Early Menarche and Internalizing and Externalizing in Adulthood: Explaining the Persistence of Effects

Abstract

Purpose.

Earlier ages at menarche are associated with elevations in internalizing and externalizing that persist into adulthood. The present paper examines whether early pubertal timing precipitates experiences during adolescence that account for long-term elevations in depressive symptoms and antisocial behavior among early maturing girls.

Method.

Using data from National Longitudinal Study of Adolescent and Adult Health (Add Health), the study examines significant post-menarcheal life events that might mediate associations of age at menarche with depressive symptoms and antisocial behavior in adulthood: teenage criminal arrest, teenage pregnancy and childbearing, high school drop-out, and different forms of post-pubertal physical and sexual traumatic assault.

Results.

Results indicate that earlier menarche was associated with greater likelihood of post-menarcheal discontinued education, physical and sexual assault, and teenage pregnancy and childbearing. Discontinued education, physical assault, and sexual assault mediated associations of pubertal timing with adult depressive symptoms; sexual assault mediated associations of pubertal timing with adult antisocial behavior.

Conclusion.

Earlier menarche seems to precipitate post-pubertal stressful events that, in turn, account for higher rates of psychological problems in adulthood. These results suggest that the adolescent experiences of early maturing girls channel them into life paths where stress, adversity, and other risks to psychological well-being are more likely to be a continuing facet of daily life.

Decades of research have established early pubertal timing as one of the most robust predictors of adolescent vulnerability for girls. Girls who mature ahead of peers show a greater prevalence and severity of both internalizing and externalizing disorders during adolescence. Long-term studies on the duration of these effects are limited, though recent research indicates they may persist into adulthood ¹. These findings dovetail with retrospective studies showing earlier ages of menarche among women in inpatient psychiatric treatment ², as well as greater rates of internalizing among women who described their puberty as earlier than peers^3,4.

Typically, the association between earlier puberty and psychological distress is attributed to a maturational disparity⁵. Because physical development outpaces cognitive, social, and emotional development, early maturing girls are left with fewer skills and resources to navigate the biological ^6,7 social^8,9,10 and environmental¹¹ changes of adolescence than later-developing peers. However, although these proximal factors might explain puberty’s association with psychopathology during adolescence, they are unlikely to explain persistence of pubertal timing effects beyond adolescence, when upheavals in social lives, circumstances, and physical development have passed.

One possible explanation is that early puberty precipitates psychological difficulties during adolescence that are linked with a greater propensity for subsequent stressful events. These events, in turn, create and perpetuate long-term risk into adulthood. To illustrate, early maturing girls experience higher rates of depression and externalizing during adolescence; these vulnerabilities may explain why early maturing girls also have higher rates of adolescent criminal arrest¹², teenage childbearing ¹³, academic problems, such as truancy and high school drop-out¹⁴, and lower educational attainment than later developing peers¹⁵. Early maturing girls are also more likely to report traumatic experiences during adolescence, such as sexual harassment ¹⁶, rape ^17,18 and physical assault either by romantic partners ¹⁹ or other individuals ²⁰. Collectively, these experiences may contribute to outcomes in adulthood such as unstable employment, legal difficulties, traumatic revictimization, or relationship instability, making it more likely that stress and adversity will be a continuing aspect of daily life.

Negative experiences during adolescence are often considered more disruptive than those at other life stages, because adolescence is a foundational time when youth make choices and commitments that shape their futures. Within the adolescent externalizing literature, these sorts of detrimental experiences have been referred to as “snares,” because they entrap people within trajectories from which it may be difficult to deviate ^21,22. Although typically discussed with regard to externalizing, snares may be relevant for internalizing as well, as stressful events tend to hold a high degree of multifinality and are associated with vulnerability in multiple psychological domains²³.

It is important to note that while snares may be related to adolescent mental health difficulties, they do not have to be. A second possible pathway is that early puberty may precipitate snares in adolescence independent of adolescent well-being. It is unlikely that some potential snares - for example, criminal arrest-would occur independent of adolescent symptomatology, but is possible that others (i.e., early pregnancy or childbearing) are independent. In this scenario, snares would still have the potential to foster enduring risk: adult internalizing and externalizing have been linked both to episodic, acute stressors (e.g., physical and sexual assault) and to chronic, prolonged stressors (e.g., ongoing financial hardship, earlier childbearing)^24,25.

