Background
Advanced puberty timing has been subject to increasing interest and concern worldwide during recent years. There were evidences that early puberty was on the rise among girls in many parts of the world, such as Gambia [
1], America [
2‐
4], Western Europe [
5], and for boys were inconsistent [
3,
6,
7]. Early menarche in girls was a risk factor for the occurrence of morbid obesity, hypertension as well as breast and endometrial cancer [
8‐
11]. Recent study provided strong evidence that the younger girls were at menarche, the greater was their risk of premature and early menopause [
12]. Although there were less research in boys, a review suggests that early puberty was also a strong risk factor for detrimental psychosocial outcomes [
13]. Factors affecting early puberty can be categorized in two distinct ways: genetically determinant [
14,
15] and non-genetically determinant [
16,
17].
Smoking, in some populations, has been a widely spread non-genetic exposure, both during pregnancy and childhood. And cigarette smoking exposure included three forms: prenatal tobacco smoke (PTS), childhood environment tobacco smoke (ETS), and both PTS and ETS. In Australia, up to 80% of indigenous women smoke during pregnancy in some communities [
18]. Worldwide, almost half of children were exposed to ETS [
19]. Results on associations between three forms of smoking exposure and puberty timing from the past decade and more years were inconsistent. The study by Fukuda 2013 [
20] reported earlier menarche in relatively young daughters with PTS exposure. However, finding by Zhang 2014 [
21] showed that PTS had no effect on age of menarche of daughters. Kolasa 1998 [
22] reported an earlier age at menarche related to ETS exposure, while Shrestha 2011 [
23] found no association between age at menarche and ETS.
Two existing reviews studied the relationship between PTS exposure and puberty. Håkonsen 2014 [
24] studied on relationship between PTS and reproductive health of adolescent including pubertal development. This review qualitatively summarized the results without meta-analysis, and it concluded that results for girls were conflicting and the number of studies for boys was sparse. Yermachenko 2015 [
25] conducted a meta-analysis based on both cross-sectional studies and cohort studies to study the association between PTS exposure and age of menarche. It suggested that pregnancy smoking may decrease age at menarche.
In summary, evidences about relationship between PTS or ETS with puberty timing were inconsistent and have not been reviewed systematically so far; therefore, it is necessary to conduct a systematic review and meta-analysis to identify the associations between PTS and/or ETS and puberty timing in both girls and boys. This systematic review seeks to address this association and highlight where more research might be needed.
Methods
Selection criteria
Inclusion criteria: we included all the cohort studies examining relationship between PTS and/or ETS with puberty timing. In this study, since PTS exposure related to pregnant women, we determined “participants” as children, adolescents, and pregnant women; “exposures” as PTS and/or ETS; “control” as not exposing to PTS and/or ETS; “outcome measures” as the number of early puberty events and the age at puberty events.
Exclusion criteria: (1) not relevant to early puberty timing or precocious puberty; (2) other language except English and Chinese; (3) repetitive research (different articles published from the same study were considered as one study).
Search strategy
We searched publications from inception to May 2017 by an electronic search among seven databases including PubMed, ISI Web of science, OVID, EBSCO, VIP Database for Chinese Technical Periodicals, WangFang Data and Chinese National Knowledge Infrastructure databases, using both the MeSH terms and free terms “puberty” or “puberty timing” or “pubertal timing” or “pubertal development” or “precocious puberty” or “sexual precocity” or “sexual prematurity” or “sexual maturation” or “menarche” or “Tanner stages” or “thelarche” or “pubarche” or “spermarche” or “spermatorrhea” or “nocturnal emission” or “testis”, in combination with “maternal exposure” or “prenatal exposure” or “prenatal smoking” or “prenatal tobacco smoke” or “PTS” or “in utero exposure” or “passive smoking” or “environmental tobacco smoke” or “ETS.” All the retrieved publications were imported into reference-managing software (EndNote, version X7, Thomson Scientific, Stamford, CT, USA) to complete the duplicate check.
