Erythrocytosis: The most consistent signal

Among the non-MACE findings emerging from TRAVERSE, erythrocytosis represented the most consistent and clinically established adverse effect of testosterone therapy. Haematocrit levels ≥ 54% were observed in 17.0% of men receiving TRT compared with only 3.3% in the placebo group (p < 0.001) [5]. This effect has been repeatedly documented across randomised trials, observational studies, and virtually all testosterone formulations, and its consistency across study designs gives it a degree of clinical certainty that most other TRT-associated signals do not yet possess.

The biological mechanisms are well characterised. Testosterone acts through at least three partially independent pathways to expand red blood cell mass: it directly stimulates erythroid progenitor cells in the bone marrow via androgen receptor-mediated genomic effects; it increases renal erythropoietin production; and it suppresses hepcidin, the master regulator of iron homeostasis, thereby enhancing ferroportin-mediated iron release and increasing the availability of substrate for erythropoiesis [12]. Testosterone aromatisation to oestradiol may additionally stimulate haematopoietic stem cell proliferation, though this pathway appears to be a contributing rather than an essential mechanism [12]. The convergence of these pathways explains both the consistency and the dose-dependence of the erythrocytosis signal. The clinical magnitude of haematocrit elevation is greater than the TRAVERSE threshold figure of 17% might suggest. In a Swiss single-centre retrospective cohort of 247 patients followed for a median of 2.9 years, Neidhart and colleagues found that 57% of men reached a haematocrit above 0.46, 23% exceeded 0.50, and 5% exceeded the 0.54 threshold used in TRAVERSE, with 54% of patients reaching their peak haematocrit only after the first year of treatment, an observation with direct implications for monitoring duration [12]. The risk of erythrocytosis is not uniform across formulations. Intramuscular testosterone cypionate and enanthate produce the largest haematocrit increments (+ 4.0% points on average), while intranasal gel produces no statistically significant change; in a head-to-head randomised comparison of 39 men, polycythemia (Hct > 52%) developed in 4.2% of men receiving intramuscular testosterone cypionate and in none of those receiving nasal testosterone over four months [13, 14]. Long-acting testosterone undecanoate shows a haematocrit increment intermediate between the two extremes, consistent with its more stable pharmacokinetic profile.

Individual patient risk is not determined by formulation alone. In the Neidhart cohort, baseline haematocrit above 0.41 (OR 4.71, 95% CI 1.21–18.29) and higher BMI (OR 1.13–1.15 per unit) were independent predictors of clinically significant haematocrit elevation, while smoking showed a trend toward association that did not reach significance in that sample [12]. A clinically important and underappreciated interaction has recently been described in men with type 2 diabetes: in a retrospective cohort of 5,235 men with T2DM and hypogonadism drawn from the Clalit Healthcare Services database, concomitant use of SGLT-2 inhibitors was associated with a markedly elevated risk of haematocrit exceeding 54% during TRT (OR 4.85, 95% CI 3.06–7.69, versus OR 1.88 without SGLT-2 inhibitors; p = 0.02), a finding attributable to the additive erythropoietic effects of both drug classes [15]. Given the high prevalence of SGLT-2 inhibitor use in the population of men most likely to receive TRT, this interaction deserves explicit discussion during treatment planning.

The cardiovascular consequences of TRT-induced erythrocytosis remain an area of active investigation, but the available evidence is concerning. In a large matched cohort study of 5,842 men, those who developed polycythemia (Hct > 52%) during the first year of TRT had significantly higher rates of both MACE and VTE compared with matched controls who did not develop polycythemia, suggesting that erythrocytosis is not merely a laboratory finding but a mechanistically plausible intermediate on the pathway to adverse vascular events [14]. The mechanism, increased blood viscosity promoting a prothrombotic milieu, is biologically coherent and consistent with the VTE signal discussed in Sect.  4.3.

From a practical standpoint, current evidence supports haematocrit monitoring at 3–6 months during the first year of TRT and annually thereafter, with dose reduction or temporary treatment discontinuation when haematocrit exceeds 54% or haemoglobin exceeds 18.5 g/dL [12, 15]. Therapeutic phlebotomy may be considered in selected cases where treatment continuation is clinically important. Importantly, given that more than half of patients reach peak haematocrit after the first year, annual monitoring should not be interpreted as a licence to reduce vigilance. In men with borderline baseline haematocrit, OSA, COPD, or concurrent SGLT-2 inhibitor therapy, we recommend more frequent early monitoring (at one month, then three months) and a lower threshold for formulation change toward transdermal or intranasal preparations. These monitoring recommendations are reflected in Table 2.

