Hostname: page-component-5cf477f64f-h6p2m Total loading time: 0 Render date: 2025-03-31T09:30:41.405Z Has data issue: false hasContentIssue false
  • English
  • Français

The causal role of male pubertal timing for the development of externalizing and internalizing traits: results from Mendelian randomization studies

Published online by Cambridge University Press:  28 March 2025

Lars Dinkelbach ,
Triinu Peters ,
Corinna Grasemann ,
Anke Hinney  and
Raphael Hirtz
Lars Dinkelbach*
Affiliation:
Department of Pediatrics III, University Hospital Essen, University of Duisburg-Essen, Essen, Germany Institute of Sex- and Gender-sensitive Medicine, University Hospital Essen, University of Duisburg-Essen, Essen, Germany
Triinu Peters
Affiliation:
Institute of Sex- and Gender-sensitive Medicine, University Hospital Essen, University of Duisburg-Essen, Essen, Germany Section of Molecular Genetics in Mental Disorders, University Hospital Essen, University of Duisburg-Essen, Essen, Germany Center for Translational Neuro- and Behavioral Sciences, University Hospital Essen, University of Duisburg-Essen, Essen, Germany
Corinna Grasemann
Affiliation:
Department of Pediatrics, Division of Rare Diseases and CeSER, St. Josef-Hospital, Ruhr-University Bochum, Bochum, Germany
Anke Hinney
Affiliation:
Institute of Sex- and Gender-sensitive Medicine, University Hospital Essen, University of Duisburg-Essen, Essen, Germany Section of Molecular Genetics in Mental Disorders, University Hospital Essen, University of Duisburg-Essen, Essen, Germany Center for Translational Neuro- and Behavioral Sciences, University Hospital Essen, University of Duisburg-Essen, Essen, Germany
Raphael Hirtz
Affiliation:
Center for Child and Adolescent Medicine, Helios University Hospital Wuppertal, Witten/Herdecke University, Wuppertal, Germany
*
Corresponding author: Lars Dinkelbach; Email: [email protected]
Rights & Permissions [Opens in a new window]

Abstract

Background

Preexisting epidemiological studies suggest that early pubertal development in males is associated with externalizing (e.g. conduct problems, risky behavior, and aggression) and internalizing (e.g. depression and anxiety) traits and disorders. However, due to problems inherent to observational studies, especially of residual confounding, it remains unclear whether these associations are causal. Mendelian randomization (MR) studies take advantage of the random allocation of genes at conception and can establish causal relationships.

Methods

In this study, N = 76 independent genetic variants for male puberty timing (MPT) were derived from a large genome-wide association study (GWAS) on 205,354 participants and used as an instrumental variable in MR studies on 17 externalizing and internalizing traits and psychopathologies utilizing outcome GWAS with 16,400–1,045,957 participants.

Results

In these MR studies, earlier MPT was significantly associated with higher scores for the overarching phenotype of ‘Externalizing Traits’ (b = −0.03, 95% CI [−0.06, −0.01]). However, this effect was likely driven by an earlier age at first sexual contact (b = −0.17, 95% CI [−0.21, −0.13]), without evidence for an effect on further externalizing phenotypes. Regarding internalizing phenotypes, earlier MPT was associated with higher levels of the ‘Depressed Affect’ subdomain of neuroticism (b = −0.04, 95% CI [−0.07, −0.01]). Late MPT was related to higher scores of internalizing traits in early life (b = 0.04, 95% CI [0.01, 0.08]).

Conclusions

This comprehensive MR study supports a causal effect of MPT on specific traits and behaviors. However, no evidence for an effect of MPT on long-term clinical outcomes (depression, anxiety disorders, alcohol dependency, cannabis abuse) was found.

Keywords

adolescencedepressionmale puberty timingmental health
Type
Original Article
Information
Psychological Medicine , Volume 55 , 2025 , e101
Check for updates
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
© The Author(s), 2025. Published by Cambridge University Press

Background

Early pubertal timing has been suggested as a risk factor for externalizing (e.g. oppositional defiant disorder, antisocial behavior, risky sexual behavior, and substance abuse) and internalizing traits and disorders (e.g. depression or anxiety) in several cross-sectional epidemiological and clinical studies with small but consistent effects in both males and females (Ullsperger & Nikolas, Reference Ullsperger and Nikolas2017). However, these associations are subject to numerous confounders. In boys exposed to a negative social microenvironment (e.g. parental rejection, peer victimization), early puberty is associated with increased rule-breaking behavior (Vijayakumar, Youssef, et al., Reference Vijayakumar, Youssef, Bereznicki, Dehestani, Silk and Whittle2023) and a higher burden of depressive symptoms (Benoit, Lacourse, & Claes, Reference Benoit, Lacourse and Claes2013). Additionally, adverse family environments may contribute to an earlier puberty onset (Pham, DiLalla, Corley, Dorn, & Berenbaum, Reference Pham, DiLalla, Corley, Dorn and Berenbaum2022), introducing a potential confounding pathway. Thus, the true causal effect of earlier puberty timing on negative mental health outcomes may have been overestimated. Additional confounders of the relationship between pubertal maturation and mental health include the macroenvironment (i.e. the neighborhood income) (Niu, Sheffield, & Li, Reference Niu, Sheffield and Li2023), ethnicity (Hamlat, Stange, Abramson, & Alloy, Reference Hamlat, Stange, Abramson and Alloy2014), and family interaction patterns (Li, Zhao, Zhang, & Zhang, Reference Li, Zhao, Zhang and Zhang2023; Ye & Rudolph, Reference Ye and Rudolph2023) – many of which are difficult to control in conventional observational studies. Therefore, the causal relationship between early pubertal timing and negative mental health outcomes remains unclear.

Mendelian randomization (MR) studies can overcome limitations inherent to observational studies and are able to distinguish between residual confounding and causal relationships (Ebrahim & Davey Smith, Reference Ebrahim and Davey Smith2008; Lawlor et al., Reference Lawlor, Harbord, Sterne, Timpson and Davey Smith2008). In brief, MR studies take advantage of the random allocation of genetic variants at conception. If a genetic variant is associated with an exposure (e.g. puberty timing), this genetic variant can serve as an unconfounded predictor to analyze the effect of the exposure on the outcome of interest (e.g. psychopathologies) under certain assumptions (Ebrahim & Davey Smith, Reference Ebrahim and Davey Smith2008; Lawlor et al., Reference Lawlor, Harbord, Sterne, Timpson and Davey Smith2008). Since most confounders in the relationship between puberty timing and mental health arise from environmental and social factors, which are unlikely to be linked to genetic variant allocation, the MR approach may be particularly advantageous for investigating this connection.

Several MR studies analyzed the effects of pubertal timing on psychopathologies in females and discovered an influence of early age at menarche on depression (Hirtz et al., Reference Hirtz, Hars, Naaresh, Laabs, Antel, Grasemann and Peters2022) and substance abuse (Pan et al., Reference Pan, Li, Cheng, Chen, Zhang, Zhang and Zhang2023). In males, one MR study analyzed the effect of pubertal timing on substance addiction and found no effect on tobacco, alcohol, or cannabis addiction (Pan et al., Reference Pan, Li, Cheng, Chen, Zhang, Zhang and Zhang2023). However, conclusions from this MR analysis are limited since both the exposure and outcome associations were drawn from a single sample, the UK Biobank, violating the non-sample overlap assumption of two-sample MR studies (Burgess, Davies, & Thompson, Reference Burgess, Davies and Thompson2016). In addition, this study relied on a Genome-Wide Association Study (GWAS) assessing pubertal timing by single measures of physical maturation (such as late vs. early voice break [VB] and late vs. early facial hair growth) rather than a combined measure of male pubertal timing (Pan et al., Reference Pan, Li, Cheng, Chen, Zhang, Zhang and Zhang2023).

In males, a recent meta-analysis of GWAS based on 205,354 European males combined different indicators of pubertal timing in boys (e.g. the timing of VB and facial hair growth) to study the genetic associations of male puberty timing (MPT) in general (Hollis et al., Reference Hollis, Day, Busch, Thompson, Soares, Timmers and Perry2020). Thus, this large-scale GWAS is less dependent on a specific assessment of MPT and provides a more reliable tool to investigate the causal effect of male pubertal timing on psychopathologies. In a previous MR study, this GWAS was utilized to analyze the effect of MPT on depression but did not find robust evidence for a relationship between both (Hirtz et al., Reference Hirtz, Grasemann, Hölling, von Holt, Albers, Hinney and Peters2023). However, further psychopathologies in males have not been addressed by MR studies so far, despite reasonable evidence for extensive associations of MPT with externalizing and internalizing traits and disorders in epidemiological studies (Ullsperger & Nikolas, Reference Ullsperger and Nikolas2017).

