Age at menarche and chemical exposure: per- and polyfluoroalkyl substances (PFAS), dichloro-diphenyl-trichloroethane (DDT), dichloro-diphenyl-dichloroethylene (DDE), and polychlorinated biphenyls (PCBs)

Abstract Context Humans are now exposed to a multitude of chemicals throughout the life course, some of which may affect growth and development owing to their endocrine-like activity. Objective To assess the relationship of suspect toxicants to maturation, specifically to age at menarche. Methods We conducted two systematic reviews of age at menarche and PFOA, PFOS, PCBs and DDE/DDT based on publications indexed by pubmed. Results 16 unique reports were identified. Most studies of PFOA and PFOS reported either no association or delays in the age at menarche; only one reported an earlier age. Studies of DDT and DDE were more mixed. Reports on PCBs varied by PCB congener group with an equal number of them reporting delays and no association but one an acceleration. Sources of variation in results include the timing of exposure assessment (prenatal vs. postnatal), level of the toxicant, and sample size. No obvious pattern to the variation in results could be tied to those sources of variation. Conclusion The absence of consistent evidence from multiple reports of earlier age at menarche suggests that these toxicants may not be responsible for accelerated sexual maturation in girls. However, human populations naturally vary in the variety and levels of exposure, making the comparison of studies difficult. Further, studies vary in methodology, complicating aggregation of results and generalisations.


Introduction
Humanity has created modern environments containing many factors unknown to our species during its evolution towards our current form and functional systems.those new factors in the environment may be presenting adaptive challenges to human physiology and well-being and exist as possible examples of mismatch between our evolutionary history and our contemporary environment.
One method of evaluating the potential challenge of new environmental factors is through the study of growth and development, both prenatal and postnatal (Schell 1986).thus, when faced with a novel environmental factor and growth is reduced and/or development is delayed compared to appropriate expectations, we may conclude that the environmental factor is not beneficial to human health and well-being or that the adaptive capability is not adequate to negate the negative factor (Johnston 2012;tanner 1986).
Most human populations are exposed to a variety of chemicals in large or small quantities depending on the environments in which they have lived (Sexton et al. 2004).exposure occurs throughout the lifecycle in both the prenatal and postnatal periods and from a variety of sources (e.g., air and water pollution and/or ingestion of foods with insecticides and pesticides).in the US alone, many new chemicals are introduced every year.there were more than 60,000 substances in circulation after the toxic Substances control Act (tScA) was enacted in 1976, which increased to over 80,000 when the tScA was amended in 2016 (Bergeson 2019).Of these 80,000, many have not been tested for safety (Bergeson 2019).there is concern not only about the tested chemicals but the untested ones also.the Stockholm convention on Persistent Organic Pollutants (POPs) is a multilateral environmental agreement among over 190 nations to protect human health and the environment from persistent organic pollutants.
Per-and polyfluoroalkyl substances (PFAS) are a class of endocrine disrupting chemicals that have been used in industrial and consumer products since the 1940s but have been examined only more recently in the literature (AtSDR 2023).As they do not break down, PFAS remain in the environment and the blood of people and animals worldwide.while they are no longer in production, perfluorooctanoic acid (PFOA) and perfluorooctane sulphonic acid (PFOS) are still found in human populations, and are the two most commonly studied PFAS (AtSDR 2023).
A common concern regarding these POPs is that they can mimic hormones that regulate human growth, development, and reproduction as well as other functions.thus, many are considered endocrine disrupting chemicals (eDcs).changes in the pace of development can have detrimental effects on adolescent and adult health (copeland et al. 2010;Mendle et al. 2018).Numerous reports in the popular press have attributed changes to the pace of maturation, usually advances, to a variety of possible factors including eDcs.However, these reports have created more controversy than clarity.to determine if common eDcs can disrupt development, we have systematically reviewed published peer reviewed reports of tests of associations between specific eDcs and age at menarche.this review supplements one published earlier that encompassed several markers of maturation in both boys and girls but did not include the perfluorinated compounds that now are of great concern to environmental health scientists and those concerned with the modification of our biology by environmental factors.