Table 2 Key Clinical Messages Post-TRAVERSEHypertension

A modest but statistically significant increase in systolic blood pressure, approximately 1.6 mmHg above placebo, was observed in the TRAVERSE trial among men receiving testosterone therapy [5]. This finding was considered clinically important enough to prompt a 2025 class-wide FDA labelling update requiring all testosterone products to include explicit warnings regarding potential blood pressure elevation (FDA, 2025, https://www.fda.gov/drugs/drug-safety-and-availability/fda-updates-labeling-testosterone-products-regarding-blood-pressure-increases). The regulatory response is proportionate: while 1.6 mmHg is a small absolute increment, it acquires clinical relevance when considered in the context of populations already carrying substantial cardiometabolic burden.

The mechanisms underlying this effect are likely multifactorial. Testosterone may promote sodium and water retention and has complex effects on endothelial function and coronary vasodilation that can be directionally opposite depending on the dose and the vascular bed involved [16]. One particularly well-characterised contributor is the erythrocytosis pathway: as haematocrit rises, blood viscosity increases, augmenting peripheral vascular resistance and amplifying the pressor response to testosterone. Olesen and colleagues demonstrated this interaction directly in a randomised trial, showing that rising haematocrit levels significantly amplified the blood pressure response to TRT in men with opioid-induced androgen deficiency [17]. This mechanistic link between erythrocytosis and hypertension reinforces the argument that several of the non-MACE signals associated with testosterone therapy are physiologically interconnected rather than independent adverse events.

The direction of the blood pressure effect is not, however, uniformly adverse. In men with metabolic syndrome, obesity, or type 2 diabetes, populations in whom hypogonadism is frequently functional rather than organic, TRT has been associated with reductions in both systolic and diastolic blood pressure in multiple interventional studies, including a 15-month trial of parenteral testosterone undecanoate and a randomised study in men with metabolic syndrome and recent-onset diabetes where testosterone produced beneficial blood pressure effects beyond those achieved by diet and exercise alone [18]. This apparent paradox likely reflects the fact that, in these patients, correcting the hypogonadal state improves the underlying metabolic milieu, reducing visceral adiposity, insulin resistance, and sympathetic tone, to an extent that outweighs any direct pressor effect of testosterone.

The epidemiological data confirm that hypertension and testosterone prescribing are closely intertwined, but the causal direction is not straightforward. In a large retrospective cohort of 20,299 men from North-West London primary care, hypertension was present in 25.7% of men who were subsequently prescribed TRT, compared with 17.8% of those who were not (OR 1.60, 95% CI 1.35–1.90, p < 0.0001), suggesting that hypogonadism and hypertension frequently co-occur, most plausibly because both are downstream consequences of obesity and metabolic syndrome, rather than that TRT causes hypertension in a previously normotensive man [19]. In a propensity-weighted provincial cohort of 6,949 men receiving TRT followed over a median of more than two decades in Manitoba, Canada, both TRT-treated men and untreated testosterone-deficient men showed a similarly elevated MACE risk (HR 1.27, 95% CI 1.16–1.39 and HR 1.21 respectively), a finding that underscores the contribution of the underlying hypogonadal state, and its associated cardiometabolic burden, including hypertension, to cardiovascular risk, independent of treatment [20].

From a clinical standpoint, the practical implication is one of careful patient stratification rather than categorical avoidance. In a normotensive man with organic hypogonadism and no metabolic comorbidities, the 1.6 mmHg increment observed in TRAVERSE is unlikely to alter his cardiovascular trajectory meaningfully. In a man with borderline or treated hypertension, chronic kidney disease, or established cardiovascular disease, even a small upward pressure shift may be clinically consequential over time, and blood pressure should be measured at baseline and at each monitoring visit. In men with metabolic syndrome or obesity-related functional hypogonadism, the priority should be metabolic optimisation before TRT initiation, not only because testosterone normalisation may occur spontaneously with weight loss, but because the blood pressure effects of TRT in this group may be genuinely beneficial once the metabolic substrate is adequately addressed.

