Analysis of two major Mendelian randomization studies revealed that endogenous testosterone production in men is positively correlated with blood clots, heart attack, and heart failure. Data from at least 200,000 participants were included in the analysis. A proposed mechanism for such a risk is the testosterone conversion into estrogen, in turn contributing to thromboembolism. Testosterone can also increase platelet aggregation via the thromboxane A2 pathway.
Data from these “natural experiments” overall follow the observed increased risk of deep venous thrombosis and heart disease in men who are over-supplemented with exogenous testosterone. On the contrary, low Testosterone levels are also associated with visceral adiposity, low muscle mass, and insulin resistance. From a cardiovascular perspective, future clinical research is needed to identify the balancing point of how much or little testosterone men should have.
GT
Also see:
BMJ
Mendelian Randomisation
March 2019
Objective
To determine whether endogenous testosterone has a causal role in thromboembolism, heart failure, and myocardial infarction.
Design
Two sample mendelian randomisation study using genetic variants as instrumental variables, randomly allocated at conception, to infer causality as additional randomised evidence.
Setting
Reduction by Dutasteride of Prostate Cancer Events (REDUCE) randomised controlled trial, UK Biobank, and CARDIoGRAMplusC4D 1000 Genomes based genome wide association study.
Participants
- These findings provide another strand of evidence consistent with the cardiovascular warnings about TRT issued by regulators.
Further evidence is needed to clarify whether our findings are relevant to the higher rates of these diseases in men than in women, or suggest that agents that lower testosterone would be protective. Additional research is also required comparing the effects of endogenous testosterone with those of exogenous testosterone.
Our study suggests that endogenous testosterone could have a role in thromboembolism, heart failure, and myocardial infarction in men. It might be worth considering whether existing treatments that modulate endogenous testosterone could be used for these conditions.
- Taken together, these findings suggest a common factor could underlie thromboembolism, heart failure, and myocardial infarction, and explain higher rates of these conditions in men than in women. Several effective treatments for cardiovascular disease including statins, digoxin, and some antihypertensives, such as spironolactone, reduce endogenous testosterone. Whether testosterone contributes to the mechanism of action of these treatments is not known, however the established targets of many other treatments for ischaemic heart disease do not seem to have clear genetic validation.
We used a new method to obtain unconfounded estimates of the effects of endogenous testosterone on thromboembolism, heart failure, and myocardial infarction. We used a validated myocardial infarction classification algorithm and validated the genetic variants as instrumental variables in the UK Biobank. Several statistical techniques and a validation study for myocardial infarction were also applied. However, our study had several limitations.
Mendelian randomisation has stringent assumptions. We checked for potential confounders of the genetic variant and disease outcome associations. We also used a Q statistic to test statistically for pleiotropic effects that might indicate violations of the exclusion restriction assumption. In addition, the genetic predictors of serum testosterone are not independent of sex hormone binding globulin because all the autosomal gene regions associated with testosterone concentrations at genome wide significance levels are also associated with sex hormone binding globulin.
Therefore, variants predicting testosterone from the SHBG gene region are open to the pleiotropic effects of sex hormone binding globulin. However, the estimates from JMJD1C variants could be least biased by sex hormone binding globulin because JMJD1C is probably relevant to male fertility and might have functional relevance to testosterone. Compared with estimates from the JMJD1C gene region, the observed reverse estimates from the SHBG gene region are consistent with the antagonistic relation between the bioavailability of testosterone and sex hormone binding globulin.
Furthermore, we used genetic predictors of serum testosterone derived from a sample of men to estimate testosterone in women, therefore the estimates for women should be interpreted with caution. However, because levels of testosterone are higher in men than in women, the stronger associations we found in men than in women are consistent with testosterone as the causal mechanism.
Although the largest available sources of genetic associations with thromboembolism and heart failure were used, the relatively low number of participants with heart failure led to imprecise estimates and wide 95% confidence intervals. The response rate of around 5.5% for the UK Biobank resulted in the recruitment of generally healthier participants, which might have biased towards the null. This could explain the discrepancy between the estimates from the UK Biobank and the myocardial infarction case-control study, although the difference might also have been because of a lack of power in the UK Biobank. Therefore, our positive estimates for thromboembolism and heart failure could be underestimated because of survivor bias in the UK Biobank.
Endogenous testosterone decreases with poor health, and the testosterone genome wide association study did not adjust for health status. Therefore, the estimates for genetic variants on testosterone in the genome wide association study could be biased towards the null and might be imprecise; however, such bias is probably minimal. Genetic associations with testosterone were estimated in men aged 50-75. We assume that these genetic associations reflect differences in testosterone concentrations that are also present across the age range (40-69) of UK Biobank participants. In addition, the UK Biobank has implemented high quality procedures to capture and classify events, but complete diagnostic accuracy is impossible. However, misclassification is unlikely to be related to genetic makeup, so any random misclassification of the outcomes would probably bias towards the null.
Our estimates represent average causal effects across the population and so might not be relevant for all subgroups of the population. We also considered venous and arterial thromboembolism together because they might share common risk factors, however, considering venous and arterial thromboembolism separately gives a similar interpretation. Finally, our study compared groups with genetically predicted higher and lower levels of endogenous testosterone to make inferences about the expected effect of raising testosterone. However, there are several qualitative differences between the comparisons that could limit the applicability of our findings to assess the effect of increasing testosterone levels. Specifically, genetic variants are associated with small but lifelong changes in endogenous testosterone levels that occur from modulating a particular biological pathway, whereas testosterone supplementation typically occurs later in life and increases testosterone levels by a relatively large amount.
From a clinical perspective, our study suggests that lifelong endogenous testosterone could have a role in thromboembolism, heart failure, and possibly myocardial infarction, particularly among men.
- These findings provide another strand of evidence consistent with the cardiovascular warnings about TRT issued by regulators.
Further evidence is needed to clarify whether our findings are relevant to the higher rates of these diseases in men than in women, or suggest that agents that lower testosterone would be protective. Additional research is also required comparing the effects of endogenous testosterone with those of exogenous testosterone.
Our study suggests that endogenous testosterone could have a role in thromboembolism, heart failure, and myocardial infarction in men. It might be worth considering whether existing treatments that modulate endogenous testosterone could be used for these conditions.
