Liquid-based cytology.

Oral contraceptives and diabetes mellitus: an update There has long been interest in the possible relationship between oral contraceptive (OC) use and diabetes mellitus. In 1991, we reported our findings (in this Journal) on 45 women who had been referred to hospital for diabetes during follow-up in the Oxford-Family Planning Association (Oxford-FPA) contraceptive study. No association was found with OC use.1 We nonetheless thought it would be of interest to comment briefly on the findings for this disease up to the time that individual follow-up of the study participants ceased in July 1994 (follow-up of cancer registrations and death notifications is still continuing). The Oxford-FPA study methods have been described in detail elsewhere.2 In brief, the study includes 171032 white women who, when recruited between 1968 and 1974, were married and aged between 25 and 39 years. At entry, 56% were using OCs, 25% a diaphragm and 19% an intrauterine device. These women (save for certain subgroups – see Vessey and Painter3) were followed up annually and information was collected about changes in contraceptive methods, pregnancies and their outcome, hospital referrals and deaths. Women who at entry to the study reported that they were suffering from diabetes were excluded from the present analyses. There were 81 cases remaining. Only the first hospital referral (inpatient or outpatient) was taken into account in the analyses. As expected, hospital referral was strongly positively related to age and body mass index (BMI). In addition, referral was three times as common in women of lower social class (IV–VI) as in women of upper social class (I–II), a difference only partly explained by BMI. Analyses of hospital referral rates in relation to OC use were therefore adjusted for age, BMI and social class. Our first analysis compared women ever using OCs with those never doing so. The rate ratio was 0.8 with a 95% confidence interval (CI) ranging from 0.5 to 1.3. Rate ratios for hospital referral in relation to total duration of OC use were as follows (95% CIs are given in parentheses): never used, 1.0 (reference category); 1–48 months, 0.9 (0.3–2.1); 49–96 months, 0.7 (0.3–1.7); 97–144 months, 0.9 (0.5–1.7); 145 months or more, 0.6 (0.2–1.6). Corresponding rate ratios in relation to interval since last use of OCs were as follows: never used, 1.0 (reference category); current–48 months, 0.7 (0.3–1.4); 49–96 months, 0.7 (0.3–1.7); 97–144 months, 0.6 (0.2–1.5); 145 months or more, 1.5 (0.7–2.8). The data were too few to enable analyses to be done by type of OC, but it should be noted that preparations containing 50 μg oestrogen made up 67% of OC exposure. OCs containing a greater amount of oestrogen provided only 2% of exposure. We recognise the shortcomings of our data, which include the small number of affected women and the associated fact that only those referred to hospital with diabetes were identified. Nonetheless we believe that our case finding has been unbiased with respect to OC use. Furthermore, as we have pointed out previously,1 if such a bias existed it might be expected to lead to hospital referral of more OC users than nonusers. In conclusion, the final results of the Oxford-FPA study with respect to diabetes mellitus offer further support to the view that OC use does not increase the risk of clinical diabetes mellitus, a finding in keeping with most other studies.4–6 Martin Vessey, FRCP, FRS Emeritus Professor of Public Health, Unit of Health Care Epidemiology, University of Oxford Old Road Campus, Headington, Oxford OX3 7LF, UK. E-mail: [email protected]