The Present Study:

Our primary research aim is to understand mechanisms that might explain long-term mental health problems in early maturing girls. We leverage a nationally representative sample to investigate the hypothesis that early puberty precipitates post-pubertal stressful events, including: (1) teenage pregnancy and childbearing; (2) sexual assault; (3) physical assault; (4) teenage criminal arrest; and (5) high school drop-out. We consider direct associations between earlier development and these events; determine if these experiences mediate puberty’s association with adult depressive symptoms and antisocial behavior; and assess whether the potential impact of snares on adult psychopathology is linked to or independent of adolescent mental health.

Method

Data are drawn from the National Longitudinal Study of Adolescent and Adult Health (Add Health) ²⁶, a nationally representative, four-wave study. Of the 20,745 participants interviewed at Wave I, 10,480 were females (M = 15.8 years). Follow-up interviews were completed in 1995–1996 (Wave II; M = 16.1 years), 2001–2002 (Wave III; M = 21.7 years) and 2007–2009 (Wave IV; M = 28.7 years). Current analyses comprise N = 7802 female participants for whom data was available on age at menarche and who also had valid survey weights at Wave IV (to account for the clustered nature of the sample) ²⁷. This project was considered exempt from IRB Review by Cornell University (Protocol #1106002285).

Measures.

Age at menarche.

Participants reported at Waves I and II the month and year of their first menstrual cycle, and age at first menstruation at Wave III. Test-retest reliability across waves was α = .80. First reported age at menarche was used to avoid telescoping bias²⁸. This was most often Wave I report (89.7%).

Psychological Outcomes.

Depressive symptoms.

At Waves I and IV, participants reported past-week depressive symptoms using the self-report Center for Epidemiological Studies - Depression Scale (CES-D)²⁹.

Although the full CES-D was given at Wave I, a ten-item short form was given at Wave IV. The correlation between the full CES-D and the short form was r=.96; to keep measurement consistent, the short version was used to quantify symptoms at Waves I and IV. For each item, participants reported whether they had experienced that symptom, where 0 = never or rarely, 1 = sometimes, 2 = a lot of the time, and 3 = most of the time or all of the time.

Antisocial behavior.

At Waves I and IV, participants reported past-year frequency of antisocial behaviors, with 0 = never, 1 = 1 or 2 times, 2 = 3 or 4 times, and 3 = 5 or more times. Behaviors assessed at both time points included property damage, stealing something worth more than $50, stealing something worth less than $50, breaking into a building, and selling drugs. Wave I additionally included running away from home, lying to parents, driving a car without the owner’s permission, shoplifting, and being loud and rowdy in public, while Wave IV included deliberately writing a bad check, using someone else’s ATM card without permission, and buying or selling stolen property. Frequency of behaviors endorsed at each time point was summed.

Snares.

Snares were defined as post-pubertal events predicted by age at menarche, coded dichotomously as having occurred (1) or not occurred (0). Care was taken to ensure all snares occurred only post-menarcheally, by confirming that the age at which the snare first occurred was later than the participant’s age at menarche; if a participant reported experiencing a snare prior to menarche, she was coded as not experiencing this snare. This coding ensures that pre-menarcheal adversities, if present, cannot account for observed findings. It also avoids confounding continuity in adversity across development with the effects of post-menarcheal events.

Teenage criminal arrest.

At Waves III and IV, participants were asked “have you ever been arrested or taken into custody by the police?” and the age at which this occurred. This was used to create a dichotomous variable representing post-menarcheal criminal arrest at age nineteen or younger.

Discontinued education.

At Wave III, participants reported highest grade of school completed and at Wave IV, they reported highest level of education achieved to date. These items were used to construct a dichotomous variable representing post-menarcheal discontinuation of high school.

Teenage pregnancy and childbirth.

