Worldwide Secular Trends in Age at Pubertal Onset Assessed by Breast Development Among Girls: A Systematic Review and Meta-analysis

Key Points

Questions  Has the age at onset of thelarche in girls changed within the past 4 decades?

Findings  This systematic review and meta-analysis found that age at pubertal onset, with thelarche assessed by physical or clinical examination of the breast, decreased by a mean of almost 3 months per decade from 1977 to 2013.

Meaning  In most textbooks, thelarche among girls younger than 8 years is considered pathologic and warrants further investigations; therefore, a younger age at thelarche in girls in the general population will change current diagnostic decision-making in girls suspected to have puberty disorders.

Importance  The initial clinical sign of pubertal onset in girls is breast gland development (thelarche). Although numerous studies have used recalled age at menarche (first menstruation) to assess secular trends of pubertal timing, no systematic review has been conducted of secular trends of thelarche.

Objectives  To systematically evaluate published data on pubertal timing based on age at thelarche and evaluate the change in pubertal onset in healthy girls around the world.

Data Sources  A systematic literature search was performed in PubMed and Embase of all original peer-reviewed articles published in English before June 20, 2019.

Study Selection  Included studies used clinical assessment of breast development in healthy girls and used adequate statistical methods, including the reporting of SEs or CIs. The quality of the articles was evaluated by assessing study design, potential sources of bias, main characteristics of the study population, and methods of statistical analysis.

Data Extraction and Synthesis  In accordance with PRISMA guidelines, all articles were assessed for eligibility independently by 2 authors. Weighted regression analysis was performed using a random-effects model.

Main Outcomes and Measures  Studies examining age at thelarche (development of Tanner breast stage 2) in healthy girls.

Results  The literature search resulted in a total of 3602 studies, of which 30 studies fulfilled the eligibility criteria. There was a secular trend in ages at thelarche according to race/ethnicity and geography. Overall, the age at thelarche decreased 0.24 years (95% CI, −0.44 to −0.04) (almost 3 months) per decade from 1977 to 2013 (P = .02).

Conclusions and Relevance  The age at thelarche has decreased a mean of almost 3 months per decade from 1977 to 2013. A younger age at pubertal onset may change current diagnostic decision-making. The medical community needs current and relevant data to redefine “precocious puberty,” because the traditional definition may be outdated, at least in some regions of the world.

Puberty is an important period of life with marked physical and psychological changes. Subsequent to the reactivation of the hypothalamus-pituitary-gonadal axis after a period of quiescence during childhood, the hormonal changes during this transitional period ultimately lead to attainment of complete adult reproductive capacity. The precise mechanisms underlying the reactivation of the hypothalamus-pituitary-gonadal axis are not fully known; however, genetic, nutritional, stress-related, and environmental factors are known to influence the onset of puberty.¹

Although menarche is a commonly studied marker of female puberty, it is a late pubertal phenomenon, and often recorded using self-reporting by adult women, which may introduce recall bias. In contrast, the most important initial clinical milestone of pubertal onset is thelarche,² which is the development of glandular breast tissue clinically classified according to the Tanner scale stages. Stage 1 (B1) represents the prepubertal state with no glandular breast tissue and stages 2 to 5 (B2-B5) represent the pubertal stages with glandular breast tissue.³ Initial elevation of both the breast and papilla combined with an enlargement of the areola diameter traditionally defines the Tanner breast stage 2 (B2).³ Originally, the Tanner scale was based on visual inspection only. Today, clinical evaluation including inspection as well as breast palpation is considered the criterion standard whereby prepubertal and pubertal girls can be reliably distinguished.⁴

Although numerous studies have used recalled age at menarche to assess secular trends of pubertal timing, fewer have assessed secular trends in the onset of thelarche in girls. There is a marked indication of younger median ages of thelarche^5,6 compared with the traditional 11 years,⁷ which calls for confirmation in a large systematic review with meta-analysis.

To our knowledge, to date, no systematic review exists concerning timing of the initial pubertal milestone of thelarche. The purpose of this systematic review is therefore to document and evaluate the overall secular changes in age at pubertal onset in healthy girls measured by age at thelarche assessed by clinical examination from 1977 to 2019.

