Nejmoa 2215025

The n e w e ng l a n d j o u r na l of m e dic i n e
Original Article
Cardiovascular Safety of Testosterone-

Replacement Therapy
A.M. Lincoff, S. Bhasin, P. Flevaris, L.M. Mitchell, S. Basaria, W.E. Boden,
G.R. Cunningham, C.B. Granger, M. Khera, I.M. Thompson, Jr., Q. Wang,
K. Wolski, D. Davey, V. Kalahasti, N. Khan, M.G. Miller, M.C. Snabes, A. Chan,
E. Dubcenco, X. Li, T. Yi, B. Huang, K.M. Pencina, T.G. Travison, and S.E. Nissen,
for the TRAVERSE Study Investigators*
A BS T R AC T
BACKGROUND
The cardiovascular safety of testosterone-replacement therapy in middle-aged and The authors’ full names, academic de-
older men with hypogonadism has not been determined. grees, and affiliations are listed in the Ap-
pendix. Dr. Nissen can be contacted at
nissens@ccf.org or at the Department of
METHODS Cardiovascular Medicine, Cleveland Clinic,
In a multicenter, randomized, double-blind, placebo-controlled, noninferiority trial, 9500 Euclid Ave., Cleveland, OH 44195.
we enrolled 5246 men 45 to 80 years of age who had preexisting or a high risk of *A list of the TRAVERSE Study Investiga-
cardiovascular disease and who reported symptoms of hypogonadism and had two tors is provided in the Supplementary
fasting testosterone levels of less than 300 ng per deciliter. Patients were randomly Appendix, available at NEJM.org.
assigned to receive daily transdermal 1.62% testosterone gel (dose adjusted to main- Drs. Lincoff and Bhasin contributed equally
tain testosterone levels between 350 and 750 ng per deciliter) or placebo gel. The to this article.
primary cardiovascular safety end point was the first occurrence of any component This article was published on June 16,
of a composite of death from cardiovascular causes, nonfatal myocardial infarction, 2023, at NEJM.org.
or nonfatal stroke, assessed in a time-to-event analysis. A secondary cardiovascular DOI: 10.1056/NEJMoa2215025
end point was the first occurrence of any component of the composite of death Copyright © 2023 Massachusetts Medical Society.
from cardiovascular causes, nonfatal myocardial infarction, nonfatal stroke, or coro-

nary revascularization, assessed in a time-to-event analysis. Noninferiority required
an upper limit of less than 1.5 for the 95% confidence interval of the hazard ratio
among patients receiving at least one dose of testosterone or placebo.
RESULTS
The mean (±SD) duration of treatment was 21.7±14.1 months, and the mean follow-
up was 33.0±12.1 months. A primary cardiovascular end-point event occurred in
182 patients (7.0%) in the testosterone group and in 190 patients (7.3%) in the
placebo group (hazard ratio, 0.96; 95% confidence interval, 0.78 to 1.17; P<0.001
for noninferiority). Similar findings were observed in sensitivity analyses in which
data on events were censored at various times after discontinuation of testosterone
or placebo. The incidence of secondary end-point events or of each of the events of
the composite primary cardiovascular end point appeared to be similar in the two
groups. A higher incidence of atrial fibrillation, of acute kidney injury, and of pul-
monary embolism was observed in the testosterone group.
CONCLUSIONS
In men with hypogonadism and preexisting or a high risk of cardiovascular disease,
testosterone-replacement therapy was noninferior to placebo with respect to the in-
cidence of major adverse cardiac events. (Funded by AbbVie and others; TRAVERSE
ClinicalTrials.gov number, NCT03518034.)
n engl j med nejm.org 1

