Concerns that the findings of a differential VTE risk between second- and third-generation OCPs were not a result of true association but of bias and confounding led to a fierce and lengthy debate in the literature90- 99 and to many reanalyses of the original data.11,74- 77,82,83 Prescription bias, "attrition of susceptibles" or "starter effect" bias, and recall and
"switching pill" bias are among the many biases and confounding factors that have been implicated as accounting for the differential risk. Arguments supporting and refuting the presence of these biases have been extensively exchanged.11,23,28,37,52,57,61,63,74- 76,82,84,85,87,100,101 Independent reviews85,102,103 of all available data have concluded that the biases may partially account for, but do not seem to entirely explain, the differential risk. In addition, 2 meta-analyses71,72 have concluded that there is a small but real differential risk of VTEs between third- and second-generation OCPs. An interesting but concerning finding of one of the meta-analyses71 is that results differ depending on the study funding sources: a significant differential risk between third- and second-generation OCPs was found by non–industry-sponsored studies, whereas industry-sponsored studies showed a point risk estimate of approximately 2.0 but with wide confidence intervals that included the unity.