This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline.⁸ The systematic review protocol was registered at PROSPERO.org (registration number CRD42018103344) prior to initiation of the review process, on June 20, 2019.

We used PubMed and Embase to perform a systematic literature search for original peer-reviewed English-language articles published before June 20, 2019. We combined medical subject headings and generic terms in the literature search which were separated into 3 search strands by using “OR” to distinct the keywords. Only articles with Embase status were a part of the Embase search. The different search strands were: (1) puberty OR pubertal OR trend* OR timing, (2) “Tanner and Marshall” OR “breast size” OR “breast development” OR “breast growth” OR nipple OR areola OR thelarche OR B2 OR “breast stage” OR “breast stages,” and (3) girl* OR woman* OR female* OR child* OR adolescence*. The 3 search strands were combined with “AND,” which resulted in 1845 articles in PubMed and 1762 articles in Embase. After exclusion of duplicates, 1842 articles remained from the PubMed search and 1760 from the Embase search. The complete search specification is provided in Figure 1 and the exclusion reasons have been detailed in the eAppendix in the Supplement. There were no criteria regarding the specific time period of data collection in each study. Furthermore, we searched the reference lists of all key articles, and consulted experts in the field. However, this approach did not result in the capture of any additional articles. No attempt was made to retrieve articles from unpublished literature.

Inclusion eligibility criteria for the systematic review were:

1. Types of studies eligible were original cross-sectional and longitudinal studies. In addition, the healthy control group from original case-control studies (eg, girls with obesity vs controls or girls with chronic disease vs controls) were eligible for inclusion if requirements 2, 3, and 4 were fulfilled.

2. Types of study participants eligible were healthy girls with no report of any disease that could interfere with pubertal onset.

3. When considering methods of pubertal evaluation, a distinction between glandular breast tissue and fat tissue can be problematic, especially in girls with obesity.⁹ Physical or clinical examination by trained pediatricians with validation of pubertal staging assessment was therefore an inclusion criterion. Furthermore, the trained pediatricians had to use the 5 Tanner stages,³ and self-assessment was an exclusion criterion, owing to the risk of misclassification of fat tissue as being glandular tissue.⁹

4. When considering data reporting and statistical methods used, to enable evaluation of the secular change in age at thelarche, either the reported mean (SEM) or median (with 95% CI) age at onset of thelarche (Tanner breast stage 2) had to be reported. If the SEM was not reported, it was possible to calculate SEM if the 95% CI was reported. Because of the importance of SEM or 95% CI, the statistical methods used to calculate the age at onset of Tanner breast stage 2 had to be either probit analysis, logistic analysis when the data were cross-sectional, or interval censored analysis when the study was longitudinal, as Tanner stage is known to change between 2 examinations. This criterion was evaluated in the qualitative synthesis. In the absence of a calculated SEM or 95% CI in the eligible published article, the authors were contacted directly for information about SEM or 95% CI.

Criteria for exclusion from the systematic review were (1) study groups with family history of a disease that could be a potential influencing factor in association with onset of puberty, (2) studies of primarily severely malnourished children or those with pathologic obesity because body mass index (BMI) plays an important role in the onset of puberty, and (3) studies examining adopted girls.

To evaluate the quality of the articles included in our review, we used a modified version of the completeness of reporting scoring system by Bonzini et al.¹⁰ The completeness of reporting was evaluated using the following 3 criteria: (1) study design, (2) main characteristics of the study population, and (3) methods of statistical analysis. Two of us (C.E.-L. and A.S.B.) independently assigned scores to the articles and later met to compare, discuss, and resolve any discrepancies regarding their independent evaluation of each individual article.

Descriptive information (Table)^11-48 was recorded from each publication on the primary outcomes: age (SEM or 95% CI) of Tanner breast stage 2 onset and the specific examination year period. Secondary outcomes, if provided in the article, were: sample size, study type, BMI, country and area, socioeconomic background, race/ethnicity, explicit description of palpation (yes or no), response rate, and sampling frame. For each study, the year of examination was defined as the median year of the study period reported by the authors.

