You are here

  • Home
  • Archive
  • Volume 79, Issue 6
  • Effects of testosterone and sex hormone binding globulin on lung function in males and females: a multivariable Mendelian Randomisation study

Article Text

Article menu
Respiratory epidemiology
Original research
Effects of testosterone and sex hormone binding globulin on lung function in males and females: a multivariable Mendelian Randomisation study
Free
  1. Diana A van der Plaat1,
  2. Alexandra Lenoir2,3,
  3. Shyamali Dharmage4,
  4. James Potts1,
  5. Francisco Gómez Real5,6,
  6. Seif O Shaheen7,
  7. Debbie Jarvis1,
  8. Cosetta Minelli1,
  9. Bénédicte Leynaert8
  1. 1 National Heart and Lung Institute (NHLI), Imperial College London, London, UK
  2. 2 Centre Hospitalier Universitaire Vaudois, Lausanne, Switzerland
  3. 3 Gesundheitsamt Fürstenfeldbruck, Fürstenfeldbruck, Switzerland
  4. 4 Allergy and Lung Health Unit, The University of Melbourne School of Population and Global Health, Melbourne, Victoria, Australia
  5. 5 Department of Clinical Science, University of Bergen, Bergen, Norway
  6. 6 Department of Gynecology and Obstetrics, Haukeland University Hospital, Bergen, Norway
  7. 7 Wolfson Institute of Population Health, Queen Mary University of London, Barts and The London School of Medicine and Dentistry, London, London, UK
  8. 8 Université Paris-Saclay, UVSQ, Université Paris-Sud, Inserm, Équipe d'Épidémiologie Respiratoire Intégrative, CESP, INSERM, Villejuif, France
  1. Correspondence to Dr Diana A van der Plaat, National Heart and Lung Institute (NHLI), Imperial College London, London, UK; d.van-der-plaat{at}imperial.ac.uk

Abstract

Background Observational studies suggest that total testosterone (TT) and sex hormone-binding globulin (SHBG) may have beneficial effects on lung function, but these findings might be spurious due to confounding and reverse causation. We addressed these limitations by using multivariable Mendelian randomisation (MVMR) to investigate the independent causal effects of TT and SHBG on lung function.

Methods We first identified genetic instruments by performing genome-wide association analyses of TT and SHBG in the large UK Biobank, separately in males and females. We then assessed the independent effects of TT and SHBG on forced expiratory volume in 1 s (FEV1), forced vital capacity (FVC) and FEV1/FVC using one-sample MVMR. We addressed pleiotropy, which could bias MVMR, using several methods that account for it. We performed subgroup MVMR analyses by obesity, physical activity and menopausal status, and assessed associations between TT and SHBG with lung function decline. Finally, we compared the MVMR results with those of observational analyses in the UK Biobank.

Findings In the MVMR analyses, there was evidence of pleiotropy, but results were consistent when accounting for it. We found a strong beneficial effect of TT on FVC and FEV1 in both males and females, but a moderate detrimental effect of SHBG on FEV1 and FEV1/FVC in males only. Subgroup analyses suggested stronger effects of TT among obese and older males. The observational analyses, in line with previous studies, agreed with MRMV for TT, but not for SHBG.

Interpretation These findings suggest that testosterone improves lung function in males and females, while SHBG has an opposite independent effect in males.

  • Respiratory Measurement
  • COPD epidemiology

Data availability statement

All data relevant to perform the multivariable Mendelian randomisation analyses are included as online supplemental information (G-X and G-Y in online supplemental table S2). Individual participant data may be obtained from a third party and are not publicly available. Please consult the data access policy of the UK Biobank at https://www.ukbiobank.ac.uk/.

https://doi.org/10.1136/thorax-2023-220374

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    WHAT IS ALREADY KNOWN ON THIS TOPIC

    • A few epidemiological studies have shown that higher testosterone levels are associated with better lung function in men and experimental studies have highlighted possible protective effects of testosterone in the lungs. Our recent observational analysis of the UK Biobank study also showed a significant beneficial association of sex hormone-binding globulin (SHBG) levels with forced vital capacity (FVC) and forced expiratory volume in 1 s (FEV1), both in males and females. However, whether testosterone and SHBG have a causal effect on lung function in both sexes remains unknown, as associations from observational studies may arise from confounding or reverse causation.

    WHAT THIS STUDY ADDS

    • By using multivariable Mendelian randomisation, which addresses the limitations of observational studies, our study in the large UK Biobank provides strong evidence that higher testosterone levels lead to higher FEV1 and FVC levels in both men and women from the general population. Conversely, SHBG has a negative effect on FEV1 and FEV1/FVC in men.

    HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

    • Our findings stress the need for further research on testosterone supplementation in people at risk of accelerated lung function decline and suggest the need for active screening for lung function impairment in people with low testosterone levels.

    Introduction

    Sex hormones influence lung physiology and respiratory disease. Clinical studies have reported low levels of testosterone in male patients with chronic obstructive pulmonary disease (COPD).1 This was initially thought to reflect endocrine dysfunctions due to the disease or to the long-term use of corticosteroids in patients with COPD.1–3 However, recent laboratory-based research has provided stronger evidence for an overall beneficial effect of androgens on respiratory health.4–8 In line with this hypothesis, epidemiological studies have reported positive associations of lung function with levels of testosterone, testosterone precursors (dehydroepiandrosterone (DHEA)), testosterone metabolites and androgen receptor expression in the airway epithelium in men,2 9 10 with some positive associations also reported in women.11–15 One study also reported higher forced vital capacity (FVC) and forced expiratory volume in 1 s (FEV1) in men with higher levels of sex hormone-binding globulin (SHBG), which binds androgens and oestrogens with high affinity regulating their bioavailability to tissues and target cells.9