- Recent Endocrine Society clinical practice guidelines recommend against TRT in men with stroke, myocardial infarction, or thrombophilia.
To our knowledge, the new TRAVERSE trial is the first TRT trial with adequate power to assess cardiovascular events. The trial is designed to evaluate major adverse cardiac events (non-fatal myocardial infarction, non-fatal stroke, or death due to cardiovascular causes) in 6,000 men over 5 years; it is unlikely to provide definitive evidence about specific cardiovascular diseases and will take several years to complete.
When the role of TRT is hotly debated but experimental evidence is limited, mendelian randomisation using genetic variants as instrumental variables can support causal inferences about the effects of modifiable risk factors. Mendelian randomisation is less susceptible to confounding than traditional observational studies because genetic variants are randomly allocated at conception.
Therefore, mendelian randomisation is at the interface of experimental and observational studies, and can be used to obtain evidence in support of a potential causal effect or of potential targets of interventions. However, mendelian randomisation studies give the effects of lifetime exposures and so the numerical effect estimates provide a guide rather than the exact level of an intervention required.
A previous adequately powered mendelian randomisation study found preliminary evidence that endogenous testosterone is detrimental for ischaemic heart disease and ischaemic stroke, especially in men.
To clarify the role of testosterone in other types of cardiovascular disease, which might have different causes, we assessed the effect of endogenous testosterone on additional cardiovascular conditions in the UK Biobank participants: thromboembolism because of evidence from randomised controlled trials and specific warnings by the US Food and Drug Administration and Health Canada, heart failure because it can be a sequela of heart attack; and myocardial infarction for specificity.
We validated our findings, when possible, using publicly available consortiums. We also considered whether the associations of genetic predictors of endogenous testosterone with the cardiovascular diseases studied varied by sex because men have higher levels of endogenous testosterone than women.
We found that when using variants in the JMJD1C gene region, genetically predicted serum testosterone was positively associated with thromboembolism in the UK Biobank. This finding is consistent with the results of a large population based case-control study of venous thromboembolism in the UK and a small meta-analysis of randomised controlled trials of thromboembolism.
Genetically predicted serum testosterone was also associated with heart failure in men, however, it was not associated with myocardial infarction in the UK Biobank. Genetically predicted serum testosterone was nominally positively associated with myocardial infarction in the larger CARDIoGRAMplusC4D 1000 Genomes based genome wide association study.
Sex hormone binding globulin was positively associated with thromboembolism and inversely associated with myocardial infarction, which makes any estimates using variants from the SHBG gene region open to pleiotropic effects and difficult to interpret.
We found no previous mendelian randomisation study that assessed the effect of serum testosterone on thromboembolism or heart failure. One small mendelian randomisation study (n=1454) in men that used rs1799941 from the SHBG gene to predict testosterone found no association of testosterone with myocardial infarction. This finding corresponds with our conclusion that there is no clear association of testosterone predicted by SHBG gene region variants with myocardial infarction, and with the results for ischaemic heart disease in our previous mendelian randomisation study.
The effects of testosterone supplementation in men have not been widely studied. This is because in 2003 the Institute of Medicine recommended that large scale trials of TRT should not be conducted until any benefits of testosterone over and above established treatments had been demonstrated in small trials. Moreover, higher rates of cardiovascular disease in men than in women have previously been attributed to the protective effects of oestrogen in women, therefore, more investigations have focused on the role of oestrogen rather than that of testosterone in cardiovascular disease.
Several large trials examining the effects of oestrogen on cardiovascular disease have been conducted in men and women. A trial of the cardiovascular disease effects of testosterone in women has been performed, but such a trial examining testosterone in men has only just started. However, before the exclusive focus on the role of oestrogen as the key sex hormone relevant to cardiovascular disease had been firmly established, testosterone was shown to cause thrombosis in male mice.
Testosterone raises oestrogen in men, which is a known cause of thromboembolism. Testosterone also increases platelet aggregation through thromboxane A2, which could underlie any effects on thromboembolism. In mice, testosterone induces cardiac myocyte hypertrophy and antiandrogens improve cardiac function and reduce mortality.
Testosterone also raises endothelin, which causes ischaemic heart disease. We found an association between endogenous testosterone and a higher risk of thromboembolism, heart failure, and myocardial infarction, particularly in men. These results extend and complement previous findings of an association between endogenous testosterone and a higher risk of ischaemic heart disease and ischaemic stroke, particularly in men.
- Taken together, these findings suggest a common factor could underlie thromboembolism, heart failure, and myocardial infarction, and explain higher rates of these conditions in men than in women. Several effective treatments for cardiovascular disease including statins, digoxin, and some antihypertensives, such as spironolactone, reduce endogenous testosterone. Whether testosterone contributes to the mechanism of action of these treatments is not known, however the established targets of many other treatments for ischaemic heart disease do not seem to have clear genetic validation.
We used a new method to obtain unconfounded estimates of the effects of endogenous testosterone on thromboembolism, heart failure, and myocardial infarction. We used a validated myocardial infarction classification algorithm and validated the genetic variants as instrumental variables in the UK Biobank. Several statistical techniques and a validation study for myocardial infarction were also applied. However, our study had several limitations.
Mendelian randomisation has stringent assumptions. We checked for potential confounders of the genetic variant and disease outcome associations. We also used a Q statistic to test statistically for pleiotropic effects that might indicate violations of the exclusion restriction assumption. In addition, the genetic predictors of serum testosterone are not independent of sex hormone binding globulin because all the autosomal gene regions associated with testosterone concentrations at genome wide significance levels are also associated with sex hormone binding globulin.
Therefore, variants predicting testosterone from the SHBG gene region are open to the pleiotropic effects of sex hormone binding globulin. However, the estimates from JMJD1C variants could be least biased by sex hormone binding globulin because JMJD1C is probably relevant to male fertility and might have functional relevance to testosterone. Compared with estimates from the JMJD1C gene region, the observed reverse estimates from the SHBG gene region are consistent with the antagonistic relation between the bioavailability of testosterone and sex hormone binding globulin.