Data screening and extraction
Two reviewers (YC, WL) independently screened all the retrieved literatures by title, abstracts, and then full texts using above inclusion criteria. Cross-checking was implemented for accuracy, and differences were resolved by discussing with the third reviewer (QL) to reach an agreement.
Data extracted from included studies by using a pre-designed extraction form were as follows: (1) general information, including authors, publication year, research area; (2) study design; (3) participants characteristics, and sample size; (4) outcomes, mode, and level of tobacco exposure; (5) other factors affecting outcomes.
Risk of bias assessment
Two reviewers (YC, WL) independently evaluated the methodology quality of each eligible study according to a pre-established assessment form based on Newcastle-Ottawa Scale [
26], which distributed a score of total 9 points for each study following the criteria: 4 items for selection, 1 item for comparability, and 3 items for outcome assessment. In selection and outcome categories, at most one score can be awarded to a study for each item, but for comparability, two scores can be awarded. Studies were divided into three grades by total scores, including grade A (scored 7–9, high quality), grade B (scored 4–6, moderate quality), and grade C (scored 0–3, low quality) [
27]. Differences were resolved by consulting with a third reviewer (QL).
Statistical analysis
Meta-analysis was conducted by Comprehensive Meta-Analysis (Version 2.0, Biostat, Inc., USA). The outcomes of continuous and dichotomous variables were estimated as standardized mean difference (SMD) and relative risk (RR) with 95% confidence intervals (CIs), respectively. Heterogeneity among the results of the included studies was checked by chi-square based Q test and I2 test. When p < 0.05 or I2 > 50%, heterogeneity was considered and random effects model was used. Otherwise, Manter–Haenszal fixed-effects model was used. Results went for statistically significant when p value less than 0.05. We qualitatively described the main findings of included studies whose data cannot be extracted or which cannot be included in meta-analysis.
We conducted subgroup analysis based on exposure time (PTS and ETS), different PTS exposure levels in girls (1–9cigarettes/day, 10–19 cigarettes/day, and ≥ 20 cigarettes/day), cohort category (prospective cohort study and retrospective cohort study), and different definitions of early menarche. For stabilization of the results, we used the leave-one-out approach to conduct sensitivity analysis of all the outcome analyses. Since the amount of included studies did not reach the quantity requirement, we did not estimate the publication bias [
28].
Discussions
For all we know, this is the first systematic review and meta-analysis of the relationship between PTS and/or ETS exposure and puberty timing of both girls and boys. We found two reviews about PTS and puberty; one of which conducted a qualitative description, reporting a hypothesis of increasing risk of puberty onset in boys and girls with PTS exposure. The other meta-analysis about PTS and age of menarche in girls suggested that PTS exposure may accelerate onset of age of menarche. Findings in these two reviews were relatively coincident with our results of PTS and age of menarche in girls or voice break onset in boys. Comparing with the previous two reviews, our study has a wider scope of tobacco exposure mode (both PTS and ETS) and included both genders. Besides, not only age of menarche, but also number of girls with early puberty events, number of boys with early puberty events, age at various puberty events were analyzed in this study. In addition, studies conducted in both developing and developed countries were included in the review, involving Asian population.
In present study, we found that PTS was possibly and negatively associated with age of menarche in girls, which suggested that mothers smoking during pregnancy may accelerate the onset of menarche of daughters. But this result was not stable enough, when removing out Maisonet 2010 [
38], it turned to no association. Given the high heterogeneity, studies in PTS group were analyzed by cohort type in the subgroup analysis, and the heterogeneity dropped from 67.069 to 0.000% in prospective cohort subgroup, which suggested that study design may be one of the main sources of heterogeneity in PTS group. Prospective cohort study has a better demonstrated effectiveness which shows that the conclusion of early menarche with PTS is well founded.