The New England Journal of Medicine
Downloaded from nejm.org at HINARI-Iraq on July 3, 2023. For personal use only. No other uses without permission.
Copyright © 2023 Massachusetts Medical Society. All rights reserved.
T
he cardiovascular effects of tes- data and safety monitoring committee reviewed
tosterone-replacement therapy in middle- unblinded safety data. The clinical database was
aged and older men with hypogonadism maintained by Labcorp Drug Development and
have not been determined. Retrospective cohort transferred to C5Research for independent sta-
studies involving men receiving testosterone- tistical analyses.
replacement therapy have shown conflicting The first author wrote the first draft of the
results, with some showing increased and others manuscript, which was reviewed and approved by
decreased cardiovascular risk.1-5 Small random- all the authors. The sponsor reviewed the manu-
ized trials similarly have not shown a consistent script and provided suggested revisions, but the
association of testosterone treatment with cardio- final decision regarding content was reserved for
vascular risk, although none were designed to the academic authors, with no restrictions on the
systematically assess cardiovascular outcomes and right to publish. The first and last authors vouch
all were inadequately powered and had a short for the completeness and accuracy of the data and
duration.6-10 for the fidelity of the trial to the protocol and
In response to concerns and conflicting data the statistical analysis plan.
regarding the cardiovascular safety of testosterone-
replacement therapy, the Food and Drug Admin- Trial Population
istration (FDA) issued a guidance on March 3, Men with preexisting cardiovascular disease or an
2015, that required manufacturers of approved elevated cardiovascular risk were eligible if they
testosterone products to conduct clinical trials to were 45 to 80 years of age; reported one or more
determine whether testosterone-replacement thera- symptoms of hypogonadism, including decreased
py is associated with an increased risk of cardio- sexual desire or libido, decreased spontaneous
vascular events.11 The Testosterone Replacement erections, fatigue or decreased energy, low or
Therapy for Assessment of Long-term Vascular depressed mood, loss of axillary or pubic body
Events and Efficacy Response in Hypogonadal Men hair or decreased frequency of shaving, or hot
(TRAVERSE) trial was designed to determine the flashes; and had two fasting serum testosterone
effects of testosterone-replacement therapy on the levels of less than 300 ng per deciliter (10.4 nmol
incidence of major adverse cardiac events among per liter) in blood samples obtained between 5:00
middle-aged and older men with hypogonadism a.m. and 11:00 a.m. and measured at a central
and either preexisting or a high risk of cardio- laboratory with the use of liquid chromatography–
vascular disease. tandem mass spectrometry.
Cardiovascular disease was defined as clinical
or angiographic evidence of coronary artery dis-
Me thods
ease, cerebrovascular disease, or peripheral arte-
Trial Design and Oversight rial disease. Increased cardiovascular risk was
We conducted this phase 4, randomized, double- defined as the presence of three or more of the
blind, placebo-controlled, noninferiority, event- following risk factors: hypertension, dyslipidemia,
driven trial at 316 clinical-trial sites in the United current smoking, stage 3 chronic kidney disease,
States (see the Supplementary Appendix, available diabetes, elevated high-sensitivity C-reactive pro-
with the full text of this article at NEJM.org). tein level, age of 65 years or older, or an Agatston
The trial design has been published previously.12 coronary calcium score that was above the 75th
The trial was funded by a consortium of testos- percentile for age and race.
terone manufacturers led by AbbVie and was Exclusion criteria included congenital or ac-
overseen by the Cleveland Clinic Coordinating quired severe hypogonadism (testosterone level,
Center for Clinical Research (C5Research) with <100 ng per deciliter [3.5 nmol per liter]), a his-
support from a contract research organization tory of prostate cancer or prostate nodules, an
(Labcorp Drug Development). The trial protocol, elevated screening prostate-specific antigen (PSA)
available at NEJM.org, was designed by the ex- level, thrombophilia, and uncontrolled heart
ecutive committee and sponsor. National and failure. Patients could not be enrolled within 4
institutional regulatory and ethical authorities months after an acute coronary syndrome, stroke,
approved the protocol, and all the patients pro- or coronary or peripheral revascularization or
vided written informed consent. An independent within 6 months after treatment with testoster-
2 n engl j med nejm.org

