Early menarche and childbirth accelerate aging-related outcomes and age-related diseases: Evidence for antagonistic pleiotropy in humans

Reviewer #1 (Public review):

Summary:

The present study aims to associate reproduction with age-related disease as support of the antagonistic pleiotropy hypothesis of ageing predominantly using Mendelian Randomization. The authors found evidence that early-life reproductive succes is associated with advanced ageing.

Strengths:

Large sample size. Many analyses

Weaknesses:

Still a number of doubts with regard to some of the results and their interpretation.

Reviewer #2 (Public review):

Summary:

The authors present an interesting paper where they test the antagonistic pleiotropy theory. Based on this theory they hypothesize that genetic variants associated with later onset of age at menarche and age at first birth may have a positive effect on a multitude of health outcomes later in life, such as epigenetic aging and prevalence of chronic diseases. Using a mendelian randomization and colocalization approach, the authors show that SNPs associated with later age at menarche are associated with delayed aging measurements, such as slower epigenetic aging and reduced facial aging and a lower risk of chronic diseases, such as type 2 diabetes and hypertension. Moreover, they identify 128 fertility-related SNPs that associate with age-related outcomes and they identified BMI as a mediating factor for disease risk, discussing this finding in the context of evolutionary theory.

Strengths:

The major strength of this manuscript is that it addresses the antagonistic pleiotropy theory in aging. Aging theories are not frequently empirically tested although this is highly necessary. The work is therefore relevant for the aging field as well as beyond this field, as the antagonistic pleiotropy theory addresses the link between fitness (early life health and reproduction) and aging.

The authors addressed the remarks on the previous version very well. Addressing the two points below would further increase the quality of the manuscript.

(1) In the previous version the authors mentioned that their results are also consistent with the disposable soma theory: "These results are also consistent with the disposable soma theory that suggests aging as an outcome tradeoff between an organism's investment in reproduction and somatic maintenance and repair."

Although the antagonistic pleiotropy and disposable soma theories describe different mechanisms, both provide frameworks for understanding how genes linked to fertility influence health. The antagonistic pleiotropy theory posits that genes enhancing fertility early in life may have detrimental effects later. In contrast, the disposable soma theory suggests that energy allocation involves a trade-off, where investment in fertility comes at the expense of somatic maintenance, potentially leading to poorer health in later life.

To strengthen the manuscript, a discussion section should be added to clarify the overlap and distinctions between these two evolutionary theories and suggest directions for future research in disentangling their specific mechanisms.

(2) In response to the question why the authors did not include age at menopause in addition to the already included age at first child and age at menarche the following explanation was provided: "Our manuscript focuses on the antagonistic pleiotropy theory, which posits that inherent trade-off in natural selection, where genes beneficial for early survival and reproduction (like menarche and childbirth) may have costly consequences later. So, we only included age at menarche and age at first childbirth as exposures in our research."

It remains, however, unclear why genes beneficial for early survival and reproduction would be reflected only in age at menarche and age at first childbirth, but not in age at menopause. While age at menarche marks the onset of fertility, age at menopause signifies its end. Since evolutionary selection acts directly until reproduction is no longer possible (though indirect evolutionary pressures persist beyond this point), the inclusion of additional fertility-related measures could have strengthened the analysis. A more detailed justification for focusing exclusively on age at menarche and first childbirth would enhance the clarity and rigor of the manuscript.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Public review):

Summary:

The present study aims to associate reproduction with age-related disease as support of the antagonistic pleiotropy hypothesis of ageing, predominantly using Mendelian Randomization. The authors found evidence that early-life reproductive success is associated with advanced ageing.

Strengths:

Large sample size. Many analyses.

Weaknesses:

There are some errors in the methodology, that require revisions.

In particular, the main conclusions drawn by the authors refer to the Mendelian Randomization analyses. However, the authors made a few errors here that need to be reconsidered:

(1) Many of the outcomes investigated by the authors are continuous outcomes, while the authors report odds ratios. This is not correct and should be revised.

