Testosterone: Effects & Evidence Base — what is actually supported

Testosterone is not a universal “cure-all,” but a hormone whose levels and effects strongly depend on your baseline situation, age, existing conditions, and the specific goal. For many clinical questions, the evidence base is relatively solid—usually from randomized controlled trials (RCTs) and their systematic evaluation. Where the data are more indirect (e.g., animal models) or rely on biomarkers rather than clinical endpoints, conclusions must be more cautious.

What testosterone is about—and what affects levels first?

Short answer: Testosterone levels depend first on sleep, body weight, movement, and metabolism—not on individual supplements. Lifestyle changes can shift hormone status in measurable ways; in cases of severe overweight, bariatric study syntheses show that testosterone levels often respond markedly after surgery. For clinical questions, systematic reviews are usually more robust than anecdotal single reports.

Testosterone acts through multiple layers in the body: production (mainly in the testes, regulated via the hypothalamic–pituitary–gonadal axis), conversion/availability (e.g., via binding proteins such as SHBG), and target-tissue response. That’s why “testosterone works” cannot be treated as one single, constant statement. What you experience depends on whether baseline levels were low, whether there is an underlying disease, and what you want to improve (libido, muscle gain, energy levels, fertility, or medical therapy goals).

The first lever is often lifestyle

Before thinking about testosterone therapy or supplementation, lifestyle levers are usually the cleaner hypothesis: adequate sleep, regular physical activity, and weight regulation. This isn’t a “lifestyle instead of medicine” claim—it’s methodologically sensible. For many metabolic and hormonal relationships, clinical data exist, whereas effects of single interventions on complex endpoints (e.g., myocardial infarction) are much harder to demonstrate cleanly in randomized designs.

Bariatrics as a “hormone-level experiment” for metabolic response

The link between metabolism and hormone status becomes particularly intuitive with bariatric interventions. In a systematic review and meta-analysis on bariatric surgery, Diniz et al., 2025, PMID 40185956 show that different bariatric surgery types can affect testosterone values (meta-analysis data, Obes Surg, 2025). Even though this is not “therapy” in the sense of a targeted testosterone prescription, it supports the core idea: weight and metabolic changes shift hormonal baselines, sometimes in clinically relevant magnitudes.

Clinical decisions need an evidence hierarchy

As soon as you move to concrete decisions (e.g., whether testosterone therapy is cardiovascularly defensible, or what prostate relevance emerges), systematic reviews and meta-analyses are often better starting points than forum opinions or single observational examples. The evidence landscape is heterogeneous: different populations, different testosterone formulations, different measurement timepoints. That’s exactly why it’s important to separate the type of evidence—what the next section covers.

Evidence hierarchy: keep RCTs, observational studies, and animal data cleanly separated

Short answer: The most reliable statements about efficacy and safety come from RCTs and the meta-analyses derived from them. Observational data can show associations, but cannot prove that testosterone is the cause. Animal models can support mechanisms, but they translate to humans only to a limited extent—especially for endpoints such as stroke.

Why study design matters so much

When studies show that testosterone is associated with an outcome (e.g., risk, disease course, or markers), the key question is: Is this a causal effect or a reflection of baseline status? Observational studies are often too “blurred” for this because they cannot fully eliminate confounders. This isn’t criticism of observational studies as a method—it’s a reminder of their limits: correlation ≠ causation.

RCTs and meta-analyses: often the best bridge to safety

For many safety questions, RCTs are the methodological standard. RCTs randomize (within inclusion criteria), reducing systematic differences between groups. When you pool these RCTs, statistical precision increases and adverse events and endpoints can be interpreted more appropriately.

One example is cardiovascular safety: Braga et al., 2025, PMID 40694252 report in a meta-analysis of RCTs that the long-term cardiovascular safety of testosterone replacement in middle-aged and older men, on average, is not associated with a clear safety problem (meta-analysis, Am J Cardiovasc Drugs, 2025). Important caveat: “no clear safety problem in the studied average” is different from “absolutely risk-free for everyone in every situation.”

Animal data: mechanisms yes, “clinical efficacy in humans” no

Animal models can help test plausibility. But even strong animal data are not a direct license to infer clinical efficacy. For stroke-related questions, there are meta-analyses from animal models that group mechanisms involving exogenous estrogen, progesterone, and testosterone—e.g., Kung et al., 2026, PMID 41612411 (meta-analysis, Biol Sex Differ, 2026). This is conceptually useful for hypotheses, but remains methodologically indirect.

Avoid common misinterpretation

The typical mistake looks like this: “If a biomarker changes or an animal model shows a mechanism, then testosterone improves clinical endpoints.” That’s exactly why it’s always worth asking: Which endpoints were measured (surrogate vs. clinical)? Which study designs were used (RCT vs. observational vs. animal)? Below, this is made concrete across multiple areas—from cardiovascular outcomes to the prostate.

What is well studied for testosterone in humans: safety and clinical endpoints

Short answer: For the question “Is testosterone replacement long-term cardiovascularly problematic?” there are RCT-based meta-analyses that, on average, show no clear safety problem (Braga et al., 2025, PMID 40694252). For other clinical topics, the emphasis is often more on prognosis than on a “therapy effect”—for example in advanced liver disease (Zhang et al., 2026, PMID 41551464).

