To the Editor:
Sallmén et al1 investigated bias in studying trends in the capacity to conceive given unprotected intercourse using the methodology of time-to-pregnancy (TTP) studies.2 They argued that without effective contraception or induced abortion, some highly fertile couples conceive accidental pregnancies, leaving a relatively subfertile subset to conceive nonaccidental, and therefore TTP-eligible, pregnancies. With rising use of effective contraception and abortion, this effect declines, and so the more recent pregnancies arise from a more fertile group; contraception or abortion now keep the most fertile couples in the “at-risk” population.
We do not dispute the biases illustrated by their stylized calculations, but we disagree that the bias is “probably impossible” to investigate. Information on accidental pregnancies can, in fact, be obtained. It is standard methodology to use this information, but Sallmén and colleagues ignored this.2,3 If the proportion of accidental pregnancies does not fall over time, then logically the bias cannot be present.2,3 The standard sensitivity analysis, including accidental pregnancies in a TTP analysis with imputed TTP = 1,2 overcompensates the bias. Bias associated with changing rates of induced abortion can also be investigated by comparing the observed trend in fertility with national data on the annual abortion rate.4
Furthermore, reality is much more complex than the authors’ stylized calculations. The issue is whether these underlying trends actually occur. One of the cited studies found that increasing fertility was not accompanied by a decreasing proportion of accidental pregnancies, nor was the observed trend altered by excluding TTP values of zero and one.4 The other TTP study cited by Sallmén and colleagues included accidental pregnancies,5 a practice that reverses the direction of bias, yet this study also showed increasing fertility.
Couple behavior cannot simply be extrapolated from efficacy of contraception and use of legal abortion. In our 4 Danish and 2 British datasets, in the periods before 1970, 1970–1984, and after 1984, the percentages of accidental pregnancies were, respectively, 6.8%, 6.6%, and 2.9% (Denmark) and 12.5%, 11.2%, and 17.5% (Britain) (detailed data available on request). The predicted fall between the first 2 periods is not seen in the data.
Finally, some small points. Why would changing family size affect analysis of first births? Changing age at first birth can be readily adjusted for statistically. TTP eligibility is based on behavior affecting biology (contraceptive practice),2 not a psychologic construct such as intended, planned, or wanted.
Department of Epidemiology and Public
Imperial College of Science, Technology
Department of Biostatistics
University of Copenhagen
Tina Kold Jensen
Institute of Public Health
Department of Environmental Medicine
University of Southern Denmark
1. Sallmén M, Weinberg CR, Baird DD, et al. Has human fertility declined over time?—why we may never know. Epidemiology
2. Joffe M, Key J, Best N, et al. Studying time to pregnancy using a retrospective design. Am J Epidemiol
3. Weinberg CR, Baird DD, Wilcox AJ. Sources of bias in studies of time to pregnancy. Stat Med
4. Joffe M. Time trends in biological fertility in Britain. Lancet
5. Akre O, Cnattingius S, Bergström R, et al. Human fertility does not decline: evidence from Sweden. Fertil Steril