Five days of treatment changed the apparent story of testosterone in a group of Chinese men with newly diagnosed diabetes. Their blood glucose came under control, and their testosterone rose sharply. Then the researchers added a familiar diabetes drug to half of them. A month later, the two groups had parted ways.
The drug was metformin, one of the most extensively studied medicines for type 2 diabetes. In the men given metformin, testosterone was lower than in those who continued on insulin alone. It was a result that attracted attention because metformin is also discussed as a possible longevity drug. But a much larger American trial, conducted in men at risk of diabetes, had not found the same hormonal decline. Which result should we believe?
For now, the evidence suggests that metformin can lower circulating testosterone in some men under particular conditions, but a consistent long-term testosterone-lowering effect has not been established. The disagreement is not merely scientific noise. It reveals how difficult it is to separate the effect of a drug from the metabolic illness it treats.
The month-long experiment that started the argument
In a 2021 study published in Diabetes Care, investigators recruited 70 men with newly diagnosed type 2 diabetes who had not previously received glucose-lowering drugs. Their HbA1c was above 9%, indicating substantial hyperglycaemia. All 70 received five days of intensive insulin treatment to bring glucose towards normal before the groups were randomly divided. The men then continued insulin for a month, with 35 also receiving metformin and 35 not receiving it [1].
During those initial five days, the average total testosterone concentration rose from 12.07 to 15.35 nmol/L (approximately 348 to 443 ng/dL). This happened before the experimental difference in metformin exposure. It is a powerful reminder that the metabolic state itself was influencing the hormone measurement.
By the end of the next month, total testosterone was 12.7 nmol/L in the metformin group and 15.3 nmol/L in the insulin-only group. The difference was 2.6 nmol/L, or roughly 75 ng/dL. Free and bioavailable testosterone were also lower in the metformin group, and the reported between-group comparisons were statistically significant [1].
Because glucose control had initially been improved for everyone, the researchers reasonably suspected an effect beyond simple correction of hyperglycaemia. Still, the study was short, involved a narrowly defined group with recently uncontrolled diabetes, and had no metformin placebo. It did not determine whether the effect persists for years, whether it changes sexual health, or whether someone with well-controlled diabetes would respond similarly. The investigators themselves called for longer studies.
Three months later, another study seemed to agree
A second paper, from the same general research group, examined 80 men with type 2 diabetes receiving insulin. Forty were using metformin; forty others who were not using it were selected as matched controls. After three months, testosterone had risen in the controls but not comparably in the metformin users. The mean change in total testosterone was −0.82 nmol/L with metformin versus +0.99 nmol/L without it, a contrast of about 1.81 nmol/L [2].
At first glance, two studies pointing in the same direction seem reassuring. But the second was not another randomised test of whether to prescribe metformin. It was a secondary analysis of a real-world insulin-treatment study, using propensity-score matching to compare existing users with non-users. Matching can balance measured characteristics. It cannot rule out differences in disease history, clinical decisions or other factors that were not captured.
There is also a subtler distinction in what happened. The comparison could be read as metformin lowering testosterone, but it might equally describe a drug-associated failure to recover as glucose control improved. Those are biologically different propositions. Neither can be separated cleanly from the other by a single before-and-after comparison.
The trial of 886 men that complicates the picture
The most instructive counterweight comes from the US Diabetes Prevention Program. In a secondary analysis published in 2016, investigators studied 886 men with excess weight and impaired glucose tolerance who had been randomly assigned to metformin, placebo or an intensive lifestyle intervention. The participants were not the same as the Chinese study population: these were men at risk of diabetes, rather than men starting treatment for severe, newly diagnosed diabetes [3].
After one year, mean total testosterone had changed by −0.12 nmol/L in the metformin group and −0.27 nmol/L in the placebo group. Those small changes did not support a meaningful testosterone-lowering effect from metformin compared with placebo. By contrast, men assigned to intensive lifestyle change gained an average of 1.15 nmol/L in total testosterone, while also losing substantially more weight [3].
The numbers should not be overread in the opposite direction. This was a secondary analysis, with testosterone measured at baseline and one year, so it could miss a transient early fall. Participants had impaired glucose tolerance, not necessarily established diabetes. Nonetheless, with hundreds of men and a placebo comparison, the study makes the claim that metformin routinely suppresses male testosterone much harder to defend.
| Study and population | Testosterone finding | Design limitation |
|---|---|---|
| 70 men with newly diagnosed type 2 diabetes; 1 month | 12.7 vs 15.3 nmol/L with added metformin versus insulin alone | Small, short randomised comparison; no metformin placebo |
| 80 insulin-treated men with type 2 diabetes; 3 months | Change −0.82 vs +0.99 nmol/L in metformin users and matched non-users | Non-randomised secondary analysis; residual confounding |
| 886 men with impaired glucose tolerance; 1 year | Change −0.12 vs −0.27 nmol/L with metformin versus placebo | Larger randomised parent trial, but secondary hormone analysis |
| 106 men with obesity, low testosterone and no diabetes; 1 year | No clear testosterone decline with metformin alone relative to placebo | Hormones were secondary outcomes |
A year-long trial asked a different question
A further comparison published in Metabolism in 2022 followed 106 men aged 18–50 with obesity and low testosterone, but without diabetes. The trial was double-blind and allocated participants to metformin, testosterone, both treatments or placebo for a year. Its principal outcome was insulin resistance, rather than symptoms of androgen deficiency [4].