We used weighted regression analysis to evaluate the change in age at onset of Tanner breast stage 2, based on 30 included studies. P < .05 was considered statistically significant. In studies exclusively reporting 95% CIs, this value was converted to SEM, using standard methods, by subtracting the age at onset of Tanner breast stage 2 from the upper 95% CI (CI_U) limit divided by 1.96 (SEM = [(CI_U − age at B2)/1.96]). The SEM, mean age at onset of Tanner breast stage 2, and the mean examination year were all included in the weighted regression analysis.

Metaregression analyses (weighted regression analyses)^49,50 were performed using a random-effects model. The model allows for within-study variation and between-study variation. The model is Y_i = a_i + b × t_i + ε_i, where Y_i is the estimated time at entry to Tanner breast stage 2 in the study and a_i is a random effect accounting for the between-study variation that is the result of geographical variations in time at entry into Tanner breast stage 2 and differences in confounding factors influencing the time to Tanner breast stage 2 onset (eg, BMI). The variable t_i is the mean examination time of each study and ε_i is the SE of the estimated time at entry to Tanner breast stage 2 in each study. The parameter of interest is b, which is the association of calendar year with time to Tanner breast stage 2. The estimation was carried out using the metafor package⁵¹ in R (R Project for Statistical Computing).⁵² Analyses were repeated, stratified by region. Finally, in sensitivity analyses we repeated the analyses based on the 18 studies in which thelarche staging was based on clinical examination in which palpation was directly specified.

Our systematic electronic database search yielded 1760 unduplicated studies from Embase and 1842 from PubMed. Two of us (C.E.-L. and A.S.B.) scanned titles and abstracts independently to assess eligibility and retrieved 102 articles from the Embase search and 162 articles from the PubMed search for full-text reading. Among these, 74 articles appeared in both the PubMed search and the Embase search and were therefore excluded from the Embase search list. Furthermore, 122 studies did not fulfill the other eligibility criteria (87 of the 122 excluded had no information of age at onset of Tanner breast stage 2 and 35 were excluded for other reasons). The exclusion reasons, including specific study references, have been detailed in the eAppendix in the Supplement. This process left 68 for the qualitative synthesis, of which 30 studies were excluded as they did not use the required statistical method for calculation of the mean age at onset of Tanner breast stage 2. The quantitative analysis was based on the remaining 38 articles, of which 30 were included in the metaregression analyses (clinical examination with palpation implied [n = 12] or clinical examination with palpation specified [n = 18]). The 8 studies not included in meta-analyses were excluded as they did not report an SE or 95% CI. The main characteristics of all 38 included studies are listed in the Table.^11-48 Most of the included studies were cross-sectional (30 [79%]) with a random selection from schools (22 [58%]), hospitals (6 [16%]), and the general population (4 [11%]). Sample sizes ranged from 138 to 20 654 and studies were reported between 1977 and 2013. The studies represented populations from the whole world, but most were from Europe or North America (Table).^11-48

Median ages at Tanner Breast stage 2 (B2+) ranged from 9.8 to 10.8 years in Europe, from 9.7 to 10.3 years in the Middle East, from 8.9 to 11.5 years in Asia, from 8.8 to 10.3 years in the United States, and from 10.1 to 13.2 years in Africa (Table).^11-48

Results of meta-analyses on the included data (30 studies) showed a significant overall decrease in age at onset of Tanner breast stage 2 by 0.24 years per decade (95% CI, −0.44 to −0.04; P = .02), equivalent to almost 3 months per decade within this period (Figure 2). In sensitivity analyses restricted to the studies in which palpation of the breast tissue was specifically described in detail (n = 18), the downward trend was confirmed. There was an overall significant decrease in age at onset of Tanner breast stage 2 by 0.26 years per decade (95% CI, −0.49 to −0.02; P = .03), equivalent to more than 3 months per decade within this period (eFigure 1 in the Supplement). Specific analyses according to each region separately were limited by low statistical power and results were not significant (eFigure 2 in the Supplement).