    In agreement with previous cross-sectional studies, our recent observational analysis of the UK Biobank study (UKB) showed a positive association of total testosterone (TT) with FEV1 and FVC in men.13 This was not observed in women, but a longitudinal analysis showed that women with higher TT levels at baseline had reduced decline in FEV1 and FVC over the follow-up. In this study, we also reported a significant positive association of SHBG levels with FEV1 and FVC, both in men and women, with participants with the greatest increase in SHBG over 4 years of follow-up also having higher FVC at the end of follow-up. Although the underlying mechanisms are unclear, there is increasing evidence that SHBG might have biological effects that are independent of those explained by its binding to sex steroids.16–19 SHBG could itself function as a hormone or as a signal transduction factor. A recent Mendelian randomisation (MR) study reported a positive causal effect of SHBG on asthma, supporting the hypothesis of beneficial effects of SHGB in the lungs.20

    Overall, these findings, mostly based on cross-sectional studies, suggest a protective effect of both TT and SHBG on lung function. However, they might be spurious and arise from confounding by factors such as smoking, physical activity and diet, or from reverse causation, whereby it is lung function affecting hormone levels rather than vice-versa (eg, hormones influenced by hypoxaemia1–3). The MR approach addresses these limitations by using genetic instruments as proxies for the risk factor of interest. MR is thereby not affected by traditional confounding or reverse causation and thus provides stronger evidence of a possible causal effect and its direction. MR has been shown to not always confirm observational evidence21 22, and it, therefore, represents an important complemental approach to observational studies when experimental studies (randomised controlled trials, RCTs) are not feasible. The validity of MR relies on some assumptions, the most problematic being the absence of pleiotropy, where a genetic instrument affects the outcome through additional pathways independent from the risk factor of interest.23 While pleiotropy can bias MR findings, methods have been developed to both detect and control for it.24

    Here, we used one-sample MR to investigate the presence of a causal effect of TT and SHBG on FEV1, FVC and FEV1/FVC, separately in males and females, using data from the large UKB. As levels of SHBG and TT are partly determined by the same genes, we used multivariable MR (MVMR), where TT and SHBG are modelled together to estimate their effects independent of each other. First, genetic instruments for MVMR were selected by performing a genome-wide association (GWA) analysis of TT and SHBG, stratified by sex in UKB. Then, several MVMR methods were used to identify and address pleiotropy. As secondary analyses, we performed subgroup analyses to investigate possible effect modification by obesity, physical activity and menopausal status in females and age above 50 years in males. We also assessed the effect of TT and SHBG on lung function decline using 8-year follow-up data on a subset of the UKB study population. Finally, we compared the results of our main MVMR analyses with observational estimates of the association of TT and SHBG with lung function obtained in UKB.

    Material and methods

    We first performed a GWA analysis to identify single-nucleotide polymorphisms (SNPs) associated with TT and SHBG to be used as genetic instruments. We then performed a one-sample MVMR to investigate the effect of TT and SHBG on lung function. We applied different MVMR methods to test the robustness of the findings. Using MVMR, we also investigated possible effect modification in population subgroups, and we assessed the effects of TT and SHBG on lung function decline. Finally, for comparison, we performed an observational analysis of the association of TT and SHBG with lung function. All analyses were performed using UKB data and were stratified by sex. A figure describing this workflow (online supplemental figure S1) and further details on the methods are provided in online supplemental methods.

    Supplemental material

    UKB data

    We included subjects from UKB, which is a large population-based study in the UK.25 Details on the study design of UKB25 and genotyping/imputation methods26 have been published elsewhere.

    Within UKB, we created three datasets (figure 1), which were subsequently stratified by sex: (1) ‘discovery hormone dataset’ (2/3 of sample: N=323 144), used in the GWA analyses of TT and SHBG; (2) ‘replication hormone dataset’ (1/3 of sample: N=161 572), used for the replication of GWA findings of TT and SHBG and for the MVMR analyses; (3) ‘lung function dataset’ (N=341 826), used for the MVMR analyses and for the observational analyses. The ‘lung function dataset’ used to obtain the lung function effect estimates included only subjects with previously defined good-quality spirometry data (see online supplemental methods). In the MVMR analyses, the overlap between the hormone replication dataset and the lung function datasets was about 70%.

    Figure 1
    Figure 1

    Flow diagram of sample selection from UK Biobank (UKB) for the one-sample MVMR analyses. We first undertook GWA, separately by sex, in the ‘hormone discovery dataset’ to identify genetic instruments (SNPs) for TT and SHBG), with replication in the ‘hormone replication dataset’. The MVMR estimates of the effects of TT and SHBG on lung function were then derived in UKB, separately by sex, from the SNP-hormone (G-X; hormone replication dataset’) and the SNP-lung function (G-Y; ‘lung function dataset’) estimates. The ‘lung function dataset included only UKB subjects with previously defined good-quality spirometry data. GWA, genome-wide association analyses; MVMR, multivariable Mendelian randomisation; SHBG, sex hormone-binding globulin; SNP, single-nucleotide polymorphism; TT, total testosterone

    For lung function, we considered three measures: FEV1 (mL), FVC (mL) and FEV1/FVC (%). The lung function estimates were based only on subjects with spirometry data that passed the quality control criteria developed for the UK BiLEVE study,27 who created UKB variables for ‘best’ measures. As lung function ‘best measure’ was derived for white participants, and therefore, only these were included in the lung function dataset, while the hormone datasets also included a very small proportion of participants from other ethnicities (5.4%).

    GWA analysis for genetic instrument selection

    To identify genetic instruments for TT and SHBG, we performed GWA analyses stratified by sex in the ‘discovery hormone dataset’ on (natural) log-transformed TT and SHBG values. We used linear mixed models implemented in BOLT-LMM,28 which account for relatedness and fine-scale population structure, adjusted for age, genotyping batch/array and centre. We selected independent SNPs (linkage disequilibrium (LD), r2<0.05) with p<5×10−8 and tested them for replication in the ‘replication hormone dataset’. We considered SNPs as replicated if the direction of their effect was the same as in the discovery dataset, and their Bonferroni-corrected one-side p value was statistically significant.

