Microsimulation reveals that medically assisted reproduction is unlikely to compensate for cohort fertility decline due to increasing maternal ages

STUDY QUESTION Can medically assisted reproduction (MAR) compensate for completed cohort fertility (CCF) decline within coresidential unions due to increasing maternal ages among Dutch women born during 1974–1984? SUMMARY ANSWER MAR is unlikely to compensate for cohort fertility decline within coresidential unions due to continued increase in maternal ages in our sample cohort of Dutch women born during 1974–1984, even under ideal conditions. WHAT IS KNOWN ALREADY The pregnancy- and live birth rates for both expectant management and MAR decline at older reproductive ages. Some infertile couples can conceive naturally without undergoing treatment by trying to conceive for a longer period of time, which complicates estimating the contribution of MAR to cohort fertility. STUDY DESIGN, SIZE, DURATION We developed a microsimulation model, which includes MAR, that simulates the reproductive life courses of women. We simulate a sample of 1 000 000 women representing the 1974–1984 Dutch birth cohort. Sample sizes for the input parameters varied from hundreds to thousands to hundreds of thousands depending on the parameter and source (surveys, clinical studies, panel data, population registers). PARTICIPANTS/MATERIALS, SETTING, METHOD Our Monte Carlo microsimulation model uses probability distributions and parameters based on representative data sources to determine a woman’s transitions through union and reproductive events across her reproductive life course. We assess the contribution of various components of the MAR process to CCF within coresidential unions and estimate the net contributions of MAR to CCF with counterfactual simulations. MAIN RESULTS AND THE ROLE OF CHANCE Increases in the maximum female age at MAR treatment, the time to starting IUI treatment, and the number of IUI cycles did not noticeably increase CCF. Out of the hypothetical policy levers that were adjusted, increasing the share of eligible women who took up MAR from 0% to 100% increased CCF linearly by 0.06 children. Increasing the waiting time to ART treatment after infertility diagnosis from 1 to 12 months reduced CCF by 0.01. Increasing the number of reimbursed IVF/ICSI cycles from 0 to 6 increased CCF by 0.05. The increase tapered off after cycle number 2, and levelled off after cycle number 4. Minimum ages at which MAR treatment was started that were between 20 and 30 resulted in near-identical patterns of change in CCF from the adjustment in the potential policy levers, whereas both the CCF levels and rate of increase or decline were reduced at age 35 years, and further at age 40 years. The main influences on the lower CCF levels and slower rate or increase or decline were the reduction in MAR pregnancy rates at advanced reproductive ages, the increasing probability of intrauterine mortality with age, and time spent trying to conceive naturally before starting MAR treatment. The net contribution of MAR (0.043 children per woman, 2.5% to non-MAR CCF) was 33% smaller than the observed MAR birth share of CCF (0.064 children per woman, 3.7% of simulated CCF including MAR). The relative contribution of MAR was strongest at ages 35–39 (6.0%) and at age 40 years and above (8.6%), and weakest at ages 20–24 (0%) and 25–29 years (1.0%). In absolute number of children born, most of the contribution of MAR occurred at ages 30–34 (0.016 children) and ages 35–39 years (0.015). The Monte Carlo variation in the simulation model was measured to be around ±0.001 based on 10 simulations of 1 000 000 women. Parameter uncertainty was accounted for where possible by allowing parameters to vary within nonparametrically bootstrapped values generated from the sample data. LIMITATIONS, REASONS FOR CAUTION Our model cannot fully capture the complexity of the fertility and MAR processes with regards to factors like infertility diagnosis, partner characteristics, and health status. Due to missing information, an extensive number of parameters, and limited sample sizes when estimating some of the parameters; the model output deviates slightly from the reference data. We cannot establish causality due to endogeneity in the modelling. We are not entirely confident in the zero contribution of IUI, because clinical data suggest that IUI pregnancy rates are higher than expectant management pregnancy rates, but the quality of that clinical data is low and sparse. The scope of the study is limited to births within coresidential unions, because information on unions and births outside of coresidential union was insufficient for modelling purposes. WIDER IMPLICATIONS OF THE FINDINGS While MAR remains an important tool to address infertility, our results suggest that it is unlikely to compensate for population fertility decline attributable to postponement of childbearing to older reproductive ages. Failing to account for the counterfactual case that some couples who undergo MAR may conceive naturally given longer expectant management may lead to overestimation of the contribution MAR can make to population fertility.