Thank you for your observation. We have revised the manuscript to ensure that the results for continuous outcomes are appropriately reported using beta coefficients, which indicate the change in the outcome per unit increase in exposure. This will accurately reflect the nature of the analysis and provide a clearer interpretation of continuous outcomes (lines 56-109).

(2) Some of the odds ratios (for example the one for osteoporosis) are really small, while still reaching the level of statistical significance. After some checking, I found the GWAS data used to generate these MR estimates were processed by the program BOLT-LLM. This program is a linear mixed model program, which requires the transformation of the beta estimates to be useful for dichotomous outcomes. The authors should check the manual of BOLT-LLM and recalculate the beta estimates of the SNP-outcome associations prior to the Mendelian Randomization analyses. This should be checked for all outcomes as it doesn't apply to all.

Thank you for your detailed feedback. We have reviewed all the GWAS data used in our MR analyses and confirmed that all GWAS of continuous traits have already been processed using the BOLT-LMM, including age at menarche, age at first birth, BMI, frailty index, father's age at death, mother's age at death, DNA methylation GrimAge acceleration, age at menopause, eye age, and facial aging. Most of the dichotomous outcomes have not been processed by BOLT-LMM, including late-onset Alzheimer's disease, type 2 diabetes, chronic heart failure, essential hypertension, cirrhosis, chronic kidney disease, early onset chronic obstructive pulmonary disease, breast cancer, ovarian cancer, endometrial cancer, and cervical cancer, except osteoporosis. We have reprocessed the GWAS beta values of osteoporosis and re-conducted the MR analysis (lines 74-75; lines 366-373).

(3) The authors should follow the MR-Strobe guidelines for presentation.

Thank you for your suggestion to follow the MR-STROBE guidelines for the presentation of our study. We appreciate the importance of adhering to these standardized guidelines to ensure clarity and transparency in reporting Mendelian Randomization (MR) analyses. We confirm that the MR components of our research are structured and presented following the MR-STROBE checklist. In addition to the MR analyses, our study also integrates Colocalization analysis, Genetic correlation analysis, Ingenuity Pathway Analysis (IPA), and population validation to provide a more comprehensive understanding of the genetic and biological context. While these analyses are not strictly covered by MR-STROBE guidelines, they complement the MR results by offering additional validation and mechanistic insights.

We have structured our manuscript to separate these complementary analyses from the core MR results, maintaining alignment with MR-STROBE for the MR-specific components. The additional analyses are discussed in dedicated sections to highlight their unique contributions and avoid conflating them with the MR findings.

(4) The authors should report data in the text with a 95% confidence interval.

Thank you for your feedback. We have added the 95% confidence intervals for the reported data within the main text to enhance clarity and provide comprehensive context (lines 56-109). Additionally, the complete analysis data, including all detailed results, can be found in Table S3.

(5) The authors should consider correction for multiple testing

Thank you for your comment regarding the need to consider correction for multiple testing. We agree that correcting for multiple comparisons is an important step to control for the possibility of false-positive findings, particularly in studies involving large numbers of statistical tests. In our study, we carefully considered the issue of multiple testing and adopted the following approach:

Context of Multiple Testing: The tests we conducted were hypothesis-driven, focusing on specific relationships (e.g., genetic correlation, colocalization, and Mendelian Randomization). These analyses are based on priori hypotheses supported by existing literature or biological relevance.

Statistical Methods: Where applicable, we applied appropriate measures to account for multiple tests. For instance, in Mendelian Randomization, sensitivity analyses serve to validate the robustness of the results.

We believe that the methodology and corrections applied in our study appropriately address concerns about multiple testing, given the hypothesis-driven nature of our analyses and the rigorous steps taken to validate our findings. If you feel that additional corrections are required for specific parts of the analysis, we would be happy to further clarify or revise as needed.