To our knowledge, this is the first systematic review with meta-analysis that has evaluated worldwide secular changes in the timing of pubertal onset based on age at thelarche rather than menarche. Meta-analyses based on 30 eligible studies suggest a significant downward trend of 0.24 years in age at thelarche onset per decade from 1977 to 2013.

These results are clinically significant for several reasons: assessment of breast development is widely used to distinguish between precocious and normally timed puberty, and secular changes of thelarche are therefore of great importance as new cutoff limits will affect clinical management of girls referred for evaluation of puberty disorders.⁵³ A Finnish study found that the number of girls presenting with premature thelarche constituted more than 33% of the girls presenting with precocious puberty.⁵³ This scenario represents a potential challenge because of the possibility of redundant evaluation of girls who experience earlier onset of puberty in accordance with the secular change. Recent diagnostic practice includes brain magnetic resonance imaging in girls younger than 8 years with pubertal signs, and biochemical evidence of central hypothalamus-pituitary-gonadal axis activation. However, a strict criterion of 8 years as the lower limit of normal puberty would imply that a large proportion of healthy girls would in fact need to undergo brain magnetic resonance imaging if they were referred for evaluation. The medical community needs current and relevant data to redefine “precocious puberty,” because the traditional definition may be outdated, at least in some regions of the world, and may lead to unnecessary brain magnetic resonance imaging in healthy girls.

In addition, changes in age at pubertal onset may serve as a sensitive indicator of environmental influences on human health. Different studies have already indicated potential mechanisms involved in early pubertal onset. It is known that age at menarche is associated with BMI; higher BMI correlates with earlier onset of menarche. Higher BMI also correlates with earlier breast gland development.⁵⁴ The ongoing global obesity epidemic⁵⁵ could partially explain the observed change in age at pubertal onset assessed as age at thelarche.⁵⁶ Another potential factor could be exposure to increasing amounts of endocrine-disrupting chemicals mimicking and/or antagonizing endogenous sex steroid action, synthesis, or degradation. A recently published study observed an association between peripubertal methylparaben concentration and earlier onset of breast gland development.⁵⁷ In addition, the banned but persistent chemicals such as DDT (dichlorodiphenyl-trichloroethane) and DDE (dichlorodiphenyldichloroethylene) have been associated with earlier age of puberty.⁵⁸ Researchers are faced with an overarching challenge: to evaluate the effects of the simultaneous exposure to hundreds or thousands of individual chemicals that may work in synergy and exert pronounced clinical adverse effects larger than the sum of the effect of each compound assessed individually, even when the individual compounds are present at concentrations below their biological activity (“cocktail effect”).^59,60 Further research regarding endocrine-disrupting chemicals is especially needed in association with pubertal onset. Monitoring changes in normal pubertal timing is of increasing importance because a younger age at pubertal onset may change the clinical standards for defining precocious puberty. Furthermore, screening programs identifying early puberty can potentially lead to preventive health measures because early pubertal onset has been associated with deleterious long-term health outcomes.

The strength of this review is primarily that it systematically, to our knowledge, includes all published evidence from 1977 to 2019 fulfilling predefined criteria using a systematic and transparent search of the literature. Reliable assessment of pubertal staging is essential for the correct assessment of change in onset of any given stage. The criterion standard is clinical assessment including inspection and/or palpation by trained pediatricians with validation of assessment. Thus, the decision to limit the review to original studies of healthy girls in whom assessment was performed by a trained person and not by self-assessment is considered of major strength and important to counteract biased findings because of differential recall often introduced by self-reports or interview information. Information recall bias (outcome ascertainment) was thus nullified in this review as we excluded articles in which breast development was based on the recall of the girls in questionnaires or by interview.

This study also has several potential limitations. We assessed all articles for major potentially confounding factors and excluded studies reporting puberty in girls with a majority of malnutrition or obesity, foreign adoption, or with a family history of disease. We cannot account for information that was not reported or unknown. For example, many studies did not evaluate BMI and we cannot exclude that BMI distribution was skewed within a study.