Objective

Here, we aim to utilize multiple MR studies based on a current large GWAS to analyze whether male pubertal timing has a causal effect on the development of (1) externalizing traits, extraversion, antisocial behavior, and aggression, (2) risky behavior, including general risk tolerance, smoking, and risky sexual behavior (number of sexual partners, age at first sexual contact), (3) substance abuse, i.e. alcohol and cannabis abuse, (4) internalizing traits including neuroticism, and (5) internalizing psychiatric entities such as depression and anxiety disorders.

Methods

MR assumptions

To address the above-mentioned research questions, 17 two-sample MR studies were conducted. In MR studies, genetic variants or single nucleotide polymorphisms (SNPs) are used as instrumental variables to analyze the effect of the exposure variable (here, MPT) on the outcome variable (here, externalizing and internalizing traits and psychopathologies). SNPs are valid instrumental variables if the three main assumptions of MR studies are met: (a) the relevance assumption, meaning that the genetic variants have to be strongly related to the exposure, (b) the independence assumption, meaning that genetic variants are not associated with confounders of the exposure–outcome association, and (c) the exclusion restriction assumption, meaning that genetic variants are only related to the outcome via pathways mediated by the exposure (Haycock et al., Reference Haycock, Burgess, Wade, Bowden, Relton and Davey Smith2016; Lawlor et al., Reference Lawlor, Harbord, Sterne, Timpson and Davey Smith2008). In two-sample MR studies, the genotype-exposure associations and genotype-outcome associations are derived from two distinct GWAS analyses. Consequently, the two-sample MR approach relies on two additional assumptions: (e) nonoverlapping exposure and outcome GWAS, as sample overlap can result in a bias of unknown direction and extent (Burgess et al., Reference Burgess, Davies and Thompson2016) and (f) the assumption that both the exposure and outcome GWAS are derived from the same population (Woolf et al., Reference Woolf, Mason, Zagkos, Sallis, Munafò and Gill2024).

Study design and sensitivity analyses

Of the first three MR assumptions, only the relevance criterion (a) can be tested directly (Haycock et al., Reference Haycock, Burgess, Wade, Bowden, Relton and Davey Smith2016). This assumption was ensured by calculating F-statistics for each genetic variant. SNPs with F-statistics < 10 are considered weak instruments (Lawlor et al., Reference Lawlor, Harbord, Sterne, Timpson and Davey Smith2008). For MPT, only genome-wide significant (p < 5 × 10−8) SNPs were utilized as instrumental variables (Hollis et al., Reference Hollis, Day, Busch, Thompson, Soares, Timmers and Perry2020), all surpassing the threshold for strong instruments (median F-statistic was 43.38, range 29.92–231.07). Violations of the second and third (b and c) assumptions arise when causal pathways between genetic variants and the outcome do not involve the exposure phenotype (i.e. the genetic variants are causally linked to the outcome either directly or via another phenotype), typically stemming from pleiotropy. Pleiotropy denotes a scenario where a genetic variant is linked to two or more seemingly unrelated phenotypes. To detect or correct violations of the second and third (b and c) assumptions, the following steps i–iii. were taken:

  1. i. Identification and exclusion of horizontally pleiotropic outliers by the MR-PRESSO framework (Verbanck, Chen, Neale, & Do, Reference Verbanck, Chen, Neale and Do2018). The primary endpoint for all MR analyses was the effect estimate (calculated by the inverse variance weighted [IVW] method) after the exclusion of horizontally pleiotropic outliers.

  2. ii. Calculation of a set of so-called robust methods that are less susceptible to or correct for different kinds of pleiotropy (Burgess, Bowden, Fall, Ingelsson, & Thompson, Reference Burgess, Bowden, Fall, Ingelsson and Thompson2017). As sensitivity analyses, the following set of robust methods were calculated after the exclusion of outliers: Weighted median- and mode-based estimates (Bowden, Davey Smith, Haycock, & Burgess, Reference Bowden, Davey Smith, Haycock and Burgess2016; Hartwig, Davey Smith, & Bowden, Reference Hartwig, Davey Smith and Bowden2017), MR Egger (Bowden, Davey Smith, & Burgess, Reference Bowden, Davey Smith and Burgess2015) and MR-RAPS (Zhao, Wang, Hemani, Bowden, & Small, Reference Zhao, Wang, Hemani, Bowden and Small2020). Details on these methods are given in Supplemental Methods 1.

  3. iii. Correction for pleiotropy was performed using multivariable MR (MVMR) analyses (Burgess & Thompson, Reference Burgess and Thompson2015). Body mass index (BMI) was selected as the primary adjustment factor due to its strong genetic overlap with pubertal timing and its association with internalizing and externalizing problems, indicating a potential pleiotropic pathway (Day et al., Reference Day, Bulik-Sullivan, Hinds, Finucane, Murabito, Tung and Perry2015; Drosopoulou et al., Reference Drosopoulou, Sergentanis, Mastorakos, Vlachopapadopoulou, Michalacos, Tzavara and Tsitsika2021; Hollis et al., Reference Hollis, Day, Busch, Thompson, Soares, Timmers and Perry2020; Hoyt, Niu, Pachucki, & Chaku, Reference Hoyt, Niu, Pachucki and Chaku2020). To assess whether body weight-related pleiotropy affected our results, effect estimates of MPT on psychological outcomes were calculated with adjustment for BMI (IVW BMIcorr) via MVMR. For MVMR, a GWAS on BMI in up to 374,756 males was used (Pulit et al., Reference Pulit, Stoneman, Morris, Wood, Glastonbury, Tyrrell and Ji2019).

Despite efforts to identify outcome GWAS without sample overlap, this study included several GWAS for relevant outcomes that used data from the UK Biobank, which also contributed to the GWAS on MPT. Thus, to assess the impact of a violation of the ‘nonoverlap assumption’ (e), beta weights and standard errors for VB were derived from 23andMe participants (Hollis et al., Reference Hollis, Day, Busch, Thompson, Soares, Timmers and Perry2020) and used as an alternative instrumental variable to calculate an effect estimate without sample overlap (IVW VB).

To correct for multiple comparisons, the p-values of primary endpoints were corrected by the false discovery rate (FDR) (Benjamini & Hochberg, Reference Benjamini and Hochberg1995). Scatter plots (visualizing the effect of each SNP on MPT and the outcome of interest), funnel plots (drawing the MR estimate of each SNP against the inverse of its standard error as a measure of its precision), and forest plots (illustrating the MR estimate for each SNP individually) were drawn to visually screen for violations of MR assumptions or distortion of effects by single genetic variants (Burgess et al., Reference Burgess, Bowden, Fall, Ingelsson and Thompson2017). Leave-one-out analyses were conducted to assess whether the results were driven by single variants. Details on the software packages used can be found in the Supplemental Methods.

The MR analyses of this study were performed and reported in accordance with the STROBE-MR statement (Skrivankova et al., Reference Skrivankova, Richmond, Woolf, Yarmolinsky, Davies, Swanson and Dimou2021). Figure 1 gives an overview of the study design.

Figure 1. Illustration of the analytic approach and outcome variables assessed. Further details on the outcome phenotypes are given in Tables 1 and 2. BMI=body mass index; IVW=inverse-variance weighted; MPT=male puberty timing; SNP=single-nucleotide polymorphism; VB=voice break.

Study population

Tables 1 and 2 give an overview of the GWAS and sample sizes utilized in this study (Harder et al., Reference Harder, Nguyen, Pasman, Mosing, Hägg and Lu2022; Ip et al., Reference Ip, Van der Laan, Krapohl, Brikell, Sánchez-Mora, Nolte and Zafarmand2021; Jami et al., Reference Jami, Hammerschlag, Ip, Allegrini, Benyamin, Border and Lu2022; Johnson et al., Reference Johnson, Demontis, Thorgeirsson, Walters, Polimanti, Hatoum and Clarke2020; Karlsson Linnér et al., Reference Karlsson Linnér, Biroli, Kong, Meddens, Wedow, Fontana and Hammerschlag2019; Karlsson Linnér et al., Reference Karlsson Linnér, Mallard, Barr, Sanchez-Roige, Madole, Driver and Dick2021; Mills et al., Reference Mills, Tropf, Brazel, van Zuydam and Vaez2021; Nagel, Jansen, et al., Reference Nagel, Jansen, Stringer, Watanabe, De Leeuw, Bryois and Muñoz-Manchado2018; Otowa et al., Reference Otowa, Hek, Lee, Byrne, Mirza, Nivard and Wolen2016; Tielbeek et al., Reference Tielbeek, Johansson, Polderman, Rautiainen, Jansen, Taylor and Tiemeier2017; Van den Berg et al., Reference Van den Berg, de Moor, Verweij, Krueger, Luciano, Arias Vasquez and Amin2016; Walters et al., Reference Walters, Polimanti, Johnson, McClintick, Adams, Adkins and Bertelsen2018; Williams et al., Reference Williams, Poore, Tanksley, Kweon, Courchesne-Krak, Londono-Correa and Waldman2023; Wray et al., Reference Wray, Ripke, Mattheisen, Trzaskowski, Byrne, Abdellaoui and Andlauer2018). The selection of outcome GWAS was motivated by (i) the sample size and novelty of the sample, (ii) the availability of summary statistics, and (iii) the potential sample overlap between exposure and outcome GWAS. However, the latter criterion was not always met due to the excessive use of consortia data to achieve large sample sizes. Supplemental Table S1 gives a detailed description of how the respective phenotypes were obtained, the SNP-based heritability of each outcome GWAS, and the covariates included in the GWAS.

