Variations in reproductive events across life: a pooled analysis of data from 505 147 women across 10 countries

Abstract

Study question: How has the timing of women's reproductive events (including ages at menarche, first birth, and natural menopause, and the number of children) changed across birth years, racial/ethnic groups and educational levels?

Summary answer: Women who were born in recent generations (1970-84 vs before 1930) or those who with higher education levels had menarche a year earlier, experienced a higher prevalence of nulliparity and had their first child at a later age.

What is known already: The timing of key reproductive events, such as menarche and menopause, is not only indicative of current health status but is linked to the risk of adverse hormone-related health outcomes in later life. Variations of reproductive indices across different birth years, race/ethnicity and socioeconomic positions have not been described comprehensively.

Study design, size, duration: Individual-level data from 23 observational studies that contributed to the International Collaboration for a Life Course Approach to Reproductive Health and Chronic Disease Events (InterLACE) consortium were included.

Participants/materials, setting, methods: Altogether 505 147 women were included. Overall estimates for reproductive indices were obtained using a two-stage process: individual-level data from each study were analysed separately using generalised linear models. These estimates were then combined using random-effects meta-analyses.

Main results and the role of chance: Mean ages were 12.9 years at menarche, 25.7 years at first birth, and 50.5 years at natural menopause, with significant between-study heterogeneity (I2 > 99%). A linear trend was observed across birth year for mean age at menarche, with women born from 1970 to 1984 having menarche one year earlier (12.6 years) than women born before 1930 (13.5 years) (P for trend = 0.0014). The prevalence of nulliparity rose progressively from 14% of women born from 1940-49 to 22% of women born 1970-84 (P = 0.003); similarly, the mean age at first birth rose from 24.8 to 27.3 years (P = 0.0016). Women with higher education levels had fewer children, later first birth, and later menopause than women with lower education levels. After adjusting for birth year and education level, substantial variation was present for all reproductive events across racial/ethnic/regional groups (all P values < 0.005).

Limitations, reasons for caution: Variations of study design, data collection methods, and sample selection across studies, as well as retrospectively reported age at menarche, age at first birth may cause some bias.

Wider implications of the findings: This global consortium study found robust evidence on variations in reproductive indices for women born in the 20th century that appear to have both biological and social origins.

Study funding/competing interest(s): InterLACE project is funded by the Australian National Health and Medical Research Council project grant (APP1027196). GDM is supported by the Australian National Health and Medical Research Council Principal Research Fellowship (APP1121844).

Keywords: age at menarche; age at menopause; first birth; number of children; reproductive events.

Figure 1

Mean ages at (A) menarche (n = 493 395), (B) first birth (among parous women, n = 384 925) and (C) natural menopause (among postmenopausal women, n = 172 125) in the InterLACE Consortium. Mean ages were estimated in each study (accounting for design weights if available), and the estimates from each study were combined using random-effects meta-analysis. Median (interquartile range, IQR) ages were also presented for each study. Data on age at first birth were only included in the analysis from women aged ≥40 years at last follow-up and data on age at natural menopause only from women aged ≥55 years at last follow-up.

Figure 2

Mean age at menarche (n = 493 395). Data are stratified by (A) year of birth (P for trend = 0.0014) and (B) racial/ethnic/regional group. Mean age at menarche was stratified by birth year and race/ethnicity/region in each study and mutually adjusted (i.e. birth year and race/ethnicity/region) using linear regression models (accounting for design weights if available), and the adjusted estimates from each study were combined using random-effects meta-analysis for each category of covariates.

Figure 3

Proportion of women with no children (nulliparity) among women aged ≥40 years at last follow-up (n = 453 515). Data are stratified by (A) year of birth (P for trend = 0.37), (B) racial/ethnic/regional group and (C) education level (P for trend = 0.0395). Proportion of nulliparity was stratified by birth year, race/ethnicity/region and education level in each study (accounting for design weights if available), and the unadjusted estimates from each study were combined using random-effects meta-analysis for each category of variables.

Figure 4

Mean age at first birth among parous women aged ≥40 years at last follow-up (n = 384 925). Data are stratified by (A) year of birth (P for trend = 0.87), (B) racial/ethnic/regional group and (C) education level (P for trend = 0.0031). Mean age at first birth was stratified by birth year, race/ethnicity/region and education level in each study and mutually adjusted (i.e. birth year, race/ethnicity/region and education level) using linear regression models (accounting for design weights if available), and the adjusted estimates from each study were combined using random-effects meta-analysis for each category of covariates.

Figure 5

Mean age at natural menopause among postmenopausal women aged ≥55 years at last follow-up (n = 172 125). Data are stratified by (A) year of birth (P for trend = 0.26), (B) racial/ethnic/regional group and (C) education level (P for trend = 0.012). Mean age at natural menopause was stratified by birth year, race/ethnicity/region and education level in each study and mutually adjusted (i.e. birth year, race/ethnicity/region and education level) and additionally adjusted for smoking status and BMI using linear regression models (accounting for design weights if available), and the adjusted estimates from each study were combined using random-effects meta-analysis for each category of covariate.