Cardiovascular Safety of Testosterone Ther apy
one or androgenic steroids. Detailed eligibility cri- tary Appendix. Effects of testosterone therapy on
teria are provided in the Supplementary Appendix. noncardiovascular outcomes in this trial are not
reported in this article (see Table S1 in the Sup-
Trial Intervention plementary Appendix).
Patients were randomly assigned in a 1:1 ratio to
receive daily transdermal 1.62% testosterone gel Statistical Analysis
or matching placebo gel provided in metered-dose In this noninferiority trial, we used a Cox pro-
pumps. Randomization was stratified according portional-hazards regression model to estimate
to the presence or absence of preexisting cardio- the hazard ratio and its two-sided 95% confi-
vascular disease. To avoid unblinding, the pa- dence interval for a primary end-point event with
tients and trial team remained unaware of the testosterone as compared with placebo, with ad-
post-baseline testosterone levels measured at the justment for preexisting cardiovascular disease.
central laboratory. Dose adjustments to main- The trial was designed to conclude after 256 pri-
tain testosterone levels between 350 and 750 ng mary composite end-point events had occurred.
per deciliter (12.1 to 26.0 nmol per liter) or to This number of events would provide 90% power
respond to a hematocrit greater than 54% were to detect an upper limit of less than 1.5 for the
managed centrally by an automated algorithm 95% confidence interval of the hazard ratio
(additional details are provided in the Supple- (one-sided alpha level, 2.5%). Because the statis-
mentary Appendix). Patients who were randomly tical analysis plan did not include a provision for
assigned to placebo underwent sham adjustments correcting for multiplicity for tests of secondary
to maintain blinding. or other end points, results are reported as point
Testosterone or placebo was discontinued in estimates and 95% confidence intervals. The
patients with testosterone levels that exceeded widths of the confidence intervals have not been
750 ng per deciliter or with a hematocrit that adjusted for multiplicity, so the intervals should
exceeded 54% even after adjustment to the low- not be used to infer definitive treatment effects
est dose, as well as in patients who had a new for secondary end points.
diagnosis of prostate cancer or were deemed to The full-analysis population consisted of all
be at risk for suicide; otherwise, the assigned in- the patients who had undergone randomization,
tervention was to be continued for the duration of and the safety population consisted of all the
the trial. The investigators received specific guide- patients who had undergone randomization and
lines regarding care or referral to a urologist for had received at least one dose of testosterone or
patients with an elevated PSA level. placebo. The primary analysis involved the safe-
ty population and was repeated in a supportive
End Points analysis involving the full-analysis population.
The primary safety end point was the first oc- The principal sensitivity analysis included major
currence of any component of major adverse car- adverse cardiac events that occurred during the
diac events, a composite of death from cardiovas- period from randomization to 365 days after the
cular causes, nonfatal myocardial infarction, or last dose, with censoring of data on events that
nonfatal stroke in a time-to-event analysis. A occurred more than 365 days after the last dose.
secondary cardiovascular end point was the first Additional sensitivity and supportive analyses
occurrence of any component of the composite based on different event censoring times and a
of death from cardiovascular causes, nonfatal restricted mean survival time at 3 years are de-
myocardial infarction, nonfatal stroke, or coro- scribed in the Supplementary Appendix.
nary revascularization in a time-to-event analy- Under the assumption of an annual event rate
sis. Tertiary end points included death from any for the primary end point of 1.5% in the placebo
cause, hospitalization or an urgent visit for heart group, with an accrual period of 3.5 years and
failure, peripheral arterial revascularization, and an annualized loss to follow-up rate of 2%, the
venous thromboembolic events. An independent sample size was estimated to be approximately
clinical-events committee whose members were 5400 patients to observe the required 256 events.
unaware of the trial-group assignments adjudi- However, to achieve similar power for the prin-
cated all cardiovascular end points. Definitions cipal sensitivity analysis, the required sample size
of the end points are provided in the Supplemen- was estimated to be approximately 6000 patients,

Table 1. Baseline Characteristics of Patients in the Full-Analysis Population.*
Testosterone Group Placebo Group