Table 1. Overview of the outcome GWAS included in the MR analyses for externalizing traits

Note: This table gives an overview of the exposure and outcome GWAS included in the MR analyses in the present study. ‘Mainly adult’ indicates that most of the sample consisted of adults (see Supplemental Table S1 for details).

a In the initial GWAS by Karlsson Linnér et al. (Reference Karlsson Linnér, Mallard, Barr, Sanchez-Roige, Madole, Driver and Dick2021) on ‘Externalizing Traits’ up to 1,373,240 participants from 65 cohorts were included. The largest cohorts came from 23andMe (N max 599,289) and the UK Biobank (N max 403,349). As the company 23andMe Holding Co. restricts access to the data of their samples, the GWAS was rerun based on 1,045,957 participants, excluding the 23andMe sample. This reanalysis yielded similar factor loadings and genetic correlations (Williams et al. Reference Williams, Poore, Tanksley, Kweon, Courchesne-Krak, Londono-Correa and Waldman2023) and was used in this study for the outcome ‘Externalizing Traits’.

Table 2. Overview of the outcome GWAS included in the MR analyses for internalizing traits

Note: This table gives an overview of the exposure and outcome GWAS included in the MR analyses in the present study. ‘Mainly adult’ indicates that the majority of the sample consisted of adults (see Supplemental Table S1 for details); PGC=Psychiatric Genomics Consortium.

a The GWAS by the EAGLE (EArly Genetics and Lifecourse Epidemiology) consortium covered data from 64,561 children and adolescents from 22 cohorts (Jami et al. Reference Jami, Hammerschlag, Ip, Allegrini, Benyamin, Border and Lu2022). As participants were longitudinally assessed, the GWAS was based on 251,152 observations.

b To avoid sample overlap with the exposure GWAS, the publicly available GWAS on ‘Depression’, excluding UK Biobank participants, was used for the purpose of this study.

Instrumental variable

For MPT, the GWAS by Hollis et al. (Reference Hollis, Day, Busch, Thompson, Soares, Timmers and Perry2020) based on 205,354 men of European ancestry, was used. This GWAS comprised data from the UK Biobank and 23andMe in which MPT was operationalized as recalled age at VB (in the 23andMe cohort) or, for UK Biobank participants, having a relatively (compared to peers) early/late (or about average) onset of VB and onset of facial hair growth (Hollis et al., Reference Hollis, Day, Busch, Thompson, Soares, Timmers and Perry2020). Separate GWAS for the five phenotypes (age at VB, early/late VB, and early/late facial hair growth) were calculated, standardized, and then meta-analyzed using Multi-Trait Analysis of GWAS (MTAG) to derive the composite phenotype ‘Male Puberty Timing’ (Hollis et al., Reference Hollis, Day, Busch, Thompson, Soares, Timmers and Perry2020; Turley et al., Reference Turley, Walters, Maghzian, Okbay, Lee, Fontana and Furlotte2018). In total, 76 independent (defined by distance of at least 1 MB) genome-wide significant (p < 5 × 10−8) target SNPs for MPT were identified and utilized as instrumental variables for the current MR analyses. These SNPs explained 2.2% of the variance in MPT. Polygenic risk scores derived from these target SNPs significantly predicted longitudinal self-reported Tanner staging in up to 2,403 boys, further supporting the validity of this operationalization of MPT (Hollis et al., Reference Hollis, Day, Busch, Thompson, Soares, Timmers and Perry2020).

To evaluate if using UK Biobank data led to phenotype-exposure-related sample overlap that confounded our MR results, we conducted sensitivity analyses with ‘Age at VB’ data from 23andMe Inc., which included 55,871 participants (Hollis et al., Reference Hollis, Day, Busch, Thompson, Soares, Timmers and Perry2020). Of the 76 target SNPs for MPT, 36 had F-statistics > 10 for VB had and thus qualified as instrumental variables for VB.

Results

The effect of MPT on externalizing traits and disorders

For ‘Externalizing Traits’, the IVW estimate revealed a negative effect of MPT on externalizing traits, i.e. earlier pubertal timing was related to higher scores for externalizing traits (IVW, MR estimate per-one-year increase in MPT (b) = −0.03, 95% CI [−0.06, −0.01], see Figure 2). The Q-statistic indicated significant heterogeneity (based on IVW: df=57, Q-statistic=139.70, p = 6.78 × 10−9). Sensitivity analyses: The weighted median, MR-RAPS, the IVW estimate without exclusion of pleiotropic SNPs (MR-PRESSO raw), and the MR analysis without sample overlap (IVW VB) confirmed the significant results (b from −0.05 to −0.03, p-values from 0.002 to 0.007; see Figure 3A for details). The weighted-mode method and the MVMR considering BMI-related pleiotropy (IVW BMIcorr) led to similar effect estimates. However, wider confidence intervals resulted in nonsignificant findings (Weighted Mode, b = −0.04, 95% CI [−0.09, 0.01], p = 0.099; IVW BMIcorr, b = −0.02, 95% CI [−0.05, 0.01], p = 0.151).

Figure 2. Results of the univariable Mendelian randomization (MR) analyses on the effect of male puberty timing on externalizing and internalizing traits and disorders. As primary endpoints, the inverse-variance weighted (IVW) effect estimates are illustrated after MR-PRESSO correction for pleiotropic outliers. The MR estimate (b) represents the average effect of a one-year increase in MPT on the outcome of interest. Negative effect estimates indicate a higher risk of developing psychopathology (for clinical endpoints), increased expression of a specific trait or behaviors, or a younger age at first sexual contact/depression onset with earlier male puberty timing. *For reasons of comparability, ‘Age at First Sexual Contact’ and ‘Age at Onset of Depression’ are inversely coded in the current MR analyses, e.g. negative effect estimates indicate earlier ‘Age at First Sexual Contact’ or earlier ‘Age at Onset of Depression’ with earlier MPT. The outcome variables ‘Early Life Aggression’ and ‘Early Life Internalizing Traits’ are exclusively based on data from children and adolescents, whereas the other outcomes mainly involve adult cohorts (see Tables 1 and 2 and Supplemental Table S1 for a detailed description of cohorts and phenotypes). Results of the sensitivity analyses for FDR-corrected outcomes with a p-value < 0.05 are illustrated in Figure 3, while those for all other outcomes are given in Supplementary Figures S21–S33.

