Balancing risks and benefits of oral contraception.

We publish in this issue an authoritative and timely opinion on the effects of third-generation oral contraceptives on the incidence of venous thromboembolism (VTE) and acute myocardial infarction (AMI) among users (Spitzer, 1997). The author is a highly experienced and leading epidemiologist in the field, and his conclusions are disturbing. He details what seems to be serious misinterpretations of epidemiological data by certain regulatory authorities, especially in the UK. Serious social consequences were set in train in many countries as a result of their public pronouncements questioning the safety of third-generation, as compared with second-generation progestagens in oral contraceptives (OCs). Coping with the situation that emerged from the 'non-epidemic', to use his term, has raised totally new matters which urgently need clarification. These are concerned with decision-taking and public relations. Unless these matters are resolved, society may face further rounds of unnecessary abortions, discontinued use of a highly safe group of OCs, perhaps many unwanted pregnancies and apparently unnecessary crises in public confidence. Epidemiology is a far-from-easy subject, especially when statistical parameters such as odds ratios reach low levels, as in datasets comparing the safety of OCs. Interpretations demand a high awareness of the effects of bias and confounders that can mislead deductions drawn from a too-simple analysis of available statistical evidence. Expert advice is essential to enable realistic conclusions to be reached when doubts arise about the correct interpretation of new data. Balanced interpretations are only possible after the passage of time, and we have today some advantages of hindsight concerning the events of two years ago. The role of regulatory authorities is made more difficult when experts disagree, which may still be the case today as we await publication of further studies which may add new aspects to this particular field. Perhaps some clarification is emerging about certain aspects of epidemiological analyses. The effects of confounders and biases such as preferential prescribing, duration of use effects and prudent doctor bias in conducting and interpreting trials on OCs, are well described by Spitzer (1997) and we have no need to reiterate them here. Yet disagreements between analysts on the relative risks of second-and third-generation progesta-gens might be as great today as they were two years ago (Farmer and Lawrenson, 1997; Poulter et al., 1997; Vandebroucke et al., 1997). Firmer conclusions about safety can be drawn from preliminary abstracts published this year. These report similar rates of absolute incidence, morbidity …