key: cord-0977138-hyvpz6m0
authors: Schooling, C. M.; Zhao, J. V.
title: Investigating the association of testosterone with survival in men and women using a Mendelian randomization study in the UK Biobank
date: 2021-07-07
journal: Sci Rep
DOI: 10.1038/s41598-021-93360-z
sha: c39aebb8b9f0b61dfada4a1956c1dfb38317abd5
doc_id: 977138
cord_uid: hyvpz6m0

Life expectancy in the developed West is currently stagnated and remains shorter in men than women. Well-established evolutionary biology theory suggests lifespan trades-off against reproductive success, possibly sex-specifically. We examined whether a key driver of reproductive success, testosterone, affected survival using a Mendelian randomization longevity study in the UK Biobank to obtain unbiased estimates, along with control exposures. We applied published genetic instruments for testosterone to obtain inverse variance weighted estimates of associations with survival to (i.e., age at) recruitment, in 167,020 men and 194,174 women. We similarly obtained estimates for a positive control (smoking initiation), and a negative control (absorbate), a marker of vitamin C metabolism. Testosterone was associated with poorer survival (0.10 years younger at recruitment per effect size of testosterone, 95% confidence interval (CI) 0.004 to 0.20). As expected, smoking initiation was also associated with poorer survival (0.37 years younger, 95% CI 0.25 to 0.50), but not absorbate (0.01 years younger, 95% CI − 0.09 to 0.11). Several aspects of a healthy lifestyle (low animal fat diet) and several widely used medications (statins, metformin, dexamethasone and possibly aspirin) may modulate testosterone. Explicitly designing interventions sex-specifically based on these insights might help address stagnating life expectancy and sexual disparities.

| (2021) 11:14039 | https://doi.org/10.1038/s41598-021-93360-z www.nature.com/scientificreports/ Here we conducted a Mendelian randomization (MR) study of the effects of testosterone on survival in men and women. MR studies take advantage of the random allocation of genetic material at conception to obtain less confounded estimates 27 . Specifically, MR studies use genetic variants as proxies of the exposure, and compare outcomes according to genetically predicted differences in exposure, i.e., use an instrumental variable design with genetic instruments. We used existing genetic predictors of testosterone and the control exposures [28] [29] [30] and applied them to a study of longevity. Longevity studies compare characteristics of survivors to older ages with characteristics of younger people because lifetime harmful factors inevitably become less common with increasing age 31 . In contrast, observational prospective cohort studies comparing mortality incidence after recruitment are open to selection bias from any deaths prior to recruitment having already depleted the susceptibles 32 which may attenuate estimates even to the extent of suggesting no benefit of proven healthy interventions 33 . As such, longevity studies, take advantage of the changing structure of risk factors with age to obtain estimates of effects on mortality free from selection bias 34 . In addition, as this is a hypothesis driven study, we also included positive and negative control exposures. Smoking initiation was used as a positive control outcome because smoking is very well-known to be addictive and to substantially reduce survival to old age 35 . A marker of vitamin C metabolism, absorbate, was used as a negative control outcome because vitamin C is no longer thought to affect health 36 .

In total 167,020 men and 194,174 women from the UK Biobank were included. Almost all the genetic variants used as instruments for the exposures (testosterone (125 variants in men and 254 variants in women), smoking initiation (361) and absorbate (1)) were available for the outcome. Two variants for smoking initiation were replaced by highly correlated (r 2 > 0.9) variants (rs2587507 replaced by rs745571, and rs2359180 by rs768576) obtained from LDlink (https:// ldlink. nci. nih. gov/). The average F-statistic for testosterone was 128.6 in men and 83.6 in women, and for smoking initiation was 44.9. As expected, smoking initiation was associated with younger age at recruitment (Table 1 ), corresponding to poorer survival. Similarly, as expected absorbate was unrelated to age at recruitment (Table 1) . Overall, testosterone was associated with younger age at recruitment, i.e., poorer survival (Table 1) , with slightly larger estimates in men than women (Table S1 ). Estimates were similar in sensitivity analysis using age at recruitment as a categorical variable (age 60 years or less) (Table S2 ) and using sensitivity analysis (Table S1 ).

Despite observational studies suggesting that testosterone might promote longer life [10] [11] [12] [13] [14] , our study with greater robustness to confounding and selection bias, consistent with theoretical expectations from evolutionary biology 24, 25 , suggests that testosterone reduces survival. Our study also showed smoking initiation reduced survival as would be expected 35 .