Characteristic (N = 2601) (N = 2603)
Mean age — yr 63.3±7.9 63.3±7.9
Age ≥65 yr — no. (%) 1241 (47.7) 1211 (46.5)
Race or ethnic group — no. (%)†
White 2070 (79.6) 2084 (80.1)
Black 445 (17.1) 432 (16.6)
Other 86 (3.3) 87 (3.3)
Hispanic or Latinx 409 (15.7) 439 (16.9)
Body-mass index‡ 35.0±5.7 34.8±6.0
Median testosterone level (IQR) — ng/deciliter 227 (189–258) 227 (188–258)
Cardiovascular risk category — no. (%)
Preexisting cardiovascular disease 1410 (54.2) 1437 (55.2)
Increased cardiovascular risk 1191 (45.8) 1166 (44.8)
History of coronary artery disease — no. (%) 1158 (44.5) 1160 (44.6)
History of cerebrovascular disease — no. (%) 304 (11.7) 318 (12.2)
History of peripheral arterial disease — no. (%) 158 (6.1) 153 (5.9)
Cardiovascular risk factors — no. (%)
Diabetes, type 1 or type 2 1788 (68.7) 1844 (70.8)
Hypertension 2423 (93.2) 2402 (92.3)
Dyslipidemia 2344 (90.1) 2332 (89.6)
Current smoker 527 (20.3) 534 (20.5)
High-sensitivity C-reactive protein level ≥2 mg/dl 1607 (61.8) 1589 (61.0)
Stage 3 chronic kidney disease 418 (16.1) 393 (15.1)
Elevated coronary calcium score 29 (1.1) 28 (1.1)
Previous testosterone use — no. (%) 5 (0.2) 10 (0.4)
Medication — no. (%)
Lipid-lowering therapy 2185 (84.0) 2180 (83.7)
Aspirin 1571 (60.4) 1550 (59.5)
Phosphodiesterase-5 inhibitor 170 (6.5) 189 (7.3)
Prostate-specific antigen level — ng/ml 0.91±0.65 0.94±0.68
Hematocrit — % 42±4 42±4
Lipid levels — mg/dl
HDL cholesterol 41.9±11.2 41.7±10.9
LDL cholesterol 80.2±34.0 79.3±33.9
Median triglycerides (IQR) 154.6 (108.1–227.6) 157.7 (112.5–226.7)
* Plus–minus values are means ±SD. The full-analysis population consisted of all patients who had undergone random-
ization. To convert the values for high-density lipoprotein (HDL) cholesterol and low-density lipoprotein (LDL) choles-
terol to millimoles per liter, multiply by 0.02586. To convert the values for triglycerides to millimoles per liter, multiply
by 0.01129. To convert the values for serum total testosterone to nanomoles per liter, divide by 28.84. IQR denotes
interquartile range.
† Race and ethnic group were reported by the patients.
‡ The body-mass index is the weight in kilograms divided by the square of the height in meters.

on the assumption of a testosterone or placebo

A Serum Testosterone Levels over Time
discontinuation rate of 20% in the first year and 600
10% in each subsequent year. To reach the in-
550
tended event rate for the primary end point, the
500
protocol allowed the executive committee and
450
sponsor to limit enrollment in the cohort with
Median Level (ng/dl)

cardiovascular risk factors if that cohort exceeded 400 Testosterone
70% of the total enrollment or if the pooled pri- 350

mary event rate fell below projections. All statis- 300
tical analyses were performed with the use of 250
SAS software, version 9.4. 200 Placebo
150
R e sult s 100
0
Patients 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48
The first patient was enrolled on May 23, 2018. Months since Randomization
A total of 5246 patients underwent randomiza-
tion, and 20 patients were excluded from the full- B Change in Serum Testosterone Levels over Time
analysis population owing to duplicate enrollment 350
325
involving a total of 42 patient identification 300
numbers (Fig. S1). Of the 5204 patients in the 275
250
full-analysis population, 2601 were assigned to
Median Change (ng/dl)
225
receive testosterone and 2603 were assigned to 200
receive placebo. The safety population included 175
150 Testosterone
5198 patients who had received at least one dose 125
of testosterone or placebo (2596 patients in the 100
75
testosterone group and 2602 in the placebo group). 50
In April 2019, blinded data from the first 25
0
2669 patients showed a pooled primary event −25
Placebo
rate below the projected rate of 1.5% per year. −50
Therefore, on May 31, 2019, the executive com- 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48
mittee and sponsor discontinued enrollment of Months since Randomization