    In total, we replicated 450 SNPs (online supplemental tables S1 and S2) and Manhattan plots in online supplemental figures S2 and S3). For TT, we replicated 92 SNPs in males and 63 in females, which explained 8.1% and 3.6% of the variance in TT levels, respectively. For SHBG, we replicated 213 SNPs in males and 153 in females, which explained 20.5% and 13.0% of SHBG variance.

    MVMR analyses

    We performed one-sample MVMR analyses, with estimates of the causal effects of TT and SHBG on lung function derived from the SNP-hormone (G-X) and the SNP-lung function (G-Y) association estimates, using different MVMR methods (see online supplemental methods for details and background information).24 29

    As G-X estimates, we used the effect estimates of our GWA replication analysis and removed SNPs that were in LD (r2>0.1) between the SNP lists of TT and SHBG. In total, 178 independent SNPs were included in the male MVMR analyses, and 174 in the female analyses. To obtain the G-Y estimates, again stratified by sex, we estimated the effect of each SNP on FEV1, FVC and FEV1/FVC in UKB using BOLT-LMM and adjusting for age, age2, height, genotyping batch/array and centre. G-X and G-Y estimates are reported in online supplemental table S2.

    Supplemental material

    The validity of MR heavily relies on the absence of ‘horizontal’ pleiotropy, which occurs when a genetic instrument affects the outcome through pathways that are independent from the risk factor of interest.23 We, therefore, applied six MVMR methods, of which two assume no pleiotropy: two-stage least squares (2SLS) and fixed-effect inverse-variance weighted (FE-IVW) meta-analysis, while the others account for pleiotropy but make different assumptions about it: random-effects IVW (RE-IVW) meta-analysis, robust regression, weighted median estimation and MR-Egger regression (see online supplemental methods for details and references).

    Possible pleiotropy was detected and reported in the online supplemental tables (Q test, Sargan test and MR-Egger intercept p value). To assess the validity of our genetic instruments, we took two approaches to identify and remove potential pleiotropic SNPs which were subsequently excluded in sensitivity analyses: (1) a statistical approach using MVMR Pleiotropy RESidual Sum and Outlier (MVMR-PRESSO)30 to detect outliers and (2) using PhenoScanner (www.phenoscanner.medschl.cam.ac.uk) to identify SNPs associated with potential sources of pleiotropy identified a priori: smoking, weight-related traits, diabetes, insulin, leptin and adiponectin (n=36 SNPs). In addition, to better understand the relevance of potential pleiotropic pathways, we further investigated MVMR-PRESSO outliers by searching for association of these SNPs with any other traits in previous GWA studies (p<5×10−8) using PhenoScanner; and used leave-one-out RE-IVW analysis to identify the most influential pleiotropic SNPs.

    MVMR secondary analyses

    We performed three subgroup analyses to investigate effect modification by potential modifiers: (1) obesity, as several studies have reported interactions between sex hormones and obesity on lung function and asthma22 31; (2) moderate physical activity, as RCTs in patients with COPD have suggested that testosterone supplementation might be more effective when combined with exercise training,3 13 32 and our previous observational analysis13 highlighted a stronger positive effect of testosterone levels in physically active men; (3) menopausal status in females and age 50 years in males, as the effects of the SHBG and testosterone might vary with age and age-related changes in hormonal milieu.

    We also performed MVRM to assess the effect of TT and SHBG on lung function decline using a subset of the UKB study population with available longitudinal data. We included 6812 males and 7470 females with a mean follow-up time of 8.3 years. We obtained the G-Y estimates stratified by sex by estimating the effect of each SNP on yearly decline of FEV1, FVC and FEV1/FVC in UKB using the same model as in the main analysis (online supplemental methods). Not all the 450 SNPs passed QC, and therefore, less SNPs were included in the MVMR analyses (see online supplemental table S2) for the G-Y estimates of the included SNPs).

    Observational analyses

    For comparison, we also performed cross-sectional observational analyses in UKB of the association of log-transformed TT and SHBG values with the three lung function outcomes, separately by sex, adjusted for the same covariates as in our previous study in UKB.13 This analysis is different as both TT and SHBG were included in the same multiple linear regression models in order to estimate their independent association with lung function.13 Moreover, our previous analysis was performed on a much smaller samples size (n=10 581) as it was based on longitudinal data only.

    Results presentation

    To help interpretation of the results and figures (based on log-transformed hormone values), we calculated the average regression coefficient (beta) across MVMR methods and multiplied it by ln(1.1) to display the effect of a 10% increase in SHBG or TT on lung function. The effect estimates reported in online supplemental tables represent the effect on lung function of a one-unit increase in log-transformed SHBG and TT.

    Results

    The characteristics of the participants included in the GWA and MVMR analyses are reported in table 1. Results of the GWA for genetic instrument selection and derivation of G-X and G-Y estimates for MVMR analysis are described in online supplemental methods.

    View this table:
    Table 1

    Population characteristics per dataset

    MVMR: testosterone

    In males, the MVMR analysis did not show consistent evidence across the various MVMR methods of an effect of TT on any lung function outcome but there was strong evidence of pleiotropy (figure 2, online supplemental table S3). After removing 13 SNPs identified by MVMR-PRESSO as pleiotropic (outliers), we found consistent evidence for a large beneficial effect of higher TT being associated with higher FVC and FEV1 across all methods, although there was evidence of residual pleiotropy. A 10% increase in TT was associated with 15.5 mL higher FVC and 17.0 mL higher FEV1, and therefore, a small beneficial effect on FEV1/FVC (figure 3 and online supplemental table S4A). The leave-one-out analysis carried out to investigate the effect of the 13 pleiotropic SNPs, identified 2 SNPs (rs56196860 in FKBP4 and rs62062271 in MAPT) as the most influential pleiotropic SNPs that were driving the results towards the null (online supplemental figures S4 and S5). Investigation of these SNPs in PhenoScanner showed that rs56196860 was not associated with any other trait at genome-wide significance level (p<5×10−8), but rs62062271 was associated with several human blood-cell traits in the GWA study by Astle et al (online supplemental table S5).33 Of note, by lowering the significance level to p<0.05 in PhenoScanner, revealed that 9 out of the 13 pleiotropic SNPs were associated with blood cell traits, including rs56196860 which was associated with platelet-related outcomes.

