small degrees of systematic bias (i.e., a larger degree of bias is needed to change the direction of the association). However, the PR estimates for hysterectomy (both for transgender men and gender nonconforming individuals) and PSA testing among transgender women were sensitive to smaller degrees of bias. All associations were sensitive to moderate and large degrees of bias.

Importantly, the main conclusion of this study is unchanged by this error: “This study provides evidence that BRFSS data provide non-representative estimates of transgender and gender non- conforming individuals who are asked questions about their reproductive health, and that these outcomes are vulnerable to bias.”1

1. Tordoff DM, Andrasik M, Hajat A. Misclassification of sex assigned at birth in the behavioral risk factor surveillance system and transgender reproductive health: a quantitative bias analysis. Epidemiology. 2019;30:669–678.