Our findings differ from previous observational studies which suggest that endogenous testosterone might improve health and lifespan [10] [11] [12] [13] [14] . However, these studies are open to both confounding by health status 18, 19 and selection bias from inevitably only recruiting survivors 32 . Our findings are more consistent with previous MR studies suggesting exogenous testosterone increases cardiovascular disease 37, 38 , and corresponding warnings from regulators (https:// www. fda. gov/ drugs/ drug-safety-and-avail abili ty/ fda-drug-safety-commu nicat ion-fdacauti ons-about-using-testo stero ne-produ cts-low-testo stero ne-due). Our findings are also consistent with the observation that in one of the few primate species with no age or sex difference in testosterone levels, there is also no sex difference in age-dependant survival.

Several potential mechanisms could underlie our findings. Experimental studies suggest testosterone increases coronary plaque volume 39 and coagulation 40 while impairing endothelial function [41] [42] [43] . Testosterone increases vulnerability to hormone related cancers, including prostate cancer 28 , breast cancer 28 and endometrial cancer 28 . Testosterone is also increasingly acknowledged to be an immune-suppressant 44 , potentially increasing susceptibility to cancer 45 , as well as to infectious diseases 44 , such as coronavirus disease 19. Testosterone may also induce impulsive behaviour 46 . Relevance to interventions. Given the importance of testosterone to reproductive success, testosterone is likely responsive to environmental indicators of suitable conditions for procreation, and hence a modifiable target of intervention. Serendipitously, many aspects of a healthy lifestyle and several widely used interventions reduce testosterone. A healthy diet, i.e., low-fat, high fiber, high soy, and low animal fat reduces testosterone in men 47-49 and women 50 . Exercise and weight loss reduce testosterone in women 51-53 , but possibly less so in men 54,55 , although severe caloric restriction (50% energy requirements) reduces testosterone in men 56 . One of the www.nature.com/scientificreports/ most effective medications for preventing and treating cardiovascular disease, i.e., statins, reduces testosterone 57 . The first-line treatment for diabetes, metformin, may reduce testosterone in women 58 , but not men 59 . Aspirin may also reduce testosterone 60 . Another essential medicine, dexamethasone reduces testosterone 61, 62 . As such, many current interventions to promote healthy aging do happen to reduce testosterone, particularly in women. Explicitly, searching sex-specifically for interventions based on this insight might facilitate the search for new means of promoting healthy aging by identifying at an early stage any potential interventions that are likely to be unsuccessful, and where interventions should be sex-specific.

Limitations. Although this study used an innovative approach to generate an unbiased assessment of the effect of testosterone on survival, it has several limitations. First, MR has stringent assumptions including that the genetic instruments predict the exposures, hence our use of published instruments [28] [29] [30] , that the instruments are independent of exposure outcome confounders, which stems from the use of genetic instruments not related to confounders such as health status, lifestyle and socioeconomic position, and that the instruments only affect the outcome via the exposure, hence our use of sensitivity analysis and control exposures. Second, we assumed a linear relation of survival with age when it is exponential. However, at the relatively young age (mean age 57 years, intended recruitment age 40 to 69 years) considered here such an approximation is less of an issue than in old age. A sensitivity analysis using survival to age 60 years as the outcome gave a similar interpretation for genetically predicted testosterone (Table S2) . Third, the samples used for exposures and outcome overlap. However, that would be expected to bias towards the null 63 , so our estimates of associations may be conservative. Moreover, the potential bias is less of concern as we did not use weak instruments to predict exposures 63 . Fourth, our study gives lifetime effects of endogenous testosterone up to age ~ 57 years rather than the effect of an intervention. However, the estimates for testosterone can be contextualized by the estimates for smoking. Fifth, longevity studies have been criticized for not taking account of cohort effects 34 . However, there is no reason to think that genetics are susceptible to cohort effects in the UK Biobank. Sixth, canalization, i.e., compensation for genetic effects might exist. However, this would likely bias towards the null. Seventh, the UK Biobank is not a population representative study. However, we used a study of longevity rather than a traditional prospective cohort study to avoid selection bias. Eighth, the relatively young age at recruitment, and correspondingly high survival means this study lacks power, so we focused on major determinants of healthy aging, but were not able to ascertain if the effects of testosterone were greater in men than women. Ninth, we were unable to replicate this study using a genome wide association study of life span or parental life span. Such studies typically do not include the X chromosome, which is important for testosterone. They also typically recruit in middle-age, i.e., recruit survivors, instead of taking account of those who have already died before recruitment, so are open to selection bias. Tenth, the effects of testosterone may vary by age. However, causes are generally expected to act consistently, although they may not be relevant in all situations 64 . Testosterone in men falls with age 20 , so these findings could be relatively less relevant to older men, but absolute risk of death increases with age.

