https://orcid.org/0000-0001-8145-8774
Correction
20 Nov 2024: Dang C, Li J, Yu X, Liu J, Liu P, et al. (2024) Correction: Association of birth weight, childhood obesity, and age at menarche with the risk of ovarian dysfunction: A mendelian randomization study. PLOS ONE 19(11): e0314370. https://doi.org/10.1371/journal.pone.0314370 View correction
Figures
Abstract
Background
Observational studies have revealed associations between birth weight, childhood obesity, age at menarche, and ovarian dysfunction. However, these studies are susceptible to unavoidable confounding factors, leading to ongoing debates regarding their conclusions and making causal relationships challenging to infer. In light of these challenges, Mendelian randomization was employed in this study to investigate the causal relationships between birth weight, childhood obesity, age at menarche, and ovarian dysfunction.
Methods
This study employed a two-sample Mendelian randomization approach using genetic variation as instrumental variables to investigate causal relationships. Genetic variation data were sourced from summary data of genome-wide association studies in European populations. Instrumental variables were selected based on the principles of Mendel’s three assumptions. The study utilized the inverse variance weighted method to assess the relationships between birth weight, childhood obesity, age at menarche, and ovarian dysfunction. Supplementary analyses were conducted using MR-Egger regression, the weighted median method, and the weighted median mode to complement the IVW results. Furthermore, the study conducted heterogeneity, horizontal pleiotropy, and sensitivity analyses to evaluate the robustness of the results.
Results
Based on the inverse variance weighted method, it was found that there exists a causal relationship between childhood obesity (OR = 1.378, 95% CI: 1.113∼1.705, p = 0.003), age at menarche (OR = 0.639, 95% CI: 0.468∼0.871, p = 0.005), and ovarian dysfunction, while no causal relationship was observed between birth weight and ovarian dysfunction. Heterogeneity tests, multiplicity tests, and leave-one-out sensitivity analyses did not detect any heterogeneity or multiplicity effects in the estimated impact of these three exposure factors on the risk of ovarian dysfunction.
Conclusions
This study represents the first evidence suggesting a potential causal relationship between childhood obesity, age at menarche, and ovarian dysfunction. Childhood obesity was found to increase the risk of ovarian dysfunction, while a later age at menarche was associated with a reduced risk of ovarian dysfunction.
Citation: Dang C, Li J, Yu X, Liu J, Liu P, Yang X (2024) Association of birth weight, childhood obesity, and age at menarche with the risk of ovarian dysfunction: A mendelian randomization study. PLoS ONE 19(7): e0306365. https://doi.org/10.1371/journal.pone.0306365
Editor: Rajakumar Anbazhagan, National Institute of Child Health and Human Development (NICHD), NIH, UNITED STATES OF AMERICA
Received: September 1, 2023; Accepted: June 14, 2024; Published: July 18, 2024
Copyright: © 2024 Dang et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All data are in the manuscript and supporting information files.
Funding: This work was supported by the National Natural Science Foundation of China (No.82104917), the Natural Science Foundation of Shandong Province (No.ZR2021MH079). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors declare no conflict of interest.
Introduction
Ovarian reserve refers to a woman’s capacity for the growth, development, and formation of potentially fertilizable oocyte cells within the ovaries. It encompasses both the quantity of eggs produced within a woman’s ovaries and the potential quality of those eggs. Ovarian dysfunction, on the other hand, refers to a condition where the number of eggs retained within a woman’s ovaries does not meet the threshold criteria, leading to a decline in her reproductive capacity [1]. Ovarian reserve and ovarian aging have become prominent areas of focus in current clinical practice and research.
There is a hypothesis suggesting that birth weight is influenced by genetic factors from both parents as well as the intrauterine environment [2]. Further research has indicated that girls born with small gestational age (SGA) tend to have a reduced ovulation rate during pregnancy [3], but limited sample sizes may introduce bias into these findings. Childhood obesity is a significant public health concern, with reported rates as high as 12.7% in the United States [4]. Previous clinical studies and meta-analyses have shown that adult obesity has a negative impact on ovarian function [5, 6]. However, there has been limited research investigating whether childhood obesity similarly affects ovarian function. Age at menarche can serve as a reflection of a female’s hormonal levels during puberty and is influenced by genetic, epigenetic, and environmental factors [7]. Some studies have suggested that age at menarche is not correlated with ovarian function [8], but the limited sample sizes and potential methodological flaws in traditional observational studies raise questions about these findings. Therefore, further research is needed to clarify the relationships between birth weight, childhood obesity, age at menarche, and ovarian dysfunction.