Methods
this paper uses two systematic reviews to gather relevant peer-reviewed studies for assessment.the systematic review targeting studies assessing DDt, DDe, and PcBs is described elsewhere (west et al. 2021).this second systematic review targets papers assessing effects of per-and polyfluoroalkylnated substances (PFAS), specifically perfluorooctane sulphonate (PFOS) and perfluorooctanoic acid (PFOA).Specific criteria are used on the NcBi website's PubMed search engine (https://pubmed.ncbi.nlm.nih.gov/).within the advanced search box on the PubMed search engine, the following search term pattern was utilised: "term(a), " "term(b), " and "term(c)" (Figure 1) are used to obtain both the acronym and the full name of the pollutant being searched for."term (a)" is the full-term word, such as "perfluorooctane sulfonate, " "term(b1)" in this example would be "perfluorooctane" and "term(b2)" would be "sulfonate, " to capture the other permutations of the term.Further term(b) subterms would not be necessary in this case."term(c)" is used to capture the acronym of the toxicant, which in this example is PFOS."term(d)" is then split into multiple parts to capture the various permutations of the outcome phrase in all fields.the search phrases used are "sexual maturation, " "menarche, " "tanner stages, " "puberty, " "sexual development, " "pubertal development, " and "first menstruation." For example, "term(d)" would be "sexual maturation, " "term(d1)" would be "sexual, " and "term(d2)" would be "maturation." this technique is used to capture articles that refer to maturation with different terms.Articles published between 1 January 1997 and 14 July 2022 were searched.these search term combinations were used to cast the broadest possible search, capturing as many appropriate papers as possible for assessment based on specified inclusion criteria.
there were 199 papers identified through database searching, with nine additional records identified through other sources (Figure 2). in order to avoid double counting results from reviews of studies, we excluded review articles.After duplicates were removed, the remaining 98 records were screened for relevance.Papers deemed not relevant were animal studies, papers not in english, and any papers that did not assess the toxicants included in this second systematic search.to be included, exposure must have been measured at the individual level with biomarkers, and not based on proxy measurements such as distance from a source of toxicant exposure.Of the remaining 24 papers, 17 did not assess age at menarche and were excluded, leaving seven articles relevant to both PFOA and/or PFOS.All PFOA and PFOS levels were reported as medians (ng/ml).Samples' toxicant levels are reported as ng/g in figures for DDe/DDt and PcBs, but as ng/ml in figures for PFOA and PFOS.Some studies report either mean or median levels and the manner of presentation is noted in the figures.

Results
Basic descriptors of the 17 reports of 26 different analyses reviewed in this report are presented in table 1.Most studies are not of populations highly exposed to a specific pollutant and sampled individuals without regard to exposure.Only two studies were of populations outside of europe and North America.Studies varied in the number and type of covariates.the most common covariates were the child's age at assessment, maternal age at menarche, race/ethnicity, and BMi.ten analyses of the 26 included some measure of socio-economic position.

Per-and polyfluoroalkylnated substances (PFAS) compounds: PFOA and PFOS
Over a dozen different perfluorinated compounds have been tested for associations with health outcomes.two of these, PFOA and PFOS, are more commonly investigated than the others, and most often tested for association with age at menarche.this review found seven studies assessing age at menarche and PFOA exposure.two of three studies that assessed PFOA exposure postnatally (Figure 3) found delays; the third study found no association with age at menarche   lopez-espinosa et al. 2011;Di Nisio et al. 2020).those finding delays had far higher levels of measured PFOA than the one study finding no association.Only one of the four studies that assessed exposures prenatally found a delay with age at menarche.All studies assessing PFOA in the prenatal period had PFOA levels below 5 ng/ml (christensen et al. 2011;Kristensen et al. 2013;ernst et al. 2019;Marks et al. 2021). in short, none of the studies focused on PFOA found associations with earlier ages at menarche.control for other influences on age at menarche, particularly other toxicants, is absent in most studies.However, some have included other PFAS and/or eDcs in their multivariate models (Figure 4).including other PFAS in the multivariate model produces mixed effects.Delays are reported in two studies (Figure 4, panel "Multiple PFAS"), one with a large sample size and high levels of PFOA and the other with a far smaller sample size and far lower level of PFOA, while two other studies with PFAS in the model report no associations.(christensen et al. 2011;lopez-espinosa et al. 2011;Kristensen et al. 2013;carwile et al. 2021).One study that included both other PFAS and eDcs in the multivariate model found no association (Marks et al. 2021).this replicated results with the same sample but in which the multivariate model included only other PFAS (christensen et al. 2011) (Figure 5).
there are six studies assessing age at menarche and PFOS exposure.when stratifying studies of PFOS and age at menarche by timing of exposure assessment, one of the two studies with exposures assessed postnatally found a delay (lopez-espinosa et al. 2011; carwile et al. 2021) and this study had a far higher mean level of PFOS than the one study finding no association.when PFOS exposure was assessed prenatally, three of the four prenatal studies report no association with age at menarche while one reports an earlier age (christensen et al. 2011;Kristensen et al. 2013;ernst et al. 2019;Marks et al. 2021).All studies assessing levels in the prenatal period had levels comparable to the one postnatal study finding an association. the two studies finding an association, one measuring the toxicant in the prenatal period and the other in the postnatal one, report associations in opposite directions despite having comparable levels of exposure though very different sample sizes.Overall, four of the six studies report no association between age at menarche and PFOS.
comparing the studies of PFOS presented earlier but now in terms of multivariate adjustment for other PFASs or eDcs