Figure 3. Results of the sensitivity analyses for outcomes with significant results in primary endpoints. The MR estimate (b) represents the average effect of a one-year increase in MPT per one standard deviation change in the outcome of interest. Fig 3A: Results of the MR analyses of the effect of male puberty timing on the outcome ‘Externalizing Traits’. For this outcome, 65 of the 76 target SNPs for MPT were available in the outcome GWAS by Williams et al. (Reference Williams, Poore, Tanksley, Kweon, Courchesne-Krak, Londono-Correa and Waldman2023). The MR-PRESSO global test for pleiotropy was significant (RSSobs=380.38, p < 3 × 10−4), and seven outliers were identified by MR-PRESSO. Thus, the instrumental variable used for further analyses consisted of 58 SNPs. a: Eggers-intercept did not show evidence for significant directional pleiotropy (intercept=0.0018, se=0.0010, p = 0.072). The I 2-statistic was 0.762, indicating a possible violation of the NOME assumption. Therefore, the MR Egger result should be interpreted with caution. Fig 3B: MR analyses on ‘Age at First Sexual Contact’. For this outcome, all the 76 target SNPs for male puberty timing were covered in the outcome GWAS by Mills et al. (Reference Mills, Tropf, Brazel, van Zuydam and Vaez2021). The MR-PRESSO global test for pleiotropy was significant (RSSobs=241.57, p < 3 × 10−4), and MR-PRESSO identified five outliers. b: Eggers-intercept did not show evidence for significant directional pleiotropy (intercept=0.0002, se=0.0017, p = 0.905). The I 2-statistic was 0.777, again indicating a possible violation of the NOME assumption. Fig 3C: MR analyses on ‘Early Life Internalizing Traits’. For this outcome, 71 of the 76 target SNPs for ‘Male Puberty Timing’ were available in the outcome GWAS by Jami et al. (Reference Jami, Hammerschlag, Ip, Allegrini, Benyamin, Border and Lu2022). The MR-PRESSO global test for pleiotropy was not significant (RSSobs=84.24, p = 0.169), and MR-PRESSO identified no outlier. c: Eggers-intercept did not show evidence for significant directional pleiotropy (intercept=−0.002, se=0.0014, p = 0.089). The I 2-statistic was 0.785, indicating a possible violation of the NOME assumption. Fig 3D: MR analyses on the ‘Depressed Affect’ subdomain of neuroticism. Concerning the ‘Depressed Affect’ subdomain of neuroticism, 72 target SNPs were included in the outcome GWAS by Nagel, Jansen, et al. (Reference Nagel, Jansen, Stringer, Watanabe, De Leeuw, Bryois and Muñoz-Manchado2018). The MR-PRESSO global test for pleiotropy was significant (RSSobs=140.03, p < 3 × 10−4), and one outlier was identified, which was subsequently excluded from further analyses. d: Eggers-intercept did not show evidence for significant directional pleiotropy (intercept=−0.0012, se=0.0012, p = 0.314). The I 2-statistic was 0.780, indicating a possible violation of the NOME assumption for the conduction of MR Egger. Description of effect estimates: ‘IVW’ represents the inverse-variance weighted (IVW) effect estimate after adjusting for pleiotropic outliers identified by MR-PRESSO. The ‘MR-PRESSO raw’ estimate represents the IVW estimate without excluding pleiotropic outliers. For the outcome ‘Early Life Internalizing Traits’ no pleiotropic outlier was detected, and thus, in this instance, the ‘MR-PRESSO raw’ effect estimate corresponds to the IVW estimate and therefore is not shown. The ‘IVW BMIcorr’ estimate represents the IVW effect estimate (after exclusion of pleiotropic outliers by MR-PRESSO) of male puberty timing on the specified outcome in an MVMR analysis considering BMI as an additional exposure. The ‘IVW VB’ effect estimate is the IVW effect estimate (after the exclusion of pleiotropic outliers by MR-PRESSO) for the exposure ‘Age at Voice Break’ on the respective outcome.

For ‘Age at First Sexual Contact’, the MR analysis revealed a significant effect of MPT on the age at first sexual contact; earlier MPT was associated with an earlier age at first sexual contact (IVW, b = −0.17, 95% CI [−0.21, −0.13], see Figure 2). The Q-statistic revealed significant heterogeneity (IVW: df=70, Q-statistic=138.83, p = 1.9×10−6). All sensitivity analyses, including those addressing heterogeneity, confirmed this effect (b from −0.20 to −0.09; see Figure 3B for details).

For the remaining externalizing outcomes (‘Alcohol Dependency’, ‘Risk Tolerance’, ‘Extraversion’, ‘Cannabis Abuse’, ‘Number of Sexual Partners’, ‘Ever Smoker’, ‘Early Life Aggression’, and ‘Antisocial Behavior’), the IVW effect estimate remained nonsignificant (see Figure 2). For ‘Risk Tolerance’, ‘Number of Sexual Partners’, ‘Cannabis Abuse’, and ‘Ever Smoker’, some sensitivity analyses indicated significant effects, but these no longer appeared in the remaining or additionally performed sensitivity analyses (see Figures S23 and S25–S27 for details).

The effect of MPT on internalizing traits and disorders

For ‘Early Life Internalizing Traits’, the MR analysis revealed a significant effect of MPT: later maturation leads to more internalizing traits in children and adolescents (IVW, b = 0.04, 95% CI [0.01, 0.08]; see Figure 2). The Q-statistic indicated no significant heterogeneity (IVW: df=70, Q-statistic=81.88, p = 0.157). The above finding was confirmed by all sensitivity analyses (b from 0.04 to 0.13), except for the weighted mode method (b = 0.01, 95% CI [−0.08, 0.10]; see Figure 3C).

Concerning the ‘Depressed Affect’ subdomain of neuroticism, the MR analysis revealed that earlier MPT is associated with an increased ‘Depressed Affect’ (IVW, b = −0.04, 95% CI [−0.07, −0.01]; see Figure 2). The Q-statistic was significant, indicating heterogeneity (IVW: df=70, Q-statistic=123.84, p = 7.6×10−5). Sensitivity analyses: The weighted median method, the IVW effect estimate without exclusion of outliers (MR-PRESSO raw), and the analysis excluding sample overlap (IVW VB) led to significant negative effect estimates (b from −0.04 to −0.02). The weighted mode method, MR-RAPS, and the MVMR (IVW BMIcorr) method provided similar but less precise effect estimates with wider confidence intervals and, thus, nonsignificant hypotheses testing (b ranging from −0.03 to −0.02; see Figure 3D for details). In contrast, the MR Egger method provided an effect estimate close to zero and wide confidence intervals (b = 0.00, 95% CI [−0.08, −0.08]). However, the MR Egger method might have been biased by a possible violation of the NOME assumption (I 2 = 0.780; see Supplemental Methods 1 for details). The funnel plot illustrating the precision of each genetic variant and their MR effect estimate revealed one outlier with a strong negative effect estimate (see Figure S14). However, even after excluding this outlier, leave-one-out analyses confirmed an effect of MPT on ‘Depressed Affect’ (see Figure S16).

For the outcome ‘Age at Onset of Depression’, a significant effect of ‘MPT’ was observed (IVW, b = −0.05, 95% CI [−0.10, 0.0], p = 0.034 without FDR-correction), suggesting that earlier maturation in males is linked to a younger age at onset of depression. However, after FDR correction, this effect must be considered nonsignificant according to conventional criteria (p = 0.115; see Figure 2). Sensitivity analyses, including MR Egger, MR-RAPS, MR-PRESSO raw, and IVW VB, supported an association between MPT and ‘Age at Onset of Depression’. In contrast, the weighted median and weighted mode method yielded nonsignificant results with smaller (weighted median, b = −0.02, 95% CI [−0.09, 0.04]) and even positive (weighted mode, b = 0.01, 95% CI [−0.14, 0.17]) effect estimates, suggesting a potential bias of the IVW results by pleiotropy (see Figure S21 for details). Leave-one-out analyses further underscored the instability of this finding, as the exclusion of individual SNPs rendered the association between MPT and ‘Age at Onset of Depression’ nonsignificant (see Figure S20 for details).

For all other internalizing outcomes, MR analyses yielded nonsignificant results (Figure 2). This pattern was confirmed across various sensitivity analyses (Figures S22–S33).

Discussion

Epidemiological studies indicate that an early onset of puberty is a risk factor for a broad spectrum of externalizing and internalizing behaviors and disorders. However, correlational relationships in epidemiological studies are subject to numerous confounders. Here, several MR studies were performed to provide evidence for a causal contribution of MPT to the development of character traits, behaviors, and psychopathologies. While earlier MPT led to earlier ‘Age at First Sexual Contact’, more ‘Externalizing Traits’, and increased ‘Depressed Affect’, later MPT was associated with more ‘Internalizing Traits’ in early life (childhood, adolescence, and young adulthood). However, the effect sizes for these associations were small, and all other tested associations yielded nonsignificant findings.

Externalizing traits and disorders

There was evidence that early puberal timing in males leads to an earlier initiation of sexual activity, i.e. a younger age at first sexual contact. This observation is consistent with a genetic correlation between the age of VB and the age at first sexual contact (Mills et al., Reference Mills, Tropf, Brazel, van Zuydam and Vaez2021) as well as increased risky sexual behaviors in early-maturing boys in epidemiological studies (Ullsperger & Nikolas, Reference Ullsperger and Nikolas2017). Our analyses also revealed an association between an early onset of puberty in boys and the development of externalizing traits. The outcome ’Externalizing Traits’ was operationalized via a meta-analysis that combined seven problematic behaviors and psychiatric entities (including smoking, cannabis, and problematic alcohol use, risk tolerance as well as the number of sexual partners, and the age at first sexual contact; see Supplemental Table S1 and (Karlsson Linnér et al., Reference Karlsson Linnér, Mallard, Barr, Sanchez-Roige, Madole, Driver and Dick2021)). However, except for the age at first sexual contact, MR analyses exploring the relationship between MPT and specific entities (‘Alcohol Dependency’, ‘Cannabis Abuse’, ‘Ever Smoker’, ‘Risk Tolerance’, and ‘Number of Sexual Partners’) by individual GWAS found no significant effects of MPT on these outcomes (see Figure 2). While an additional influence of MPT on the broader phenotype ‘Externalizing Traits’ beyond individual outcomes cannot be ruled out, the relatively larger effect size observed between MPT and ‘Age at First Sexual Contact’, compared to its effect on ‘Externalizing Traits’, suggests that the association with the combined outcome is primarily driven by its effect on age at first sexual contact. Therefore, it remains uncertain whether MPT exerts any further influence on externalizing traits beyond this specific outcome.