Mendelian randomization (MR) employs single nucleotide polymorphisms (SNPs) as instrumental variables (IVs) to infer causal relationships between exposures and outcomes. By adhering to the genetic principle of "random allocation of parental alleles to offspring", it achieves a quasi-randomized grouping effect similar to that of a randomized controlled trial, while mitigating the influence of external environmental factors. This allows MR to address the limitations of observational studies [9]. In this study, a two-sample Mendelian randomization approach was employed. Leveraging publicly available large-scale Genome-Wide Association Study (GWAS) databases, it aimed to investigate the causal relationships between birth weight, childhood obesity, age at menarche, and ovarian dysfunction.
Material and methods
Research design
In determining where the GWAS summary data for the exposure variable and the outcome variable are assumed to be mutually independent, we conducted association analyses using the TwoSampleMR package in the R programming language. Specifically, we utilized birth weight, childhood obesity, and age at menarche as exposure variables, with ovarian dysfunction serving as the outcome variable. The objective was to investigate whether a causal effect exists between these exposure variables and the outcome variable, with the specific design outlined in Fig 1. MR analysis adheres to the following three key assumptions [10]:
- Instrumental variables have a strong correlation with the exposure variables (birth weight, childhood obesity, age at menarche).
- Instrumental variables are independent of both observed and unobserved confounding factors.
- Instrumental variables solely influence the outcome (ovarian dysfunction) through the exposure variables.
Data source
The summary data for birth weight were sourced from the IEU OpenGWAS project database (https://gwas.mrcieu.ac.uk/), encompassing 143,677 participants. Data for childhood obesity came from the Early Growth Genetics (EGG) Consortium (http://egg-consortium.org/), comprising 13,848 participants (5,530 cases and 8,318 controls) [11]. Age at menarche summary data was obtained from the ReproGen Consortium (https://reprogen.org/), involving 182,416 participants [12]. Ovarian dysfunction summary data were acquired from the Finngen database, with a total of 16,379,685 SNPs in 119,170 individuals (https://gwas.mrcieu.ac.uk/datasets/finn-b-E4_OVARDYS/). The exposure and outcome summary data were derived from two separate databases to prevent potential data sample overlap. The samples included both males and females and were sourced from European populations. As our study is based on publicly available databases, ethical committee approval was not required.
Instrumental variable selection
- IVs Selection: To obtain strongly associated exposure data, birth weight and age at menarche were screened with a significance threshold of P < 5 × 10^-8 [13]. However, under this genome-wide significance threshold, most of the childhood obesity-related SNPs failed to be identified. Therefore, for childhood obesity, a more lenient P-value threshold of P < 1 × 10^-6 was set to select associated SNPs. (2) Independence Criterion: The linkage disequilibrium (LD) among SNPs for each risk factor was assessed using PLINK clustering. SNPs with LD coefficient r^2 greater than 0.001 and located within a physical distance of less than 10,000 kb were excluded. This step ensured the mutual independence of SNPs to eliminate the influence of genetic pleiotropy on outcomes [14, 15]. (3) Statistical Strength Criteria: The effectiveness of instrumental variables was assessed using the F-statistic, calculated as F = β^2/SE^2 (where β represents the effect size of the allele and SE is the standard error). Instrumental variables with F-values less than 10 were excluded, ensuring that the instrumental variables were unrelated to unmeasured confounding factors [16]. Finally, the "harmonise_data" function from the Two Sample MR package was employed to align the directions of effect alleles for exposure and outcome variables, remove palindromic and incompatible SNPs, and consolidate P-values while excluding SNPs highly correlated with ovarian dysfunction (P < 5 × 10^-8) [17].
Mendelian randomization analysis
In this study, the inverse variance weighted (IVW) method [18] was employed as the primary analytical approach for establishing causal relationships. This method, assuming the validity of all instrumental variables, calculates weighted estimates by taking the reciprocal of their variances as weights. It provides the most accurate results when there is no heterogeneity or horizontal pleiotropy present. Additionally, MR-Egger regression, the weighted median (WME) method, and weighted mode (WM) were used as supplementary analyses to complement the IVW results.