Furthermore, we used genetic predictors of serum testosterone derived from a sample of men to estimate testosterone in women, therefore the estimates for women should be interpreted with caution. However, because levels of testosterone are higher in men than in women, the stronger associations we found in men than in women are consistent with testosterone as the causal mechanism.
Although the largest available sources of genetic associations with thromboembolism and heart failure were used, the relatively low number of participants with heart failure led to imprecise estimates and wide 95% confidence intervals. The response rate of around 5.5% for the UK Biobank resulted in the recruitment of generally healthier participants, which might have biased towards the null. This could explain the discrepancy between the estimates from the UK Biobank and the myocardial infarction case-control study, although the difference might also have been because of a lack of power in the UK Biobank. Therefore, our positive estimates for thromboembolism and heart failure could be underestimated because of survivor bias in the UK Biobank.
Endogenous testosterone decreases with poor health, and the testosterone genome wide association study did not adjust for health status. Therefore, the estimates for genetic variants on testosterone in the genome wide association study could be biased towards the null and might be imprecise; however, such bias is probably minimal. Genetic associations with testosterone were estimated in men aged 50-75. We assume that these genetic associations reflect differences in testosterone concentrations that are also present across the age range (40-69) of UK Biobank participants. In addition, the UK Biobank has implemented high quality procedures to capture and classify events, but complete diagnostic accuracy is impossible. However, misclassification is unlikely to be related to genetic makeup, so any random misclassification of the outcomes would probably bias towards the null.
Our estimates represent average causal effects across the population and so might not be relevant for all subgroups of the population. We also considered venous and arterial thromboembolism together because they might share common risk factors, however, considering venous and arterial thromboembolism separately gives a similar interpretation. Finally, our study compared groups with genetically predicted higher and lower levels of endogenous testosterone to make inferences about the expected effect of raising testosterone. However, there are several qualitative differences between the comparisons that could limit the applicability of our findings to assess the effect of increasing testosterone levels. Specifically, genetic variants are associated with small but lifelong changes in endogenous testosterone levels that occur from modulating a particular biological pathway, whereas testosterone supplementation typically occurs later in life and increases testosterone levels by a relatively large amount.
From a clinical perspective, our study suggests that lifelong endogenous testosterone could have a role in thromboembolism, heart failure, and possibly myocardial infarction, particularly among men.
- These findings provide another strand of evidence consistent with the cardiovascular warnings about TRT issued by regulators.
Further evidence is needed to clarify whether our findings are relevant to the higher rates of these diseases in men than in women, or suggest that agents that lower testosterone would be protective. Additional research is also required comparing the effects of endogenous testosterone with those of exogenous testosterone.
Our study suggests that endogenous testosterone could have a role in thromboembolism, heart failure, and myocardial infarction in men. It might be worth considering whether existing treatments that modulate endogenous testosterone could be used for these conditions.
- Thromboembolism (OR per unit increase in log transformed testosterone (nmol/L) 2.09, p<0.05)
- Heart failure (OR 7.81, p<0.05)
- But not myocardial infarction (1.17, 0.78-1.75).
Associations were less obvious in women.
In the validation study, genetically predicted testosterone (based on JMJD1C gene region variants) was positively associated with myocardial infarction (OR 1.37, p<0.05).
No excess heterogeneity was observed among genetic variants in their associations with the outcomes. However, testosterone genetically predicted by potentially pleiotropic variants in the SHBG gene region had no association with the outcomes.
Conclusions
- Endogenous testosterone was positively associated with thromboembolism, heart failure, and myocardial infarction in men.
- Rates of these conditions are higher in men than women.
- Endogenous testosterone can be controlled with existing treatments and could be a modifiable risk factor for thromboembolism and heart failure.
More from the publication
Testosterone sales increased 12-fold globally from 2000 to 2011, with noticeable increases in North America. In the United States, older men have probably been targeted in addition to men requiring medically indicated treatment for low testosterone levels.
Testosterone replacement therapy (TRT) prescriptions in the US decreased by 50% between 2013 and 2016, but remain well above the levels needed to treat pathological hypogonadism. Since the 1970s, use of anabolic steroids has spread from athletes to the general population, with a lifetime prevalence rate of 6.4% for men.
Observational studies of the association of measured endogenous testosterone with overall and specific cardiovascular diseases, including thromboembolism, myocardial infarction, and heart failure, are difficult to interpret. These studies are inherently open to confounding by obesity and ill health, which could reduce testosterone and are well established causes of cardiovascular diseases.
Therefore, it is unclear from observational studies whether testosterone has a role or is an indicator of poor general health, particularly because testosterone might be affected by some cardioprotective treatments. Pharmacoepidemiology studies of drugs are open to subtle time related biases, such as immortal time bias, which are not always taken into consideration in studies of TRT.
Studies of TRT using self comparisons to avoid confounding and time related biases have suggested that TRT could increase myocardial infarction and possibly cardiovascular events. Randomised placebo controlled trials of TRT are limited in size and scale. Therefore, systematic reviews and meta-analyses of randomised controlled trials are usually too small to be definitive overall or for any specific types of cardiovascular disease, although adverse effects of TRT on thromboembolism have been found.
- Recent Endocrine Society clinical practice guidelines recommend against TRT in men with stroke, myocardial infarction, or thrombophilia.
To our knowledge, the new TRAVERSE trial is the first TRT trial with adequate power to assess cardiovascular events. The trial is designed to evaluate major adverse cardiac events (non-fatal myocardial infarction, non-fatal stroke, or death due to cardiovascular causes) in 6,000 men over 5 years; it is unlikely to provide definitive evidence about specific cardiovascular diseases and will take several years to complete.
When the role of TRT is hotly debated but experimental evidence is limited, mendelian randomisation using genetic variants as instrumental variables can support causal inferences about the effects of modifiable risk factors. Mendelian randomisation is less susceptible to confounding than traditional observational studies because genetic variants are randomly allocated at conception.
Therefore, mendelian randomisation is at the interface of experimental and observational studies, and can be used to obtain evidence in support of a potential causal effect or of potential targets of interventions. However, mendelian randomisation studies give the effects of lifetime exposures and so the numerical effect estimates provide a guide rather than the exact level of an intervention required.
A previous adequately powered mendelian randomisation study found preliminary evidence that endogenous testosterone is detrimental for ischaemic heart disease and ischaemic stroke, especially in men.