     Microsimulation reveals that medically assisted reproduction is unlikely to compensate for cohort fertility decline due to increasing maternal ages

     Human Reproduction

    Rolf Granholm, Astrid Cantineau, Annemieke Hoek, Gert Stulp

     Download manuscript here

Abstract

STUDY QUESTION Can medically assisted reproduction (MAR) compensate for completed cohort fertility (CCF) decline within coresidential unions due to increasing maternal ages among Dutch women born during 1974–1984?

SUMMARY ANSWER MAR is unlikely to compensate for cohort fertility decline within coresidential unions due to continued increase in maternal ages in our sample cohort of Dutch women born during 1974–1984, even under ideal conditions.

WHAT IS KNOWN ALREADY The pregnancy- and live birth rates for both expectant management and MAR decline at older reproductive ages. Some infertile couples can conceive naturally without undergoing treatment by trying to conceive for a longer period of time, which complicates estimating the contribution of MAR to cohort fertility.

STUDY DESIGN, SIZE, DURATION We developed a microsimulation model, which includes MAR, that simulates the reproductive life courses of women. We simulate a sample of 1 000 000 women representing the 1974–1984 Dutch birth cohort. Sample sizes for the input parameters varied from hundreds to thousands to hundreds of thousands depending on the parameter and source (surveys, clinical studies, panel data, population registers).

PARTICIPANTS/MATERIALS, SETTING, METHOD Our Monte Carlo microsimulation model uses probability distributions and parameters based on representative data sources to determine a woman’s transitions through union and reproductive events across her reproductive life course. We assess the contribution of various components of the MAR process to CCF within coresidential unions and estimate the net contributions of MAR to CCF with counterfactual simulations.

MAIN RESULTS AND THE ROLE OF CHANCE Increases in the maximum female age at MAR treatment, the time to starting IUI treatment, and the number of IUI cycles did not noticeably increase CCF. Out of the hypothetical policy levers that were adjusted, increasing the share of eligible women who took up MAR from 0% to 100% increased CCF linearly by 0.06 children. Increasing the waiting time to ART treatment after infertility diagnosis from 1 to 12 months reduced CCF by 0.01. Increasing the number of reimbursed IVF/ICSI cycles from 0 to 6 increased CCF by 0.05. The increase tapered off after cycle number 2, and levelled off after cycle number 4. Minimum ages at which MAR treatment was started that were between 20 and 30 resulted in near-identical patterns of change in CCF from the adjustment in the potential policy levers, whereas both the CCF levels and rate of increase or decline were reduced at age 35 years, and further at age 40 years. The main influences on the lower CCF levels and slower rate or increase or decline were the reduction in MAR pregnancy rates at advanced reproductive ages, the increasing probability of intrauterine mortality with age, and time spent trying to conceive naturally before starting MAR treatment. The net contribution of MAR (0.043 children per woman, 2.5% to non-MAR CCF) was 33% smaller than the observed MAR birth share of CCF (0.064 children per woman, 3.7% of simulated CCF including MAR). The relative contribution of MAR was strongest at ages 35–39 (6.0%) and at age 40 years and above (8.6%), and weakest at ages 20–24 (0%) and 25–29 years (1.0%). In absolute number of children born, most of the contribution of MAR occurred at ages 30–34 (0.016 children) and ages 35–39 years (0.015). The Monte Carlo variation in the simulation model was measured to be around ±0.001 based on 10 simulations of 1 000 000 women. Parameter uncertainty was accounted for where possible by allowing parameters to vary within nonparametrically bootstrapped values generated from the sample data.

LIMITATIONS, REASONS FOR CAUTION Our model cannot fully capture the complexity of the fertility and MAR processes with regards to factors like infertility diagnosis, partner characteristics, and health status. Due to missing information, an extensive number of parameters, and limited sample sizes when estimating some of the parameters; the model output deviates slightly from the reference data. We cannot establish causality due to endogeneity in the modelling. We are not entirely confident in the zero contribution of IUI, because clinical data suggest that IUI pregnancy rates are higher than expectant management pregnancy rates, but the quality of that clinical data is low and sparse. The scope of the study is limited to births within coresidential unions, because information on unions and births outside of coresidential union was insufficient for modelling purposes.

WIDER IMPLICATIONS OF THE FINDINGS While MAR remains an important tool to address infertility, our results suggest that it is unlikely to compensate for population fertility decline attributable to postponement of childbearing to older reproductive ages. Failing to account for the counterfactual case that some couples who undergo MAR may conceive naturally given longer expectant management may lead to overestimation of the contribution MAR can make to population fertility.