    Supplemental material

    Supplemental material

    Supplemental material

    Figure 2
    Figure 2

    Results of the MVMR analysis of the effect of SHBG and TT on (A) FVC, (B) FEV1 and (C) FEV1/FVC separately in males and females. Results expressed as the effect of a 10% increase in SHBG or TT on lung function. MVMR, multivariable Mendelian randomisation; FE-IVW, fixed effect inverse variance weighted meta-analysis; FEV1, forced expiratory volume in 1 s; FVC, forced vital lung capacity; RE-IVW, random effect Inverse variance weighted meta-analysis; SHBG, sex hormone-binding globulin; TT, total testosterone; 2SLS, two-stage least squares.

    Figure 3
    Figure 3

    Results of the MVMR analysis after excluding pleiotropic SNPs identified by MVMR-PRESSO, for the effect of SHBG and TT on: (A) FVC, (B) FEV1 and (C) FEV1/FVC, separately in males and females. Results expressed as the effect of a 10% increase in SHBG or TT on lung function. FE-IVW, fixed effect inverse variance weighted meta-analysis; FEV1, forced expiratory volume in 1 s; FVC, forced vital capacity; MVMR, multivariable Mendelian randomisation; RE-IVW, random effect Inverse variance weighted meta-analysis; SHBG, sex hormone-binding globulin; SNPs, single-nucleotide polymorphisms; TT, total testosterone; 2SLS, two-stage least squares.

    In females, we found a beneficial effect of TT on FVC and FEV1 (figure 2, online supplemental table S3), with 4.0 mL higher FVC and 3.8 mL higher FEV1. There was a strong evidence of pleiotropy, but the results remained similar after removing 10 SNPs identified as outliers (figure 3, online supplemental table S4A).

    MVMR and SHBG

    In males, the MVMR analyses showed a detrimental effect of SHBG on FEV1/FVC (figure 2, online supplemental table S3), with about 0.07% lower FEV1/FVC. A detrimental effect on FEV1 (−10.5 mL) was also found, but only with the robust and weighted median MVMR methods. None of the MVMR methods showed an effect of SHBG on FVC. Again, there was strong evidence of pleiotropy, and the analyses removing 13 outlier SNPs showed a detrimental effect for SHBG on both FEV1/FVC and FEV1 across all MVMR methods (figure 3, online supplemental table S4A). Of note, it is interesting that the detrimental effect of SHBG was accompanied by a beneficial effect of TT on both FEV1 and FEV1/FVC (figure 3).

    In females, we did not find consistent evidence of an effect of SHBG on FVC or FEV1/FVC (figure 2, online supplemental table S3), while there was evidence of a small detrimental effect on FEV1 (−2.0 mL). However, when repeating the analysis excluding 10 outlier SNPs (which reduced but did not eliminate the heterogeneity), no significant effect was found in any of the MVMR analyses (figure 3, online supplemental table S4A).

    For the analyses of both TT and SHBG and in both sexes, when we removed potential sources of pleiotropy identified a priori (36 SNPs), the results did not change (online supplemental table S4B).

    MVMR: secondary analyses

    In males, the subgroup analyses showed a stronger effect of TT on FVC and FEV1 in obese males (figure 4A–B, online supplemental table S6A), although a statistically significant interaction with obesity was only found for one MVMR method (2SLS; FVC: +32.7 mL in obese vs+3.8 mL in non-obese, p=0.010). Furthermore, the beneficial effect of TT and detrimental effect of SHBG on FEV1/FVC (TT: +0.12%; SHBG: −0.11%) were restricted to males older than 50, with no effect found in younger males (figure 4C, online supplemental table S6D). We found no evidence of modifying effects of obesity, moderate physical activity or menopause in females (online supplemental table S6).

    Supplemental material

    Figure 4
    Figure 4

    Subgroup analysis showing significant interaction by obesity (A–B) and age (C) in males. Outlier SNPs and ‘weak’ SNPs (F-statistic≤10) were excluded. Results expressed as the effect of a 10% increase in SHBG or TT on lung function. *Nominal significant interaction (p values for interaction <0.05). BMI, body mass index; FEV1, forced expiratory volume in 1 s; FE-IVW, fixed effect inverse variance weighted meta-analysis; FVC, forced vital capacity; RE-IVW, random effect Inverse variance weighted meta-analysis; TT, total testosterone; SHBG, sex hormone-binding globulin; SNP, single-nucleotide polymorphisms; 2SLS, two-stage least squares.

    In the MVMR analyses assessing the effect on lung function decline, we did not find a significant effect of TT or SHBG on the decline of any of the three lung function outcomes (online supplemental table S7).

    Supplemental material

    Observational analyses

    In the observational analyses, a 10% increase in TT was independently associated with higher FVC (+10.4 mL) and FEV1 (+9.3 mL) in males (figure 2, online supplemental table S8). Beneficial statistically significant associations, although of small magnitude, were also observed between TT and FEV1 in females (+0.87 mL) and between TT and FEV1/FVC in both males (+0.04%) and females (+0.03%). The same analysis showed beneficial association of SHBG with higher FVC in both males (+6.6 mL) and females (+6.0 mL). SHBG was also beneficially, but less strongly, associated with FEV1, and thus detrimentally associated with FEV1/FVC.

