Effects of changes in smoking status on risk estimates for myocardial infarction among women recruited for the Royal College of General Practitioners' Oral Contraception Study in the UK

Study objective-To determine whether changes in smoking status among women recruited for the Royal College of General Practitioners' Oral Contraception Study affected previous risk estimates for myocardial infarction. Design-(1) Postal survey between November 1994 and July 1995 of women still under general practitioner observation. Validation of the smoking information supplied by the women on the questionnaire by comparison with that reported by the general practitioner at recruitment to the main study. (2) Nested case-control study of 103 cases of myocardial infarction, matched with 309 controls, to see if different risk estimates were obtained when smoking status at recruitment or smoking status at time of event were used in the analysis. Setting-650 general practices throughout the United Kingdom. Participants-10 073 women who responded to the questionnaire (85.4% of 11 797 sent out). Main results-There was good agreement between smoking information recorded by the general practitioner at recruitment and that supplied retrospectively by respondents to the questionnaire. The risk estimates for myocardial infarction associated with use of combined oral contraceptives (COCs) were almost identical irrespective of whether smoking status at recruitment or at time of event was used for the statistical adjustment. This was because few women stopped smoking while also using COCs. In fact, fewer regular smokers who have ever used COCs reported stopping smoking than never users. The risk estimates for myocardial infarction associated with smoking were smaller when smoking habits at recruitment was used than when smoking habits at time of event was used. Conclusions-Previous results from the Oral Contraception Study regarding the effects of COCs are unlikely to hare been biased by changes in the smoking habits of the cohort, but the effects of smoking have probably been underestimated. There is still a need for effective health education regarding the risks associated with smoking, particularly among users of COCs.