patients who qualified on the basis of cardiovas-
Figure 1. Median Serum Testosterone Levels and Changes in Serum
cular risk factors and thereafter enrolled only Testosterone Levels over Time.
those with preexisting cardiovascular disease. Shown are median serum testosterone levels (Panel A) and changes from
Subsequently, on the basis of blinded assessments baseline in serum testosterone levels over time since randomization (Pan-
of pooled accrual rates of major adverse cardiac el B). Blood samples were collected 24 hours (±2 hours) after the patient re-
events, enrollment was terminated on February 1, ceived a dose of testosterone or placebo, preferably in the morning. Verti-
cal bars represent interquartile ranges. To convert serum total
2022, and end-of-trial visits began after May 31,
testosterone levels to nanomoles per liter, divide by 28.84.
2022. At the completion of follow-up, 372 adju-
dicated primary end-point events had occurred,
and 306 primary end-point events qualified for the
principal sensitivity analysis. middle-aged and older men with hypogonadism
Baseline characteristics are summarized in and a high risk of cardiovascular events (Table S2).
Table 1. The trial included 2847 patients with The mean (±SD) durations of treatment and
preexisting cardiovascular disease and 2357 pa- follow-up, respectively, were 21.8±14.2 and
tients at elevated cardiovascular risk. The baseline 33.1±12.0 months in the testosterone group
median serum testosterone level was 227 ng per and 21.6±14.0 and 32.9±12.1 months in the pla-
deciliter; interquartile range, 188 to 258 (7.8 nmol cebo group. A total of 61.4% of the patients in
per liter; interquartile range, 6.5 to 8.9) in the com- the testosterone group discontinued testoster-
bined trial groups. The demographic character- one, and 61.7% of the patients in the placebo
istics were representative of the population of group discontinued placebo (Fig. S2); patients

A Primary Cardiovascular Composite Safety End Point: Safety B Primary Cardiovascular Composite Safety End Point: Principal
Population Sensitivity Analysis Population
100 15 Hazard ratio, 0.96 (95% CI, 0.78–1.17) 100 15 Hazard ratio, 1.02 (95% CI, 0.81–1.27)
P<0.001 for noninferiority P<0.001 for noninferiority
Cumulative Incidence (%)

10 10 Placebo
75 75
Placebo
50 5 Testosterone 50 5 Testosterone
0 0
25 25
0 6 12 18 24 30 36 42 48 0 6 12 18 24 30 36 42 48
0 0
0 6 12 18 24 30 36 42 48 0 6 12 18 24 30 36 42 48
Months since Randomization Months since Randomization
No. at Risk No. at Risk
Placebo 2602 2507 2323 2088 1792 1568 1337 598 33 Placebo 2602 2507 2323 1842 1390 1070 829 347 16
Testosterone 2596 2504 2339 2120 1829 1605 1380 653 39 Testosterone 2596 2504 2339 1850 1408 1089 850 370 20
C Secondary Cardiovascular Composite Safety End Point: Safety D Death from Cardiovascular Causes: Safety Population
Population
Placebo

15 8
75 75 6
10
4 Placebo
Testosterone
50 5 50
2 Testosterone
0 0
25 25
0 6 12 18 24 30 36 42 48 0 6 12 18 24 30 36 42 48
0 0
0 6 12 18 24 30 36 42 48 0 6 12 18 24 30 36 42 48
Placebo 2602 2488 2289 2048 1747 1522 1293 575 31 Placebo 2602 2533 2360 2130 1845 1624 1390 619 34
E Nonfatal Myocardial Infarction: Safety Population F Nonfatal Stroke: Safety Population