While the results showed that there is no statistical significance on association between ETS and age of menarche, which may not necessarily mean that there is no relationship between them. As fewer studies reported ETS and puberty timing, the exact conclusion cannot be given exactly, which calls for more high-quality studies to confirm the relationship.
There was no strong evidence that PTS or ETS is associated with the number of girls with early menarche. The possible reason may be the differences in cut-offs for early menarche. There were four definitions for early menarche in the five studies [
32,
35,
37,
41,
48]: ≤ 12, < 12, ≤ 11, < 11 years respectively. When early menarche defined as ≤ 11 years, the number of girls with early menarche in PTS group was significantly more than that in control group, and the heterogeneity also decreased when we conducted subgroup analysis by definition of early menarche.
Results from three studies on the association between PTS and puberty development in boys showed consistent results. Fried 2001 [
39] reported an earlier age of voice break and shaving onset among boys exposed to PTS, which was based on a small cohort without confounder adjustment. The large retrospective study by Ravnborg 2011 [
46] showed that exposed boys experienced an earlier voice break without confounder adjustment. The latest prospective study by Håkonsen 2012 [
29] reported tendencies toward earlier age at acne, voice break, regular shaving, and first nocturnal emission, which indicated that boys exposed to PTS had earlier onset of puberty. However, no statistically significant differences were found with several important potential confounders adjusted. And no data provided can be used in meta-analysis in above studies. Therefore, the association between PTS exposure and puberty in boys could not be inferred. More studies reporting unified and complete outcome measures need to be conducted in boys to confirm the association.
The mechanism by which smoking exposure influence puberty timing is not clear enough yet. While 4000 chemicals contained over cigarettes, nicotine reduced blood flow to the placenta and fetus in pregnant smokers [
18], and heavy metal cadmium led a retardation of trophoblastic outgrowth and development of placental [
19]. Studies of animals and humans suggested that PTS exposure alter production of sex hormones and gonadotrophins [
22‐
24], which are all crucial chemicals in puberty onset. For females, PTS exposure may potentially impact primordial follicle number at puberty and uterine volume [
20]. And both PTS and ETS exposure to nicotine resulted in delayed ovarian dysfunction in adult female offspring [
21]. For males, several large studies reported that moderate or heavy smoking in pregnancy reduced the testis size and sperm count of male offspring in adult by 20–40% [
49,
50].
Studies included in the systematic review were conducted in four continents, namely, North America, Europe, Oceania, and Asia with multiethnic. To a certain extent, the results of this systematic review remind parents of the tobacco exposure effect on their children.
We carried out the research in strict accordance with the criterion of systematic review. However, there are still some limitations. First, though we tried to contact with the original authors to obtain necessary data of the included studies, six studies failed to obtain data required for meta-analysis. Among them, three [
29,
39,
46] reported different puberty events of boys, respectively, and three [
34,
45,
47] reported age of menarche did not provide sufficient data for meta-analysis. Failure to merge data above may lead to inaccuracy of results. Besides, one potential study in Czech was excluded, which may lead to selection bias.
Second, the assessment of methodology quality showed that scores of comparability and adequacy of follow-up of cohorts were relatively low; therefore, future research should pay more attention to these above aspects. Third, the main outcome measure of girls was menarche onset. Other outcomes cannot be conducted in meta-analysis due to few reports of tanner stage, inconsistent standard of division of staging and early onset. Meanwhile, reports of exposure levels were also limited and had different division standard, which made it difficult to find the dose-effect relationship.
Fourth, heterogeneity that comes from criteria of smoking exposure levels, race diversity, regional diversity, and sample size difference could not be analyzed in the current analysis. Fifth, few included studies reported the association of puberty development and PTS or ETS in boys, therefore need more studies to confirm this association. Finally, 4 out of 16 prospective cohort studies we included have not used the onset of outcome as the end of follow-up, so that children who have not occurred onset of puberty were excluded from the data analysis in these studies, which may lead to the loss of information and affect the accuracy of the results.