To clarify the role of testosterone in other types of cardiovascular disease, which might have different causes, we assessed the effect of endogenous testosterone on additional cardiovascular conditions in the UK Biobank participants: thromboembolism because of evidence from randomised controlled trials and specific warnings by the US Food and Drug Administration and Health Canada, heart failure because it can be a sequela of heart attack; and myocardial infarction for specificity.
We validated our findings, when possible, using publicly available consortiums. We also considered whether the associations of genetic predictors of endogenous testosterone with the cardiovascular diseases studied varied by sex because men have higher levels of endogenous testosterone than women.
We found that when using variants in the JMJD1C gene region, genetically predicted serum testosterone was positively associated with thromboembolism in the UK Biobank. This finding is consistent with the results of a large population based case-control study of venous thromboembolism in the UK and a small meta-analysis of randomised controlled trials of thromboembolism.
Genetically predicted serum testosterone was also associated with heart failure in men, however, it was not associated with myocardial infarction in the UK Biobank. Genetically predicted serum testosterone was nominally positively associated with myocardial infarction in the larger CARDIoGRAMplusC4D 1000 Genomes based genome wide association study.
Sex hormone binding globulin was positively associated with thromboembolism and inversely associated with myocardial infarction, which makes any estimates using variants from the SHBG gene region open to pleiotropic effects and difficult to interpret.
We found no previous mendelian randomisation study that assessed the effect of serum testosterone on thromboembolism or heart failure. One small mendelian randomisation study (n=1454) in men that used rs1799941 from the SHBG gene to predict testosterone found no association of testosterone with myocardial infarction. This finding corresponds with our conclusion that there is no clear association of testosterone predicted by SHBG gene region variants with myocardial infarction, and with the results for ischaemic heart disease in our previous mendelian randomisation study.
The effects of testosterone supplementation in men have not been widely studied. This is because in 2003 the Institute of Medicine recommended that large scale trials of TRT should not be conducted until any benefits of testosterone over and above established treatments had been demonstrated in small trials. Moreover, higher rates of cardiovascular disease in men than in women have previously been attributed to the protective effects of oestrogen in women, therefore, more investigations have focused on the role of oestrogen rather than that of testosterone in cardiovascular disease.
Several large trials examining the effects of oestrogen on cardiovascular disease have been conducted in men and women. A trial of the cardiovascular disease effects of testosterone in women has been performed, but such a trial examining testosterone in men has only just started. However, before the exclusive focus on the role of oestrogen as the key sex hormone relevant to cardiovascular disease had been firmly established, testosterone was shown to cause thrombosis in male mice.
Testosterone raises oestrogen in men, which is a known cause of thromboembolism. Testosterone also increases platelet aggregation through thromboxane A2, which could underlie any effects on thromboembolism. In mice, testosterone induces cardiac myocyte hypertrophy and antiandrogens improve cardiac function and reduce mortality.
Testosterone also raises endothelin, which causes ischaemic heart disease. We found an association between endogenous testosterone and a higher risk of thromboembolism, heart failure, and myocardial infarction, particularly in men. These results extend and complement previous findings of an association between endogenous testosterone and a higher risk of ischaemic heart disease and ischaemic stroke, particularly in men.
- Taken together, these findings suggest a common factor could underlie thromboembolism, heart failure, and myocardial infarction, and explain higher rates of these conditions in men than in women. Several effective treatments for cardiovascular disease including statins, digoxin, and some antihypertensives, such as spironolactone, reduce endogenous testosterone. Whether testosterone contributes to the mechanism of action of these treatments is not known, however the established targets of many other treatments for ischaemic heart disease do not seem to have clear genetic validation.
We used a new method to obtain unconfounded estimates of the effects of endogenous testosterone on thromboembolism, heart failure, and myocardial infarction. We used a validated myocardial infarction classification algorithm and validated the genetic variants as instrumental variables in the UK Biobank. Several statistical techniques and a validation study for myocardial infarction were also applied. However, our study had several limitations.
Mendelian randomisation has stringent assumptions. We checked for potential confounders of the genetic variant and disease outcome associations. We also used a Q statistic to test statistically for pleiotropic effects that might indicate violations of the exclusion restriction assumption. In addition, the genetic predictors of serum testosterone are not independent of sex hormone binding globulin because all the autosomal gene regions associated with testosterone concentrations at genome wide significance levels are also associated with sex hormone binding globulin.
Therefore, variants predicting testosterone from the SHBG gene region are open to the pleiotropic effects of sex hormone binding globulin. However, the estimates from JMJD1C variants could be least biased by sex hormone binding globulin because JMJD1C is probably relevant to male fertility and might have functional relevance to testosterone. Compared with estimates from the JMJD1C gene region, the observed reverse estimates from the SHBG gene region are consistent with the antagonistic relation between the bioavailability of testosterone and sex hormone binding globulin.
Furthermore, we used genetic predictors of serum testosterone derived from a sample of men to estimate testosterone in women, therefore the estimates for women should be interpreted with caution. However, because levels of testosterone are higher in men than in women, the stronger associations we found in men than in women are consistent with testosterone as the causal mechanism.
Although the largest available sources of genetic associations with thromboembolism and heart failure were used, the relatively low number of participants with heart failure led to imprecise estimates and wide 95% confidence intervals. The response rate of around 5.5% for the UK Biobank resulted in the recruitment of generally healthier participants, which might have biased towards the null. This could explain the discrepancy between the estimates from the UK Biobank and the myocardial infarction case-control study, although the difference might also have been because of a lack of power in the UK Biobank. Therefore, our positive estimates for thromboembolism and heart failure could be underestimated because of survivor bias in the UK Biobank.
Endogenous testosterone decreases with poor health, and the testosterone genome wide association study did not adjust for health status. Therefore, the estimates for genetic variants on testosterone in the genome wide association study could be biased towards the null and might be imprecise; however, such bias is probably minimal. Genetic associations with testosterone were estimated in men aged 50-75. We assume that these genetic associations reflect differences in testosterone concentrations that are also present across the age range (40-69) of UK Biobank participants. In addition, the UK Biobank has implemented high quality procedures to capture and classify events, but complete diagnostic accuracy is impossible. However, misclassification is unlikely to be related to genetic makeup, so any random misclassification of the outcomes would probably bias towards the null.