Second, our systematic review resulted in the inclusion of many relatively small studies (35% with <500 girls). This could potentially affect precision, as smaller studies have less power to reflect the general population when compared with large studies.

Third, 8 of the eligible studies were excluded from the meta-analyses based on the lack of a reported SE of the mean or median (SEM) and 95% CI. Fourth, all studies were identified using a search strand in PubMed and Embase. Experts were contacted and references lists of key articles were searched, which could be a potential limitation, but no additional references were captured. No attempt was made to retrieve articles from unpublished literature.

Before the study selection process began, different potential biases were identified. The main bias regarding the calculation of the age at onset of Tanner breast stage 2 was associated with the method used for the pubertal staging. Palpation and/or inspection of the breast tissue by a trained examiner (and not self-assessment of pubertal stage) was defined as an inclusion criterion owing to the risk of misclassification by self-assessment or by untrained persons.

Interrater variation in palpation and inspection of Tanner staging could also be a potential bias. It is very unlikely that the different clinically trained examiners throughout the included studies performed the examinations in precisely the same way. This variation is therefore included in the random effect in the statistical model and does not result in overestimation or underestimation of the secular change.

To confirm whether palpation implied in the clinical or physical examination (n = 18) led to variation in our results, we performed a sensitivity analyses based on studies describing palpation only. The results of these analyses confirm a significant downward trend in age at onset of thelarche, validating our analyses.

Abnormal BMI in most of the study population or presence of disease that could interfere with timing of puberty were also potential biases that could interfere with the onset of puberty. These features were defined a priori as exclusion criteria. The decision to limit the review to studies of healthy girls with most within the normal BMI range in which thelarche staging was assessed by clinical examination is considered of major importance to counteract biased findings. Disease and extremes in BMI are known to interfere with pubertal onset.

Another potential bias is publication bias, and therefore there is a potential risk that studies showing unaltered timing of puberty (negative findings) may not be published. It was not feasible to assess publication bias, as we did not retrieve unpublished studies.

Because of the importance of the 95% CI or SEM, potential large studies or well-performed studies not reporting 95% CI or SEM have been excluded. This could be a potential bias associated with the evaluation of the potential secular change in age at onset of Tanner breast stage 2.

The quality of the evaluation of the potential secular change within specific countries can be compromised by not knowing the socioeconomic background or the BMI profile within the studies performed in the same country because these are known factors potentially associated with age at pubertal onset within and between countries.¹ Evaluation of the precise change in age at the onset of Tanner breast stage 2 within different countries is beyond the scope of our present review.

In this comprehensive systematic review with meta-analysis, we show for the first time, to our knowledge, that age at pubertal onset based on thelarche has decreased by almost 3 months per decade from 1977 to 2013. Geographical variations indicate earliest onset in the United States (8.8-10.3 years) and latest onset in Africa (10.1-13.2 years).

Accepted for Publication: November 26, 2019.

Corresponding Author: Camilla Eckert-Lind, MB, Department of Growth and Reproduction, Rigshospitalet, University of Copenhagen, Blegdamsvej 9, DK-2100 Copenhagen, Denmark (camilla.eckert-lind@regionh.dk).

Published Online: February 10, 2020. doi:10.1001/jamapediatrics.2019.5881

Author Contributions: Ms Eckert-Lind and Dr Petersen had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Eckert-Lind, Busch, Bräuner, Juul.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Eckert-Lind, Busch, Biro, Bräuner, Juul.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Eckert-Lind, Petersen.

Obtained funding: Eckert-Lind, Juul.

Administrative, technical, or material support: Biro, Juul.

Supervision: Busch, Petersen, Bräuner, Juul.

Conflict of Interest Disclosures: Ms Eckert-Lind and Dr Bräuner reported receiving grants from the Research Council at Rigshospitalet during the conduct of the study. No other disclosures were reported.

Funding/Support: During the preparation of this systematic review, Ms Eckert-Lind was supported by a scholarship from the Research Council at Rigshospitalet.

Role of the Funder/Sponsor: Rigshospitalet had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.