Consistent with well-established evolutionary biology theory 24-26 testosterone appeared to reduce survival. Serendipitously, many existing strategies for promoting healthy aging reduce testosterone, particularly in women. Explicitly using an established theory as a guide might facilitate the search for new interventions, and draws attention to the importance of sex-specific interventions.

Data sources. The UK Biobank recruited half a million people intended to be aged 40 to 69 years from across Great Britain in 2006-10 65 . Average age is about 57 years and just over half the participants are women. We extracted overall and sex-specific genetic associations with age at recruitment from publicly available UK Biobank summary statistics in white British (http:// www. neale lab. is/ uk-bioba nk). These specific summary statistics were generated from linear regression adjusted for sex and the first 20 principal components, after excluding poor quality samples. Where relevant genetic information was not available from these summary statistics, we similarly generated corresponding information from the UK Biobank individual data.

We used published sex-specific genetic predictors of testosterone (125 variants for men and 254

for women) 28 . We used published genetic predictors of smoking initiation (361 variants) 29 , and of vitamin C metabolism (1 variant) 30 .

Outcome. In this study, we used age at recruitment to the UK Biobank as a proxy for survival, as there is no particular reason, apart from prior death (i.e., lack of survival), why genetics should generate differences in willingness to participate by age. So, harmful exposures would be expected to be associated with younger age at recruitment.

We obtained the F-statistics for the genetic instruments using an established approximation (square of genetic variant on exposure divided by its variance) 66 . As an F-statistic of 10 or less is usually taken as indicating a weak instrument. We aligned genetic associations on the same allele for exposures and outcome. We included palindromic and non-biallelic genetic variants because the genetic associations with exposures and outcome were obtained in the same study or have been used in the same study. We used proxies, as necessary. We meta-analyzed genetic variant specific Wald estimates (ratio of genetic variant on outcome to genetic variant on exposure) using inverse variance weighting (IVW) with multiplicative random effects. 67 . We used estimates with different assumptions as sensitivity analysis. The weighted median has a majority valid assumption and assumes > 50% of the genetic variants are valid instruments, while the contamination mixture method has a contrasting plurality valid assumption 67 . MR-Egger detects pleiotropic effects acting on the outcome other than via the exposure, but assumes that the genetic instruments do not act via confounders of exposure on outcome 66 . To avoid pleiotropic effects of sex hormone binding globulin, we used established genetic predictors of bio-available testosterone in men and total testosterone in women, as previously 68 . Bio-available testosterone may also be a more sensitive indicator that total testosterone of active testosterone in men 69 . Where necessary we combined estimates for men and women using IVW metaanalysis. Given this is a hypothesis driven study, with a positive and a negative control, we used a statistical significance level of 0.05. All statistical analysis was conducted using R version 3.6.1 (The R Foundation for Statistical Computing, Vienna, Austria). The MendelianRandomization R package was used for the MR estimates and the metafor R package for meta-analysis. Estimates of genetic associations were taken from publicly available UK Biobank summary statistics, except for the unavailable associations which were based on individual level genetic associations from the UK Biobank obtained under application #42468. All UK Biobank data was collected with fully informed consent. 

Sex differences in adult lifespan and aging rates of mortality across wild mammals

Women live longer than men even during severe famines and epidemics

Clinical advances in sex-and gender-informed medicine to improve the health of all: A review

Gender and longevity: Why do men die earlier than women? Comparative and experimental evidence

Could androgens be relevant to partly explain why men have lower life expectancy than women

Sex-biased reduction in reproductive success drives selective constraint on human genes

Is the long-term decline in cardiovascular-disease mortality in high-income countries over? Evidence from national vital statistics

Estrogen replacement therapy and coronary heart disease: A quantitative assessment of the epidemiologic evidence

Commentary: Hormone replacement therapy and coronary heart disease: Four lessons

Association of sex hormones with incident 10-year cardiovascular disease and mortality in women

Extreme concentrations of endogenous sex hormones, ischemic heart disease, and death in women

Low testosterone levels predict all-cause mortality and cardiovascular events in women: A prospective cohort study in German primary care patients

Clinical review: Endogenous testosterone and mortality in men: a systematic review and meta-analysis

Mortality is linked to low serum testosterone levels in younger and middle-aged men

The association of reproductive hormone levels and all-cause, cancer, and cardiovascular disease mortality in men

Sex hormones and cause-specific mortality in the male veterans: The Vietnam Experience Study

Androgens shorten the longevity of women: Sopranos last longer

Basal endogenous steroid hormones, sex hormone-binding globulin, physical fitness, and health risk factors in young adult men

Hormones and health outcomes in aging men

Age-specific reference ranges for serum testosterone and androstenedione concentrations in women measured by liquid chromatography-tandem mass spectrometry

Safety and efficacy of LibiGel® for treatment of hypoactive sexual desire disorder in postmenopausal women (BLOOM)

Adverse events associated with testosterone administration

A study to evaluate the effect of testosterone replacement therapy (TRT) on the incidence of major adverse cardiovascular events MACE) and efficacy measures in hypogonadal men (TRAVERSE)

Life history evolution, reproduction, and the origins of sex-dependent aging and longevity

Issues in evolutionary medicine

Evolutionary public health: Introducing the concept

Mendelian randomization": Can genetic epidemiology contribute to understanding environmental determinants of disease?

