More twins expected in low-income countries with later maternal ages at birth and population growth

Abstract

Study question: How are the changing maternal age structure and population growth expected to shape future twinning rates in low-income countries?

Summary answer: With maternal age at birth projected to shift toward older ages, twinning rates are also estimated to increase in most low-income countries by 2050 and even more by 2100.

What is known already: Many of the sub-Saharan African and South Asian countries are undergoing, and projected to further experience, the shift of maternal age at birth to older ages. Advanced maternal age is a well-established predictor of multiple births at the individual level, but currently, it is unknown how the changes in maternal age distribution are associated with the changes in twinning rates at the population level in low-income countries.

Study design, size, duration: We first estimated age-specific twinning probability based on Demographic Health Surveys and World Fertility Surveys data. We then scaled up the age-specific twinning probability at the population level to estimate changes in the number of twin births in 2050 and 2100 attributable to the estimated shifts in maternal age toward older ages as projected by the UN World Population Prospects (WPP).

Participants/materials, setting, methods: We analyzed ∼3.19 million births that occurred within 10 years before the interview. Majority of the births in our data took place between 1980 and 2015 across 39 countries, where the uptake of medically assisted reproduction (MAR) is known to have been low during the observation period. We estimated country fixed-effects models to obtain country-specific twinning rates and age-specific twinning probability. We applied these estimates to the future number of births projected by the UN WPP, to estimate the number of twin births in 2050 and 2100.

Main results and the role of chance: With maternal age at birth projected to shift toward older ages, twinning rates are also estimated to increase in most countries by 2050 compared to 2010 (increases from 0.3% to 63% depending on countries), and even more in all studied countries by 2100 (increases from 3.5% to 79%). Due to its large population size, India will continue to have among the largest share of twin births despite its estimated decline of twin births by 10.5% by 2100. Nigeria, due to its not only large and growing population size but also high twinning rate, is expected to contribute the second largest number of twin births.

Limitations, reasons for caution: Although the accuracy in maternal recall of multiple births tends to be high, our use of data based on recalled births from the past nonetheless imply a potential bias in our estimation of twinning rates.

Wider implications of the findings: The present study suggests that, even if the spread of MAR remains slow in many low-income countries, twinning rates and number of twin births are expected to grow as an increase in maternal age at birth and population growth continue. Our findings call for more public health attention and societal support to be paid to twins and their families in low-income countries, given that twins are at higher risk of developmental challenges and health disadvantages.

Study funding/competing interest(s): D.S.L. was supported by the European Union (ERC, BIOSFER, 101071773). K.J.B. was supported by a Pro Futura Scientia XIV Fellowship awarded by the Swedish Collegium for Advanced Study and Riksbankens Jubileumsfond. There are no competing interests to declare.

Trial registration number: N/A.

Keywords: Demographic Health Surveys; Southern Asia; UN World Population Prospects; maternal age structure; sub-Saharan Africa; twinning rate.

Figure 1.

Twinning rates estimated based on the Demographic Health Surveys data. Posterior distribution of country fixed-effects taken from a regression model.  ${\mathrm{twin}\mathrm{ }\mathrm{birth}}_{i,c}=\mathrm{ }\beta {\mathrm{Maternal}\mathrm{ }\mathrm{age}\mathrm{ }\mathrm{at}\mathrm{ }\mathrm{childbirth}}_i+{\gamma }_c+{\epsilon }_{i,c}$ where ${\mathrm{twin}\mathrm{ }\mathrm{birth}}_{i,c}$ is a binary outcome of either twin or singleton, of a mother i from a country c in year y. $\beta$ refer to the relationship between the outcome with maternal age at childbirth of the mother i at the given birth, categorized at 5-year intervals (15–19, 20–24, 25–29, 30–34, 35–39, 40–44, 45–49). IQR, interquartile range (Q1–Q3).

Figure 2.

Probability of twinning by maternal age at birth, relative to the probability at age 19 years and below. Each distribution shows the posterior distribution of relative change in the probability of twinning, estimated based on the regression model ${\mathrm{twin}\mathrm{ }\mathrm{birth}}_{i,c}=\beta {\mathrm{Maternal}\mathrm{ }\mathrm{age}\mathrm{ }\mathrm{at}\mathrm{ }\mathrm{childbirth}}_i+{\gamma }_c+{\epsilon }_{i,c}$ where ${\mathrm{twin}\mathrm{ }\mathrm{birth}}_{i,c}$ is a binary outcome of either twin or singleton, of a mother i from a country c in year y. $\beta$ refer to the relationship between the outcome with maternal age at childbirth of the mother i at the given birth, categorized at 5-year intervals (15–19, 20–24, 25–29, 30–34, 35–39, 40–44, 45–49).

Figure 3.

Predicted percent change in twinning rates compared to 2010. Values used to make the figure can be found in Supplementary Table S4.

Figure 4.

Bubble chart showing the projected number of twin births for 2100, by the projected number of births (as proxy of population growth) and baseline twinning rates estimated from the Demographic Health Surveys (also see  Fig. 1).