Internalizing traits and disorders

To our surprise, we observed that later male puberty leads to more internalizing traits in children and adolescents (‘Early Life Internalizing Traits’; i.e. depressive symptoms, emotional problems, and anxiety). This observation contrasts with most previous epidemiological studies suggesting a link between early (but not late) onset of puberty and the development of internalizing problematic traits in adolescent boys (Ullsperger & Nikolas, Reference Ullsperger and Nikolas2017). Additionally, a meta-analysis of the effects of late puberty timing in males did not indicate an association with the occurrence of psychopathologies (Ullsperger & Nikolas, Reference Ullsperger and Nikolas2017). However, this conclusion rests on relatively few studies, as noted by the authors of this meta-analysis (Ullsperger & Nikolas, Reference Ullsperger and Nikolas2017). Further, a recent large combined epidemiological and MR study, including the most comprehensive set of confounders to date, found no evidence of a relationship between (either early or late) MPT and major depressive disorder (MDD) in males (Hirtz et al., Reference Hirtz, Grasemann, Hölling, von Holt, Albers, Hinney and Peters2023). This finding suggests that previous studies may have been affected by confounding, which could also extend to other domains of internalizing traits. The ‘gendered deviation’ hypothesis postulates that those experiencing the most extreme deviations in puberty timing are at the highest risk for mental health problems (Sontag, Graber, & Clemans, Reference Sontag, Graber and Clemans2011; Ullsperger & Nikolas, Reference Ullsperger and Nikolas2017). As girls generally mature earlier than boys, late-maturing boys are particularly at risk of suffering psychosocial stress, as late-maturing boys are not only the last of their sex but also the last of their whole age group to go through puberty (Sontag et al., Reference Sontag, Graber and Clemans2011; Ullsperger & Nikolas, Reference Ullsperger and Nikolas2017). So far, epidemiological studies have suggested little evidence for this hypothesis (Ullsperger & Nikolas, Reference Ullsperger and Nikolas2017). In contrast, our observation of more internalizing symptoms with later maturation in boys would be consistent with this hypothesis. However, it must be noted that the outcome GWAS by Jami et al. (Reference Jami, Hammerschlag, Ip, Allegrini, Benyamin, Border and Lu2022) relied on a heterogeneously defined phenotype of ‘Early Life Internalizing Traits’ that was assessed using multiple sources (e.g. parental, self, or teacher ratings) and different measurement instruments, across a broad age range (boys and girls between 3 and 18 years). As individual-level data or age-/instrument-specific GWAS were not available to us, our analysis cannot disentangle whether the effect found here is age-specific (e.g. later MPT leading to more internalizing problems, early MPT leading to less internalizing problems, or both) or nonlinear (e.g. extreme early or late puberty leading to more internalizing problems). Additionally, it remains unclear whether our finding is applicable to all measurement approaches included to define the outcome phenotype. Thus, the final interpretation of our results, and the extent to which confounding factors may account for discrepancies with existing epidemiological studies, must be determined by future, well-designed epidemiological studies that are able to analyze age-specific and nonlinear effects and complement our approach.

Long-term effects of MPT

Although numerous cross-sectional studies suggest a relationship between earlier onset of puberty and (clinical) endpoints (i.e. the development of alcohol dependency, cannabis abuse, major depression, or anxiety disorders), this was not shown by the MR analyses conducted in the present study. The clinical outcomes in this study were assessed in adults considering acute or past diagnoses or symptoms, thereby serving as an estimate of the lifetime risk for these conditions. The ‘maturational disparity’ hypothesis postulates that early maturing children and adolescents are confronted with a mismatch between rapid physical changes and slowly evolving cognitive resources to cope with these changes (Benoit et al., Reference Benoit, Lacourse and Claes2013; Ullsperger & Nikolas, Reference Ullsperger and Nikolas2017) which is also supported on a neurophysiological level (Vijayakumar, Whittle, & Silk, Reference Vijayakumar, Whittle and Silk2023). However, whether this mismatch attenuates with increasing age, and thus the long-term impact of early puberty on problematic behaviors, still needs to be investigated. In line, recent longitudinal studies have shown that the effects of early puberty in males on antisocial behavior, depression, externalizing traits, and substance abuse decrease (and in some cases even reverse) with increasing age (Dimler & Natsuaki, Reference Dimler and Natsuaki2021; Hoyt et al., Reference Hoyt, Niu, Pachucki and Chaku2020). Accordingly, the MR analyses conducted here also showed a relationship between early onset of puberty and age at first sexual contact but not the total number of sexual partners in adulthood. Also, while there was a notable trend towards a younger age of onset of depressive disorders, this did not apply to the lifetime risk for the development of a MDD. Thus, these results indicate that mental health consequences associated with ‘maturational disparity’ are transient, possibly confined to adolescence and early adulthood. On a genetic level, a recent GWAS in 94,154 participants with depression found a significant, albeit not perfect, genetic correlation (r g = −0.49) between age at onset of depression and lifetime MDD risk (Harder et al., Reference Harder, Nguyen, Pasman, Mosing, Hägg and Lu2022). Similarly, strong but not perfect correlations were found between internalizing traits in children and adolescents and the lifetime risk for depression and anxiety disorders (r g = 0.72 and 0.76; (Jami et al., Reference Jami, Hammerschlag, Ip, Allegrini, Benyamin, Border and Lu2022). Thus, even though much of the genetic liability of psychiatric disorders is shared across the lifespan, their manifestation exhibits age-related specificity. Further GWAS focusing on specific age groups (e.g. in children, adolescents, and young adults) are needed to gain a deeper understanding of these complex relationships across the lifespan and could inform future MR studies.

The only lifetime trait affected by MPT was the ‘Depressed Affect’ subdomain of neuroticism, with early puberty timing in males leading to increased scores for ‘Depressed Affect’. It should be noted that the effect of MPT on ‘Depressed Affect’ was not consistently found in sensitivity analyses, potentially related to issues of pleiotropy. The ‘Depressed Affect’ subdomain is derived from four of twelve items of a neuroticism scale (for details, see Supplemental Table S1) by clustering genetic correlations on a single-item level (Nagel, Watanabe, Stringer, Posthuma, & Van Der Sluis, Reference Nagel, Watanabe, Stringer, Posthuma and Van Der Sluis2018). In contrast, the other subdomain of this scale, ‘Worry’, exhibited a nonsignificant effect estimate in the opposite direction in the present study. Consequently, our MR studies revealed a nonsignificant effect of MPT on the broad trait ‘Neuroticism’. Despite strong genetic correlations between ‘Depressed Affect’ and depressive symptoms as well as anxiety (Nagel, Watanabe, et al., Reference Nagel, Watanabe, Stringer, Posthuma and Van Der Sluis2018), no effect of MPT on these outcomes was seen here. Thus, further studies on the relevance of these subdomains of ‘Neuroticism’ are needed. The effect of early MPT on increased levels of ’Depressed Affect’ is noteworthy, given our other finding that late MPT is associated with more pronounced ’Early Life Internalizing Traits,’ suggesting age-specific effects of MPT on internalizing traits. Similarly, sex differences in internalizing psychopathologies, such as depression and anxiety disorders, emerge only during adolescence (Dalsgaard et al., Reference Dalsgaard, Thorsteinsson, Trabjerg, Schullehner, Plana-Ripoll, Brikell and Timmerman2020), indicating that the causal pathways for prepubertal versus postpubertal onset may differ, with hormonal influences or puberty timing being potential contributing factors (Naninck, Lucassen, & Bakker, Reference Naninck, Lucassen and Bakker2011; Rutter, Caspi, & Moffitt, Reference Rutter, Caspi and Moffitt2003).

Limitations

Our study has several limitations, some of which are inherent to the MR approach. Therefore, our results should be viewed as complementary to existing evidence from observational studies. First, MR studies are subject to the risk of horizontal pleiotropy. Although we conducted sensitivity analyses using multiple approaches to detect and correct for pleiotropic effects, including robust methods such as MR-Egger, MR-PRESSO, weighted median/mode, and multivariable MR analyses, it is not possible to eliminate the risk of undetected pleiotropic pathways. Nonetheless, the consistency of findings across various approaches strengthens the validity of our conclusions.