Sensitivity analysis
Heterogeneity testing [19] assesses the presence of differences among various IVs. It utilizes the P-value from Cochran’s Q test to evaluate heterogeneity, with P > 0.05 indicating the absence of heterogeneity. If heterogeneity is detected, the MR pleiotropy residual sum and outlier (MR-PRESSO) test is employed to assess potential outliers [20], eliminate them, and then reanalyze the data. Multiplicity testing [21] verifies the reliability of MR analysis results. MR-Egger intercept is used to detect horizontal pleiotropy, with P > 0.05 indicating the absence of horizontal pleiotropy and, thus, the reliability of the MR analysis results. Sensitivity testing [22] is conducted using a "leave-one-out" approach, sequentially removing each SNP. If the MR results derived from the remaining SNPs do not exhibit significant differences from the overall result, it demonstrates the robustness of the MR results.
Results
Validity of instrumental variables
Following the selection process, the study obtained 2,264, 365, and 2,454 SNPs strongly correlated with birth weight, childhood obesity, and age at menarche, respectively. After pruning for linkage disequilibrium, these numbers were reduced to 50, 8, and 68 SNPs, respectively. Subsequently, palindrome and incompatible SNPs were identified and removed. Additionally, a search was conducted in the PhenoScanner database to identify any SNPs associated with potential confounding factors. In the final selection, a total of 42 SNPs for birth weight, 7 SNPs for childhood obesity, and 54 SNPs for age at menarche were included as instrumental variables. Moreover, all SNPs had F-statistics exceeding 10, indicating the effectiveness of the selected instrumental variables. Weak instrument bias was not expected to affect the causal inference results in this MR analysis. Basic information for some of the instrumental variables is provided in Table 1, while comprehensive details for all instrumental variables can be found in S1 Data.
Association of birth weight, childhood obesity, and age at menarche with the overall risk of PCOS
As both heterogeneity and pleiotropy tests yielded negative results, the IVW analysis results were considered the primary reference in this study. The MR analysis results indicated statistical significance with P-values of 0.003 for childhood obesity and 0.005 for age at menarche. These results suggest a causal relationship between childhood obesity, age at menarche, and the occurrence of ovarian dysfunction. Furthermore, childhood obesity showed a positive association with ovarian dysfunction (OR = 1.378, 95% CI: 1.113∼1.705), while age at menarche exhibited a negative association (OR = 0.639, 95% CI: 0.468∼0.871). These findings are illustrated in Figs 2 and 3. Additionally, there was no causal relationship between birth weight and ovarian dysfunction, with a P-value of 0.780, indicating a lack of statistical significance (P > 0.05).
Scatter plot for causal associations of (A) childhood obesity and (B) age at menarche with ovarian dysfunction risk based on four MR methods.
Sensitivity analysis of MR analysis
An examination of the heterogeneity revealed that the funnel plot for childhood obesity did not exhibit the typical symmetrical shape (Fig 4). To eliminate subjective factors, further analysis computed Cochran’s Q-values with a distribution set at 10,000. Cochran’s Q tests indicated the absence of heterogeneity among SNPs. Moreover, the results from the pleiotropy test showed that the MR-Egger regression intercept p-values were all greater than 0.05, indicating the absence of horizontal pleiotropy (Table 2). In the sensitivity analysis using the "leave-one-out" approach, the removal of any individual SNP did not significantly alter the IVW analysis results, and the remaining SNPs yielded results similar to those of the entire dataset (Fig 5). Additionally, the MR-PRESSO analysis did not detect any significantly influential SNPs, indicating the absence of SNPs with a substantial impact on causal associations.
Funnel plot of the MR analysis of (A) childhood obesity and (B) age at menarche on ovarian dysfunction.
Leave-one-out regression analysis of (A) childhood obesity and (B) age at menarche on ovarian dysfunction.
Discussion
Main findings and interpretation
This study employed a two-sample Mendelian randomization approach using GWAS data to investigate the causal relationships between birth weight, childhood obesity, age at menarche, and ovarian dysfunction. The results indicate that childhood obesity increases the risk of ovarian dysfunction in women (OR = 1.378, 95% CI: 1.113∼1.705), and an earlier age at menarche is associated with a greater risk of ovarian dysfunction (OR = 0.639, 95% CI: 0.468∼0.871). However, there is no causal relationship between birth weight and ovarian dysfunction.