Placebo
4
75 75 2
3 Placebo
2 Testosterone Testosterone
50 50 1
1
0 0
25 25
0 6 12 18 24 30 36 42 48 0 6 12 18 24 30 36 42 48
0 0
0 6 12 18 24 30 36 42 48 0 6 12 18 24 30 36 42 48
Placebo 2602 2515 2335 2104 1815 1590 1357 603 33 Placebo 2602 2525 2347 2111 1819 1600 1366 610 34
received the assigned regimen for 67.5% and trial before end-of-trial visits began, and an ad-
67.3% of the potential treatment time in the ditional 947 patients (18.2%) did not attend their
testosterone and placebo groups, respectively. A end-of-trial visit after that date. Withdrawal rates
total of 1082 patients (20.8%) withdrew from the were similar in the two trial groups, and outcome

Figure 2 (facing page). Time-to-Event Analysis for the A primary safety end-point event (the first ad-
Primary and Secondary Cardiovascular Safety End Points. judicated major adverse cardiac event) occurred
Panel A shows the cumulative incidence of the primary in 182 patients (7.0%) in the testosterone group
cardiovascular composite safety end point, which was and in 190 patients (7.3%) in the placebo group
defined as death from cardiovascular causes, nonfatal (hazard ratio, 0.96; 95% confidence interval [CI],
myocardial infarction, or nonfatal stroke. Panel B
0.78 to 1.17; P<0.001 for noninferiority) (Fig. 2A
shows the cumulative incidence of the primary cardio-
vascular composite safety end point in the primary and Table 2). In the principal sensitivity analysis
sensitivity analysis population in which data on end- (with censoring of data on events that occurred
point events that occurred more than 365 days after >365 days after the last dose), a primary safety
discontinuation of testosterone or placebo were cen- end-point event occurred in 154 patients (5.9%)
sored. Panel C shows the cumulative incidence of the
in the testosterone group and in 152 patients
secondary cardiovascular composite safety end point
of death from cardiovascular causes, nonfatal myocar- (5.8%) in the placebo group (hazard ratio, 1.02;
dial infarction, nonfatal stroke, or coronary revascular- 95% CI, 0.81 to 1.27; P<0.001 for noninferiority)
ization; Panel D, the cumulative incidence of death due (Fig. 2B). Similar findings were observed in sen-
to cardiovascular causes; Panel E, the cumulative inci- sitivity analyses in which data on events were
dence of nonfatal myocardial infarction; and Panel F,
censored more than 30 days after the last dose
the cumulative incidence of nonfatal stroke. The defini-
tions of all end points are provided in the Supplemen- or after interruption of testosterone or placebo,
tary Appendix. In each case, the cumulative incidence in the confirmatory analysis involving the full-
was estimated with the Kaplan–Meier method, and the analysis population, and among prespecified sub-
hazard ratio was calculated with the Cox proportional- groups; analysis of the restricted mean survival
hazards regression model with adjustment for preex-
time at 3 years met the criteria for noninferiority
isting cardiovascular disease (yes or no) as a covariate.
Because the statistical analysis plan did not include a (Figs. S3 and S4 and Table S5).
provision for correcting for multiplicity when conduct- No apparent clinically meaningful differences
ing tests for secondary or other end points, results are in the incidence of secondary cardiovascular
reported as point estimates and 95% confidence inter- end-point events were observed between the trial
vals. The widths of the confidence intervals have not
groups (Table 2 and Fig. 2C through 2F). A higher
been adjusted for multiplicity, so the intervals should
not be used to infer definitive treatment effects for incidence of pulmonary embolism, a component
secondary end points. The insets show the same data of the adjudicated tertiary end point of venous
on an enlarged y axis. thromboembolic events (Table 2), occurred in the
testosterone group than in the placebo group
(0.9% vs. 0.5%).
data were available for 82.7% of the possible
follow-up time (observed person-time divided by Adverse Events
total person-time, on the assumption of no with- Prostate cancer occurred in 12 patients (0.5%) in
drawals)13 in the testosterone group and 81.7% of the testosterone group and in 11 patients (0.4%)
the possible follow-up time in the placebo group. in the placebo group (P = 0.87); these cases were
adjudicated as high-grade prostate cancer (Glea-
Trial End Points son score, 4+3 or higher, indicating an interme-
The mean (±SD) daily dose of testosterone was diate or higher risk of progression) in 5 patients
65±22 mg. Serum testosterone levels that were and 3 patients, respectively (P = 0.51). The in-
measured approximately 24 hours after the patient crease in PSA levels from baseline was greater in
had received a dose of testosterone or placebo patients in the testosterone group than in those
over the course of the trial are shown in Figure 1 in the placebo group (0.20±0.61 ng per milliliter
and Table S3. At 12 months, the median increase vs. 0.08±0.90 ng per milliliter, respectively;
from baseline in serum testosterone levels was P<0.001). The change in mean systolic blood
148 ng per deciliter; interquartile range, 34 to pressure from baseline through 6 months was
312 (5.1 nmol per liter; interquartile range, 1.2 0.3 mm Hg (95% CI, −0.3 to 0.9) in the testos-
to 10.8) in the testosterone group, as compared terone group and −1.5 mm Hg (95% CI, −2.0 to
with a median increase of 14 ng per deciliter; −0.9) in the placebo group (P<0.001) (Table S6).
interquartile range, −21 to 56 (0.5 nmol per liter; Investigator-reported adverse events are summa-
interquartile range, −0.7 to 1.9) in the placebo rized in Table 3. Nonfatal arrhythmias warrant-
group. Estradiol levels are summarized in Table S4. ing intervention occurred in 134 patients (5.2%)