    Supplemental material

    Discussion

    Our MVMR showed a beneficial effect of testosterone on FVC and FEV1 in males, with a more moderate beneficial effect in females, independent from the effect of SHBG. While this provides causal support for previous observational evidence, our findings for SHBG contrast with the observational evidence in showing a detrimental effect of SHBG on FEV1 and FEV1/FVC in males, with no effect of SHBG in females.

    The MVMR finding of a beneficial effect of testosterone on FVC in males is consistent with observational findings from previous studies and from our own analysis.2 9 10 13 While in the initial analysis there was no consistent evidence of an effect across MVMR methods, this appeared to be driven by pleiotropy, where genetic instruments affect the outcome through alternative independent pathways (phenotypes). Moreover, it suggested ‘negative’ pleiotropy, where the effect of a genetic instruments on the alternative phenotype is in the opposite direction compared with its effect on testosterone. After removing pleiotropic SNPs identified by MVMR-PRESSO, we found a strong beneficial effect of TT and a weaker detrimental effect of SHBG on FVC in males. This suggests that a few SNPs associated with higher levels of testosterone are also associated with other factors that negatively impact lung function. Further investigation of the identified pleiotropic SNPs suggests that pleiotropy might occur through the effect of these SNPs on blood-cell subtypes, and indeed several blood-cell types have been previously associated with lung function and its decline.34 35 Heterogeneity remained, although reduced, when we removed outlier SNPs, thus suggesting the presence of some unresolved pleiotropy. However, when we removed a priori considered potential sources of pleiotropy the results did not change, suggesting the presence of other unknown sources of pleiotropy. Moreover, the consistency of the findings across the different MVMR methods accounting for pleiotropy provides reassurance of the robustness of our results.

    The MVMR analysis also suggests a moderate beneficial effect of testosterone on FVC in females. Significant beneficial effects of testosterone were found in all female analyses, including the subgroup analyses.34 Few studies have investigated the associations between circulating testosterone and lung function in females. Recent cross-sectional analyses showed no significant, or modest detrimental, associations between free testosterone levels and FVC in women.10 13 However, the longitudinal analysis over an 8-year follow-up by Lenoir et al 13 showed a slower decline in FVC in females with higher levels of total and free testosterone at baseline. This is consistent with another study showing higher FVC in women with higher levels of DHEAs, a major precursor of testosterone in women.12

    In addition to its action on lung development and maturation early in life, testosterone might increase FEV1 through direct bronchodilator effects, and FVC through direct effects on respiratory or skeletal muscle strength, or it could affect both lung function measures by reducing inflammatory responses and subsequent airway remodelling.3 4 7 8 14 In our cross-sectional MVMR analyses, we found that TT had some beneficial effect on FVC, whereas we found no significant effect of TT in the longitudinal analysis with FVC decline. A first possible explanation might be a beneficial effect of TT in attaining higher maximum peak FVC during lung growth. However, at least in females, the slower FVC decline observed by Lenoir et al 13 in females with higher testosterone might rather point towards a protective effect of testosterone in sustaining lung capacity in adulthood, rather than a beneficial effect during lung growth in early life. An alternative hypothesis might be a positive short-term effect of testosterone. Such a ‘short-term’ effect was also proposed by Lenoir et al 13 as an alternative explanation for the positive cross-sectional association between TT and FVC but lack of association with FVC decline in men. In the European Community Respiratory Health Survey (ECRHS), Pesce et al 12 found that low level of the major precursor of testosterone in women (DHEA-S) was associated with FVC but not FVC decline and suggested that this might point towards a short-term, maybe reversible, effect of DHEA-S deficiency on pulmonary vital capacity. Similarly, a short-term positive effect of testosterone (possibly through broncho-relaxing properties or anabolic effects on muscle strength) might explain why we observe strong associations in the cross-sectional analysis, but not in the longitudinal analysis. We further cannot exclude a small effect of TT on lung function decline that we could not observe after a short follow-up. Longitudinal studies, with lung function trajectories from childhood to adulthood, would be useful to investigate whether the effect of TT is mostly on lung growth, and whether this takes place similarly in males and females.

    Testosterone deficiency has been frequently reported in patients with COPD.1 3 32 However, to date, small clinical trials of male patients with COPD failed to show any protective effect of testosterone supplementation on lung function, although it was associated with an increase in lean body mass and a reduced risk of respiratory hospitalisations.3 32 36 Moreover, a recent phase 2A trial in men and postmenopausal women with COPD showed that a selective androgen receptor modulator (GSK2881078) increased leg strength, at least in men, and lean body mass in both men and women, but no changes in lung function were observed.37 As suggested by Mohan et al,9 it is possible that testosterone might have a beneficial effect in preserving lung function through mechanisms independent from those driving the lower testosterone levels in patients with COPD. It is also possible that the duration of the intervention trials was too short to observe a protective effect on lung function measures. In an additional analysis (data not shown), we assessed the association of TT and SHBG with airflow obstruction (FEV1/FVC<lower limit of normal) and did not find a clear association in either males or females, in line with the null association observed for the FEV1/FVC ratio.

    There is an increasing interest in a possible role of SHBG in several diseases.16–20 38 SHBG binds to circulating sex hormones; in particular, 40%–65% of testosterone is bound to SHBG, which reduces the concentration of unbound (‘free’) testosterone.39 In our observational analysis, which was adjusted for testosterone, SHBG showed positive associations with FVC in males and females, consistent with previous observational studies.9 13 In contrast, our MVMR analysis showed significant independent detrimental effects of SHBG on FVC, FEV1 and FEV1/FVC in males after removing pleiotropic SNPs. This detrimental effect of SHBG was always accompanied by a beneficial effect of TT. Therefore, one might speculate that SHBG has a detrimental effect on lung function through its binding to testosterone.36

    In females, we found no evidence for a detrimental effect of SHBG on lung function. Of note, oestradiol also binds to SHBG, and in contrast to androgens, oestrogens may also have proinflammatory proprieties and might detrimentally impact lung function.4 7 40 Therefore, the lack of effect of SHBG on lung function in females might be partly explained by SHBG binding, not only to testosterone, but also to oestradiol in females, and the balance between the two may be driving the final effect.