Second, the exposure (MPT) was determined based on self-reported information on the timing of voice breakage and facial hair growth, which might be subject to recall bias.

Third, GWAS for 16 of the 17 outcomes analyzed in this study were based on sex-pooled effect estimates rather than sex-specific analyses, primarily because sex-specific GWAS were not available. Current evidence suggests marginal sex differences in the autosomal genetic architecture of most traits, behaviors, and psychopathologies (Martin et al., Reference Martin, Khramtsova, Goleva, Blokland, Traglia, Walters and Børglum2021), justifying their use as outcomes in MR with sex-specific exposure. Accordingly, this approach has been applied in various MR studies, including testosterone levels in men or age at menarche concerning MDD (Hirtz et al., Reference Hirtz, Hars, Naaresh, Laabs, Antel, Grasemann and Peters2022; Syed, He, & Shi, Reference Syed, He and Shi2020). Although we do not expect significant differences based on sex, given the genetic similarity of the traits and the inclusion of sex as a covariate in the outcome GWAS (see Supplemental Table S1), future research using sex-specific GWAS may help to further refine our understanding of any potential sex-specific effects.

Fourth, the sample sizes of the outcome GWAS varied widely, leading to large differences in the power of the conducted MR studies. As most of the outcome GWAS were based on continuous outcomes and the power estimation of MR studies on continuous outcomes is dependent on further information, e.g. on the exposure–outcome association on the phenotypic level (Brion, Shakhbazov, & Visscher, Reference Brion, Shakhbazov and Visscher2013), which was not available to us, we decided against power calculations based on arbitrary assumptions. However, despite reasonable sample sizes in most of the studies included, minor effects of MPT may have been overlooked, particularly when less reliable outcome measures were considered, such as individual rather than composite outcome phenotypes.

Fifth, two-sample MR studies as utilized here are based on linear models and, therefore, do not account for potential nonlinear relationships.

Sixth, the MR assumption that the exposure influences the outcome, rather than vice versa, could not be directly assessed. Although a reverse relationship appears unlikely, given that most outcome phenotypes were measured in adulthood, bidirectional MR studies to confirm this could not be performed due to restricted access to the summary statistics from the GWAS on MPT.

Seventh, all participants of the GWAS utilized in this study were of white European ancestry. The focus on this population was driven by the higher availability of data rather than scientific interest. However, it remains unclear whether the results of our study are transferable to populations of other ancestry.

Conclusions

The MR analyses conducted here support a causal effect of MPT on the development of specific psychological traits and behaviors. Specifically, earlier MPT was associated with an earlier age at first sexual contact, while later pubertal timing was associated with increased internalizing traits in children and adolescents. However, all effect sizes for these effects were small. Further, the effects of MPT might be transient as no effects on adulthood extraversion or neuroticism or lifetime risk for psychiatric disorders were found. In line with this conclusion, there was a notable trend towards earlier age at the onset of depression with early maturation. However, this trend did not perpetuate to the lifetime risk of developing a MDD. To further explore the causal relationships between puberty timing and the onset of psychopathologies, particularly during the critical periods of late adolescence and early adulthood, additional age- and sex-specific GWAS followed by MR analyses are needed. MR analyses should, wherever possible, be combined with observational studies to compensate for the advantages and limitations of each of these analytical approaches.

Abbreviations

BMI

body mass index

FDR

false discovery rate

GWAS

genome-wide association study

IVW

inverse-variance weighted

MDD

major depressive disorder

MPT

male puberty timing

MR

Mendelian randomization

MVMR

multivariable Mendelian randomization

SNP

single-nucleotide polymorphism

Supplementary material

The supplementary material for this article can be found at http://doi.org/10.1017/S0033291725000352.

Data availability statement

All GWAS summary statistics utilized in this study were publicly available or available upon request (see Supplemental Table S1 details on how the GWAS summary statistics were obtained).

Acknowledgments

We sincerely thank the three anonymous reviewers for their valuable comments, which have strengthened the article. The figures were created with BioRender.com. We acknowledge support by the Open Access Publication Fund of the University of Duisburg-Essen.

Author contribution

LD, TP, and RH conceptualized the study. LD conducted the statistical analyses. TP and RH supervised the statistical analyses. LD wrote the first draft of the article. TP, CG, AH, and RH critically revised the article.

Funding statement

This work was supported by a fellowship from the University Medicine Essen Clinician Scientist Academy (UMEA) of the Medical Faculty of the University Duisburg-Essen (supported by the German Research Foundation (DFG)) to LD (FU 356/12-2).

Competing interests

LD, TP, CG, AH, and RH have nothing to disclose.

Ethical standards

All GWAS utilized in the current MR studies were conducted after approval of the respective local ethics committee and obtained written informed consent of the participants or their legal representatives.