Table 2. Adjudicated Cardiovascular End Points in the Safety Population.*
Testosterone Group Placebo Group Hazard Ratio

End Point (N = 2596) (N = 2602) (95% CI)†
number of patients (percent)

Primary cardiovascular composite end point: major 182 (7.0) 190 (7.3) 0.96 (0.78–1.17)
adverse cardiac events‡
Secondary cardiovascular composite end point§ 269 (10.4) 264 (10.1) 1.02 (0.86–1.21)
Components of primary and secondary composite
end points
Death from cardiovascular causes 87 (3.4) 103 (4.0) 0.84 (0.63–1.12)
Nonfatal myocardial infarction 68 (2.6) 62 (2.4) 1.10 (0.78–1.56)
Nonfatal stroke 36 (1.4) 38 (1.5) 0.94 (0.60–1.49)
Coronary revascularization 144 (5.5) 121 (4.6) 1.20 (0.95–1.53)
Tertiary end points
Death from any cause 144 (5.5) 148 (5.7) 0.98 (0.78–1.23)
Hospitalization or urgent visit for heart failure 55 (2.1) 50 (1.9) 1.11 (0.76–1.62)
Peripheral arterial revascularization 30 (1.2) 33 (1.3) 0.92 (0.56–1.51)
Venous thromboembolic events¶ 44 (1.7) 30 (1.2) 1.46 (0.92–2.32)
Components of venous thromboembolic events‖
Pulmonary embolism 24 (0.9) 12 (0.5) —
Deep venous thrombosis 16 (0.6) 13 (0.5) —
Other peripheral venous thrombosis 11 (0.4) 12 (0.5) —
* The safety population consisted of all patients who had undergone randomization and received at least one dose of
testosterone or placebo.
† The analysis plan did not include a provision for correcting for multiplicity when conducting tests for secondary or oth-
er end points, so results are reported as point estimates and 95% confidence intervals. The widths of the confidence
intervals have not been adjusted for multiplicity, so the intervals should not be used to infer definitive treatment effects
for secondary end points.
‡ The composite end point of major adverse cardiac events in the primary analysis was the first occurrence of death from
cardiovascular causes, nonfatal myocardial infarction, or nonfatal stroke. P<0.001 for the noninferiority of testosterone
to placebo with regard to this end point.
§ The secondary composite cardiovascular end point was the first occurrence of the components of the primary end
point (major adverse cardiac events) plus coronary revascularization procedures (percutaneous coronary intervention
or coronary-artery bypass graft surgery).
¶ Venous thromboembolic events include pulmonary embolism, deep venous thrombosis, and other peripheral venous
thrombosis. Other peripheral venous thrombosis must have been confirmed by imaging and included thrombosis
in the calf, portal, subclavian, or mesenteric veins. Superficial thrombophlebitis alone was not considered a venous
thromboembolic event.
‖ There was no prespecified plan to calculate hazard ratios for the individual components of venous thromboembolic
events.
in the testosterone group and in 87 patients (3.3%)terone-replacement therapy in middle-aged and