Our estimates represent average causal effects across the population and so might not be relevant for all subgroups of the population. We also considered venous and arterial thromboembolism together because they might share common risk factors, however, considering venous and arterial thromboembolism separately gives a similar interpretation. Finally, our study compared groups with genetically predicted higher and lower levels of endogenous testosterone to make inferences about the expected effect of raising testosterone. However, there are several qualitative differences between the comparisons that could limit the applicability of our findings to assess the effect of increasing testosterone levels. Specifically, genetic variants are associated with small but lifelong changes in endogenous testosterone levels that occur from modulating a particular biological pathway, whereas testosterone supplementation typically occurs later in life and increases testosterone levels by a relatively large amount.
From a clinical perspective, our study suggests that lifelong endogenous testosterone could have a role in thromboembolism, heart failure, and possibly myocardial infarction, particularly among men.
- These findings provide another strand of evidence consistent with the cardiovascular warnings about TRT issued by regulators.
Further evidence is needed to clarify whether our findings are relevant to the higher rates of these diseases in men than in women, or suggest that agents that lower testosterone would be protective. Additional research is also required comparing the effects of endogenous testosterone with those of exogenous testosterone.
Our study suggests that endogenous testosterone could have a role in thromboembolism, heart failure, and myocardial infarction in men. It might be worth considering whether existing treatments that modulate endogenous testosterone could be used for these conditions.
- 13,691 had thromboembolism (6,208 men – 7,483 women),
- 1,688 had heart failure (1186 – 502), and
- 12,882 had myocardial infarction (10,136 – 2746).
In men, endogenous testosterone genetically predicted by variants in the JMJD1C gene region was positively associated with:
- Thromboembolism (OR per unit increase in log transformed testosterone (nmol/L) 2.09, p<0.05)
- Heart failure (OR 7.81, p<0.05)
- But not myocardial infarction (1.17, 0.78-1.75).
Associations were less obvious in women.
In the validation study, genetically predicted testosterone (based on JMJD1C gene region variants) was positively associated with myocardial infarction (OR 1.37, p<0.05).
No excess heterogeneity was observed among genetic variants in their associations with the outcomes. However, testosterone genetically predicted by potentially pleiotropic variants in the SHBG gene region had no association with the outcomes.
Conclusions
- Endogenous testosterone was positively associated with thromboembolism, heart failure, and myocardial infarction in men.
- Rates of these conditions are higher in men than women.
- Endogenous testosterone can be controlled with existing treatments and could be a modifiable risk factor for thromboembolism and heart failure.
More from the publication
Testosterone sales increased 12-fold globally from 2000 to 2011, with noticeable increases in North America. In the United States, older men have probably been targeted in addition to men requiring medically indicated treatment for low testosterone levels.
Testosterone replacement therapy (TRT) prescriptions in the US decreased by 50% between 2013 and 2016, but remain well above the levels needed to treat pathological hypogonadism. Since the 1970s, use of anabolic steroids has spread from athletes to the general population, with a lifetime prevalence rate of 6.4% for men.
Observational studies of the association of measured endogenous testosterone with overall and specific cardiovascular diseases, including thromboembolism, myocardial infarction, and heart failure, are difficult to interpret. These studies are inherently open to confounding by obesity and ill health, which could reduce testosterone and are well established causes of cardiovascular diseases.
Therefore, it is unclear from observational studies whether testosterone has a role or is an indicator of poor general health, particularly because testosterone might be affected by some cardioprotective treatments. Pharmacoepidemiology studies of drugs are open to subtle time related biases, such as immortal time bias, which are not always taken into consideration in studies of TRT.
Studies of TRT using self comparisons to avoid confounding and time related biases have suggested that TRT could increase myocardial infarction and possibly cardiovascular events. Randomised placebo controlled trials of TRT are limited in size and scale. Therefore, systematic reviews and meta-analyses of randomised controlled trials are usually too small to be definitive overall or for any specific types of cardiovascular disease, although adverse effects of TRT on thromboembolism have been found.
- Recent Endocrine Society clinical practice guidelines recommend against TRT in men with stroke, myocardial infarction, or thrombophilia.
To our knowledge, the new TRAVERSE trial is the first TRT trial with adequate power to assess cardiovascular events. The trial is designed to evaluate major adverse cardiac events (non-fatal myocardial infarction, non-fatal stroke, or death due to cardiovascular causes) in 6,000 men over 5 years; it is unlikely to provide definitive evidence about specific cardiovascular diseases and will take several years to complete.
When the role of TRT is hotly debated but experimental evidence is limited, mendelian randomisation using genetic variants as instrumental variables can support causal inferences about the effects of modifiable risk factors. Mendelian randomisation is less susceptible to confounding than traditional observational studies because genetic variants are randomly allocated at conception.
Therefore, mendelian randomisation is at the interface of experimental and observational studies, and can be used to obtain evidence in support of a potential causal effect or of potential targets of interventions. However, mendelian randomisation studies give the effects of lifetime exposures and so the numerical effect estimates provide a guide rather than the exact level of an intervention required.
A previous adequately powered mendelian randomisation study found preliminary evidence that endogenous testosterone is detrimental for ischaemic heart disease and ischaemic stroke, especially in men.
To clarify the role of testosterone in other types of cardiovascular disease, which might have different causes, we assessed the effect of endogenous testosterone on additional cardiovascular conditions in the UK Biobank participants: thromboembolism because of evidence from randomised controlled trials and specific warnings by the US Food and Drug Administration and Health Canada, heart failure because it can be a sequela of heart attack; and myocardial infarction for specificity.
We validated our findings, when possible, using publicly available consortiums. We also considered whether the associations of genetic predictors of endogenous testosterone with the cardiovascular diseases studied varied by sex because men have higher levels of endogenous testosterone than women.
We found that when using variants in the JMJD1C gene region, genetically predicted serum testosterone was positively associated with thromboembolism in the UK Biobank. This finding is consistent with the results of a large population based case-control study of venous thromboembolism in the UK and a small meta-analysis of randomised controlled trials of thromboembolism.
