PIP: Critical comment is provided on the risk of pelvic inflammatory disease (PID) with the use of IUDs, and specifically the Dalkon Shield. The reopening of the issue of IUD and PID was precipitated by the reanalysis of the Women's Health Study (WHS) data by Kronmal et al. The focus of the discussion is on the description of an epidemiological case control study, the WHS, assessing the criticisms of the study, the possible biases, and an interpretation of the conflicting analyses. The WHS was conducted in 8 health care centers in the US between 1976-78 and included hospitalized PID cases. Controls were other hospitalized patients. Burkman in 1981 and Lee et al, in 1983 published results of their analyses of the WHS data. Burkman estimated a 1.6 risk of PID among IUd users relative to all other contraceptive users and 1 or 2.1 relative to women with no history of PID. Lee estimated risk among Dalkon Shield users as 8.3 relative to women not using any contraceptive method prior to hospitalization, and 1.6 among users of other IUDs; analysis excluded women with a prior history of PID. Recently Kronmal et al, found fault with the WHS on 10 counts, which are provided in detail. Their reanalysis indicates there is no increased risk of PID for IUd users. However, the question is raised as to the appropriate reference group, -all other users of contraceptives or nonusers. Another question pertains to whether subjects should all, or just those without a prior PID history, be included. Due to the protective effect of oral contraceptives, the reference group should be nonusers of any type as analyzed by Lee et al. Lee et al. also excludes those with a prior PID history, which is the best solution to the 2nd question. The Kronmal analysis allowed those with a prior PID history, and the reference group was all women regardless of use. The Kronmal analysis also describes ways in which selection bias is involved in the higher risk of PID among Dalkon Shield users relative to nonusers. Some of the reasons are that women using the Dalkon Shield may have been less likely to seek early medical care, to have a greater likelihood of having their diseases properly diagnosed, and to identify more easily the Dalkon Shield as their IUD type. This reasoning is found to be excessive and unrealistic to explain a difference between a risk of 8.3 vs. 1.6, and the Dalkon Shield was deservedly removed as an unsafe product. Support for the IUD is better served by discussing the careful selection of users to reduce PID, meticulous insertion techniques, and close monitoring.