    The ‘free hormone’ hypothesis assumes that only unbound testosterone can bind to androgen receptors and thus exert its effect, and measurement of free testosterone has been recommended to establish clinical diagnoses, particularly in men with altered SHBG levels and in women.41 However, more recent evidence suggests that testosterone binding to SHBG is important for reaching and becoming active in target tissues,41 and that SHBG might even increase sex hormones signal transduction.42 In this study, we used MVMR with the aim to investigate the independent effects of SHBG and TT on lung function (ie, whether for a given value of TT, individuals with higher SHBG have higher lung function and vice-versa), and hence we did not consider any calculated index to reflect the proportion of TT that is not bound to SHBG.

    The large sample size allowed us to investigate possible interactions based on a priori hypotheses. We observed that the effects of TT and SHBG on FVC in men were more marked in obese than in non-obese individuals, although the effects were in the same direction in the two groups and there was no consistent evidence of interaction. We found no evidence of interaction with physical activity and menopausal status in women, while we found that the small negative effect of SHBG on FEV1/FVC was limited to older men.

    There are some limitations in our study. Like many previous observational studies on hormones, TT was not measured using a standardised Liquid chromatography–mass spectrometry (LC-MS/MS) assay, which is the gold-standard method. Regarding our findings for FEV1/FVC, a marker of obstruction, only prebronchodilator lung function measurements were available in UKB, and this does not allow us to differentiate between reversible airflow obstruction (as in asthma) and fixed obstruction (as in COPD). However, our findings mostly provided evidence of an effect of TT on FVC, while there was little evidence for an effect on FEV1/FVC. As we only used UKB data, which pronominally includes white participants, and good spirometry data were limited to participants of Caucasian ancestry, our findings should be generalised with caution. Finally, we did not find and effect on lung function decline in our secondary MVMR analyses in contrast to our previous observational study.13 However, this could be due to the greatly reduced the number of UKB participants in the longitudinal analysis (N=14 282) compared with the main MVMR analyses (N=341 826) and the relatively short follow-up time (average 8 years), and there could be a small effect that we would be underpowered to detect. Future availability of longitudinal lung function data from large samples and over a long period of time will allow further MR investigations to provide conclusive evidence on the effects of sex hormones on lung function decline.

    Conclusion

    Our findings show that higher genetically determined testosterone levels are associated with higher FVC and FEV1 in both males and females. In contrast, higher genetically determined SHBG levels are associated with lower FEV1 and FEV1/FVC in males. These findings suggest a beneficial causal effect of testosterone and an adverse causal effect of SHBG on adult lung function. Our findings stress the need for further research to investigate whether testosterone therapy might be used to slow down lung function impairment among selected groups of people at risk of accelerated lung function decline and in older age. They also suggest the need for active screening for lung function impairment in people with low-testosterone levels.

    Supplemental material

    Data availability statement

    All data relevant to perform the multivariable Mendelian randomisation analyses are included as online supplemental information (G-X and G-Y in online supplemental table S2). Individual participant data may be obtained from a third party and are not publicly available. Please consult the data access policy of the UK Biobank at https://www.ukbiobank.ac.uk/.

    Ethics statements

    Patient consent for publication

    Ethics approval

    This study involves human participants and UK Biobank has received ethics approval from the National Health Service National Research Ethics Service (Ref 11/NW/0382). Participants gave informed consent to participate in the study before taking part.