At all waves, participants reported pregnancies to date, when the pregnancy occurred, and how each pregnancy ended. This was used to create two dichotomous variables representing either a pregnancy or a childbirth at age nineteen or younger.

Physical Assault:

Across all waves, participants provided details on multiple forms of physical victimization, which were used to assess three types of post-menarcheal physical assault. First, at Wave IV, participants reported experiences of physical abuse perpetrated by a parent/caregiver, assessed through the question “before your 18th birthday, how often did a parent or adult caregiver hit you with a fist, kick you, or throw you down on the floor, into a wall, or down stairs?” and the age at which this first occurred. Participants who endorsed one or more experiences occurring for the first time after menarche were coded as experiencing post-menarcheal physical assault by a caregiver. Second, participants were asked a series of questions at all waves about past-year violence exposure, including how often they had a knife or gun pulled on them; been cut, shot or stabbed; or were “jumped” by another person. Waves III and IV also included being “beaten up” by another person. Participants who reported any of these events occurring for the first time after menarche were coded as experiencing post-menarcheal physical assault by an unknown perpetrator. Third, at Waves II and III, participants reported on relationships with romantic partners, including whether a partner had pushed, shoved, or thrown something at them. Wave III also included reports of how frequently a romantic partner had slapped, hit or kicked them. Participants who reported the onset of these events after menarche were coded as experiencing post-menarcheal physical assault by an intimate partner.

Sexual Assault:

As with physical assault, experiences of sexual assault were separated according to context. First, at Wave IV, participants were asked “prior to the age of 18, how often did a parent or other adult caregiver touch you in a sexual way, force you to touch him or her in a sexual way, or force you to have sexual relations?” and the age at which this first occurred. Participants who reported one or more experiences occurring for the first time after menarche were coded as experiencing post-menarcheal sexual assault by a caregiver. Second, at Wave IV, participants were asked two questions assessing sexual assault: “have you ever been forced, in a non-physical way, to have any type of sexual activity against your will? For example, through verbal pressure, threats of harm, or by being given alcohol or drugs?” and “have you ever been physically forced to have any type of sexual activity against your will?” and instructed to exclude any experiences with a parent or adult caregiver. For both questions, participants reported the age at which this occurred for the first or only time. Participants who reported either of these experiences occurring for the first time after menarche were coded as experiencing post-menarcheal sexual assault by a non-caregiver.

Covariates.

Analyses included demographic and familial covariates commonly associated with earlier puberty or depression/antisocial behavior: race/ethnicity (European American as reference), father absence (absent from birth, early childhood, or middle childhood; father presence as reference), and socioeconomic indicators indexed by receipt of public assistance, household income-to-needs ratio at Wave I, and maternal education (no high school degree as reference) ^30,31. Age at Wave I was also included.

Missing data.

The analytic sample included all individuals with complete data for age at menarche, but not necessarily complete data on covariates or mediators. To address missing data, data were multiply imputed using the ICE command in Stata 12.0, which is based on a regression switching protocol using chained equations³². Ten imputed datasets were generated and coefficients and standard errors were combined using the MI Estimate command. Girls with complete data on all variables did not differ in age at menarche from those who needed imputation for one or more variables. However, participants with complete data for all variables did have lower reported depression (M = 5.22 versus M = 5.80, t[7799] =−4.90, p<.001) and antisocial behavior (M = .09 versus M = 1.99, t[7787] =−4.66, p<.001) at Wave IV than those missing data for one or more variables. Both dependent variables, and all covariates and independent variables, were included in the multiple imputation model to ensure that data could be considered Missing at Random (MAR) for imputation purposes³³.

Analytic Strategy.

A series of regression models was used to assess associations of age at menarche with post-pubertal snares and levels of depression and antisocial behavior in adulthood. All models were weighted using Wave IV survey and design weights.