Reviewer #2 (Public review):

Summary:

The authors present an interesting paper where they test the antagonistic pleiotropy theory. Based on this theory they hypothesize that genetic variants associated with later onset of age at menarche and age at first birth have a positive causal effect on a multitude of health outcomes later in life, such as epigenetic aging and prevalence of chronic diseases. Using a mendelian randomization and colocalization approach, the authors show that SNPs associated with later age at menarche are associated with delayed aging measurements, such as slower epigenetic aging and reduced facial aging, and a lower risk of chronic diseases, such as type 2 diabetes and hypertension. Moreover, they identified 128 fertility-related SNPs that are associated with age-related outcomes and they identified BMI as a mediating factor for disease risk, discussing this finding in the context of evolutionary theory.

Strengths:

The major strength of this manuscript is that it addresses the antagonistic pleiotropy theory in aging. Aging theories are not frequently empirically tested although this is highly necessary. The work is therefore relevant for the aging field as well as beyond this field, as the antagonistic pleiotropy theory addresses the link between fitness (early life health and reproduction) and aging.

Points that have to be clarified/addressed:

(1) The antagonistic pleiotropy is an evolutionary theory pointing to the possibility that mutations that are beneficial for fitness (early life health and reproduction) may be detrimental later in life. As it concerns an evolutionary process and the authors focus on contemporary data from a single generation, more context is necessary on how this theory is accurately testable. For example, why and how much natural variation is there for fitness outcomes in humans?

Thank you for these insightful questions. We appreciate the opportunity to clarify how we approach the testing of AP theory within a contemporary human cohort and address the evolutionary context and comparative considerations with the disposable soma theory.

We recognize that modern human populations experience selection pressures that differ from those in the past, which may affect how well certain genetic variants reflect historical fitness benefits. Nonetheless, the genetic variation present today still offers valuable insights into potential AP mechanisms through statistical associations in contemporary cohorts. We believe that AP can indeed be explored in current populations by examining genetic links between reproductive traits and age-related health outcomes. In our study, we investigate whether certain genetic variants linked to reproductive timing—such as age at menarche and age at first birth—also correlate with late-life health risks. By identifying SNPs associated with both early-life reproductive success and adverse aging outcomes, we aim to capture the evolutionary trade-offs that AP theory suggests.

Despite contemporary selection pressures that differ from historical conditions, there remains natural genetic variation in traits like reproductive timing and longevity in humans today. This diversity allows us to apply MR to test causal relationships between reproductive traits and aging outcomes, providing insights into potential AP mechanisms. Prior studies have demonstrated that reproductive behaviors exhibit significant heritability and have identified genetic loci associated with reproductive timing (1,2). This genetic variation facilitates causal inference in modern cohorts, despite environmental and healthcare advances that might modulate these associations (3). By leveraging genetic risk scores for reproductive timing, our study captures the necessary variability to assess potential AP effects, thus providing valuable insights into how evolutionary trade-offs may continue to influence human health outcomes.

How do genetic risk score distributions of the exposure data look like?

Thank you for your question. Our study is focused on Mendelian Randomization (MR) analysis, which aims to infer causal relationships between exposures and outcomes. While genetic risk scores (GRS) provide valuable insights at an individual level, they do not directly align with our study's objective, which is centered on population-level causal inference rather than individual-level genetic risk assessment. In MR, we use genetic variants as instrumental variables to determine the causal effect of an exposure on an outcome. GRS analysis typically focuses on summarizing an individual's risk based on multiple genetic variants, which is outside the scope of our current research. Therefore, we did not perform or analyze the distribution of genetic risk scores, as our primary goal was to understand broader causal relationships using established genetic instruments.

Also, how can the authors distinguish in their data between the antagonistic pleiotropy theory and the disposable soma theory, which considers a trade-off between investment in reproduction and somatic maintenance and can be used to derive similar hypotheses? There is just a very brief mention of the disposable soma theory in lines 196-198.