Due to the original ovarian follicle pool being established within the uterus, it may be influenced by parental characteristics and the intrauterine environment [23]. Additionally, a study in 2007 [24] suggested that newborns with relatively lower birth weights have smaller ovarian volumes, raising the hypothesis that smaller ovarian volumes and ovarian dysfunction in females may originate during fetal development. However, a study in 2011 found no significant differences in follicular-phase LH, FSH, E2, AMH levels, or the response to endogenous GnRH between small for gestational age and appropriate for gestational age females [25]. This suggests that there may be no correlation between female ovarian function and birth weight. The contrasting results between these two studies could be attributed to the challenges of conducting extensive clinical follow-ups, resulting in limited sample sizes and potential biases influenced by various confounding factors.
A study conducted in 2014, which included 103 participants, found a higher prevalence of ovarian dysfunction among females who had insufficient weight during childhood [26]. Recent research has consistently indicated that childhood obesity leads to an earlier onset of puberty in females, increases the risk of developing polycystic ovary syndrome in adulthood, and affects fertility, among other factors [27, 28]. However, studies exploring the relationship between childhood obesity and adult ovarian function have been lacking. In contrast to previous research, our study discovered that females who were obese during childhood had a higher risk of developing ovarian dysfunction. This could potentially be attributed to the coordinated actions of various hormones such as leptin, insulin, and adrenaline [29]. The SNP rs9941349, located within the FTO gene intron, has been associated with abnormalities in FTO and linked to impaired oocyte maturation and premature ovarian failure. However, the precise mechanisms involved still require further elucidation through future research [30].
The earliest study from 2012 [31] found that women who experienced an earlier age at menarche had higher AMH levels during their youth, but it did not conduct longer-term research. In contrast, a study by Andrea Weghofer and colleagues in 2013 [32] suggested that an earlier age at menarche was associated with a higher risk of ovarian dysfunction. This association may be related to factors like follicular pool size and the speed of follicle recruitment. However, studies from 2018 and 2021 [8, 33] did not find any significant relationship between age at menarche and female ovarian function. The question of whether age at menarche is related to or causally linked to ovarian function in women is complex and may be influenced by various factors such as lifestyle and reproductive factors, making it challenging to provide a definitive answer through observational studies alone.
In contrast to previous observational studies, which can be susceptible to various confounding factors like genetics and immunity, and considering the limitations of conducting large-scale RCT clinical trials, this MR analysis study utilized openly accessible GWAS data. The results are transparent and publicly available. The selected data populations were exclusively of European descent, minimizing biases stemming from racial factors. This approach also aimed to reduce the inherent biases in observational studies that may arise due to residual confounding or reverse causality. Furthermore, it leverages a causal temporal framework [34]. To ensure the credibility of the study, heterogeneity tests were conducted. The P-values for both IVW and MR-Egger were greater than 0.05, indicating the absence of heterogeneity. Multiple-effect tests yielded P-values greater than 0.05, indicating that the results were not affected by multiple effects. Sensitivity analysis using the "leave-one-out" method also confirmed the stability and reliability of the findings. Our data support a causal relationship between childhood obesity, age at menarche, and ovarian dysfunction while not supporting a causal relationship between birth weight and ovarian dysfunction.
Limitation
However, our study has several limitations: (1) MR assumes a linear relationship between exposure and outcome. Therefore, it may not be applicable if the relationship between the two is nonlinear. (2) MR analysis can only explore causal relationships and cannot investigate specific biological mechanisms. (3) The outcome data used in the study come from European populations, and whether the results are representative of the entire population remains to be verified. (4) The data obtained lack more detailed cohort information such as age and gender, which limits the possibility of conducting subgroup analyses.
Conclusions
In summary, this study employed Mendelian randomization to analyze the causal relationships between birth weight, childhood obesity, age at menarche, and ovarian dysfunction. The results indicate that childhood obesity increases the risk of ovarian dysfunction, while a later age at menarche is associated with a reduced risk. This finding further confirms the positive correlation between age at menarche and ovarian function. Additionally, maintaining a healthy BMI during childhood is equally crucial for preserving ovarian function. Nevertheless, larger experimental studies are needed to elucidate the specific mechanisms through which childhood obesity and age at menarche impact ovarian function since ovarian dysfunction remains a disease with unclear etiology.