DDT and DDE
this review found eight studies assessing age at menarche and DDe-t exposure.DDe and DDt are associated with early and late menarche depending on the time of exposure assessment (Figure 7).One study assessing DDe exposure prenatally with very high levels of DDe (reported only as a median level) but with a small sample, found an association with early menarche (vasiliu et al. 2004).the other study assessing exposure prenatally found no association with lower levels of DDe (Namulanda et al. 2016).three of four studies assessing exposure postnatally had lower levels and found no association with menarche while the fourth with a similarly low level found a delay (Denham et al. 2005  studies assessing DDt exposure, the one study assessing prenatal exposure found no association (Namulanda al. 2016).Of the two studies assessing DDt exposure postnatally, one found an association with earlier age at menarche and the other with later age (Ouyang et al. 2005;Bapayeva et al. 2016).Notably, of the studies assessing DDt in the postnatal period, one with high levels of DDt found a later age at menarche and the other with much lower levels found an earlier age (Ouyang et al. 2005;Bapayeva et al. 2016).Additionally, the study with postnatal DDt assessment reporting a delay studied a population with known high exposure to multiple organochlorines (lindane, dieldrin, and endrin), although not all organochlorines were tested in the analytic model (Bapayeva et al. 2016).

PCBs
Studies of exposure to PcBs are stratified by timing of toxicant assessment and by PcB congener type (non-dioxin-like  PcBs (NDlPcBs), dioxin-like PcBs (DlPcBs) and estrogen-like PcBs (e-PcBs)).two reports with NDlPcB exposure assessed postnatally find delays in menarche (Den Hond et al. 2011;Attfield et al. 2019).the two studies reporting DlPcB exposure, one assessing prenatal levels and the other postnatal ones, find no associations with menarche (Denham et al. 2005;leijs et al. 2008).Denham et al. (2005) also used a multivariate model to consider the effect of DlPcBs in the context of other eDcs, but still found no association with age at menarche.there are no reports of earlier age at menarche in association with either NDlPcB or DlPcB exposure.
However, one study tested a group of four PcB congeners with oestrogenic activity determined by laboratory studies (wolff and toniolo 1995) that did not fit either the NDlPcB or the DlPcB definitions (Denham et al. 2005).A significantly earlier age at menarche was associated with postnatally assessed levels (Figure 8).Further, this relationship was modelled with control for age, SeS, and p,p'-DDe, HcB, mirex, and mercury.
Studies of PcBs present mixed results that are related to the groups of congeners tested.this reinforces the conclusion that while all the tested PcB groups are in the category  of potential or proven disruptors, they not act a uniform manner to delay or to accelerate sexual maturation as indexed by age at menarche.the most uniformity is in delays associated with levels of NDlPcBs assessed in the postnatal period.However, the finding of an earlier age with correction for other eDcs as well as metals and sociodemographic characteristics is noteworthy also.