In our manuscript, we test AP theory specifically by examining genetic variants associated with reproductive timing and their association with age-related health risks in later life. MR and genetic risk scores allow us to assess these associations, directly testing the hypothesis that certain alleles enhancing reproductive success might have adverse effects on aging outcomes. This gene-centered approach aligns with AP’s premise of genetic trade-offs, enabling us to observe whether alleles associated with early-life reproductive traits correlate with increased risks of age-related diseases. Distinguishing from disposable soma theory, which would predict a general trade-off in energy allocation affecting somatic maintenance and not specific genetic effects, our data focuses on how certain alleles have differential impacts across life stages. Our findings thus support AP theory over disposable soma by highlighting the effects of specific genetic loci on both reproductive and aging phenotypes. However, future research could indeed explore the intersection of these theories, for example, by examining how resource allocation and genetic predispositions interact to influence longevity in various environmental contexts.

(2) The antagonistic pleiotropy theory, used to derive the hypothesis, does not necessarily distinguish between male and female fitness. Would the authors expect that their results extrapolate to males as well? And can they test that?

Emerging evidence suggests that early puberty in males is linked to adverse health outcomes, such as an increased risk of cardiovascular disease, type 2 diabetes, and hypertension in later life (4). A Mendelian randomization study also reported a genetic association between the timing of male puberty and reduced lifespan (5). These findings support the hypothesis that genetic variants associated with delayed reproductive timing in males might similarly confer health benefits or improved longevity, akin to the patterns observed in females. This would suggest that similar mechanisms of antagonistic pleiotropy could operate in males as well.

In our study, BMI was identified as a mediator between reproductive timing and disease risk. Given that BMI is a common risk factor for age-related diseases in both males and females (6-9), it is plausible that similar mechanisms involving BMI, reproductive timing, and disease risk could exist in males. This shared mediator points to the possibility that, while reproductive timelines may differ, the pathways through which these traits influence aging outcomes may be consistent across genders.

AP theory could potentially be tested in males, as the principles of the theory may extend to analogous reproductive traits in males, such as age at puberty and testosterone levels, which could similarly influence health outcomes later in life. However, as our current study focuses specifically on female reproductive traits, testing the AP theory in males is outside the scope of this work. We acknowledge the importance of exploring these mechanisms in males, and we hope that future research will address this by investigating male-specific reproductive traits and their relationship to aging and health outcomes.

(3) There is no statistical analyses section providing the exact equations that are tested. Hence it's not clear how many tests were performed and if correction for multiple testing is necessary. It is also not clear what type of analyses have been done and why they have been done. For example in the section starting at line 47, Odds Ratios are presented, indicating that logistic regression analyses have been performed. As it's not clear how the outcomes are defined (genotype or phenotype, cross-sectional or longitudinal, etc.) it's also not clear why logistic regression analysis was used for the analyses.

Thank you for your thoughtful comments regarding the statistical analyses and the clarification of methods and variables used in the study.

Statistical Analyses Section: We have included a detailed explanation of all statistical analyses in the Methods section (lines 291–408), specifying the rationale for the choice of methods, the variables analyzed, and their relationships. Additionally, we have provided the relevant equations or statistical models used where appropriate to ensure transparency.

Beta Values and Odds Ratios: In the Results section (starting at line 56), both Beta values and Odds Ratios are presented: Beta values were used for analyses of continuous outcomes to quantify the linear relationship between predictors and outcomes. Odds Ratios (ORs) were calculated for binary or categorical disease outcomes to describe the relative odds of an outcome given specific exposures or independent variables.

Validation and Regression Analyses: For further validation of the MR results, we conducted analyses using the UK Biobank dataset (starting at line 162). Logistic regression analysis was then employed for disease risk assessments involving categorical outcomes (e.g., diseased or not).

We hope that this clarifies the methods and their applicability to our study, as well as the rationale for the presentation of Beta values and Odds Ratios. If further details or refinements are required, we are happy to incorporate them.

(4) Mendelian Randomization is an important part of the analyses done in the manuscript. It is not clear to what extent the MR assumptions are met, how the assumptions were tested, and if/what sensitivity analyses are performed; e.g. reverse MR, biological knowledge of the studied traits, etc. Can the authors explain to what extent the genetic instruments represent their targets (applicable expression/protein levels) well?