Supporting information
S1 Data. All SNP information of birth weight, childhood obesity and age at menarche.
https://doi.org/10.1371/journal.pone.0306365.s001
(XLSX)
Acknowledgments
This work benefited from the publicly available statistics of GWAS. We thank the contributors to the original GWAS database.
References
- 1. Al Rashid K, Taylor A, Lumsden MA, Goulding N, Lawlor DA, Nelson SM. Association of the functional ovarian reserve with serum metabolomic profiling by nuclear magnetic resonance spectroscopy: a cross-sectional study of ∼ 400 women. BMC Med. 2020;18(1):247. pmid:32862829
- 2. Lima MLS, Romão GS, Bettiol H, Barbieri MA, Ferriani RA, Navarro PA. Association of birth weight with functional ovarian reserve during menacme estimated by serum concentration of anti-Müllerian hormone. Sci Rep. 2019;9(1):8071. pmid:31147558
- 3. Ibáñez L, Potau N, Ferrer A, Rodriguez-Hierro F, Marcos MV, de Zegher F. Reduced ovulation rate in adolescent girls born small for gestational age. J Clin Endocrinol Metab. 2002;87(7):3391–3. pmid:12107255
- 4. Martin FZ, Fraser A, Zuccolo L. Alcohol Intake and Hypertensive Disorders of Pregnancy: A Negative Control Analysis in the ALSPAC Cohort. J Am Heart Assoc. 2022;11(19):e025102. pmid:36172928
- 5. Moslehi N, Shab-Bidar S, Ramezani Tehrani F, Mirmiran P, Azizi F. Is ovarian reserve associated with body mass index and obesity in reproductive aged women? A meta-analysis. Menopause. 2018;25(9):1046–1055. pmid:29738413
- 6. Grigoryan OR, Mikheev RK, Andreeva EN, Dedov II. Comparative analysis of ovarian reserve in women with obesity in reproductive period. Ter Arkh. 2018;90(10):30–34. pmid:30701792
- 7. La Vecchia C. Ovarian cancer: epidemiology and risk factors. Eur J Cancer Prev. 2017;26(1):55–62. pmid:26731563
- 8. Moreau J, Gatimel N, Cohade C, Parinaud J, Léandri R. Mother’s age at menopause but not own age at menarche has an impact on ovarian reserve. Gynecol Endocrinol. 2018;34(8):664–665. pmid:29345168
- 9. Carnegie R, Zheng J, Sallis HM, Jones HJ, Wade KH, Evans J, et al. Mendelian randomisation for nutritional psychiatry. Lancet Psychiatry. 2020;7(2):208–216. pmid:31759900
- 10. Lawlor DA, Harbord RM, Sterne JA, Timpson N, Davey Smith G. Mendelian randomization: using genes as instruments for making causal inferences in epidemiology. Stat Med. 2008;27(8):1133–63. pmid:17886233
- 11. Bradfield JP, Taal HR, Timpson NJ, Scherag A, Lecoeur C, Warrington NM, et al. A genome-wide association meta-analysis identifies new childhood obesity loci. Nat Genet. 2012;44(5):526–31. pmid:22484627
- 12. Xiang M, Wang Y, Gao Z, Wang J, Chen Q, Sun Z, et al. Exploring causal correlations between inflammatory cytokines and systemic lupus erythematosus: A Mendelian randomization. Front Immunol. 2023;13:985729. pmid:36741410
- 13. Perry JR, Day F, Elks CE, Sulem P, Thompson DJ, Ferreira T, et al. Parent-of-origin-specific allelic associations among 106 genomic loci for age at menarche. Nature. 2014;514(7520):92–97. pmid:25231870
- 14. Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird D, et al. The MR-Base platform supports systematic causal inference across the human phenome. Elife. 2018;7:e34408. pmid:29846171
- 15. Slatkin M. Linkage disequilibrium—understanding the evolutionary past and mapping the medical future. Nat Rev Genet. 2008;9(6):477–85. pmid:18427557
- 16. Zhuang Z, Yu C, Guo Y, Bian Z, Yang L, Millwood IY, et al. Metabolic Signatures of Genetically Elevated Vitamin D Among Chinese: Observational and Mendelian Randomization Study. J Clin Endocrinol Metab. 2021;106(8):e3249–e3260. pmid:33596318