References

Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society: Series B (Methodological), 57(1), 289–300.CrossRefGoogle Scholar
Benoit, A., Lacourse, E., & Claes, M. (2013). Pubertal timing and depressive symptoms in late adolescence: The moderating role of individual, peer, and parental factors. Development and Psychopathology, 25(2), 455–471.CrossRefGoogle ScholarPubMed
Bowden, J., Davey Smith, G., & Burgess, S. (2015). Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. International Journal of Epidemiology, 44(2), 512–525.CrossRefGoogle ScholarPubMed
Bowden, J., Davey Smith, G., Haycock, P.C., & Burgess, S. (2016). Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genetic Epidemiology, 40(4), 304–314.CrossRefGoogle ScholarPubMed
Brion, M. J. A., Shakhbazov, K., & Visscher, P. M. (2013). Calculating statistical power in Mendelian randomization studies. International Journal of Epidemiology, 42(5), 1497–1501.CrossRefGoogle ScholarPubMed
Burgess, S., Bowden, J., Fall, T., Ingelsson, E., & Thompson, S. G. (2017). Sensitivity analyses for robust causal inference from Mendelian randomization analyses with multiple genetic variants. Epidemiology (Cambridge, Mass.), 28(1), 30.CrossRefGoogle ScholarPubMed
Burgess, S., Davies, N. M., & Thompson, S. G. (2016). Bias due to participant overlap in two-sample Mendelian randomization. Genetic Epidemiology, 40(7), 597–608.Google ScholarPubMed
Burgess, S., & Thompson, S. G. (2015). Multivariable Mendelian randomization: the use of pleiotropic genetic variants to estimate causal effects. American Journal of Epidemiology, 181(4), 251–260.Google ScholarPubMed
Dalsgaard, S., Thorsteinsson, E., Trabjerg, B. B., Schullehner, J., Plana-Ripoll, O., Brikell, I., … Timmerman, A. (2020). Incidence rates and cumulative incidences of the full spectrum of diagnosed mental disorders in childhood and adolescence. JAMA Psychiatry, 77(2), 155–164.Google ScholarPubMed
Day, F. R., Bulik-Sullivan, B., Hinds, D. A., Finucane, H. K., Murabito, J. M., Tung, J. Y., … Perry, J. R. B (2015). Shared genetic aetiology of puberty timing between sexes and with health-related outcomes. Nature Communications, 6(1), 8842.CrossRefGoogle ScholarPubMed
Dimler, L. M., & Natsuaki, M. N. (2021). Trajectories of violent and nonviolent behaviors from adolescence to early adulthood: does early puberty matter, and, if so, how long? Journal of Adolescent Health, 68(3), 523–531.CrossRefGoogle ScholarPubMed
Drosopoulou, G., Sergentanis, T. N., Mastorakos, G., Vlachopapadopoulou, E., Michalacos, S., Tzavara, C., … Tsitsika, A. (2021). Psychosocial health of adolescents in relation to underweight, overweight/obese status: The EU NET ADB survey. European Journal of Public Health, 31(2), 379–384.CrossRefGoogle ScholarPubMed
Ebrahim, S., & Davey Smith, G. (2008). Mendelian randomization: can genetic epidemiology help redress the failures of observational epidemiology? Human Genetics, 123, 15–33.CrossRefGoogle ScholarPubMed
Hamlat, E. J., Stange, J. P., Abramson, L. Y., & Alloy, L. B. (2014). Early pubertal timing as a vulnerability to depression symptoms: Differential effects of race and sex. Journal of Abnormal Child Psychology, 42, 527–538.Google ScholarPubMed
Harder, A., Nguyen, T. D., Pasman, J. A., Mosing, M. A., Hägg, S., & Lu, Y. (2022). Genetics of age-at-onset in major depression. Translational Psychiatry, 12(1), 124.CrossRefGoogle ScholarPubMed
Hartwig, F. P., Davey Smith, G., & Bowden, J. (2017). Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption. International Journal of Epidemiology, 46(6), 1985–1998.CrossRefGoogle ScholarPubMed
Haycock, P. C., Burgess, S., Wade, K. H., Bowden, J., Relton, C., & Davey Smith, G. (2016). Best (but oft-forgotten) practices: the design, analysis, and interpretation of Mendelian randomization studies. The American Journal of Clinical Nutrition, 103(4), 965–978.CrossRefGoogle ScholarPubMed
Hirtz, R., Grasemann, C., Hölling, H., von Holt, B. H., Albers, N., Hinney, A., … Peters, T. (2023). No relationship between male pubertal timing and depression–new insights from epidemiology and Mendelian randomization. Psychological Medicine, 1–10.Google Scholar
Hirtz, R., Hars, C., Naaresh, R., Laabs, B. H., Antel, J., Grasemann, C., … Peters, T. (2022). Causal Effect of Age at Menarche on the Risk for Depression: Results From a Two-Sample Multivariable Mendelian Randomization Study. Frontiers in Genetics, 13.Google ScholarPubMed
Hollis, B., Day, F. R., Busch, A. S., Thompson, D. J., Soares, A. L. G., Timmers, P. R. H. J., … Perry, J. R. B (2020). Genomic analysis of male puberty timing highlights shared genetic basis with hair colour and lifespan. Nature Communications, 11(1), 1536.Google ScholarPubMed
Hoyt, L. T., Niu, L., Pachucki, M. C., & Chaku, N. (2020). Timing of puberty in boys and girls: Implications for population health. SSM-Population Health, 10, 100549.Google Scholar
Ip, H. F., Van der Laan, C. M., Krapohl, E. M. L., Brikell, I., Sánchez-Mora, C., Nolte, I. M., … Zafarmand, H. (2021). Genetic association study of childhood aggression across raters, instruments, and age. Translational Psychiatry, 11(1), 413.Google ScholarPubMed
Jami, E. S., Hammerschlag, A. R., Ip, H. F., Allegrini, A. G., Benyamin, B., Border, R., … Lu, Y. (2022). Genome-wide association meta-analysis of childhood and adolescent internalizing symptoms. Journal of the American Academy of Child & Adolescent Psychiatry, 61(7), 934–945.Google ScholarPubMed
Johnson, E. C., Demontis, D., Thorgeirsson, T. E., Walters, R. K., Polimanti, R., Hatoum, A. S., … Clarke, T. K. (2020). A large-scale genome-wide association study meta-analysis of cannabis use disorder. The Lancet Psychiatry, 7(12), 1032–1045.Google ScholarPubMed
Karlsson Linnér, R., Biroli, P., Kong, E., Meddens, S. F. W., Wedow, R., Fontana, M. A., … Hammerschlag, A. R. (2019). Genome-wide association analyses of risk tolerance and risky behaviors in over 1 million individuals identify hundreds of loci and shared genetic influences. Nature Genetics, 51(2), 245–257.CrossRefGoogle ScholarPubMed
Karlsson Linnér, R., Mallard, T. T., Barr, P. B., Sanchez-Roige, S., Madole, J. W., Driver, M. N., … Dick, D. M.. (2021). Multivariate analysis of 1.5 million people identifies genetic associations with traits related to self-regulation and addiction. Nature Neuroscience, 24(10), 1367–1376.Google ScholarPubMed
Lawlor, D. A., Harbord, R. M., Sterne, J. A. C., Timpson, N., & Davey Smith, G. (2008). Mendelian randomization: using genes as instruments for making causal inferences in epidemiology. Statistics in Medicine, 27(8), 1133–1163.Google ScholarPubMed
Li, C., Zhao, Q., Zhang, L., & Zhang, Y. (2023). Tell me what you think about: Does parental solicitation weaken the links between pubertal timing and depressive symptoms? Current Psychology, 42(6), 4326–4335.CrossRefGoogle Scholar
Martin, J., Khramtsova, E. A., Goleva, S. B., Blokland, G. A. M., Traglia, M., Walters, R. K., … Børglum, A. D. (2021). Examining sex-differentiated genetic effects across neuropsychiatric and behavioral traits. Biological Psychiatry, 89(12), 1127–1137.Google ScholarPubMed
Mills, M. C., Tropf, F. C., Brazel, D. M., van Zuydam, N., Vaez, A., & Team, BIOS Consortium Management. (2021). Identification of 371 genetic variants for age at first sex and birth linked to externalising behaviour. Nature Human Behaviour, 5(12), 1717–1730.Google ScholarPubMed
Nagel, M., Jansen, P. R., Stringer, S., Watanabe, K., De Leeuw, C. A., Bryois, J., … Muñoz-Manchado, A. B. (2018). Meta-analysis of genome-wide association studies for neuroticism in 449,484 individuals identifies novel genetic loci and pathways. Nature Genetics, 50(7), 920–927.Google Scholar
Nagel, M., Watanabe, K., Stringer, S., Posthuma, D., & Van Der Sluis, S. (2018). Item-level analyses reveal genetic heterogeneity in neuroticism. Nature Communications, 9(1), 905.CrossRefGoogle ScholarPubMed
Naninck, E. F. G., Lucassen, P. J., & Bakker, J. (2011). Sex differences in adolescent depression: do sex hormones determine vulnerability? Journal of Neuroendocrinology, 23(5), 383–392.Google ScholarPubMed
Niu, L., Sheffield, P., & Li, Y. (2023). Pubertal timing, neighborhood income, and mental health in boys and girls: Findings from the adolescent brain cognitive development study. Social Science & Medicine, 334, 116220.Google Scholar
Otowa, T., Hek, K., Lee, M., Byrne, E. M., Mirza, S. S., Nivard, M. G., … Wolen, A. (2016). Meta-analysis of genome-wide association studies of anxiety disorders. Molecular Psychiatry, 21(10), 1391–1399.CrossRefGoogle ScholarPubMed
Pan, C., Li, C., Cheng, S., Chen, Y., Zhang, J., Zhang, Z., … Zhang, F. (2023). The Effect of Secondary Sexual Characteristics Outset Time Abnormality on Addiction in Adults: a Mendelian Randomization Study. International Journal of Mental Health and Addiction, 1–18.Google Scholar
Pham, H. T., DiLalla, L. F., Corley, R. P., Dorn, L. D., & Berenbaum, S. A. (2022). Family environmental antecedents of pubertal timing in girls and boys: a review and open questions. Hormones and Behavior, 138, 105101.Google Scholar
Pulit, S. L., Stoneman, C., Morris, A. P., Wood, A. R., Glastonbury, C. A., Tyrrell, J., … Ji, Y. (2019). Meta-analysis of genome-wide association studies for body fat distribution in 694 649 individuals of European ancestry. Human Molecular Genetics, 28(1), 166–174.Google ScholarPubMed
Rutter, M., Caspi, A., & Moffitt, T. E. (2003). Using sex differences in psychopathology to study causal mechanisms: unifying issues and research strategies. Journal of Child Psychology and Psychiatry, 44(8), 1092–1115.CrossRefGoogle ScholarPubMed
Skrivankova, V. W., Richmond, R. C., Woolf, B. A. R., Yarmolinsky, J., Davies, N. M., Swanson, S. A., … Dimou, N. (2021). Strengthening the reporting of observational studies in epidemiology using Mendelian randomization: the STROBE-MR statement. Jama, 326(16), 1614–1621.Google ScholarPubMed
Sontag, L. M., Graber, J. A., & Clemans, K. H. (2011). The role of peer stress and pubertal timing on symptoms of psychopathology during early adolescence. Journal of Youth and Adolescence, 40, 1371–1382.CrossRefGoogle ScholarPubMed
Syed, A. A. S., He, L., & Shi, Y. (2020). The potential effect of aberrant testosterone levels on common diseases: a mendelian randomization study. Genes, 11(7), 721.CrossRefGoogle Scholar
Tielbeek, J. J., Johansson, A., Polderman, T. J. C., Rautiainen, M. R., Jansen, P., Taylor, M., … Tiemeier, H. (2017). Genome-wide association studies of a broad spectrum of antisocial behavior. JAMA Psychiatry, 74(12), 1242–1250.Google ScholarPubMed
Turley, P., Walters, R. K., Maghzian, O., Okbay, A., Lee, J. J., Fontana, M. A., … Furlotte, N. A. (2018). Multi-trait analysis of genome-wide association summary statistics using MTAG. Nature Genetics, 50(2), 229–237.Google ScholarPubMed
Ullsperger, J. M, & Nikolas, M. A. (2017). A meta-analytic review of the association between pubertal timing and psychopathology in adolescence: Are there sex differences in risk? Psychological Bulletin, 143(9), 903.Google ScholarPubMed
Van den Berg, S. M., de Moor, M. H. M., Verweij, K. J. H., Krueger, R. F., Luciano, M., Arias Vasquez, A., … Amin, N. (2016). Meta-analysis of genome-wide association studies for extraversion: findings from the genetics of personality consortium. Behavior Genetics, 46, 170–182.Google ScholarPubMed
Verbanck, M., Chen, C. Y., Neale, B., & Do, R. (2018). Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nature Genetics, 50(5), 693–698.CrossRefGoogle ScholarPubMed
Vijayakumar, N., Whittle, S., & Silk, T. J. (2023). Corticolimbic connectivity mediates the relationship between pubertal timing and mental health problems. Psychological Medicine, 53(16), 7655–7665.Google ScholarPubMed
Vijayakumar, N., Youssef, G., Bereznicki, H., Dehestani, N., Silk, T. J., & Whittle, S.. (2023). The social determinants of emotional and behavioral problems in adolescents experiencing early puberty. Journal of Adolescent Health.Google ScholarPubMed
Walters, R. K., Polimanti, R., Johnson, E. C., McClintick, J. N., Adams, M. J., Adkins, A. E., … Bertelsen, S. (2018). Transancestral GWAS of alcohol dependence reveals common genetic underpinnings with psychiatric disorders. Nature Neuroscience, 21(12), 1656–1669.Google ScholarPubMed
Williams, C. M., Poore, H., Tanksley, P. T., Kweon, H., Courchesne-Krak, N. S., Londono-Correa, D., … Waldman, I. D. (2023). Guidelines for evaluating the comparability of down-sampled GWAS summary statistics. Behavior Genetics, 53(5), 404–415.Google ScholarPubMed
Woolf, B., Mason, A., Zagkos, L., Sallis, H., Munafò, M. R, & Gill, D. (2024). MRSamePopTest: introducing a simple falsification test for the two-sample mendelian randomisation ‘same population’assumption. BMC Research Notes, 17(1), 27.Google ScholarPubMed
Wray, N. R., Ripke, S., Mattheisen, M., Trzaskowski, M., Byrne, E. M., Abdellaoui, A., … Andlauer, T. M. F (2018). Genome-wide association analyses identify 44 risk variants and refine the genetic architecture of major depression. Nature Genetics, 50(5), 668–681.CrossRefGoogle ScholarPubMed
Ye, Z., & Rudolph, K. D. (2023). Talk it out or tuck it away: The contribution of maternal socialization of coping to depression in youth with early pubertal maturation. Developmental Psychology.Google ScholarPubMed
Zhao, Q., Wang, J., Hemani, G., Bowden, J., & Small, D. S. (2020). Statistical inference in two-sample summary-data Mendelian randomization using robust adjusted profile score.CrossRefGoogle Scholar
Figure 0