Non-MACE safety signals: a differentiated assessment

The non-MACE safety signals in TRAVERSE require careful interpretation. The trial data and the observational data tell a subtly different story, and conflating them as has sometimes been done in clinical guidelines and regulatory documents leads to either overestimation or underestimation of risk depending on which source is emphasised.

Atrial fibrillation (AF)

The TRAVERSE trial identified a numerical excess of atrial fibrillation in the testosterone arm, 3.1% versus 2.4% in the placebo group (HR 1.33, 95% CI 0.97–1.83, p = NS), that did not reach statistical significance but represented an absolute risk difference of 0.7% points over a median follow-up of 3.2 years [5]. The failure to reach significance is almost certainly a reflection of statistical power rather than true absence of effect: AF is a relatively rare outcome, and TRAVERSE was not designed or powered to detect small increments in arrhythmic events as a primary endpoint.

The aggregate evidence from randomised trials, while numerically reassuring, is constrained by the same limitation. A meta-analysis of 26 randomised controlled trials enrolling 10,941 hypogonadal men found a pooled relative risk of 1.38 (95% CI 0.54–3.53, p = 0.36) for incident AF with TRT versus placebo, a point estimate directionally consistent with TRAVERSE but with confidence intervals too wide to exclude clinically meaningful harm [16]. Critically, only 4 of the 26 included trials contributed AF data, confirming that the existing randomised evidence base is substantially underpowered for this outcome and that the absence of a statistically significant signal should not be interpreted as evidence of safety.

Observational data provide important contextual information, though they are subject to substantial confounding. A large prospective cohort study of 173,498 men followed for 11 years within the UK Biobank demonstrated a complex relationship between endogenous androgen levels and incident AF: low serum testosterone was associated with a modestly increased AF risk (HR 1.06, 95% CI 1.00–1.12), whilst elevated sex hormone-binding globulin (SHBG) conferred a more pronounced risk increment (HR 1.44, 95% CI 1.34–1.54), suggesting that the androgenic milieu modulates atrial electrophysiology through multiple pathways [21]. This finding is physiologically important: it implies that the relationship between testosterone and AF risk is not linear, and that both testosterone deficiency and the hormonal dysregulation associated with elevated SHBG may independently promote arrhythmic substrate. In a large retrospective primary care cohort of 20,299 men, those subsequently prescribed TRT had significantly higher baseline prevalence of ischaemic heart disease (OR 1.57, p = 0.002), obstructive sleep apnoea (OR 3.47, p < 0.0001), previous DVT or pulmonary embolism (OR 2.24, p = 0.002), and prior stroke (OR 1.82, p = 0.03), all established risk factor for AF, and were on average six years older with a higher BMI than untreated men [19]. This profile of pre-treatment comorbidity is a critical source of residual confounding in all observational analyses of TRT and AF, and it renders direct causal inference from observational data alone methodologically untenable.

Pharmacovigilance data from the FDA Adverse Event Reporting System (FAERS), encompassing 28,921 TRT-related reports submitted between 2004 and 2022, identified a disproportionality signal for ventricular fibrillation (reporting odds ratio 2.51, 95% CI 1.49–4.24) and cardiac arrest (ROR 1.88, 95% CI 1.51–2.35), while characterising both events as rare in absolute terms [22]. These signals are hypothesis-generating rather than confirmatory, given the well-recognised limitations of spontaneous reporting systems, including underreporting, selective reporting, and inability to establish causation, but they are consistent with the biological plausibility of testosterone-mediated effects on ventricular electrophysiology.

The mechanistic underpinning for a testosterone–arrhythmia relationship is biologically coherent. Testosterone inhibits L-type calcium channels, modulates cardiac potassium currents, and may promote atrial structural remodelling through androgenic receptor-mediated pathways [23]. Crucially, the direction of these effects may depend on the concentration and duration of exposure: the acute anti-ischaemic and vasodilatory effects of testosterone in men with coronary artery disease are well documented at physiological levels [24], whereas chronic exposure at the upper end of the normal range, which is where many men on TRT are maintained, may have different electrophysiological consequences, particularly in atrial tissue already subject to fibrotic remodelling from hypertension, obesity, or sleep apnoea.