Following principles of mediation ³⁴, we established (1) direct associations between age at menarche and Wave IV depressive symptoms/antisocial behavior; (2) direct associations between age at menarche and each potential snare; and (3) the degree to which the association between menarche and depression/antisocial behavior could be explained by the inclusion of these snares in the model. Age at menarche was entered both as a linear and as a quadratic term, to consider the potential for nonlinear effects. Our first set of models, the long-term effects models, separately regressed Wave IV depressive symptoms and antisocial behavior on age at menarche and covariates, to establish the degree to which timing of puberty prospectively predicted psychological outcomes in adulthood. These models restate findings from Mendle et al. (2018) and are included only to provide parameter estimates necessary to calculate the magnitude of subsequent mediation effects; results are in the article supplement. Our second set of models, the snares models, assessed direct associations of age at menarche with each post-menarcheal life event. Finally, in our third set of models, the mediation models, depressive symptoms and antisocial behavior at Wave IV were regressed on age at menarche, covariates, and life events shown to be significantly associated with menarche, using separate models for each outcome.

Mediation models were divided into tests of two hypothetical pathways. To test the adolescence-independent pathway, we included age at menarche, snares, and covariates in the model but not indices of adolescent mental health. To test the adolescence-dependent pathway, we incorporated adolescent levels of depressive symptoms or antisocial behavior. Snares that significantly mediate the association between age at menarche and adult outcomes in the first model, but not in the second, suggest that age at menarche is linked with adult outcomes primarily because these snares are connected to mental health challenges experienced during adolescence. Comparing these models can establish whether continued adolescent mental health symptoms or post-pubertal life experiences best explain duration of psychological vulnerability; although this paper contains many analyses, these models are the ones most relevant to primary study hypotheses (Tables 3 and 4). A variant of the product of coefficients method suitable for combining coefficients from ordinary least squares (OLS) and logistic regression models was used to calculate the magnitude and significance of the indirect effects of each significant life event ^34,35.

Table 3.

Adolescence-Independent Pathway Model. Significant life events, age at menarche, and depressive symptoms and antisocial behavior in adulthood

	Depressive Symptoms				Antisocial Behavior
	B	SE	95% Confidence Int.		b	SE	95% Confidence Int.
			Lower	Upper			Lower	Upper
Intercept	9.24	2.35	4.60	13.89	.52	.15	.22	.82
Menarche	−.48	.38	−1.22	.26	−.01	.01	−.03	.007
Menarche, quadratic	.02	.02	−.01	.05	-	-	-	-
Age at Wave I	−.001	.04	−.07	.07	−.02	.01	−.03	−.01
Snares
H.S. Graduation	−1.19	.30	−1.78	−.60	−.01	.06	−.12	.10
Teenage pregnancy	.09	.17	−.25	.43	.11	.06	−.004	.23
Teenage childbearing	−.12	.25	−.62	.39	−.08	.08	−.25	.08
Sexual assault - caregiver	.82	.53	−.24	1.88	.27	.18	−.09	.62
Sexual assault - nonfamily	1.50	.19	1.13	1.88	.22	.04	.14	.31
Physical assault - caregiver	1.23	.26	.71	1.75	.04	.06	−.07	.15
Physical assault - nonfamily	.78	.17	.45	1.11	.07	.03	.003	.13
Maternal Education Level
Maternal H.S. Graduation	−.49	.24	−.96	−.03	.04	.04	−.05	.13
Maternal College Grad.	−.94	.25	−1.44	−.44	.03	.04	−.06	.10
Father Presence
Father Absent Since Birth	.06	.24	−.42	.54	.04	.05	−.06	.15
Father Left 0-5 years	.43	.27	−.10	.95	.08	.06	−.05	.20
Father Left 6-13 years	.22	.26	−.29	.73	.003	.05	−.09	.09
Race/Ethnicity
African American	.30	.25	−.18	.79	.04	.05	−.05	.14
Hispanic	.54	.23	.10	.99	.01	.04	−.07	.09
Other Race	.60	.27	.07	1.14	−.04	.05	−.13	.05
Family Income to Needs	−.03	.02	−.07	.004	.002	.003	−.005	.01
AFDC Receipt	.71	.29	.13	1.30	.03	.09	−.14	.21

Note: Estimates significant at p<.05 are in bold. Analyses conducted in STATA’s design-based weighting program, to account for the clustered nature of the sample; this program does not provide R-squared values

Table 4.

