Respond

Abstract

We thank niels Keiding and Thomas Scheike for their commentary pertaining to our article on selection bias due to parity-conditioning in studies of time trends in fertility. We agree that a clear postponement of parenthood (labeled “survivor selection” by Keiding and Scheike) has occurred over time in Western countries, and that this is explained by behavioral and volitional rather than biological factors. However, postponement of parenthood is an entirely different concept from the bias due to parity-conditioning that we presented. in contrast to behavior-driven postponement of parenthood, parity-conditioning bias is a design-driven selection bias with a strong biological component and strength that varies according to sampling frame. The key issues behind parity-conditioning bias are the definition of a study population with a wide age range at a cross-section of time, restriction by classes of parity, heterogeneity in biological fertility, and differential success in at-risk cycles before the cross-section. Thus, parity-conditioning bias is a design-driven phenomenon that is independent of behavioral or societal developments. inappropriate restriction on parity classes may cause selection bias, whereas postponement of parenthood may, if anything, distort the estimation of the time trend in fecundity. Both these phenomena should be considered in studies of time trends in fecundity. We took the Swedish study as an example, to demonstrate the potential for parity-conditioning bias. it defined eligibility as follows: “our study population encompasses all primiparous women aged 20 years or older at the time they initiated their pregnancy attempt, who subsequently gave birth from 1983 through 2002.” 192) Thus, the researchers made restrictions due to classes of parity at two cross-sections of time. Specifically, all the women who were parous on december 31, 1982, were excluded, as were those who remained childless at the end of follow-up. Table 2 in Sallmén et al., the Lexis diagrams in Sallmén at al., and Keiding and Scheike’s frequencies of nulliparous women in denmark in 1980–2010 clearly demonstrate the difference in age group-specific exclusions at a cross-section of time in a study using a similar sampling frame. in denmark, about 12% to 3.6% of 20-year-old women were parous at cross-sections of time from 1980 to 2010. The same figures were about 91% to 86% for women aged 40. Moreover, the table shows the age group-specific frequencies of nulliparous women who would be excluded in a study with follow-up ending before menopause. in consequence, restriction by parity classes leads to an arbitrary trend over time in age at first birth. in the Swedish study, the 1945 cohort members should have had their first child at the age of 37 or over to be included, whereas the 1979 cohort members should have reproduced before 24 years old. The exclusions in the Swedish study were also differentially related to fecundity. Fecundity has a biological effect on parity. once a study population with a wide age