The overall interpretation is one of a plausible but incompletely characterised signal. The trial and observational data are directionally consistent: both point toward a modest relationship between the androgenic milieu and AF risk that is likely mediated through multiple converging pathways, direct electrophysiological effects, haemodynamic consequences of erythrocytosis-driven viscosity increases, and the cardiometabolic substrate of the population most likely to receive TRT. The signal does not currently justify withholding TRT from men with well-established hypogonadism, but it does justify incorporating AF risk into the pre-treatment evaluation. Baseline cardiac rhythm assessment, at minimum a standard 12-lead ECG, should be obtained in men with established cardiovascular disease, hypertension, obesity, or obstructive sleep apnoea before initiating TRT, and the arrhythmic signal should be discussed explicitly during the informed consent process in men with pre-existing conduction abnormalities or a history of paroxysmal AF.

Venous thromboembolism (VTE) and pulmonary embolism (PE)

Pulmonary embolism occurred in 0.9% of men receiving testosterone and 0.5% of those receiving placebo in the TRAVERSE trial, while venous thromboembolic events overall were reported at 2.0% versus 1.5% respectively, neither difference reaching statistical significance, but both directionally consistent with a pro-thrombotic signal [5, 25]. As with atrial fibrillation, the failure to reach statistical significance reflects the trial’s primary design for MACE detection rather than the absence of a biologically meaningful effect: TRAVERSE was not powered to detect small increments in rare thromboembolic events, and these findings should be interpreted as safety signals warranting further investigation rather than as evidence of null effect.

The observational literature provides a more granular picture, though it is marked by an apparent paradox that requires careful interpretation. A systematic review and meta-analysis of 14 randomised controlled trials and 10 observational studies encompassing 314,315 men found that, in RCT-derived data, TRT was not associated with a significantly increased risk of VTE overall (OR 1.42, 95% CI 0.22–9.03, p = 0.71) or pulmonary embolism specifically (OR 1.38, 95% CI 0.27–7.04, p = 0.70) [26]. Observational data from the same meta-analysis showed a paradoxical overall reduction in VTE risk (OR 0.78, 95% CI 0.61–1.00, p = 0.05), almost certainly a reflection of healthy-user bias and confounding by indication in real-world prescribing, but a statistically significant increase in deep vein thrombosis specifically (OR 1.44, 95% CI 1.04–2.00, p = 0.03; I²=0%), a finding that persisted across sensitivity analyses [26]. The discordance between the overall VTE and DVT-specific results underscores the importance of disaggregating thromboembolic outcomes when interpreting this literature: the apparent protective signal for total VTE in observational cohorts is likely artefactual, whereas the DVT signal from studies with low heterogeneity carries greater methodological credibility.

The risk appears concentrated in the early treatment period. A large population-based case-control study by Martinez and colleagues (n > 19,000) reported a 63% increase in VTE risk during the first six months of therapy (HR 1.63, 95% CI 1.12–2.37), corresponding to approximately 10 additional events per 10,000 person-years, with risk normalising beyond six months [6]. Walker and colleagues confirmed an approximately twofold increase in VTE risk during the first one to three months of therapy in a case-crossover study of nearly 40,000 men, irrespective of hypogonadism status and route of administration [7]. The individual patient data meta-analysis by Hudson and colleagues across 35 trials further corroborated this early temporal clustering of thromboembolic events [8].

The mechanistic explanation for this early risk window is most plausibly erythrocytosis-driven hyperviscosity. Testosterone stimulates erythropoiesis through increased renal erythropoietin production, direct bone marrow effects, and hepcidin suppression, all of which converge to raise haematocrit within the first months of treatment. Men who develop polycythaemia (Hct ≥ 52%) during TRT experience a significantly higher risk of both MACE and VTE during the first year, and the temporal coincidence of peak erythrocytosis risk and peak VTE risk is not incidental [25]. Androgens additionally increase platelet thromboxane A₂ receptor density and formation, providing a second, haematocrit-independent pathway to a prothrombotic milieu [26]. Formulation pharmacokinetics amplify this risk further: short-acting intramuscular regimens generating supraphysiological testosterone peaks are associated with greater haematocrit increments and higher rates of cardiovascular events and hospitalisation than transdermal systems in observational cohorts, and subcutaneous testosterone enanthate autoinjectors appear to produce smaller increases in both oestradiol and haematocrit than intramuscular cypionate [25].