Adolescence-Dependent Pathway Model. Significant life events, age at menarche, and depressive symptoms and antisocial behavior in adulthood

	Depressive Symptoms				Antisocial Behavior
	B	SE	95% Confidence Int.		b	SE	95% Confidence Int.
			Lower	Upper			Lower	Upper
Intercept	7.53	2.35	2.88	12.18	.44	.15	.14	.74
Menarche	−.35	.37	−1.08	.38	−.01	.01	−.03	.007
Menarche, quadratic	.02	.02	−.01	.04	-	-	-	-
Wave 1 Symptoms	.24	.02	.21	.27	.03	.01	.02	.04
Age at Wave I	−.06	.04	−.13	.007	−.02	.01	−.03	−.01
Snares
H.S. Graduation	−.90	.27	−1.44	−.37	.03	.05	−.08	.13
Teenage pregnancy	−.07	.17	−.40	.28	.09	.06	−.03	.20
Teenage childbearing	−.14	.25	−.64	.36	−.08	.08	−.24	.09
Sexual assault - caregiver	.81	.53	−.23	1.86	.28	.17	−.07	.62
Sexual assault - nonfamily	1.31	.18	.95	1.67	.20	.04	.11	.28
Physical assault - caregiver	1.07	.24	.59	1.55	.02	.05	−.09	.12
Physical assault - nonfamily	.59	.16	.27	.91	.03	.03	−.03	.08
Maternal Education Level
Maternal H.S. Graduation	−.33	.23	−.79	.13	.04	.04	−.05	.12
Maternal College Grad.	−.69	.24	−1.17	−.20	.01	.04	−.07	.09
Father Presence
Father Absent Since Birth	.02	.23	−.45	.48	.04	.05	−.06	.15
Father Left 0-5 years	.36	.26	−.15	.87	.07	.06	−.06	.19
Father Left 6-13 years	.16	.25	−.34	.65	−.003	.05	−.10	.10
Race/Ethnicity
African American	.19	.21	−.24	.62	.02	.05	−.07	.13
Hispanic	.45	.21	.03	.88	.02	.04	−.06	.09
Other Race	.47	.24	.003	.94	−.06	.05	−.15	.03
Family Income to Needs	−.02	.02	−.06	.01	.002	.003	−.01	.01
AFDC Receipt	.64	.29	.05	1.22	.05	.09	−.13	.22

Note: Estimates significant at p<.05 are in bold. Analyses conducted in STATA’s design-based weighting program, to account for the clustered nature of the sample; this program does not provide R-squared values. Wave 1 symptoms refers to levels of depressive symptoms at Wave 1 in model predicting depressive symptoms at Wave IV and levels of antisocial behaviors at Wave 1 in model predicting antisocial behavior at Wave IV

Results

Long-Term Effects Models.

Age at menarche significantly and linearly predicted levels of depressive symptoms at Wave IV, with earlier ages of menarche associated with greater symptoms. There was also a quadratic association, indicating the magnitude of this association was stronger at earlier ages of menarche than later ages. Age at menarche was also linearly related to antisocial behavior, with early maturing girls reporting higher frequency of antisocial behavior as adults (Table S1, in the article supplement).

Snares Models.

Logistic regression models indicated that age at menarche linearly predicted several snares (Table 2). Earlier ages at menarche significantly increased the likelihood of teenage pregnancy, teenage childbearing, onset of sexual and physical assault by a caregiver, sexual assault by a non-caregiver, and physical assault by an unknown perpetrator, after controlling for covariates. Age at menarche linearly and quadratically predicted high school graduation, suggesting that earlier ages at menarche were associated with reduced likelihood of graduation and that the risk for not graduating was most pronounced at very young ages of menarche. The effects of menarcheal age on the odds of each snare were moderate; for example, a one-year increase in age at menarche decreased the odds of teenage pregnancy by 12% and the odds of sexual assault by a caregiver by 32%. There were no significant associations between menarcheal age and teenage criminal arrest or intimate partner violence.

Table 2.