Genetically predicted serum testosterone was also associated with heart failure in men, however, it was not associated with myocardial infarction in the UK Biobank. Genetically predicted serum testosterone was nominally positively associated with myocardial infarction in the larger CARDIoGRAMplusC4D 1000 Genomes based genome wide association study.
Sex hormone binding globulin was positively associated with thromboembolism and inversely associated with myocardial infarction, which makes any estimates using variants from the SHBG gene region open to pleiotropic effects and difficult to interpret.
We found no previous mendelian randomisation study that assessed the effect of serum testosterone on thromboembolism or heart failure. One small mendelian randomisation study (n=1454) in men that used rs1799941 from the SHBG gene to predict testosterone found no association of testosterone with myocardial infarction. This finding corresponds with our conclusion that there is no clear association of testosterone predicted by SHBG gene region variants with myocardial infarction, and with the results for ischaemic heart disease in our previous mendelian randomisation study.
The effects of testosterone supplementation in men have not been widely studied. This is because in 2003 the Institute of Medicine recommended that large scale trials of TRT should not be conducted until any benefits of testosterone over and above established treatments had been demonstrated in small trials. Moreover, higher rates of cardiovascular disease in men than in women have previously been attributed to the protective effects of oestrogen in women, therefore, more investigations have focused on the role of oestrogen rather than that of testosterone in cardiovascular disease.
Several large trials examining the effects of oestrogen on cardiovascular disease have been conducted in men and women. A trial of the cardiovascular disease effects of testosterone in women has been performed, but such a trial examining testosterone in men has only just started. However, before the exclusive focus on the role of oestrogen as the key sex hormone relevant to cardiovascular disease had been firmly established, testosterone was shown to cause thrombosis in male mice.
Testosterone raises oestrogen in men, which is a known cause of thromboembolism. Testosterone also increases platelet aggregation through thromboxane A2, which could underlie any effects on thromboembolism. In mice, testosterone induces cardiac myocyte hypertrophy and antiandrogens improve cardiac function and reduce mortality.
Testosterone also raises endothelin, which causes ischaemic heart disease. We found an association between endogenous testosterone and a higher risk of thromboembolism, heart failure, and myocardial infarction, particularly in men. These results extend and complement previous findings of an association between endogenous testosterone and a higher risk of ischaemic heart disease and ischaemic stroke, particularly in men.
- Taken together, these findings suggest a common factor could underlie thromboembolism, heart failure, and myocardial infarction, and explain higher rates of these conditions in men than in women. Several effective treatments for cardiovascular disease including statins, digoxin, and some antihypertensives, such as spironolactone, reduce endogenous testosterone. Whether testosterone contributes to the mechanism of action of these treatments is not known, however the established targets of many other treatments for ischaemic heart disease do not seem to have clear genetic validation.
We used a new method to obtain unconfounded estimates of the effects of endogenous testosterone on thromboembolism, heart failure, and myocardial infarction. We used a validated myocardial infarction classification algorithm and validated the genetic variants as instrumental variables in the UK Biobank. Several statistical techniques and a validation study for myocardial infarction were also applied. However, our study had several limitations.
Mendelian randomisation has stringent assumptions. We checked for potential confounders of the genetic variant and disease outcome associations. We also used a Q statistic to test statistically for pleiotropic effects that might indicate violations of the exclusion restriction assumption. In addition, the genetic predictors of serum testosterone are not independent of sex hormone binding globulin because all the autosomal gene regions associated with testosterone concentrations at genome wide significance levels are also associated with sex hormone binding globulin.
Therefore, variants predicting testosterone from the SHBG gene region are open to the pleiotropic effects of sex hormone binding globulin. However, the estimates from JMJD1C variants could be least biased by sex hormone binding globulin because JMJD1C is probably relevant to male fertility and might have functional relevance to testosterone. Compared with estimates from the JMJD1C gene region, the observed reverse estimates from the SHBG gene region are consistent with the antagonistic relation between the bioavailability of testosterone and sex hormone binding globulin.
Furthermore, we used genetic predictors of serum testosterone derived from a sample of men to estimate testosterone in women, therefore the estimates for women should be interpreted with caution. However, because levels of testosterone are higher in men than in women, the stronger associations we found in men than in women are consistent with testosterone as the causal mechanism.
Although the largest available sources of genetic associations with thromboembolism and heart failure were used, the relatively low number of participants with heart failure led to imprecise estimates and wide 95% confidence intervals. The response rate of around 5.5% for the UK Biobank resulted in the recruitment of generally healthier participants, which might have biased towards the null. This could explain the discrepancy between the estimates from the UK Biobank and the myocardial infarction case-control study, although the difference might also have been because of a lack of power in the UK Biobank. Therefore, our positive estimates for thromboembolism and heart failure could be underestimated because of survivor bias in the UK Biobank.
Endogenous testosterone decreases with poor health, and the testosterone genome wide association study did not adjust for health status. Therefore, the estimates for genetic variants on testosterone in the genome wide association study could be biased towards the null and might be imprecise; however, such bias is probably minimal. Genetic associations with testosterone were estimated in men aged 50-75. We assume that these genetic associations reflect differences in testosterone concentrations that are also present across the age range (40-69) of UK Biobank participants. In addition, the UK Biobank has implemented high quality procedures to capture and classify events, but complete diagnostic accuracy is impossible. However, misclassification is unlikely to be related to genetic makeup, so any random misclassification of the outcomes would probably bias towards the null.
Our estimates represent average causal effects across the population and so might not be relevant for all subgroups of the population. We also considered venous and arterial thromboembolism together because they might share common risk factors, however, considering venous and arterial thromboembolism separately gives a similar interpretation. Finally, our study compared groups with genetically predicted higher and lower levels of endogenous testosterone to make inferences about the expected effect of raising testosterone. However, there are several qualitative differences between the comparisons that could limit the applicability of our findings to assess the effect of increasing testosterone levels. Specifically, genetic variants are associated with small but lifelong changes in endogenous testosterone levels that occur from modulating a particular biological pathway, whereas testosterone supplementation typically occurs later in life and increases testosterone levels by a relatively large amount.
From a clinical perspective, our study suggests that lifelong endogenous testosterone could have a role in thromboembolism, heart failure, and possibly myocardial infarction, particularly among men.
