Menarche is the culmination of a complex sequence of events involving the maturation of the reproductive axis.1,2 Early menarche is associated with increased risk of adverse health outcomes, such as cardiovascular diseases, cancers, spontaneous abortion, and premature death,3-9 whereas late menarche is associated with increased risk of fractures.10,11 Studies have found trends toward earlier menarche during the past 5 to 10 decades in the US as well as globally.12-19 In the US, studies have additionally evaluated whether this trend varied by sociodemographic factors.13,20-26 Some of them13,21-23,25 showed significant racial and ethnic differences, whereas others20,24,26 did not, and most13,20,22,24 were limited to non-Hispanic Black vs White comparisons. Furthermore, most studies20,21,23,24,26 focused on mean age at menarche, with the frequency of early or late menarche rarely evaluated. Notably, obesity is a risk factor for early-onset puberty,27-31 and the prevalence of childhood obesity has increased in the US,32,33 leading to hypotheses on the potential role of obesity in the trends toward earlier menarche. However, whether obesity is the primary factor underlying the trends in menarche remains debatable.34 Whether and to what extent the trend in menarche is attributable to changes in early-life body mass index (BMI) remains to be determined.28
The menstrual cycle is a vital sign.35 The maturation of the reproductive axis, measured as the time from menarche to established cycle regularity, is another important but understudied hallmark of early-life menstrual health. Within 1 to 2 years after menarche, irregular cycles are considered a normal process of pubertal transition.36,37 Full maturation of the reproductive axis leads to more regular menstrual function.38 Longer time to regularity has been associated with lower fecundability, longer menstrual cycles, and increased risk of metabolic conditions and all-cause mortality.39-42 Whereas the trends in time to regularity (influenced by environmental pollutants)43,44 were evaluated in Japanese14 and French45 cohorts, it is not known whether it has also changed during the past several decades in the US.
In this study, we used data from a large, mobile application–based cohort of adults in the US to evaluate temporal trends in menarche and time to regularity among members of a racially and ethnically diverse study population born between 1950 and 2005. We analyzed overall temporal trends and whether observed trends differ by sociodemographic factors. Additionally, we explored whether BMI at menarche might mediate the observed temporal trends.
The Apple Women’s Health Study is a prospective digital cohort study in the US. Users of the Apple Research app on their iPhone were eligible if they had ever menstruated at least once in life, live in the US, were at least 18 years old (19 years in Alabama and Nebraska and 21 years in Puerto Rico), and were able to communicate in English. Eligibility also required sole use of an iCloud account and an iPhone. Enrollment began on November 14, 2019, and is ongoing. Participants provided written informed consent at enrollment. This study was approved by the institutional review board at Advarra. Details were described previously.46 On enrollment, participants were asked to complete surveys of demographics as well as reproductive and medical history. For this analysis, we included participants who reported female sex assigned at birth, who were enrolled until March 20, 2023, and who provided age at menarche information. We excluded those born in 1931 to 1949 due to potential survival bias and too few individuals representing this group. A conceptual model is shown in eFigure 1 in Supplement 1. The final study population included 71 341 participants; data analysis was limited to subsets who answered the relevant questions (eFigure 2 in Supplement 1). This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline.
Birth Year and Age at Menarche
We grouped self-reported year of birth as 1950 to 1969 (n = 5223), 1970 to 1979 (n = 12 226), 1980 to 1989 (n = 22 086), 1990 to 1999 (n = 23 894), and 2000 to 2005 (n = 7912). Participants were asked the question, “At what age did you have your first menstrual period? It’s okay to estimate,” with the following response options: “7 years old or younger,” integer options between 8 and 15 years old, “16 years old or older,” “I don’t know,” or “I prefer not to answer.” Those who indicated they did not know or preferred not to answer or did not respond were excluded. We derived the following measures: (1) age at menarche (in years) (we assigned the value of 7 to those aged ≤7 years [196 (0.3%)] and the value of 16 to those aged ≥16 years [2447 (3.4%)]); (2) early menarche (yes/no) (age at menarche <11 years47); (3) very early menarche (yes/no) (age at menarche <9 years48); and (4) late menarche (yes/no) (age at menarche ≥16 years).49
Participants were asked, “After your first menstrual cycle, how long did it take for your cycle to become regular?” with the following response options: “less than 1 year,” “1-2 years,” “3-4 years,” “more than 5 years,” “after using hormones (eg, birth control pills),” “They’re not yet regular,” “I don’t know,” or “I prefer not to answer.” Those who indicated don’t know or prefer not to answer or who did not respond were considered missing. We further excluded 224 individuals with potentially inaccurate time-to-regularity information (eMethods in Supplement 1). For the remaining 61 932 participants, we categorized time to regularity as 2 years or less, 3 to 4 years, more than 5 years, not yet regular, or regular after using hormones.