Snares Models. Associations of different life events with age at menarche

	High School Graduation				Teenage Pregnancy				Teenage Childbearing
	b	SE	95% Confidence Int.		b	SE	95% Confidence Int.		b	SE	95% Confidence Int.
			Lower	Upper			Lower	Upper			Lower	Upper
Intercept	−5.27	2.01	−9.26	−1.28	.72	.48	−.23	1.67	.72	.59	−.45	1.90
Menarche	.83	.30	.23	1.43	−.12	.03	−.17	−.07	−.13	.03	−.19	−.07
Menarche, quadratic	−.03	.01	−.06	−.007	-	-	-	-	-	-	-	-
Age at Wave I	.06	.03	−.01	.13	−.01	.02	−.05	.04	−.03	.03	−.10	−.02
Maternal Education Level
Maternal H.S. Graduation	.75	.16	.43	1.07	−.16	.13	−.42	.10	−.30	.16	−.63	.04
Maternal College Grad.	1.41	.20	.99	1.82	−.53	.13	−.80	−.26	−.74	.18	−1.10	−.38
Father Presence
Father Absent Since Birth	−.78	.14	−1.05	−.51	.49	.12	.25	.73	.45	.14	.34	.73
Father Left 0-5 years	−.59	.14	−.88	−.30	.52	.12	.28	.75	.48	.14	.24	.78
Father Left 6-13 years	−.23	.17	−.58	.11	.36	.11	.13	.59	.20	.14	−.07	.48
Race/Ethnicity
African American	−.10	.21	−.52	.31	.27	.15	−.03	.57	.21	.18	−.15	.56
Hispanic	.25	.16	−.06	.56	.46	.12	.23	.69	.39	.13	.13	.64
Other Race	−.07	.19	−.44	.31	.05	.20	−.35	.43	−.14	.23	−.59	.31
Family Income to Needs	.07	.04	.001	.14	−.07	.02	−.11	−.02	−.09	.03	−.15	−.02
AFDC Receipt	−.72	.17	−1.06	−.38	.58	.13	.32	.85	.59	.16	.27	.90

	Sexual Assault - Caregiver				Sexual Assault - Nonfamily Perpetrator
	b	SE	95% Confidence Int.		b	SE	95% Confidence Int.
			Lower	Upper			Lower	Upper
Intercept	−.38	1.29	−2.94	2.19	.01	.47	−.91	.94
Menarche	−.38	.07	−.51	−.24	−.09	.03	−.14	−.04
Age at Wave I	.03	.06	−.10	.16	−.04	.02	−.09	−.002
Maternal Education Level
Maternal H.S. Graduation	−.06	.46	−1.01	.89	.38	.14	.10	.66
Maternal College Grad.	−.14	.44	−1.04	.76	.55	.15	.25	.85
Father Presence
Father Absent Since Birth	.94	.42	.10	1.79	.42	.12	.17	.66
Father Left 0-5 years	1.28	.34	.60	1.96	.49	.11	.27	.70
Father Left 6-13 years	1.17	.35	.47	1.87	.34	.15	.04	.64
Race/Ethnicity
African American	−.29	.39	−1.07	.48	−.40	.15	−.70	−.10
Hispanic	−.36	.32	−.98	.27	−.55	.13	−.81	−.28
Other Race	.08	.46	−.82	.99	−.01	.17	−.34	.31
Family Income to Needs	−.05	.07	−.18	.09	−.02	.02	−.05	.01
AFDC Receipt	.26	.42	−.56	1.09	.19	.18	−.17	.54