- These findings provide another strand of evidence consistent with the cardiovascular warnings about TRT issued by regulators.
Further evidence is needed to clarify whether our findings are relevant to the higher rates of these diseases in men than in women, or suggest that agents that lower testosterone would be protective. Additional research is also required comparing the effects of endogenous testosterone with those of exogenous testosterone.
Our study suggests that endogenous testosterone could have a role in thromboembolism, heart failure, and myocardial infarction in men. It might be worth considering whether existing treatments that modulate endogenous testosterone could be used for these conditions.
- 3,225 men of European ancestry aged 50-75 in REDUCE;
- 392,038 white British men and women aged 40-69 from the UK Biobank, and
- 171,875 participants of about 77% European descent, from CARDIoGRAMplusC4D 1000 Genomes based study for validation.
Main outcomes
Thromboembolism, heart failure, and myocardial infarction based on self reports, hospital episodes, and death.
Results
Of the UK Biobank participants:
- 13,691 had thromboembolism (6,208 men – 7,483 women),
- 1,688 had heart failure (1186 – 502), and
- 12,882 had myocardial infarction (10,136 – 2746).
In men, endogenous testosterone genetically predicted by variants in the JMJD1C gene region was positively associated with:
- Thromboembolism (OR per unit increase in log transformed testosterone (nmol/L) 2.09, p<0.05)
- Heart failure (OR 7.81, p<0.05)
- But not myocardial infarction (1.17, 0.78-1.75).
Associations were less obvious in women.
In the validation study, genetically predicted testosterone (based on JMJD1C gene region variants) was positively associated with myocardial infarction (OR 1.37, p<0.05).
No excess heterogeneity was observed among genetic variants in their associations with the outcomes. However, testosterone genetically predicted by potentially pleiotropic variants in the SHBG gene region had no association with the outcomes.
Conclusions
- Endogenous testosterone was positively associated with thromboembolism, heart failure, and myocardial infarction in men.
- Rates of these conditions are higher in men than women.
- Endogenous testosterone can be controlled with existing treatments and could be a modifiable risk factor for thromboembolism and heart failure.
More from the publication
Testosterone sales increased 12-fold globally from 2000 to 2011, with noticeable increases in North America. In the United States, older men have probably been targeted in addition to men requiring medically indicated treatment for low testosterone levels.
Testosterone replacement therapy (TRT) prescriptions in the US decreased by 50% between 2013 and 2016, but remain well above the levels needed to treat pathological hypogonadism. Since the 1970s, use of anabolic steroids has spread from athletes to the general population, with a lifetime prevalence rate of 6.4% for men.
Observational studies of the association of measured endogenous testosterone with overall and specific cardiovascular diseases, including thromboembolism, myocardial infarction, and heart failure, are difficult to interpret. These studies are inherently open to confounding by obesity and ill health, which could reduce testosterone and are well established causes of cardiovascular diseases.
Therefore, it is unclear from observational studies whether testosterone has a role or is an indicator of poor general health, particularly because testosterone might be affected by some cardioprotective treatments. Pharmacoepidemiology studies of drugs are open to subtle time related biases, such as immortal time bias, which are not always taken into consideration in studies of TRT.
Studies of TRT using self comparisons to avoid confounding and time related biases have suggested that TRT could increase myocardial infarction and possibly cardiovascular events. Randomised placebo controlled trials of TRT are limited in size and scale. Therefore, systematic reviews and meta-analyses of randomised controlled trials are usually too small to be definitive overall or for any specific types of cardiovascular disease, although adverse effects of TRT on thromboembolism have been found.
- Recent Endocrine Society clinical practice guidelines recommend against TRT in men with stroke, myocardial infarction, or thrombophilia.
To our knowledge, the new TRAVERSE trial is the first TRT trial with adequate power to assess cardiovascular events. The trial is designed to evaluate major adverse cardiac events (non-fatal myocardial infarction, non-fatal stroke, or death due to cardiovascular causes) in 6,000 men over 5 years; it is unlikely to provide definitive evidence about specific cardiovascular diseases and will take several years to complete.
When the role of TRT is hotly debated but experimental evidence is limited, mendelian randomisation using genetic variants as instrumental variables can support causal inferences about the effects of modifiable risk factors. Mendelian randomisation is less susceptible to confounding than traditional observational studies because genetic variants are randomly allocated at conception.
Therefore, mendelian randomisation is at the interface of experimental and observational studies, and can be used to obtain evidence in support of a potential causal effect or of potential targets of interventions. However, mendelian randomisation studies give the effects of lifetime exposures and so the numerical effect estimates provide a guide rather than the exact level of an intervention required.
A previous adequately powered mendelian randomisation study found preliminary evidence that endogenous testosterone is detrimental for ischaemic heart disease and ischaemic stroke, especially in men.
To clarify the role of testosterone in other types of cardiovascular disease, which might have different causes, we assessed the effect of endogenous testosterone on additional cardiovascular conditions in the UK Biobank participants: thromboembolism because of evidence from randomised controlled trials and specific warnings by the US Food and Drug Administration and Health Canada, heart failure because it can be a sequela of heart attack; and myocardial infarction for specificity.
We validated our findings, when possible, using publicly available consortiums. We also considered whether the associations of genetic predictors of endogenous testosterone with the cardiovascular diseases studied varied by sex because men have higher levels of endogenous testosterone than women.
We found that when using variants in the JMJD1C gene region, genetically predicted serum testosterone was positively associated with thromboembolism in the UK Biobank. This finding is consistent with the results of a large population based case-control study of venous thromboembolism in the UK and a small meta-analysis of randomised controlled trials of thromboembolism.
Genetically predicted serum testosterone was also associated with heart failure in men, however, it was not associated with myocardial infarction in the UK Biobank. Genetically predicted serum testosterone was nominally positively associated with myocardial infarction in the larger CARDIoGRAMplusC4D 1000 Genomes based genome wide association study.
Sex hormone binding globulin was positively associated with thromboembolism and inversely associated with myocardial infarction, which makes any estimates using variants from the SHBG gene region open to pleiotropic effects and difficult to interpret.