We considered the following self-reported variables to evaluate whether the temporal trends in age at menarche or time to regularity differ by sociodemographic factors: (1) self-identified race and ethnicity (Asian, Hispanic, non-Hispanic Black, non-Hispanic White, and other and multiple races (including American Indian or Alaska Native, Middle Eastern or North African, Native Hawaiian or other Pacific Islander, “None of these fully describe me,” and self-identified with >1 option)50; (2) subjective socioeconomic status (SES) at enrollment based on the MacArthur Scale of Subjective Social Status51 (categorized as 0-3 [low], 4-5 [medium], and 6-9 [high]), which was used as a surrogate for premenarche SES; and (3) geographic location (based on state of residence and categorized as Northeast, Midwest, South, and West).
In addition, among a subset of 9865 participants (13.8%) who retrospectively reported weight and height at menarche, we derived BMI (calculated as weight in kilograms divided by height in meters squared) for age z scores, percentiles, and categories at menarche using the Centers for Disease Control and Prevention Growth Chart.52-54 We considered BMI at menarche as a potential mediator of temporal trends in age at menarche or time to regularity.
We calculated means (SDs) for continuous variables and reported numbers (percentages) for binary or categorical variables, overall and stratified by birth year categories (χ2 tests were performed to identify differences in time to regularity by birth year categories). We summarized the percentages of time to regularity by age at menarche. We used generalized linear regression (gaussian or binomial distributions for continuous or binary categorical variables), with birth year as the exposure variable to generate P values for temporal trends.
To understand how temporal trends differ by sociodemographic factors, we performed analyses stratified by each covariate. A test of trend was performed within each level of the covariate by including birth year as the exposure variable in regression models. We also tested whether the slope of trends differed by covariates by including an interaction term between each covariate and birth year in the regression models and performing a type 3 test for significance.
We evaluated trends over time in the subset of 9865 participants with data on BMI at menarche. We performed an exploratory causal mediation analysis55,56 with nonparametric bootstrap (500 simulations) to quantify the proportions (95% CIs) of temporal trends in menarche or time to regularity mediated by BMI z score at menarche. We evaluated these temporal trends as a secondary analysis after stratifying by BMI categories at menarche or adjusting for BMI at menarche z scores.
To evaluate the robustness of our results, we performed sensitivity analyses, including evaluating the temporal trends in time to regularity when further adjusted for age at menarche, using models that mutually adjusted for race and ethnicity and SES and using multinomial logistic models for the categorical time-to-regularity variable, with 2 years or less as the referent group. Analyses were conducted in Python, version 3.6 (Python Software Foundation) and R, version 4.1.2 (R Project for Statistical Computing). All statistical tests were 2-sided with 95% CIs. P < .05 was considered statistically significant.
The Table shows the characteristics of the 71 341 participants. Among them, 2228 (3.1%) self-identified as Asian, 4918 (6.9%) as Hispanic, 3665 (5.1%) as non-Hispanic Black, 49 518 (69.4%) as non-Hispanic White, and 8461 (11.9%) as other or multiple races. A total of 21 561 (30.2%) had a high subjective SES level. The mean (SD) age at menarche was 12.2 (1.6) years, and 9174 (12.9%) had early menarche (aged <11 years). A total of 38 524 (62.2%) reached regularity within 2 years after menarche, whereas 6950 (11.2%) did not establish regularity. Characteristics of the 9865 participants with weight and height information at menarche are given in eTable 1 in Supplement 1. Compared with the full study population, these participants tend to have earlier birth years, be non-Hispanic White, and have high subjective SES.
Figure 1 shows the temporal trends of age at menarche and time to regularity. The mean (SD) age at menarche decreased from 12.5 (1.6) to 11.9 (1.5) years comparing those born in 1950 to 1969 vs 2000 to 2005 (P < .001 for trend) (Figure 1A and Table). The percentage of individuals experiencing early menarche increased from 449 (8.6%) to 1223 (15.5%), the percentage of individuals experiencing very early menarche increased from 31 (0.6%) to 110 (1.4%) for very early menarche, and the percentage of individuals experiencing late menarche decreased from 286 (5.5%) to 137 (1.7%) (P < .001 for trend) (Figure 1B and Table). From the 1950 to 1969 birth years to the 2000 to 2005 birth years, the percentage reaching regularity within 2 years decreased from 3463 (76.3%) to 4075 (56.0%), and the percentage not yet in regular cycles increased from 153 (3.4%) to 1375 (18.9%) (P < .001 for trend) (Figure 1C and Table). The mean (SD) time to regularity among those who spontaneously established regularity increased from 1.27 to 1.40 years (P < .001 for trend) (Table). eFigure 3 in Supplement 1 shows lower percentages of time to regularity less than 2 years among those with either early or late menarche (inverse U-shaped association). Further adjusting for age at menarche resulted in similar distributions of time to regularity (eFigure 4 in Supplement 1).