Using human genetics to understand the disease impacts of testosterone in men and women

Genetic predisposition to smoking in relation to 14 cardiovascular diseases

Genetic variants in a sodium-dependent vitamin C transporter gene and age-related cataract

Common gene variants, mortality and extreme longevity in humans

Using Causal Diagrams to Understand Common Problems in Social Epidemiology

From clinical trials to bedside: the use of antihypertensives in aged individuals. Part 1: Evaluation and evidence of treatment benefit

Methodological problems in genetic association studies of longevity-The apolipoprotein E gene as an example

21st-century hazards of smoking and benefits of cessation in the United States

Antioxidant supplements for prevention of mortality in healthy participants and patients with various diseases

Genetic predictors of testosterone and their associations with cardiovascular disease and risk factors: A Mendelian randomization investigation

Association of genetically predicted testosterone with thromboembolism, heart failure, and myocardial infarction: Mendelian randomisation study in UK Biobank

Testosterone treatment and coronary artery plaque volume in older men with low testosterone

Castration reduces platelet thromboxane A2 receptor density and aggregability

Androgen deprivation is associated with enhanced endothelium-dependent dilatation in adult men. Arterioscler

Vascular reactivity is impaired in genetic females taking high-dose androgens

The effects of sex steroids on plasma levels of marker proteins of endothelial cell functioning

Sexual dimorphism in innate immunity

Strengthening the immune system for cancer prevention

Single dose testosterone administration increases impulsivity in the intertemporal choice task among healthy males

Dietary supplements of soya flour lower serum testosterone concentrations and improve markers of oxidative stress in men

Low-fat high-fiber diet decreased serum and urine androgens in men

Postprandial changes in sex hormones after meals of different composition

Reducing bioavailable sex hormones through a comprehensive change in diet: The diet and androgens (DIANA) randomized trial

Effect of weight loss, with or without exercise, on body composition and sex hormones in postmenopausal women: The SHAPE-2 trial

Effect of lifestyle intervention on the reproductive endocrine profile in women with polycystic ovarian syndrome: A systematic review and meta-analysis

Reduced-calorie dietary weight loss, exercise, and sex hormones in postmenopausal women: Randomized controlled trial

Long-term effects of a randomised controlled trial comparing high protein or high carbohydrate weight loss diets on testosterone, SHBG, erectile and urinary function in overweight and obese men

Short-term exercise training inconsistently influences basal testosterone in older men: A systematic review and meta-analysis

Metabolic adaptation to caloric restriction and subsequent refeeding: The Minnesota Starvation Experiment revisited

The effect of statins on testosterone in men and women, a systematic review and meta-analysis of randomized controlled trials

The effects of metformin on endogenous androgens and SHBG in women: A systematic review and meta-analysis

Effects of metformin and short-term lifestyle modification on the improvement of male hypogonadism associated with metabolic syndrome

Aspirin inhibits androgen response to chorionic gonadotropin in humans

Six-month treatment with low-dose dexamethasone further reduces androgen levels in PCOS women treated with diet and lifestyle advice, and metformin

Dexamethasone-induced suppression of the circadian rhythm of plasma testosterone in normal adult males

Bias due to participant overlap in two-sample Mendelian randomization

Effect measure modification conceptualized using selection diagrams as mediation by mechanisms of varying population-level relevance

What makes UK Biobank special?

Assessing the suitability of summary data for two-sample Mendelian randomization analyses using MR-Egger regression: The role of the I2 statistic

A comparison of robust Mendelian randomization methods using summary data

Investigating pleiotropic effects of statins on ischemic heart disease in the UK Biobank using Mendelian randomisation

Assessment of free testosterone concentration

Teeth Sex and Testosterone: Aging in the World's Smallest Primate

JVZ and CMS had full access to the data and take full responsibility for the results. 

C.M.S. conceptualized the study, conducted the initial analysis and wrote the first draft. J.V.Z. contributed to the conceptualization, extracted data, checked the analysis and reviewed the interpretation.

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

The authors declare no competing interests.

Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1038/ s41598-021-93360-z.Correspondence and requests for materials should be addressed to C.M.S.Reprints and permissions information is available at www.nature.com/reprints.Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.