We found no previous mendelian randomisation study that assessed the effect of serum testosterone on thromboembolism or heart failure. One small mendelian randomisation study (n=1454) in men that used rs1799941 from the SHBG gene to predict testosterone found no association of testosterone with myocardial infarction. This finding corresponds with our conclusion that there is no clear association of testosterone predicted by SHBG gene region variants with myocardial infarction, and with the results for ischaemic heart disease in our previous mendelian randomisation study.
The effects of testosterone supplementation in men have not been widely studied. This is because in 2003 the Institute of Medicine recommended that large scale trials of TRT should not be conducted until any benefits of testosterone over and above established treatments had been demonstrated in small trials. Moreover, higher rates of cardiovascular disease in men than in women have previously been attributed to the protective effects of oestrogen in women, therefore, more investigations have focused on the role of oestrogen rather than that of testosterone in cardiovascular disease.
Several large trials examining the effects of oestrogen on cardiovascular disease have been conducted in men and women. A trial of the cardiovascular disease effects of testosterone in women has been performed, but such a trial examining testosterone in men has only just started. However, before the exclusive focus on the role of oestrogen as the key sex hormone relevant to cardiovascular disease had been firmly established, testosterone was shown to cause thrombosis in male mice.
Testosterone raises oestrogen in men, which is a known cause of thromboembolism. Testosterone also increases platelet aggregation through thromboxane A2, which could underlie any effects on thromboembolism. In mice, testosterone induces cardiac myocyte hypertrophy and antiandrogens improve cardiac function and reduce mortality.
Testosterone also raises endothelin, which causes ischaemic heart disease. We found an association between endogenous testosterone and a higher risk of thromboembolism, heart failure, and myocardial infarction, particularly in men. These results extend and complement previous findings of an association between endogenous testosterone and a higher risk of ischaemic heart disease and ischaemic stroke, particularly in men.
- Taken together, these findings suggest a common factor could underlie thromboembolism, heart failure, and myocardial infarction, and explain higher rates of these conditions in men than in women. Several effective treatments for cardiovascular disease including statins, digoxin, and some antihypertensives, such as spironolactone, reduce endogenous testosterone. Whether testosterone contributes to the mechanism of action of these treatments is not known, however the established targets of many other treatments for ischaemic heart disease do not seem to have clear genetic validation.
We used a new method to obtain unconfounded estimates of the effects of endogenous testosterone on thromboembolism, heart failure, and myocardial infarction. We used a validated myocardial infarction classification algorithm and validated the genetic variants as instrumental variables in the UK Biobank. Several statistical techniques and a validation study for myocardial infarction were also applied. However, our study had several limitations.
Mendelian randomisation has stringent assumptions. We checked for potential confounders of the genetic variant and disease outcome associations. We also used a Q statistic to test statistically for pleiotropic effects that might indicate violations of the exclusion restriction assumption. In addition, the genetic predictors of serum testosterone are not independent of sex hormone binding globulin because all the autosomal gene regions associated with testosterone concentrations at genome wide significance levels are also associated with sex hormone binding globulin.
Therefore, variants predicting testosterone from the SHBG gene region are open to the pleiotropic effects of sex hormone binding globulin. However, the estimates from JMJD1C variants could be least biased by sex hormone binding globulin because JMJD1C is probably relevant to male fertility and might have functional relevance to testosterone. Compared with estimates from the JMJD1C gene region, the observed reverse estimates from the SHBG gene region are consistent with the antagonistic relation between the bioavailability of testosterone and sex hormone binding globulin.
Furthermore, we used genetic predictors of serum testosterone derived from a sample of men to estimate testosterone in women, therefore the estimates for women should be interpreted with caution. However, because levels of testosterone are higher in men than in women, the stronger associations we found in men than in women are consistent with testosterone as the causal mechanism.
Although the largest available sources of genetic associations with thromboembolism and heart failure were used, the relatively low number of participants with heart failure led to imprecise estimates and wide 95% confidence intervals. The response rate of around 5.5% for the UK Biobank resulted in the recruitment of generally healthier participants, which might have biased towards the null. This could explain the discrepancy between the estimates from the UK Biobank and the myocardial infarction case-control study, although the difference might also have been because of a lack of power in the UK Biobank. Therefore, our positive estimates for thromboembolism and heart failure could be underestimated because of survivor bias in the UK Biobank.
Endogenous testosterone decreases with poor health, and the testosterone genome wide association study did not adjust for health status. Therefore, the estimates for genetic variants on testosterone in the genome wide association study could be biased towards the null and might be imprecise; however, such bias is probably minimal. Genetic associations with testosterone were estimated in men aged 50-75. We assume that these genetic associations reflect differences in testosterone concentrations that are also present across the age range (40-69) of UK Biobank participants. In addition, the UK Biobank has implemented high quality procedures to capture and classify events, but complete diagnostic accuracy is impossible. However, misclassification is unlikely to be related to genetic makeup, so any random misclassification of the outcomes would probably bias towards the null.
Our estimates represent average causal effects across the population and so might not be relevant for all subgroups of the population. We also considered venous and arterial thromboembolism together because they might share common risk factors, however, considering venous and arterial thromboembolism separately gives a similar interpretation. Finally, our study compared groups with genetically predicted higher and lower levels of endogenous testosterone to make inferences about the expected effect of raising testosterone. However, there are several qualitative differences between the comparisons that could limit the applicability of our findings to assess the effect of increasing testosterone levels. Specifically, genetic variants are associated with small but lifelong changes in endogenous testosterone levels that occur from modulating a particular biological pathway, whereas testosterone supplementation typically occurs later in life and increases testosterone levels by a relatively large amount.
From a clinical perspective, our study suggests that lifelong endogenous testosterone could have a role in thromboembolism, heart failure, and possibly myocardial infarction, particularly among men.
- These findings provide another strand of evidence consistent with the cardiovascular warnings about TRT issued by regulators.
Further evidence is needed to clarify whether our findings are relevant to the higher rates of these diseases in men than in women, or suggest that agents that lower testosterone would be protective. Additional research is also required comparing the effects of endogenous testosterone with those of exogenous testosterone.
Our study suggests that endogenous testosterone could have a role in thromboembolism, heart failure, and myocardial infarction in men. It might be worth considering whether existing treatments that modulate endogenous testosterone could be used for these conditions.