    References

    1. ↵
      2. Atlantis E ,
      3. Fahey P ,
      4. Cochrane B , et al
      . Endogenous testosterone level and testosterone supplementation therapy in chronic obstructive pulmonary disease (COPD): a systematic review and meta-analysis. BMJ Open 2013;3:e003127. doi:10.1136/bmjopen-2013-003127
    2. ↵
      2. Svartberg J ,
      3. Schirmer H ,
      4. Medbø A , et al
      . Reduced pulmonary function is associated with lower levels of endogenous total and free testosterone. The Tromsø study. Eur J Epidemiol 2007;22:107–12. doi:10.1007/s10654-006-9095-9
      OpenUrlCrossRefPubMedWeb of Science
    3. ↵
      2. Baillargeon J ,
      3. Urban RJ ,
      4. Zhang W , et al
      . Testosterone replacement therapy and hospitalization rates in men with COPD. Chron Respir Dis 2019;16:1479972318793004. doi:10.1177/1479972318793004
    4. ↵
      2. Kalidhindi RSR ,
      3. Ambhore NS ,
      4. Balraj P , et al
      . Androgen receptor activation alleviates airway hyperresponsiveness, inflammation, and remodeling in a murine model of asthma. Am J Physiol Lung Cell Mol Physiol 2021;320:L803–18. doi:10.1152/ajplung.00441.2020
      OpenUrlCrossRef
    5. ↵
      2. Fuseini H ,
      3. Yung JA ,
      4. Cephus JY , et al
      . Testosterone decreases house dust mite-induced type 2 and IL-17A-mediated airway inflammation. J Immunol 2018;201:1843–54. doi:10.4049/jimmunol.1800293
      OpenUrlAbstract/FREE Full Text
    6. ↵
      2. Sathish V ,
      3. Martin YN ,
      4. Prakash YS
      . Sex steroid signaling: implications for lung diseases. Pharmacol Ther 2015;150:94–108. doi:10.1016/j.pharmthera.2015.01.007
      OpenUrlCrossRefPubMed
    7. ↵
      2. Wang X ,
      3. Huang L ,
      4. Jiang S , et al
      . Testosterone attenuates pulmonary epithelial inflammation in male rats of COPD model through preventing NRF1-derived NF-κB signaling. J Mol Cell Biol 2021;13:128–40. doi:10.1093/jmcb/mjaa079
      OpenUrl
    8. ↵
      2. Aono K ,
      3. Matsumoto J ,
      4. Nakagawa S , et al
      . Testosterone deficiency promotes the development of pulmonary emphysema in orchiectomized mice exposed to elastase. Biochem Biophys Res Commun 2021;558:94. doi:10.1016/j.bbrc.2021.04.051
      OpenUrl
    9. ↵
      2. Mohan SS ,
      3. Knuiman MW ,
      4. Divitini ML , et al
      . Higher serum testosterone and dihydrotestosterone, but not oestradiol, are independently associated with favourable indices of lung function in community-dwelling men. Clin Endocrinol (Oxf) 2015;83:268–76. doi:10.1111/cen.12738
      OpenUrlCrossRefPubMed
    10. ↵
      2. Han Y-Y ,
      3. Yan Q ,
      4. Yang G , et al
      . Serum free testosterone and asthma, asthma hospitalisations and lung function in British adults. Thorax 2020;75:849–54. doi:10.1136/thoraxjnl-2020-214875
      OpenUrlAbstract/FREE Full Text
    11. ↵
      2. DeBoer MD ,
      3. Phillips BR ,
      4. Mauger DT , et al
      . Effects of endogenous sex hormones on lung function and symptom control in adolescents with asthma. BMC Pulm Med 2018;18:58. doi:10.1186/s12890-018-0612-x
    12. ↵
      2. Pesce G ,
      3. Triebner K ,
      4. van der Plaat DA , et al
      . Low serum DHEA-S is associated with impaired lung function in women. EClinicalMedicine 2020;23:100389. doi:10.1016/j.eclinm.2020.100389
    13. ↵
      2. Lenoir A ,
      3. Fuertes E ,
      4. Gómez-Real F , et al
      . Lung function changes over 8 years and testosterone markers in both sexes: UK Biobank. ERJ Open Res 2020;6:00070-2020. doi:10.1183/23120541.00070-2020
    14. ↵
      2. Zein JG ,
      3. McManus JM ,
      4. Sharifi N , et al
      . Benefits of airway androgen receptor expression in human asthma. Am J Respir Crit Care Med 2021;204:285–93. doi:10.1164/rccm.202009-3720OC
      OpenUrl
    15. ↵
      2. Han Y-Y ,
      3. Forno E ,
      4. Witchel SF , et al
      . Testosterone-to-estradiol ratio and lung function in a prospective study of Puerto Rican youth. Ann Allergy Asthma Immunol 2021;127:236–42. doi:10.1016/j.anai.2021.04.013
      OpenUrl
    16. ↵
      2. Arathimos R ,
      3. Millard LAC ,
      4. Bell JA , et al
      . Impact of sex hormone-binding globulin on the human phenome. Hum Mol Genet 2020;29:1824–32. doi:10.1093/hmg/ddz269
      OpenUrlCrossRef
    17. ↵
      2. Simó R ,
      3. Sáez-López C ,
      4. Barbosa-Desongles A , et al
      . Novel insights in SHBG regulation and clinical implications. Trends Endocrinol Metab 2015;26:376–83. doi:10.1016/j.tem.2015.05.001
      OpenUrlCrossRefPubMed
    18. ↵
      2. Dimou NL ,
      3. Papadimitriou N ,
      4. Gill D , et al
      . Sex hormone binding globulin and risk of breast cancer: a Mendelian randomization study. Int J Epidemiol 2019;48:807–16. doi:10.1093/ije/dyz107
      OpenUrlCrossRefPubMed
    19. ↵
      2. Zhao JV ,
      3. Schooling CM
      . Sex-specific associations of sex hormone binding globulin with CKD and kidney function: a univariable and multivariable mendelian randomization study in the UK biobank. JASN 2021;32:686–94. doi:10.1681/ASN.2020050659
      OpenUrlAbstract/FREE Full Text
    20. ↵
      2. Arathimos R ,
      3. Granell R ,
      4. Haycock P , et al
      . Genetic and observational evidence supports a causal role of sex hormones on the development of asthma. Thorax 2019;74:633–42. doi:10.1136/thoraxjnl-2018-212207
      OpenUrlAbstract/FREE Full Text
    21. ↵
      2. Markozannes G ,
      3. Kanellopoulou A ,
      4. Dimopoulou O , et al
      . Systematic review of Mendelian randomization studies on risk of cancer. BMC Med 2022;20:41. doi:10.1186/s12916-022-02246-y
    22. ↵
      2. van der Plaat DA ,
      3. Pereira M ,
      4. Pesce G , et al
      . Age at menopause and lung function: a mendelian randomisation study. Eur Respir J 2019;54:1802421. doi:10.1183/13993003.02421-2018
    23. ↵
      2. Davies NM ,
      3. Holmes MV ,
      4. Davey Smith G
      . Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians. BMJ 2018:k601. doi:10.1136/bmj.k601
    24. ↵
      2. Sheehan NA ,
      3. Didelez V ,
      4. Burton PR , et al