Thank you for your insightful comments regarding the Mendelian Randomization (MR) analysis and the evaluation of its assumptions. Below, we provide additional clarification on how the MR assumptions were addressed, sensitivity analyses performed, and the representativeness of the genetic instruments (starting at line 314):

Relevance Assumption (Genetic instruments are associated with the exposure): “We identified single nucleotide polymorphisms (SNPs) associated with exposure datasets with p < 5 × 10^-8 (10,11). In this case, 249 SNPs and 67 SNPs were selected as eligible instrumental variables (IVs) for exposures of age at menarche and age at first birth, respectively. All selected SNPs for every exposure would be clumped to avoid the linkage disequilibrium (r² = 0.001 and kb = 10,000).” “During the harmonization process, we aligned the alleles to the human genome reference sequence and removed incompatible SNPs. Subsequent analyses were based on the merged exposure-outcome dataset. We calculated the F statistics to quantify the strength of IVs for each exposure with a threshold of F>10 (12).”

Independence Assumption (Genetic instruments are not associated with confounders, Genetic instruments affect the outcome only through the exposure): Then we identified whether there were potential confounders of IVs associated with the outcomes based on a database of human genotype-phenotype associations, PhenoScanner V2 (13,14) (http://www.phenoscanner.medschl.cam.ac.uk/), with a threshold of p < 1 × 10^-5. IVs associated with education, smoking, alcohol, activity, and other confounders related to outcomes would be excluded.

Sensitivity Analyses Performed: A pleiotropy test was used to check if the IVs influence the outcome through pathways other than the exposure of interest. A heterogeneity test was applied to ensure whether there is a variation in the causal effect estimates across different IVs. Significant heterogeneity test results indicate that some instruments are invalid or that the causal effect varies depending on the IVs used. MRPRESSO was applied to detect and correct potential outliers of IVs with NbDistribution = 10,000 and threshold p = 0.05. Outliers would be excluded for repeated analysis. The causal estimates were given as odds ratios (ORs) and 95% confidence intervals (CI). A leave-one-out analysis was conducted to ensure the robustness of the results by sequentially excluding each IV and confirming the direction and statistical significance of the remained remaining SNPs.

Supplemental post-GWAS analysis: Colocalization analysis (starting at line 356), Genetic correlation analysis (starting at line 366).

Our MR analysis adheres to the guidelines for causal inference in MR studies. By combining multiple sensitivity analyses and ensuring the quality of genetic instruments, we demonstrate that the results are robust and unlikely to be driven by confounding or pleiotropy.

(5) It is not clear what reference genome is used and if or what imputation panel is used. It is also not clear what QC steps are applied to the genotype data in order to construct the genetic instruments of MR.

Starting in line 314, the steps of SNPs selection were included in the Methods part. “We identified single nucleotide polymorphisms (SNPs) associated with exposure datasets with p < 5 × 10^-8 (10,11). In this case, 249 SNPs and 67 SNPs were selected as eligible instrumental variables (IVs) for exposures of age at menarche and age at first birth, respectively. All selected SNPs for every exposure would be clumped to avoid the linkage disequilibrium (r² = 0.001 and kb = 10,000). Then we identified whether there were potential confounders of IVs associated with the outcomes based on a database of human genotype-phenotype associations, PhenoScanner V2 (13,14) (http://www.phenoscanner.medschl.cam.ac.uk/), with a threshold of p < 1 × 10^-5. IVs associated with education, smoking, alcohol, activity, and other confounders related to outcomes would be excluded. During the harmonization process, we aligned the alleles to the human genome reference sequence and removed incompatible SNPs. Subsequent analyses were based on the merged exposure-outcome dataset. We calculated the F statistics to quantify the strength of IVs for each exposure with a threshold of F>10 (12). If the effect allele frequency (EAF) was missing in the primary dataset, EAF would be collected from dsSNP (https://www.ncbi.nlm.nih.gov/snp/) based on the population to calculate the F value.” The SNP numbers of exposures for each outcome and F statistics results were listed in supplemental table S2.