From a clinical management standpoint, haematocrit monitoring is the central instrument for VTE risk mitigation. Monitoring should be performed at baseline, at three to six months, and annually thereafter, with temporary treatment interruption and dose adjustment triggered by Hct exceeding 54%, or 50% according to the more conservative AACE and ICSM thresholds [25, 26]. Men with prior VTE or known thrombophilia represent a particularly high-risk phenotype in whom TRT should only be initiated after explicit shared decision-making, with an individualised monitoring plan and specialist input regarding thromboprophylaxis. The pre-treatment evaluation should also encompass obstructive sleep apnoea, chronic kidney disease, and uncontrolled hypertension, all conditions that independently amplify erythrocytosis risk and thromboembolic susceptibility. Patients should be counselled proactively on alarm symptoms, new or unexplained dyspnoea, pleuritic chest pain, or unilateral leg swelling, with a low threshold for diagnostic evaluation during the first six months of therapy. A clinically important gap in current practice is that no widely used VTE risk stratification tool including the Wells criteria, the PERC score, or the YEARS algorithm, currently incorporates testosterone replacement therapy as a contributing risk factor, an omission that warrants attention in future guideline revisions [25, 27].

Acute kidney injury (AKI)

Acute kidney injury was the only non-MACE endpoint to reach nominal statistical significance in TRAVERSE, occurring in 2.3% of testosterone-treated men versus 1.5% of placebo recipients, an absolute risk difference of 0.8% points that, while modest, was statistically significant (p = 0.048) [5]. The caveat that this finding was not adjusted for multiple comparisons is methodologically important, but the signal has since been independently replicated in a propensity-matched real-world cohort of 4,268 hypogonadal men drawn from the TriNetX global research network, in which TRT was associated with a significantly increased risk of AKI over three years of follow-up (RR 1.53, 95% CI 1.07–2.18), while no significant increase in atrial fibrillation was observed in the same dataset (RR 1.48, 95% CI 0.93–2.37) [28]. This dissociation between the AKI and AF signals in real-world data reinforces the specificity of the renal finding and argues against a generic confounding explanation.

The AKI signal is further substantiated by converging evidence from propensity-matched surgical cohorts, which provide a natural experimental context in which renal vulnerability is acutely amplified. In a national database analysis of 7,906 men undergoing total hip arthroplasty, preoperative TRT use was associated with a significantly higher rate of AKI at one year (7.6% versus 5.6%, p < 0.001), in addition to increased rates of deep vein thrombosis at both 90 days (2.8% versus 2.0%, p = 0.023) and one year (4.3% versus 3.0%, p = 0.002) [29]. In a parallel analysis of 13,250 men undergoing total knee arthroplasty, preoperative TRT was independently associated with AKI within 90 days (OR 1.5, 95% CI 1.2–1.8, p < 0.001) and pulmonary embolism (OR 1.4, 95% CI 1.0–1.8, p = 0.041) [30], and in a third cohort of 2,738 men undergoing total shoulder arthroplasty, TRT users had a 45% higher relative risk of AKI at one year (17.5% versus 12.1%, RR 1.45, p < 0.001) [31]. The consistency of the AKI signal across three independent surgical cohorts and two medical datasets spanning different patient populations, surgical stresses, and database sources, substantially strengthens the biological credibility of the association. Notably, in a smaller cohort of men undergoing rotator cuff repair, a procedure associated with lower physiological stress than major joint arthroplasty, no significant difference in 90-day AKI was observed (0.7% versus 0.5%, p = 0.18) [32], suggesting that the magnitude of perioperative renal stress may modulate the expression of TRT-associated renal vulnerability.

