Human Fertility Decline?

Abstract

The authors respond: Joffe and his colleagues1 approach the problems of retrospective fertility studies with an optimism we wish we could share. We began our work suspecting that studies of fertility trends over time might be subject to bias due to demographic trends over time, and we had to conclude that the biases are plausibly large enough that such studies appear to be virtually impossible to interpret.2 We explored underascertainment of 2 factors: unprotected acts of intercourse and induced abortions. The former contributes heavily to distorting estimates of fecundability. Induced abortions are less likely to be forgotten but are often concealed. We used published data on changes in induced abortions and the use of contraception merely to illustrate the extent of these problems and the consequent bias, not to recreate reality. Using assumptions based on the published data, we found that the simulations could produce substantial but spurious trends in fertility over time.2 Biases produced by the 2 factors easily could have affected previously published reports on changes in fertility over time.3–5 Neither of the TTP studies3,4 included study populations representative of the full fecundability distribution, because both were limited to births and therefore selectively eliminated aborted accidental pregnancies. The Finnish study5 categorized women as fertile or infertile (our design II) but did not consider prior cycles at risk. Some of the analytic solutions proposed by Joffe et al1 seem inadequate. His study3 is not an exception in that only a small fraction of induced abortions were recorded. The strategy of assigning them a short time to pregnancy is likely to distort reality to varying degrees in different time periods. The usefulness of national data on abortions may be limited. We agree with Joffe et al1 that reality is far more complex than our simple simulations, as we stated in our paper.2 For example, we mentioned changes in desired age at first birth as a source of distortion. Even such a simple factor as desired family size can affect results based on the first intended pregnancy; the most fertile couples can reach their desired family size through accidental pregnancies, especially if their desired size is small. Simple age adjustment may not suffice, especially if changes over time in such complex social factors vary by education and other factors. Identifying biologic changes in fertility over time in the face of complex social change may ask too much of retrospective fertility studies. We remain pessimistic. Markku Sallmén Donna Baird Allen Wilcox Epidemiology Branch [email protected] Clarice Weinberg Biostatistics Branch National Institute of Environmental Health Sciences Durham, NC Marja-Liisa Lindbohm Department of Epidemiology and Biostatistics Finnish Institute of Occupational Health Helsinki, Finland