(6) A code availability statement is missing. It is understandable that data cannot always be shared, but code should be openly accessible.

We have added it to the manuscript (starting at line 410).

Reviewer #2 (Recommendations for the authors):

(1) The outcomes seem to be genotypes (lines 274-288). In MR, genotypes are used as an instrument, representing an exposure, which is then associated with an outcome that is typically observed and measured at a later moment in time than the predictors. If both exposure and outcome are genotypes it is not clear how this works in terms of causality; it would rather reflect a genetic correlation. One would expect the genotypes that function as instruments for the exposure to have a functional cascade of (age-related) effects, leading to an (age-related) outcome. From line 149 the outcomes seem to be phenotypes. Can the authors please clearly explain in each section what is analyzed, how the analyses were done, and why the analyses were done that way?

Thank you for your insightful comment. We understand the concern regarding the use of genotypes as both exposures and outcomes and the implications this has for interpreting causality versus genetic correlation. To clarify, in our study, the outcomes analyzed in the MR framework are indeed genotypes, starting from line 47. We use genotypes as instrumental variables for exposures, which are then linked to phenotypic outcomes observed at a later stage, in line with standard MR principles.

To improve the robustness of the MR results, we validated the genetic associations in the population with phenotype data from UK Biobank (lines 162-203), and the detailed methods were listed in lines 385-408.

(2) Overall, the English writing is good. However, some small errors slipped in. Please check the manuscript for small grammar mistakes like in sentences 10 (punctuation) and 33 (grammar).

Thank you for your feedback. We appreciate your careful review and attention to detail. We thoroughly rechecked the manuscript for any grammatical errors, including punctuation and sentence structure, especially in sentences 11 and 35 in revised manuscript, as suggested.

(3) There is currently no results and discussion section.

The manuscript was submitted as Short Reports article type with a combined Results and Discussion section. We have added the section title of Discussion.

(4) Why did the authors not include SNPs associated with age at menopausal onset? See for example: https://www.nature.com/articles/s41586-021-03779-7.

Thank you for your information. Our manuscript focuses on the antagonistic pleiotropy theory, which posits that inherent trade-off in natural selection, where genes beneficial for early survival and reproduction (like menarche and childbirth) may have costly consequences later. So, we only included age at menarche and age at first childbirth as exposures in our research.

(5) Can the authors include genetic correlations between menarche, age at first child, BMI, and preferably menopause?

Thank you for your suggestion. We acknowledge that including genetic correlations between age at menarche, age at first childbirth, BMI, and menopause can provide valuable context to our analysis. While our current MR study sets age at menarche and age at first childbirth as exposures and menopause as the outcome, and we have already included results that account for BMI-related SNPs before and after correction, we recognize the importance of assessing genetic correlations.

To address this, we calculated the genetic correlations between these traits to provide insight into their shared genetic architecture. This analysis helps clarify whether there is a significant genetic overlap between the two exposures and between exposure and outcome, which can inform and support the interpretation of our MR results. We appreciate your suggestion and include these calculations to enhance the robustness and comprehensiveness of our study. In the genetic correlations analysis, LDSC software was applied and the genetic correlation values for all pairwise comparisons among age at menarche, age at first birth, BMI, and age at menopause onset were calculated(15,16). The results are listed in Table S6.

(6) Line 39-40: that is not entirely true. There is also amounting evidence that socioeconomic factors cause earlier onset of menarche through stress-related mechanisms: https://doi.org/10.1016/j.annepidem.2010.08.006

Thank you so much for your information. We changed it to “Considering reproductive events are partly regulated by genetic factors that can manifest the physiological outcome later in life”.

(7) Why did the authors choose to work with studies derived from IEU Open GWAS? as it is often does not contain the most recent and relevant GWAS for a specific trait.