Figure 1. Illustration of the analytic approach and outcome variables assessed. Further details on the outcome phenotypes are given in Tables 1 and 2. BMI=body mass index; IVW=inverse-variance weighted; MPT=male puberty timing; SNP=single-nucleotide polymorphism; VB=voice break.

Figure 1

Table 1. Overview of the outcome GWAS included in the MR analyses for externalizing traits

Figure 2

Table 2. Overview of the outcome GWAS included in the MR analyses for internalizing traits

Figure 3

Figure 2. Results of the univariable Mendelian randomization (MR) analyses on the effect of male puberty timing on externalizing and internalizing traits and disorders. As primary endpoints, the inverse-variance weighted (IVW) effect estimates are illustrated after MR-PRESSO correction for pleiotropic outliers. The MR estimate (b) represents the average effect of a one-year increase in MPT on the outcome of interest. Negative effect estimates indicate a higher risk of developing psychopathology (for clinical endpoints), increased expression of a specific trait or behaviors, or a younger age at first sexual contact/depression onset with earlier male puberty timing. *For reasons of comparability, ‘Age at First Sexual Contact’ and ‘Age at Onset of Depression’ are inversely coded in the current MR analyses, e.g. negative effect estimates indicate earlier ‘Age at First Sexual Contact’ or earlier ‘Age at Onset of Depression’ with earlier MPT. The outcome variables ‘Early Life Aggression’ and ‘Early Life Internalizing Traits’ are exclusively based on data from children and adolescents, whereas the other outcomes mainly involve adult cohorts (see Tables 1 and 2 and Supplemental Table S1 for a detailed description of cohorts and phenotypes). Results of the sensitivity analyses for FDR-corrected outcomes with a p-value < 0.05 are illustrated in Figure 3, while those for all other outcomes are given in Supplementary Figures S21–S33.

Figure 4

Figure 3. Results of the sensitivity analyses for outcomes with significant results in primary endpoints. The MR estimate (b) represents the average effect of a one-year increase in MPT per one standard deviation change in the outcome of interest. Fig 3A: Results of the MR analyses of the effect of male puberty timing on the outcome ‘Externalizing Traits’. For this outcome, 65 of the 76 target SNPs for MPT were available in the outcome GWAS by Williams et al. (2023). The MR-PRESSO global test for pleiotropy was significant (RSSobs=380.38, p < 3 × 10−4), and seven outliers were identified by MR-PRESSO. Thus, the instrumental variable used for further analyses consisted of 58 SNPs. a: Eggers-intercept did not show evidence for significant directional pleiotropy (intercept=0.0018, se=0.0010, p = 0.072). The I2-statistic was 0.762, indicating a possible violation of the NOME assumption. Therefore, the MR Egger result should be interpreted with caution. Fig 3B: MR analyses on ‘Age at First Sexual Contact’. For this outcome, all the 76 target SNPs for male puberty timing were covered in the outcome GWAS by Mills et al. (2021). The MR-PRESSO global test for pleiotropy was significant (RSSobs=241.57, p < 3 × 10−4), and MR-PRESSO identified five outliers. b: Eggers-intercept did not show evidence for significant directional pleiotropy (intercept=0.0002, se=0.0017, p = 0.905). The I2-statistic was 0.777, again indicating a possible violation of the NOME assumption. Fig 3C: MR analyses on ‘Early Life Internalizing Traits’. For this outcome, 71 of the 76 target SNPs for ‘Male Puberty Timing’ were available in the outcome GWAS by Jami et al. (2022). The MR-PRESSO global test for pleiotropy was not significant (RSSobs=84.24, p = 0.169), and MR-PRESSO identified no outlier. c: Eggers-intercept did not show evidence for significant directional pleiotropy (intercept=−0.002, se=0.0014, p = 0.089). The I2-statistic was 0.785, indicating a possible violation of the NOME assumption. Fig 3D: MR analyses on the ‘Depressed Affect’ subdomain of neuroticism. Concerning the ‘Depressed Affect’ subdomain of neuroticism, 72 target SNPs were included in the outcome GWAS by Nagel, Jansen, et al. (2018). The MR-PRESSO global test for pleiotropy was significant (RSSobs=140.03, p < 3 × 10−4), and one outlier was identified, which was subsequently excluded from further analyses. d: Eggers-intercept did not show evidence for significant directional pleiotropy (intercept=−0.0012, se=0.0012, p = 0.314). The I2-statistic was 0.780, indicating a possible violation of the NOME assumption for the conduction of MR Egger. Description of effect estimates: ‘IVW’ represents the inverse-variance weighted (IVW) effect estimate after adjusting for pleiotropic outliers identified by MR-PRESSO. The ‘MR-PRESSO raw’ estimate represents the IVW estimate without excluding pleiotropic outliers. For the outcome ‘Early Life Internalizing Traits’ no pleiotropic outlier was detected, and thus, in this instance, the ‘MR-PRESSO raw’ effect estimate corresponds to the IVW estimate and therefore is not shown. The ‘IVW BMIcorr’ estimate represents the IVW effect estimate (after exclusion of pleiotropic outliers by MR-PRESSO) of male puberty timing on the specified outcome in an MVMR analysis considering BMI as an additional exposure. The ‘IVW VB’ effect estimate is the IVW effect estimate (after the exclusion of pleiotropic outliers by MR-PRESSO) for the exposure ‘Age at Voice Break’ on the respective outcome.

Supplementary material: File

Dinkelbach et al. supplementary material 1

Dinkelbach et al. supplementary material
Download Dinkelbach et al. supplementary material 1(File)
File 31.1 KB
Supplementary material: File

Dinkelbach et al. supplementary material 2

Dinkelbach et al. supplementary material
Download Dinkelbach et al. supplementary material 2(File)
File 7.6 MB