      . Mendelian randomisation and causal inference in observational epidemiology. PLoS Med 2008;5:e177. doi:10.1371/journal.pmed.0050177
    25. ↵
      2. Sudlow C ,
      3. Gallacher J ,
      4. Allen N , et al
      . UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med 2015;12:e1001779. doi:10.1371/journal.pmed.1001779
    26. ↵
      2. Bycroft C ,
      3. Freeman C ,
      4. Petkova D , et al
      . The UK Biobank resource with deep phenotyping and genomic data. Nature 2018;562:203–9. doi:10.1038/s41586-018-0579-z
      OpenUrlCrossRefPubMed
    27. ↵
      2. Wain LV ,
      3. Shrine N ,
      4. Miller S , et al
      . Novel insights into the genetics of smoking behaviour, lung function, and chronic obstructive pulmonary disease (UK BiLEVE): a genetic association study in UK Biobank. Lancet Respir Med 2015;3:769–81. doi:10.1016/S2213-2600(15)00283-0
      OpenUrl
    28. ↵
      2. Loh P-R ,
      3. Tucker G ,
      4. Bulik-Sullivan BK , et al
      . Efficient Bayesian mixed-model analysis increases association power in large cohorts. Nat Genet 2015;47:284–90. doi:10.1038/ng.3190
      OpenUrlCrossRefPubMed
    29. ↵
      2. Grant AJ ,
      3. Burgess S
      . Pleiotropy robust methods for multivariable mendelian randomization. Stat Med 2021;40:5813–30. doi:10.1002/sim.9156
      OpenUrl
    30. ↵
      2. Verbanck M ,
      3. Chen C-Y ,
      4. Neale B , et al
      . Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 2018;50:693–8. doi:10.1038/s41588-018-0099-7
      OpenUrlCrossRefPubMed
    31. ↵
      2. Han Y-Y ,
      3. Forno E ,
      4. Celedón JC
      . Sex steroid hormones and asthma in a nationwide study of U.S. adults. Am J Respir Crit Care Med 2020;201:158–66. doi:10.1164/rccm.201905-0996OC
      OpenUrlCrossRefPubMed
    32. ↵
      2. Pan L ,
      3. Wang M ,
      4. Xie X , et al
      . Effects of anabolic steroids on chronic obstructive pulmonary disease: a meta-analysis of randomised controlled trials. PLoS One 2014;9:e84855. doi:10.1371/journal.pone.0084855
    33. ↵
      2. Astle WJ ,
      3. Elding H ,
      4. Jiang T , et al
      . The allelic landscape of human blood cell trait variation and links to common complex disease. Cell 2016;167:1415–29. doi:10.1016/j.cell.2016.10.042
      OpenUrlCrossRefPubMed
    34. ↵
      2. Tan WC ,
      3. Bourbeau J ,
      4. Nadeau G , et al
      . High eosinophil counts predict decline in FEV1: results from the CanCOLD study. Eur Respir J 2021;57:2000838. doi:10.1183/13993003.00838-2020
    35. ↵
      2. Wu X ,
      3. Wang C ,
      4. Li H , et al
      . Circulating white blood cells and lung function impairment: the observational studies and mendelian randomization analysis. Ann Med 2021;53:1118–28. doi:10.1080/07853890.2021.1948603
      OpenUrl
    36. ↵
      2. Svartberg J ,
      3. Aasebø U ,
      4. Hjalmarsen A , et al
      . Testosterone treatment improves body composition and sexual function in men with COPD, in a 6-month randomized controlled trial. Respir Med 2004;98:906–13. doi:10.1016/j.rmed.2004.02.015
      OpenUrlCrossRefPubMedWeb of Science
    37. ↵
      2. Mohan D ,
      3. Rossiter H ,
      4. Watz H , et al
      . Selective androgen receptor modulation for muscle weakness in chronic obstructive pulmonary disease: a randomised control trial. Thorax 2023;78:258–66. doi:10.1136/thorax-2021-218360
      OpenUrlAbstract/FREE Full Text
    38. ↵
      2. Day FR ,
      3. Hinds DA ,
      4. Tung JY , et al
      . Causal mechanisms and balancing selection inferred from genetic associations with polycystic ovary syndrome. Nat Commun 2015;6:8464. doi:10.1038/ncomms9464
    39. ↵
      2. de Ronde W ,
      3. van der Schouw YT ,
      4. Pierik FH , et al
      . Serum levels of sex hormone-binding globulin (SHBG) are not associated with lower levels of non-SHBG-bound testosterone in male newborns and healthy adult men. Clin Endocrinol (Oxf) 2005;62:498–503. doi:10.1111/j.1365-2265.2005.02252.x
      OpenUrlCrossRefPubMed
    40. ↵
      2. Card JW ,
      3. Zeldin DC
      . Hormonal influences on lung function and response to environmental agents: lessons from animal models of respiratory disease. Proc Am Thorac Soc 2009;6:588–95. doi:10.1513/pats.200904-020RM
      OpenUrlPubMed
    41. ↵
      2. Keevil BG ,
      3. Adaway J
      . Assessment of free testosterone concentration. J Steroid Biochem Mol Biol 2019;190:207–11. doi:10.1016/j.jsbmb.2019.04.008
      OpenUrlCrossRef
    42. ↵
      2. Li H ,
      3. Pham T ,
      4. McWhinney BC , et al
      . Sex hormone binding globulin modifies testosterone action and metabolism in prostate cancer cells. Int J Endocrinol 2016;2016:6437585. doi:10.1155/2016/6437585

    Supplementary materials

    Footnotes

    • CM and BL are joint senior authors.

    • Correction notice This article has been corrected since it was published Online First. A typo in the title has been fixed.

    • Contributors DAvdP participated in the study design, analysis and interpretation of the data, and drafting of the manuscript, tables and figures, and is guarantor of the work. CM, BL and DAvdP participated in determining the study design and interpretation of the data. CM, BL, AL, SD, JP, FGR and DJ were involved in data collection and interpretation of the data. SOS, BL, CM, DJ, SD and FGR obtained funding. All authors read and approved the final version of the manuscript.

    • Funding The current study is part of the Ageing for Lungs in European Cohorts (ALEC) study (www.alecstudy.org), which has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement number 633212. This research has been conducted using the UK Biobank Resource under application number 19136, and we thank the participants, field workers and data managers for their time and cooperation. This work used the computing resources of the UK Medical Bionformatics partnership—aggregation, integration, visualisation and analysis of large, complex data (UK MED-BIO) which is supported by the Medical Research Council (grant number MR/L01632X/1).

    • Competing interests None declared.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

    Linked Articles