- 17. Hartwig FP, Davies NM, Hemani G, Davey Smith G. Two-sample Mendelian randomization: avoiding the downsides of a powerful, widely applicable but potentially fallible technique. Int J Epidemiol. 2016;45(6):1717–1726. pmid:28338968
- 18. Wu F, Huang Y, Hu J, Shao Z. Mendelian randomization study of inflammatory bowel disease and bone mineral density. BMC Med. 2020;18(1):312. pmid:33167994
- 19. Bowden J, Spiller W, Del Greco M F, Sheehan N, Thompson J, Minelli C, et al. Improving the visualization, interpretation and analysis of two-sample summary data Mendelian randomization via the Radial plot and Radial regression. Int J Epidemiol. 2018;47(6):2100. pmid:30423109
- 20. Verbanck M, Chen CY, Neale B, Do R. Publisher Correction: Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet. 2018;50(8):1196. pmid:29967445
- 21. Carter AR, Sanderson E, Hammerton G, Richmond RC, Davey Smith G, Heron J, et al. Mendelian randomisation for mediation analysis: current methods and challenges for implementation. Eur J Epidemiol. 2021;36(5):465–478. pmid:33961203
- 22. Gronau QF, Wagenmakers EJ. Limitations of Bayesian Leave-One-Out Cross-Validation for Model Selection. Comput Brain Behav. 2019;2(1):1–11. pmid:30906917
- 23. Fraser A, McNally W, Sattar N, Anderson EL, Lashen H, Fleming R, et al. Prenatal exposures and anti-Mullerian hormone in female adolescents: the Avon Longitudinal Study of Parents and Children. Am J Epidemiol. 2013;178(9):1414–23. pmid:24008900
- 24. Orbak Z, Kantarci M, Yildirim ZK, Karaca L, Doneray H. Ovarian volume and uterine length in neonatal girls. J Pediatr Endocrinol Metab. 2007;20(3):397–403. pmid:17451078
- 25. Sadrzadeh-Broer S, Kuijper EA, Van Weissenbruch MM, Lambalk CB. Ovarian reserve in young women with low birth weight and normal puberty: a pilot case-control study. Gynecol Endocrinol. 2011;27(9):641–4. pmid:20807167
- 26. Chikvaidze N, Kristesashvili J, Gegechkori M. Peculiarities of sexual development and reproductive function in young women with childhood onset weight problems. Georgian Med News. 2014;(235):11–6. pmid:25416210
- 27. Huang A, Reinehr T, Roth CL. Connections Between Obesity and Puberty: Invited by Manuel Tena-Sempere, Cordoba. Curr Opin Endocr Metab Res. 2020;14:160–168. pmid:33043167
- 28. Heras V, Castellano JM, Fernandois D, Velasco I, Rodríguez-Vazquez E, Roa J, et al. Central Ceramide Signaling Mediates Obesity-Induced Precocious Puberty. Cell Metab. 2020;32(6):951–966.e8. pmid:33080217
- 29. Shalitin S, Kiess W. Putative Effects of Obesity on Linear Growth and Puberty. Horm Res Paediatr. 2017;88(1):101–110. pmid:28183093
- 30. Sun X, Zhang Y, Hu Y, An J, Li L, Wang Y, et al. Decreased expression of m6A demethylase FTO in ovarian aging. Arch Gynecol Obstet. 2021;303(5):1363–1369. pmid:33221958
- 31. Bragg JM, Kuzawa CW, Agustin SS, Banerjee MN, McDade TW. Age at menarche and parity are independently associated with Anti-Müllerian hormone, a marker of ovarian reserve, in Filipino young adult women. Am J Hum Biol. 2012;24(6):739–45. pmid:22915230
- 32. Weghofer A, Kim A, Barad DH, Gleicher N. Age at menarche: a predictor of diminished ovarian function? Fertil Steril. 2013;100(4):1039–43. pmid:23809497
- 33. Ayyavoo C, Baxi A, Mandrupkar G, Malik S, Kannan J, Goenka D, et al. MOTHER Study: A Multicenter Observational, Retrospective Study to Determine Coorelation Between Physical CHaracteristics and Ovarian REserve Markers in Sub-feRtile Women. J Obstet Gynaecol India. 2021;71(5):510–521. pmid:34602763
- 34. Ebrahim S, Davey Smith G. Mendelian randomization: can genetic epidemiology help redress the failures of observational epidemiology? Hum Genet. 2008;123(1):15–33. pmid:18038153