Discussion
this analysis of variation in age at menarche reveals that there are some consistencies in effects reported that are evident and worthy of further investigation.However, there are also differences in the direction of effects among studies of different toxicants and among studies of the same toxicant.Some of these differences may be due to variation in the timing of exposure assessment, the level of exposure and its reporting, route of exposure, cohort effects, sample sizes, the number of other eDcs included in the analyses, and the measurement of the outcome variable -age at menarche.
Of the eight analyses of six samples testing PFOA (Figure 4), four report a significant delay in age at menarche and none report an earlier age.the delays are associated with both prenatal and postnatal exposure (Figure 3).PFOA levels associated with delays are reported in analyses with very high PFOA levels and one with far lower levels.Similarly, variation in findings is not patterned by sample size.Of studies that included other PFASs in the multivariate model; two show a delay and two find no association (Figure 4).One sample was analysed twice, once with PFASs in the multivariate model and a second time with both PFASs and other eDcs without any change in findings (Figure 4;christensen et al. 2011;Marks et al. 2021).
there are two studies that found an association of PFOS levels with age at menarche, but the one with prenatal assessment found an earlier age and the other with postnatal assessment, a later one.Both samples have similar levels of PFOS (Figure 5). the one study finding earlier age at menarche does not include other PFASs or eDcs in the multivariate model while the one finding a delay included other PFASs (Figure 6). the effect on results of control for other PFASs and eDcs is not clear given the paucity of studies with multiple controls.
For studies testing associations of age at menarche with DDe and DDt levels, there are some consistencies.Of two studies assessing DDe levels in the prenatal period, one found no association while the other, exhibiting a high level of exposure, found an earlier age.Of the four studies assessing DDe levels in postnatal life, three reported no association and one a delay, all with low levels of exposure.Of the three studies assessing DDt exposure, one found no association, one found a delay, and one found earlier age at menarche.Studies of DDe and DDt reporting an earlier or a later age at menarche are distributed in both categories of exposure timing, and there are too few studies varying in timing of exposure assessment to discern a pattern of results.Further there is no pattern of toxicant levels to the results; high levels of DDe are associated with earlier menarche and high levels of DDt with other toxicants are associated with a later age at menarche (vasiliu et al. 2004;Bapayeva et al. 2016).
For studies reporting PcBs, differences in PcB structure are related to the effects on age at menarche.Both studies of NDlPcBs are associated with delays and assessed exposure in postnatal life.Both studies of DlPcBs, one with prenatal assessment and the other with postnatal assessment, found no association with age at menarche (Figure 8) and controlling for multiple eDcs did not change the results.However, when focussing only on the PcB congeners proven to be oestrogenic, those PcBs were associated with an earlier age menarche using a multivariate model for other toxicants and factors.
in many studies of these different toxicants, multivariate models were employed to adjust for relevant influences.A common covariate is BMi.Justification is usually on theoretical grounds as some studies have found associations between weight and/or fat with the pace of sexual maturation in girls (Bleil et al. 2021;Ortega et al. 2021).However, adipose tissue is also a storage organ for lipophilic toxicants.thus, higher BMi reflects a larger storage organ for past exposure.including BMi in the multivariate model may be adjusting for exposure which would be an overcorrection.Other influences may be considered also.However, in healthy, economically well-off populations, such as those used in the studies reviewed here, there are very few other environmental influences on age at menarche beyond weight or BMi.Severe neglect, as in the orphanages in communist Romania, has also been associated with delayed sexual maturation (Himes et al. 2009).Psychosocial stress, including father absence, trauma, and assault, has also been associated with age at menarche (Belsky 2019).None of the study samples were characterised with any of these conditions.the use of multivariate models to control for sociodemographic and anatomical characteristics such as BMi occurs in many of the studies reviewed (table 1).However, control for other toxicants is rare.Humans are usually exposed to multiple chemicals throughout the lifespan and carry this burden.Further, all of the toxicants considered in this review can cross the placenta and expose the foetus.Statistical approaches, or in the case of very large samples, stratified analyses, can reduce the problem of co-exposure.Few studies have employed such means to assess the influence of other toxicants when testing the one of greatest concern.therefore, it is difficult to be certain that the focal toxicant in the study is the one responsible for the effects observed.Further, given the contrasting influences of different toxicants (e.g.oestrogenic and antiestrogenic), a finding of no effect can be the result of counterbalancing effects.Measuring other toxicants beyond the focal one and including them in the analyses should produce more definitive results.
Despite the barriers to reaching firm conclusions, overall, there is evidence of changes in age at menarche in relation to chemical exposures, although the direction is not consistent.the inconsistency cannot be attributed simply to sample size or to level of the toxicant.considering these factors did not reduce the variation greatly or produce apparent replication of findings.However, differences in results were reduced by considering the toxicant itself and when it was assessed.NDlPcBs measured in postnatal life were consistently associated with delays.Neither study of DlPcBs, one with prenatal and the other with postnatal assessment, found an association with age at menarche.DDe and DDt associations did not appear patterned by timing of assessment or by toxicant level.Of nine samples of DDt and DDe, two found an earlier age at menarche, two found a delay and five found no association.Among the studies of PFASs, all but one found a delay or no association.Similarly, all studies of PcBs found no association or a delay except for the study that tested specific oestrogenic PcBs and controlled for other toxicants; it found an earlier age at menarche.evidence for the biological plausibility of toxicant exposure influencing the pace of sexual maturation comes from numerous laboratory studies with full experimental control.Animal studies were identified using the search criteria described in the methods, with the addition of the search terms "first estrous" and "vaginal opening." these have demonstrated that these toxicants do impact sexual maturation.Mice who were dosed with PFOA showed delays in vaginal opening (vO; lau et al. 2006; yang et al. 2009), a signal of the onset of puberty in rodent studies.Rats exposed to either PFOA or PFOS also had an earlier onset of vO (Du et al. 2019).the difference in the direction of effects may be due to known differences in pharmacokinetic profiles of rats and mice, but still exhibits an impact on maturation for both toxicants (lau et al. 2006).Higher doses of lactational DDe exposure in mice showed advanced vO compared to lower doses (yamasaki et al. 2009).Similarly, DDt is associated with early vO and early first oestrous in rats (Rasier et al. 2007).However, lindane, another organochlorine that frequently co-occurs with DDt and DDe exposure, has been shown to cause delays in vO and first oestrous (chadwick et al. 1988).this may explain the variation in direction of association within toxicant groups for populations that experience co-exposure.exposure to dioxin-like PcB-126 was associated with delays in vO in rats with postnatal exposure in multiple studies (Muto et al. 2003;Shirota et al. 2006).Additionally, a study assessing the effects of Aroclor 1254, a PcBs mixture, finds that vO and first oestrous were delayed in exposed rats (lee et al. 2007).thus, the variation in effects seen in studies of human populations with multiple exposures are biologically plausible even if not all in the same direction of effect.
this review may provide some clarification on sources of variation in the methodology that contributes to variation in the literature. in addition to variation in sample size, level of toxicant, type of toxicant, and time of toxicant assessment, there are other issues that could not be resolved in this review.One vexing problem is simple differences in reporting levels of the toxicant.Some studies of the same toxicant report medians and others means.Some studies report geometric means and some means without indicating whether they are either arithmetic or geometric.Some studies group toxicants existing in multiple forms differently, making results difficult to compare.Differences in the timing of assessment were reduced in this review to either prenatal or postnatal assessment, yet there is variation as to when in each period blood was drawn for toxicant assessment.Finally, there also are differences in how menarche is modelled.Some studies predict age at menarche while others compare early maturers versus later ones with an arbitrary definition of early and late that can vary by study.Future studies may analyse data and report results in ways that allow closer comparison of results as this will further our understanding of the multitude of chemicals that are now in our environment and in ourselves.
Based on the variety of results it is difficult to label toxicant exposure as a definite cause of earlier sexual maturation.importantly, menarche is a milestone of sexual maturation that appears very late in the process, usually between tanner stages 4 and 5 (tanner 1962).thus, if menarche were earlier, it would mean that sexual ends as menis not an indicator the start of sexual maturation.A different marker of maturation is needed to establish if sexual maturation begins earlier.