	Physical Assault – Caregiver				Physical Assault – Nonfamily Perpetrator
	b	SE	95% Confidence Int.		b	SE	95% Confidence Int.
			Lower	Upper			Lower	Upper
Intercept	.73	.66	−.58	2.05	.18	.45	−.71	1.07
Menarche	−.17	.04	−.24	−.10	−.11	.03	−.16	−.06
Age at Wave I	−.05	.03	−.12	.01	.01	.02	−.03	.06
Maternal Education Level
Maternal H.S. Graduation	−.18	.15	−.48	.12	−.30	.11	−.52	−.08
Maternal College Grad.	−.34	.15	−.64	−.04	−.31	.12	−.55	−.07
Father Presence
Father Absent Since Birth	−.65	.15	.36	.95	.34	.11	.12	.56
Father Left 0-5 years	.45	.17	.11	.79	.27	.12	.03	.51
Father Left 6-13 years	.22	.21	−.21	.66	.28	.13	.02	.55
Race/Ethnicity
African American	−.01	.19	−.38	.36	.32	.09	.13	.50
Hispanic	−.21	.14	−.49	.07	.56	.11	.35	.77
Other Race	.21	.22	−.23	.64	.21	.12	−.03	.45
Family Income to Needs	−.04	.03	−.10	.02	−.01	.02	−.04	.02
AFDC Receipt	.33	.21	−.11	.76	.27	.13	.02	.52

Note: Estimates significant at p<.05 are in bold. Analyses conducted in STATA’s design-based weighting program, to account for the clustered nature of the sample; this program does not provide R-squared values.

Models were then rerun with adolescent depressive symptoms and antisocial behavior held constant, to determine if the association between pubertal timing and snares was related to adolescent mental health (Tables S2A and S2B in the supplement). Age at menarche continued to predict all snares as in the models above, with the exception of sexual assault by a non-family member when including Wave 1 depressive symptoms.

Mediation Models:

Adolescence-independent pathway.

Age at menarche did not significantly predict Wave IV depressive symptoms or antisocial behavior once the snares predicted by menarcheal age were included in models. Including snares in the model reduced the linear association between age at menarche and depressive symptoms and antisocial behavior by 44% and 50%, respectively, indicating that snares partially mediate the association of menarche and adult mental health. Specifically, the association between menarche and depressive symptoms was mediated by high school drop-out (indirect effect=.99 (.43), z=2.24 p<.05) and different types of traumatic assault, including physical assault by a caregiver (indirect effect=.21 (.07), z=3.12 p<.05), physical assault by an unidentified perpetrator (indirect effect=.09 (.03), z=4.48 p<.05), and sexual assault by a non-caregiver (indirect effect=.14 (.05), z=2.79; p<.05). Only sexual assault by non-caregivers mediated associations of age at menarche with frequency of antisocial behavior at Wave IV (indirect effect=.020(.008), z=2.56; p<.05).

Adolescence-dependent pathway.

Results did not change when adolescent symptoms were held constant, although the magnitude of indirect effects of menarche on adult depression was reduced. Figures 1 displays the indices of mediation for each snare that significantly mediated the menarche-depressive symptoms association, with and without adolescent depression held constant.

Figure 1.

Indices of mediation for all snares that significantly mediate age at menarche—adult depressive symptoms association, without (adolescence-independent pathway model) and with (adolescence-dependent pathway model) Wave 1 depressive symptoms controlled. Indices of mediation are fully standardized effect sizes for indirect effects.

Sensitivity Analyses:

Several sets of sensitivity analyses were conducted, including models run with log-transformations of depressive symptoms and antisocial behaviors (in order to account for base rates and non-normal distributions). Results were consistent across transformed and non-transformed models. We additionally re-ran mediation models on one imputed data set to generate bootstrapped standard error estimates for the indirect effects; results did not change whether using the full imputed sample without bootstrapping or one imputation with bootstrapped standard errors.

Discussion

In the present study, we examined mechanisms that explain why girls who mature earlier than peers show enduring vulnerabilities to internalizing and externalizing disorders^1–3. Results indicate that these associations may be partially explained by new life experiences that follow maturation: earlier menarche precipitates negative events post-pubertally that, in turn, account for higher rates of psychological problems in adulthood. Specifically, post-pubertal discontinued education and various forms of post-pubertal traumatic assault mediated associations of pubertal timing with depressive symptoms roughly fifteen years post-menarche. Post-pubertal sexual victimization also mediated associations of pubertal timing with frequency of adult antisocial behavior. These events influence adult mental health independent of adolescent symptomatology.