Metformin improved insulin resistance. Yet metformin alone did not produce a statistically established fall in testosterone relative to placebo. Only the combination of metformin plus testosterone significantly increased measured total and free testosterone compared with placebo, and even those comparisons came from secondary outcomes. The authors did not demonstrate a significant improvement in erectile function or quality of life with any treatment [4].
This does not disprove a temporary decline in severely hyperglycaemic men. It tells us that metabolic context matters, and that an effect in one group cannot be treated as a property of every man who swallows metformin.
Why a blood-test result is harder to interpret than it looks
Testosterone is not isolated from the rest of metabolism. Obesity, insulin resistance, severe hyperglycaemia, calorie restriction and weight change can affect its production and the amount circulating in blood. The liver also produces sex hormone-binding globulin (SHBG), which binds much of circulating testosterone. A change in SHBG can alter a measured total testosterone value without the same change in biologically available hormone.
This is one reason it helps to see total, free and bioavailable testosterone reported separately. They are related measures, not synonyms. Assay methods, timing and differences between study populations create further variation. A drug that helps reduce weight and improve glucose metabolism could influence several of these pathways simultaneously, potentially in opposing directions.
There are plausible laboratory hypotheses for a direct influence of metformin on steroid production. But cell experiments and animal findings do not establish the size or clinical consequences of an effect in humans. A 2023 systematic review of diabetes treatments and testosterone described the human metformin findings as conflicting, often based on small or observational studies, and insufficient to justify metformin as a treatment for low testosterone [5]. That assessment fits the experiments better than either a categorical warning or a promise of hormonal benefit.
One surprisingly neglected outcome is what the participants felt. Most of the research above was designed around hormone concentrations or metabolic endpoints, not whether men developed persistent low libido, infertility, fewer spontaneous erections or distressing symptoms. A difference between group averages is not a measured rate of medication-induced sexual dysfunction.
When testosterone falls, what should a patient actually do?
Erectile problems in men with diabetes can arise from nerve injury, vascular disease, medication effects, psychological factors and metabolic illness, as well as hormonal causes. Persistent fatigue and low libido likewise have many possible explanations. A single laboratory result, especially one taken during illness or at an inconsistent time of day, cannot identify metformin as the culprit.
The Endocrine Society recommends diagnosing hypogonadism only when compatible symptoms accompany consistently and unequivocally low testosterone concentrations, with confirmation using repeat morning fasting measurements. When SHBG may be abnormal, free testosterone can provide useful additional context. The cause should then be investigated rather than inferred from the name of a medicine on the prescription list [6].
A more firmly established longer-term issue with metformin is its association with low vitamin B12. In the Diabetes Prevention Program follow-up, biochemical B12 deficiency was more common after five years in metformin users than in placebo recipients (4.3% versus 2.3%). That is a reason to consider B12 status during clinical review, particularly when symptoms could have other causes; it is not evidence that B12 explains every sexual or hormonal complaint [7].
None of these findings justifies stopping prescribed metformin on the basis of this article. For someone whose diabetes is controlled by the drug, its established benefits have to be weighed against symptoms, repeat measurements and alternative explanations with the treating clinician.
The longevity question is not the diabetes question
Metformin has attracted interest beyond diabetes, particularly because of experiments concerning energy metabolism and ageing. Its ability to alter biological pathways does not mean it has been shown to extend human life when taken by healthy people. A proposed preventive use in someone without diabetes faces a different risk–benefit calculation from treatment of a diagnosed disease.
Personal note: I originally became interested in metformin because of longevity research and at one point considered taking it for that purpose. I chose not to. That is a personal decision, not evidence that others should stop or avoid metformin, particularly when it is prescribed for diabetes.
The question that survives these studies is more precise than “does metformin kill testosterone?” In newly diagnosed, severely hyperglycaemic men, adding metformin was followed by lower concentrations in a short randomised study. In a much larger population at risk of diabetes, a year of metformin did not show the same result. An observational comparison cannot close the gap between them.
Whether there is a short-lived effect in some patients, a lasting effect in a smaller subgroup, or no clinically meaningful effect in most men remains open. Good research would follow hormone concentrations, glucose control, SHBG and symptoms over time in comparable patients assigned to a clearly defined treatment. Until then, the right answer is neither a blanket reassurance nor a blanket alarm.
References
[1] Hu Y et al. Rapid Changes in Serum Testosterone in Men With Newly Diagnosed Type 2 Diabetes With Intensive Insulin and Metformin. Diabetes Care. 2021;44:1059–1061.
[2] Cai T et al. Effect of Metformin on Testosterone Levels in Male Patients With Type 2 Diabetes Mellitus Treated With Insulin. Frontiers in Endocrinology. 2021.
[3] Kim C et al. Testosterone and depressive symptoms among men in the Diabetes Prevention Program. Psychoneuroendocrinology. 2016;72:63–71.
[4] Fernández-García JC et al. Metformin, testosterone, or both in men with obesity and low testosterone: a double-blind, parallel-group, randomised controlled trial. Metabolism. 2022;136:155290.
[5] Van Cauwenberghe J et al. Effects of treatment for diabetes mellitus on testosterone concentrations: a systematic review. Andrology. 2023;11:225–233.
[6] Bhasin S et al. Testosterone therapy in men with hypogonadism: Endocrine Society clinical practice guideline. 2018.
[7] Aroda VR et al. Long-term metformin use and vitamin B12 deficiency in the Diabetes Prevention Program Outcomes Study. Journal of Clinical Endocrinology & Metabolism. 2016.