Figure 1 .
Figure 1.Advanced search format for the systematic literature review for Pubmed.

Figure 2 .
Figure 2. PRismA diagram from second systematic search.

(
Figure6), there are two with no adjustment, four with adjustment for multiple PFASs and only one with adjustment for both other PFASs and eDcs.without any adjustment, PFOS is associated with earlier menarche in one study and no association in another(ernst et al. 2019; Kristensen  et al. 2013).One study adjusting for PFASs only found later menarche(lopez-espinosa et al. 2011) and the one study adjusting for both PFASs and eDcs found no association(Marks et al. 2021).

Figure 3 .
Figure 3. PfoA and age at menarche by timing of exposure.(a) marks et al. (2021) uses the same sample as Christensen et al. (2011) while controlling for other EDCs.

Figure 4 .
Figure 4. PfoA and age at menarche by inclusion of other PfAs or EDCs.(a) marks et al. (2021) uses the same sample as Christensen et al. (2011) while controlling for other EDCs.

Figure 5 .
Figure 5. Pfos and age at menarche by timing of exposure.(a) marks et al. (2021) uses the same sample as Christensen et al. (2011) while controlling for other EDCs.

Figure 6 .
Figure 6.Pfos and age at menarche by inclusion of other PfAs or EDCs.(a) marks et al. (2021) uses the same sample as Christensen et al. (2011) while controlling for other EDCs.

Figure 7 .
Figure 7. DDE-T and age at menarche by timing of exposure (*the value is a median).

Figure 8 .
Figure 8. nDl and Dl PCBs and age at menarche by timing of exposure (* the value is a median).

Table 1 .
Description of studies analysed in the systematic reviews.
(Denham et al. 200511; croes et al. 2015;Attfield et al. 2019).Notably, only one of these studies assessed multi-toxicant exposure, and it found no association with DDe exposure(Denham et al. 2005).