The mechanistic basis for TRT-associated AKI is multifactorial. Testosterone promotes sodium and water retention through both direct renal tubular effects and RAAS activation, increasing intraglomerular pressure and fluid loading in a manner that can precipitate AKI in men with reduced renal reserve [25]. Erythrocytosis-driven hyperviscosity impairs renal microvascular perfusion, and the peak-trough variability characteristic of short-acting intramuscular formulations may generate episodic haemodynamic stress on the renal vasculature that is attenuated by more pharmacokinetically stable transdermal or long-acting preparations [25]. Importantly, a subgroup analysis of 13,598 men undergoing radical nephrectomy, a population with inherently reduced functional renal mass, found that hypogonadism itself was associated with higher rates of postoperative AKI (25.5% versus 21.6%, p = 0.005), but that, among hypogonadal men, TRT use was not associated with any adverse postoperative outcome [33]. This paradoxical finding reinforces the complexity of the testosterone–kidney relationship: the baseline hypogonadal state may itself confer renal vulnerability, and TRT may not be the sole or primary driver of the observed risk in all contexts. The most clinically important challenge in interpreting the AKI signal is its apparent context-dependence. A 2025 analysis of 10,009 men with type 2 diabetes and hypogonadism reported that TRT was associated with a significantly reduced risk of AKI (HR 0.76, 95% CI 0.65–0.89, p < 0.001), a finding diametrically opposite to the TRAVERSE signal [34]. This apparent contradiction most plausibly reflects population heterogeneity in the renal substrate: testosterone’s haemodynamic effects on renal blood flow, including vasodilation of the afferent arteriole and attenuation of the fibrotic and inflammatory milieu of diabetic nephropathy, may be nephroprotective when the primary renal threat is ischaemic and metabolic, while the same haemodynamic and erythropoietic effects may be harmful in men whose renal vulnerability is driven by reduced nephron mass, established CKD, or perioperative stress. The implication is that the direction of the renal effect of TRT is not fixed but depends critically on the underlying renal phenotype of the patient. From a clinical management standpoint, pre-treatment renal assessment is mandatory in all men being considered for TRT, and CKD stage 3–5 (eGFR < 60 ml/min/1.73 m²), a solitary kidney, rapidly progressive CKD, or a history of recurrent AKI should each be considered high-risk phenotypes requiring intensified monitoring or reconsideration of treatment initiation [25]. Renal function should be reassessed at three to six weeks after initiation, capturing the early haemodynamic response and again at three months, with annual review thereafter in stable patients and more frequent assessment in those with baseline renal impairment. A confirmed eGFR decline of ≥ 20% from baseline should trigger evaluation for intercurrent illness, nephrotoxin exposure, and reversible contributors, with dose reduction, formulation change, or temporary interruption as appropriate; persistent or recurrent renal dysfunction should prompt formal reassessment of the indication for TRT [25]. In men with pre-existing CKD or significant proteinuria, nephrology input before initiation and at the first sign of renal deterioration is a clinically prudent step that the current evidence base supports.

Fractures

A 2026 meta-analysis by Anagnostis and colleagues reported an increased risk of clinical fractures in men receiving TRT (HR 1.43, 95% CI 1.04–1.97) [35], a finding that was unexpected given the well-established anabolic effects of testosterone on bone mineral density. This paradox is further supported by large observational database studies: a PearlDiver analysis of 77,491 matched pairs reported a vertebral fracture incidence of 0.31% in TRT users versus 0.04% in controls (adjusted OR 7.70, 95% CI 5.09–11.65, p < 0.001), corresponding to an incidence rate 7.6 times higher than in untreated men [36]; a separate PearlDiver analysis of 335,753 matched pairs reported a proximal humerus fracture incidence of 0.029% versus 0.005% (adjusted OR 3.14, 95% CI 1.84–5.59, p < 0.001), with risk concentrated in the 46–55 year age group (adjusted OR 7.52, 95% CI 2.47–32.73) [37]. The apparent contradiction between BMD benefit and fracture risk increase is reconciled by a possible behavioural–musculoskeletal mismatch hypothesis: TRT-related improvements in muscle strength, energy, and perceived vitality lead to increased physical activity and fall exposure in men who were previously sedentary, before slower connective tissue and bone quality adaptations have had time to fully consolidate [36, 38]. Most fractures in TRAVERSE occurred at traumatic sites (ribs, wrists, ankles) and clustered in the early months after treatment initiation, supporting a behavioural rather than a bone quality mechanism [38]. This interpretation is consistent with the high-quality RCT data from the T-Trial Bone Study, which demonstrated a lumbar spine trabecular volumetric BMD increase of 7.5% (95% CI 4.8–10.3%, p < 0.001) at 12 months, and the T4 Bone Trial, which showed significant cortical BMD gains at tibia and radius at 24 months, genuine skeletal benefit at the tissue level that has not yet translated into demonstrable fracture risk reduction at the clinical level in any adequately powered trial [