Long-term cardiovascular safety: RCTs are central

The most common concern with testosterone therapies is: “Does it increase the risk of cardiovascular events?” The best evidence base is often from RCTs and their systematic evaluation. Braga et al., 2025, PMID 40694252 pool randomized controlled trials on long-term cardiovascular safety. The core result is: on average, testosterone replacement in middle-aged and older men is not associated with a clear safety problem (meta-analysis, Am J Cardiovasc Drugs, 2025).

What you should take from this:

• You get more of a risk message (“within the studied range of trials, there is no clear signal”) than a guarantee.
• Detailed risks may still differ by design, population, and how endpoints are defined—consistent with the nature of such meta-analyses.

Liver disease: association can reflect prognosis

With advanced liver disease, the logic shifts: a high or low testosterone value may reflect disease burden. Zhang et al., 2026, PMID 41551464 examine in a meta-analysis the relationship between serum total testosterone and prognosis in patients with advanced liver disease (meta-analysis, PeerJ, 2026). Core takeaway: testosterone levels are associated with prognosis. Methodologically, this suggests biomarkers may mirror baseline status, not necessarily that testosterone is the causal driver of the disease course.

Kidney function in gender-affirming therapy: endpoints are decisive

For trans people receiving testosterone in gender-affirming therapy, there is a meta-analysis on kidney function. Tienforti et al., 2025, PMID 40575258 pool studies and summarize effects on kidney parameters (meta-analysis, Front Endocrinol (Lausanne), 2025). Practical interpretation matters: depending on the endpoint (e.g., different kidney markers) and study duration, results can vary. Therefore, “affects” is too vague without specifying endpoints.

“Safe” means something specific in study contexts

In medicine, “safe” in studies usually means: no clear increases in risk across the endpoints examined under the conditions studied. That’s useful, but it’s not the same as “harmless for every person.” When weighing clinically, it helps to list your relevant risks and endpoints (e.g., cardiovascular history, lab values, prostate screening, kidney parameters).

Testosterone and the prostate: postoperative erectile function and links to exercise and cancer

Short answer: In men, the preoperative testosterone level may be related to the risk of postoperative erectile-function problems after nerve-sparing radical prostatectomy; the data should be viewed as risk markers (Chen et al., 2025, PMID 40699366). In prostate cancer contexts, meta-analyses point more toward interactions between exercise, PSA, and testosterone rather than a simple mechanism where “testosterone improves cancer/PSA” (Zhang et al., 2025, PMID 40842180).

Postoperative erectile function: association, not “therapy effect”

For many people, the prostate question isn’t only “cancer”—it’s also: “What happens to sexual function after surgery?” A systematic review with a meta-analysis reports that preoperative testosterone levels may be associated with the risk of postoperative erectile dysfunction after nerve-sparing radical prostatectomy (risk-marker interpretation). This comes from Chen et al., 2025, PMID 40699366 (meta-analysis, World J Urol, 2025).

Practical implications:

• This kind of result answers more “who is at higher risk?” than “testosterone causes the dysfunction.”
• Clinically, the handling is typically screening/stratification, not “raise/lower testosterone on purpose.”

Exercise, PSA, and testosterone: the lifestyle knot

In prostate cancer contexts, the question is complex because PSA is a biomarker affected by multiple factors (including inflammation, tumor biology, and treatment context). Zhang et al., 2025, PMID 40842180 synthesize studies on exercise and its influence on PSA and testosterone in relation to prostate cancer (meta-analysis, Aging Male, 2025). The core message is conceptually more: exercise changes biomarkers, and testosterone/PSA are connected—but it does not automatically follow that an isolated change in testosterone improves PSA or clinical endpoints.

Why this matters for decisions

When reading meta-analyses, always check the underlying interpretation:

• Was the research question “testosterone therapy improves clinical endpoints”?
• Or did it address “associations/interactions between biomarkers under lifestyle or disease conditions”?

In the second case, it is methodologically risky to derive an isolated causal effect of testosterone. And if you consider testosterone therapy, prostate/PSA context is therefore not a pure supplement discussion—it becomes a medical risk–benefit balancing issue.

Directly usable: think in endpoints, not “yes/no”

Remember: testosterone may be relevant, but the right question is: “Which endpoints matter for me, and what evidence exists specifically for those endpoints?” This endpoint logic is what guides the next sections on metabolism and sex hormones.

Metabolism and sex hormones: bariatrics, gender-affirming therapy, and prognostic markers

Short answer: After bariatric procedures, testosterone levels can change depending on the type of operation—interpreted mainly as a metabolic and baseline response (Diniz et al., 2025, PMID 40185956). For gender-affirming testosterone therapy, there is a meta-analysis on kidney function in trans people assigned female at birth; generalization to all groups is not automatically valid (Tienforti et al., 2025, PMID 40575258).