Collectively, these results extend beyond the documentation of associations to clarify why early maturing girls remain at elevated mental health risk into adulthood. Perhaps most importantly, they emphasize the worrisome circumstances early maturing girls encounter, both within and outside the home environment. Although pubertal timing has been linked with greater prevalence of subsequent adolescent victimization in previous studies^17–19, this literature is relatively sparse and much of it is not recent. To our knowledge, our results represent the first documentation of the onset of physical and sexual assault by caregivers post-pubertally for early maturing girls. Discontinued education was also a significant mediator of the association between menarche and adult depressive symptoms, perhaps because it places girls on a trajectory where low-paying, low-status jobs may contribute to high degrees of chronic strain, perceptions of helplessness, and ongoing decrements in mood.

Importantly, while many of our hypothesized snares were linked with early menarche, not all explained why age at menarche was associated with internalizing and externalizing in adulthood. Specifically, teenage pregnancy and childbearing were more prevalent among early maturers, but neither significantly mediated associations with psychopathology. Although pregnancy and childbearing present early maturing adolescents with challenging circumstances, understanding that these experiences - in and of themselves - do not account for persistent socioemotional vulnerability offers pragmatic insights into the health, well-being, and life course for adolescent mothers and their children.

Overall, our findings contribute to an emerging body of research suggesting sequelae of early pubertal timing linger beyond adolescence, and they further corroborate a broad existing literature on the importance of stressful life events in mental health^24,25. Yet the study is not without limitations. These include reliance on self-report measures, a single indicator of pubertal timing, relatively low base rates of some snares, and an exclusively female sample. Future research incorporating pubertal benchmarks other than menarche will be particularly important, given that menarche occurs relatively late in the pubertal process³⁶ and may not adequately index psychological or social changes that occur earlier in development. This might explain why our findings are inconsistent with another recent study of the long-term effects of early pubertal timing³⁷, which operationalized pubertal timing using indicators of maturation that occur early in development and did not find long-term socioemotional effects. In addition, our study targeted a specific research question regarding post-pubertal life events, but did not investigate the range of pre-menarcheal adversities^38,39 that might place some girls more at risk for early puberty compared to others.

Conclusion

The present study suggests long-term elevations in internalizing and externalizing symptoms among early maturing girls may be explained, at least in part, by stressful experiences that occur during their post-pubertal adolescent years.

Table 1.

Sample Characteristics

	M / %	SD
Menarche	12.16	7.42
Depression, Wave I	6.57	4.60
Depression, Wave IV	5.67	4.34
Antisocial behavior, Wave I	2.94	3.60
Antisocial behavior, Wave IV	.17	.89
Snares
Teenage criminal arrest	7.80
No high school diploma	15.26
Teenage pregnancy	29.41
Teenage childbearing	16.34
Sexual assault - caregiver	1.76
Sexual assault - nonfamily member	20.35
Physical assault - caregiver	9.84
Physical assault - nonfamily member	28.41
Physical assault - intimate partner	16.06
Covariates
Maternal Education
<HS	17.97
Mother has HS/GED	41.65
Mother has college degree	40.38
Father Absence
Always absent	13.91
Absent 0-5 years	12.36
Absent 6-13 years	11.21
Always present 0-13 years	62.52
Race/Ethnicity
European American	65.84
African American	12.06
Hispanic	16.21
Other race/ethnicity	5.89
Income-to-needs ratio	3.11	3.61
AFDC receipt	8.41

Note. All percents are drawn from final imputed data set and means are drawn from one randomly selected imputed data set. All means and percents are weighted using Add Health Wave IV survey weights. Numbers which indicate means include standard deviations as well; numbers which indicate percentages do not. HS = high school. GED = graduate equivalency diploma. AFDC = Aid to Families with Dependent Children and indicates the participant’s family received public assistance benefits in Wave I. “Depression” refers to summed CES-D short form scores. Scores of 8-10 have been suggested as a clinical cutoff. “Antisocial behavior” refers to the summed frequency of antisocial behavior.

Implications and Contributions:

Recent research suggests that earlier puberty in girls is associated with prolonged psychological problems not only in adolescence, but also into adulthood. This study suggests these associations may be partially explained by post-pubertal stressful life events, particularly traumatic assault, that perpetuate long-term vulnerability.