The temporal trends stratified by race and ethnicity are presented in Figure 2 (estimates in eTable 2 in Supplement 1). Participants who were Asian, Hispanic, non-Hispanic Black, or of other or multiple races or ethnicities had consistently earlier mean age at menarche than non-Hispanic White participants. All racial and ethnic groups had temporal trends toward earlier menarche (P < .001 for trend), but when compared with non-Hispanic White participants, the magnitude of decrease in mean age at menarche across birth year categories was larger among those self-identified as non-Hispanic Black, Asian, and other or multiple races or ethnicities (P = .003 for interaction). All racial and ethnic groups showed a decreasing proportion of time to regularity within 2 years and an increased proportion of not establishing regularity (P < .001 for trend) (eFigure 5 and eTable 2 in Supplement 1), although there was no interaction between race and ethnicity and birth year. The temporal trends stratified by SES are presented in Figure 3 (estimates in eTable 3 in Supplement 1). Compared with those with high SES, those with low SES had earlier menarche, lower proportion of time to regularity within 2 years (eFigure 6 in Supplement 2), and larger magnitude of decrease in age at menarche. The heterogeneity by race and ethnicity for the trend toward earlier menarche remained when further adjusted for SES and vice versa (eTable 4 in Supplement 1). There was no interaction between geographic region and birth year when adjusting for race and ethnicity (eTable 5 in Supplement 1).
Overall, using multinomial logistic regressions yielded P-for-trend values of <.001 for each category of time to regularity compared with ≤2 years. Among the 9865 participants who provided weight and height at menarche, the BMI z score, percentile, and prevalence of obesity increased across birth year categories (eFigure 7 and eTable 1 in Supplement 1). An exploratory mediation analysis showed that the proportion of the temporal trends toward earlier menarche mediated by BMI z score at menarche was 46% (95% CI, 35%-61%) (Figure 4). When stratified by BMI categories at menarche, the healthy and underweight group still showed a trend toward earlier menarche (eTable 6 in Supplement 1). When adjusted for BMI z scores, a trend toward earlier menarche remained (eTable 6 in Supplement 1), as did heterogeneity by race and ethnicity (eFigure 8 in Supplement 1). There was no evidence of significant mediation by BMI at menarche for the temporal trends in time to regularity (eTable 7 in Supplement 1).
This cohort study of 71 341 participants born between 1950 and 2005 in the US found temporal trends toward earlier menarche (earlier mean age, higher percentage of early menarche, and lower percentage of late menarche) and longer time from menarche to cycle regularity (lower percentage of time to regularity within 2 years, higher percentage of time to regularity within 3-4 years, and higher percentage of not establishing regularity). These trends remained across all sociodemographic groups but were stronger among certain non-White (specifically, Asian, non-Hispanic Black, and other or multiple races or ethnicities) and low SES groups. In an exploratory analysis, BMI at menarche may explain a significantly large proportion of the temporal trends toward earlier menarche.