We chose to work with studies derived from the IEU Open GWAS database after careful consideration of several sources, including the GWAS Catalog database and recently published GWAS papers. Our selection criteria focused on publicly available GWAS with large sample sizes and a higher number of SNPs to ensure robust analysis. For specific traits such as late-onset Alzheimer's disease and eye aging, we used GWAS data published in scientific articles to ensure that our research reflects the latest findings in the field.

(1) Barban, N. et al. Genome-wide analysis identifies 12 loci influencing human reproductive behavior. Nat Genet 48, 1462-1472 (2016). https://doi.org:10.1038/ng.3698

(2) Tropf, F. C. et al. Hidden heritability due to heterogeneity across seven populations. Nat Hum Behav 1, 757-765 (2017). https://doi.org:10.1038/s41562-017-0195-1

(3) Stearns, S. C., Byars, S. G., Govindaraju, D. R. & Ewbank, D. Measuring selection in contemporary human populations. Nat Rev Genet 11, 611-622 (2010). https://doi.org:10.1038/nrg2831

(4) Day, F. R., Elks, C. E., Murray, A., Ong, K. K. & Perry, J. R. Puberty timing associated with diabetes, cardiovascular disease and also diverse health outcomes in men and women: the UK Biobank study. Sci Rep 5, 11208 (2015). https://doi.org:10.1038/srep11208

(5) Hollis, B. et al. Genomic analysis of male puberty timing highlights shared genetic basis with hair colour and lifespan. Nat Commun 11, 1536 (2020). https://doi.org:10.1038/s41467-020-14451-5

(6) Field, A. E. et al. Impact of overweight on the risk of developing common chronic diseases during a 10-year period. Arch Intern Med 161, 1581-1586 (2001). https://doi.org:10.1001/archinte.161.13.1581

(7) Singh, G. M. et al. The age-specific quantitative effects of metabolic risk factors on cardiovascular diseases and diabetes: a pooled analysis. PLoS One 8, e65174 (2013). https://doi.org:10.1371/journal.pone.0065174

(8) Kivimaki, M. et al. Obesity and risk of diseases associated with hallmarks of cellular ageing: a multicohort study. Lancet Healthy Longev 5, e454-e463 (2024). https://doi.org:10.1016/S2666-7568(24)00087-4

(9) Kivimaki, M. et al. Body-mass index and risk of obesity-related complex multimorbidity: an observational multicohort study. Lancet Diabetes Endocrinol 10, 253-263 (2022). https://doi.org:10.1016/S2213-8587(22)00033-X

(10) Savage, J. E. et al. Genome-wide association meta-analysis in 269,867 individuals identifies new genetic and functional links to intelligence. Nat Genet 50, 912-919 (2018). https://doi.org:10.1038/s41588-018-0152-6

(11) Gao, X. et al. The bidirectional causal relationships of insomnia with five major psychiatric disorders: A Mendelian randomization study. Eur Psychiatry 60, 79-85 (2019). https://doi.org:10.1016/j.eurpsy.2019.05.004

(12) Burgess, S., Small, D. S. & Thompson, S. G. A review of instrumental variable estimators for Mendelian randomization. Stat Methods Med Res 26, 2333-2355 (2017). https://doi.org:10.1177/0962280215597579

(13) Staley, J. R. et al. PhenoScanner: a database of human genotype-phenotype associations. Bioinformatics 32, 3207-3209 (2016). https://doi.org:10.1093/bioinformatics/btw373

(14) Kamat, M. A. et al. PhenoScanner V2: an expanded tool for searching human genotype-phenotype associations. Bioinformatics 35, 4851-4853 (2019). https://doi.org:10.1093/bioinformatics/btz469

(15) Bulik-Sullivan, B. et al. An atlas of genetic correlations across human diseases and traits. Nat Genet 47, 1236-1241 (2015). https://doi.org:10.1038/ng.3406

(16) Bulik-Sullivan, B. K. et al. LD Score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat Genet 47, 291-295 (2015). https://doi.org:10.1038/ng.3211