Bariatrics: testosterone as a “dynamic metabolism marker”

Bariatric surgery changes energy balance, insulin sensitivity, inflammatory profiles, and hormonal axes. In practice, a common question is: “Does testosterone go up or down after weight loss?” Data from systematic reviews show: yes, testosterone can change, and the direction may depend on the surgery type. Diniz et al., 2025, PMID 40185956 provide the pooled evidence (meta-analysis, Obes Surg, 2025). This supports the idea that testosterone should not be considered in isolation: metabolic changes often drive hormonal status more than any single isolated intervention.

Gender-affirming therapy: kidney parameters are the endpoint

In gender-affirming therapy, safety questions often come to the forefront, especially for kidney parameters. Tienforti et al., 2025, PMID 40575258 pool data on kidney function under testosterone in trans people assigned female at birth (meta-analysis, Front Endocrinol (Lausanne), 2025). The core takeaway is a summary of kidney parameters—however, with a practical caveat: effect sizes and interpretation depend on endpoint definitions and study duration. For other populations (e.g., different assignments or different therapy protocols), it is not automatically transferable 1:1.

Prognosis associations are not “therapy endorsement”

A common reasoning error is: “If a marker is linked to worse prognosis, then changing it therapeutically must help.” The evidence doesn’t support that simplistically. In advanced liver disease, the link between testosterone and prognosis is plausible (Zhang et al., 2026, PMID 41551464), but that does not yet mean testosterone therapy improves prognosis. This separation between marker and cause is crucial.

That’s why the best strategy is often: lifestyle first, then indication

Before discussing testosterone therapy, in many cases lifestyle interventions are the fastest lever to improve metabolic baseline and thereby influence hormone physiology. Supplements can be reasonable for specific deficiencies, but for testosterone-related problems, the evidence logic is generally strongest for sleep, movement, and weight.

Metabolism and clinical questions: study types and key findings at a glance

Short answer: In this area, meta-analyses dominate—either summarizing changes after interventions (e.g., bariatrics) or pooling relationships between biomarkers and endpoints (e.g., prognosis/kidney markers). The main “key findings” are: surgery type affects testosterone levels (Diniz et al., 2025), kidney parameters are aggregated meta-analytically (Tienforti et al., 2025), and prognosis may be associated without proving therapy causality (Zhang et al., 2026).

Topics where the evidence is more limited or indirect (risks, mechanisms, animal models)

Short answer: For some questions (e.g., whether testosterone is linked to a particular risk- or behavior profile), the evidence is more limited to null. A meta-analysis finds no clear relationship between testosterone and risk aversion (Sánchez et al., 2026, PMID 41638539). For stroke mechanisms, much of the data comes from animal models, which remains indirect (Kung et al., 2026, PMID 41612411).

Risk aversion and behavior: no simple causal story

In biohacking communities, an idea sometimes appears: “Higher testosterone makes people more risk-seeking.” But if you look at the evidence methodologically, the story is more complicated. Sánchez et al., 2026, PMID 41638539 investigate in a meta-analysis the relationship between testosterone and risk aversion and find no relationship (meta-analytic data, Neurosci Biobehav Rev, 2026). This is a good example of how quickly hypotheses form—and how important it is to anchor them back to pooled evidence.

Stroke and sex hormones: animal models instead of clinical endpoints

For stroke mechanisms, there are meta-analyses from animal models that account for exogenous estrogen, progesterone, and testosterone. Kung et al., 2026, PMID 41612411 pool these animal data (meta-analysis, Biol Sex Differ, 2026). This can support mechanisms or refine hypotheses. But: animal models rarely answer whether testosterone helps or harms humans clinically. For endpoints like stroke events, you need human studies—and such data can vary substantially in quality and design.

What you should draw from this practically

If you ask “Does testosterone help against …?” the decisive follow-up question is: Which study data exist for that exact endpoint in humans?

• If there are only animal models: mechanism support yes, recommendation no.
• If there are only biomarker associations: marker ≠ cause.
• If there are RCTs/meta-analyses for clinical endpoints: then a risk–benefit oriented balancing is more appropriate.

Lifestyle remains the foundation here too

Especially where the evidence is indirect, you should prioritize lifestyle levers: sleep, movement, weight, and light schedules are usually the most robust basis for influencing metabolic and hormonal axes—often without needing a new “supplement mechanism” for each effect.

Bottom Line

• The best evidence for safety often comes from RCTs and systematic syntheses; for example, cardiovascular safety on average shows no clear safety problem in middle-aged/older men (Braga et al., 2025, PMID 40694252).
• Biomarker logic is not therapy logic: Associations (e.g., testosterone & prognosis in liver disease) do not automatically mean testosterone therapy helps (Zhang et al., 2026, PMID 41551464).
• Prostate questions should be understood as a medical risk–benefit assessment; preoperative testosterone levels may be risk markers for postoperative erectile dysfunction (Chen et al., 2025, PMID 40699366).
• Metabolism and weight are often the drivers of hormonal levels: bariatric interventions change testosterone values depending on surgery type (Diniz et al., 2025, PMID 40185956).
• Indirect data (animal models, mechanisms) can help thinking, but they do not replace human endpoint data (Kung et al., 2026, PMID 41612411).