Our findings of a temporal trend toward earlier menarche are consistent with some US-based studies, with similar magnitude of changes.12,13,16 Other studies indicated that age at menarche stabilized during the past 50 years, whereas evidence that the median decreased by 2.5 to 4 months in the past 25 years remains,20,57 consistent with a change from 12.2 to 11.9 years of age for those born in 1980 to 1989 vs 2000 to 2005 in our study. Despite a relatively small magnitude of change in mean age, our study is among the first to show that the percentages of early and very early menarche have also increased by almost 2-fold across birth years from 1950 to 2005, raising concerns that more individuals may be vulnerable to adverse health outcomes related to early menarche.3-6 Late menarche has decreased, which may have other health implications, such as the decreasing rates of fractures.58,59
We found that non-Hispanic Black participants had consistently earlier mean age at menarche than White participants, also similar to prior US-based studies.13,20-26 We also found that non-Hispanic Black participants had a larger magnitude of change toward earlier menarche across birth year categories compared with non-Hispanic White participants. Similarly, we found other groups (Asian and other or multiple races), rarely evaluated in previous studies of menarche, also had consistently earlier mean ages and larger magnitudes of change toward earlier menarche than non-Hispanic White participants. We found similar patterns for self-rated low SES (compared with high SES). The factors driving this widening gap of disparities remain to be explored; transethnic genome-wide association studies indicated that these disparities are unlikely to be attributed to genetic variations, suggesting they may be driven by other environmental or contextual factors that may, through racism, impact different pathways, leading to earlier menarche.60
Onset of menarche is closely related to attainment of adequate body fat via pathways such as increased insulin-like growth factor 1 and leptin that stimulate gonadotropin-releasing hormone.61 In our exploratory analysis, we found that BMI at menarche may explain 46% of the temporal trends in menarche. This finding suggests that childhood obesity, a risk factor for earlier puberty,27-30,62 which has increased in the US,32,33 could be a contributing factor to the trend toward earlier menarche. However, the remaining 54% remain unclear. Our exploratory analysis also showed that BMI may have contributed to earlier menarche among non-Hispanic White, Black, and other or multiple races, whereas the trend among Asian and Hispanic individuals remains to be further explored. Previous studies also showed that the biggest decrease in age at menarche occurred before the obesity epidemic in the US,34 suggesting that other factors need to be explored to explain these trends and disparities, including environmental factors (eg, endocrine-disrupting chemicals, metals, or air pollutants could impact pubertal timing,63-65 with disproportionally higher exposure among certain racial and ethnic minority groups),66-68 dietary patterns (eg, sugar intake via insulin-mediated pathways),69-71 psychosocial stress,72 and adverse childhood experiences.73
Our findings of temporal trends toward longer time to regularity and higher proportion of never establishing regularity in the US have not been previously reported. Because longer time to regularity has been associated with adverse outcomes,39-42 it may serve as an early-life vital sign. These temporal trends may be driven by longer time to maturation of the reproductive axis (eg, impacted by endocrine disruptors)74 or increasing ovulation disorders75 that impact cycle regularity. Although earlier studies suggest that regular menstruation should be established within 1 to 2 years after menarche,37,38,76,77 evidence remains limited on whether delays beyond 2 years warrant clinical or lifestyle intervention.78 In our study, the mean time to regularity among those who spontaneously established regularity was 1.2 to 1.5 years within the 2-year window. However, the proportions taking 3 to 4 years or never establishing regularity were increasing. A French cohort born between 1935 and 1950 showed a decrease from 64% to 53% for time to regularity within 1 year, but data are limited for the US.45 We also found differences by race and ethnicity (eg, Hispanic individuals reported a higher rate of not establishing regularity compared with their non-Hispanic White peers, consistent with a study79 showing that Hispanic individuals having the highest risk of cycle irregularity in adulthood). Our findings suggest the necessity of further studies on the postmenarche years and the need for early intervention during relevant time windows. Continued research on the association of BMI and other factors on reproductive development is needed, and findings should be conveyed to health professionals.
Strengths and Limitations
This study has several unique strengths. First, a large study size of 71 341 participants in a heterogeneous population allowed for sufficient statistical power to detect racial and ethnic differences, even for groups that were previously understudied. Second, we evaluated temporal trends in the percentages of early or late menarche in addition to mean age. Third, our study is the first, to our knowledge, to evaluate and report a temporal trend toward longer time to regularity, suggesting future research directions on this understudied early-life marker of menstrual health. Fourth, our study is the first, to our knowledge, to use digital observational cohort data evaluating BMI at menarche as a potential contributor to the observed temporal trends.
Our study also has limitations. First, the retrospective self-report may induce recall bias and misclassification, likely differential across birth year categories. However, previous validation studies80-82 showed moderate to high correlations between recalled and original age and body size at menarche. Second, BMI at menarche was only available among a subset of participants with demographic distributions different from the full study population. Third, data are limited for additional early-life factors that may contribute to these trends. Fourth, our results may not be generalizable to all US individuals who menstruate or to other populations. Potential selection bias may arise due to self-selection into the study that may be impacted by sociodemographic characteristics.
In this US cohort study of 71 341 individuals born between 1950 and 2005, we observed temporal trends toward earlier menarche and longer time to regularity. These trends appeared across all sociodemographic groups but were stronger among certain racial and ethnic groups (Asian, non-Hispanic Black, or other and multiple races or ethnicities) and low subjective SES groups. Body mass index at menarche mediated a significantly large proportion of the trends toward earlier menarche. Further awareness among health care practitioners and researchers is needed to understand the reasons for these trends and their health implications.