in the placebo group (P = 0.001); atrial fibrillation
older men with hypogonadism increases the risk
occurred in 91 patients (3.5%) and 63 patients of cardiovascular events. We included men with
(2.4%), respectively (P = 0.02), and acute kidney established cardiovascular disease or multiple
injury occurred in 60 patients (2.3%) and 40 pa- cardiac risk factors who would be most vulner-
tients (1.5%), respectively (P = 0.04). able to an increased risk of adverse cardiovascu-
lar outcomes. Among 5198 patients who received
testosterone or placebo for a mean duration of
Discussion
22 months, testosterone-replacement therapy was
We conducted this randomized, placebo-controlled noninferior to placebo with respect to the inci-
trial to address uncertainty as to whether testos- dence of major adverse cardiac events. The haz-

Table 3. Investigator-Reported Adverse Events.*
Testosterone Group Placebo Group

Event (N = 2596) (N = 2602) P Value†
number of patients (percent)

Any adverse event 1187 (45.7) 1164 (44.7) 0.47
Serious adverse event 721 (27.8) 697 (26.8) 0.42
Adverse event leading to discontinuation of testosterone 244 (9.4) 226 (8.7) 0.37
or placebo
Prespecified adverse events of special interest 196 (7.6) 167 (6.4) 0.11
Hospitalization for unstable angina 44 (1.7) 60 (2.3) 0.12
Nonfatal arrhythmia warranting intervention 134 (5.2) 87 (3.3) 0.001
Cardiovascular disease causing syncope 27 (1.0) 32 (1.2) 0.52
Transient ischemic attack 15 (0.6) 17 (0.7) 0.73
Other adverse events
Diabetes mellitus 189 (7.3) 213 (8.2) 0.22
Coronavirus disease 2019 121 (4.7) 117 (4.5) 0.78
Atrial fibrillation 91 (3.5) 63 (2.4) 0.02
Pneumonia 64 (2.5) 56 (2.2) 0.45
Acute kidney injury 60 (2.3) 40 (1.5) 0.04
Benign prostatic hyperplasia 45 (1.7) 46 (1.8) 0.92
Acute respiratory failure 52 (2.0) 37 (1.4) 0.11
Urinary retention 50 (1.9) 34 (1.3) 0.08
Cellulitis 35 (1.3) 46 (1.8) 0.22
Congestive cardiac failure 34 (1.3) 41 (1.6) 0.42
* The safety population consisted of all patients who had undergone randomization and received at least one dose of
testosterone or placebo. Events are classified according to preferred terms in the Medical Dictionary for Regulatory
Activities, version 25.0.
† P values were calculated with the use of a chi-square test.
ard ratio for a primary end-point event was 0.96 confounding and selection bias, also showed con-
(95% CI, 0.78 to 1.17), a noninferiority finding flicting findings,1-5 although one analysis from a
that was confirmed in sensitivity analyses with large health care database suggested that an age
adjustment for the influence of interruptions or of 65 years or older or preexisting cardiovascular
discontinuations of testosterone or placebo. disease might be important modifiers of in-
The FDA initiated a review of the cardiovas- creased risk with testosterone-replacement ther-
cular safety of testosterone products in 2010 after apy.2 Nearly half the patients enrolled in the
a small placebo-controlled trial was prematurely current trial were 65 years of age or older, and
discontinued because of an increased incidence more than half had preexisting cardiovascular
of cardiovascular events among patients who disease. The 372 adjudicated primary end-point
received testosterone therapy.14 However, subse- events that occurred in this trial were greater in
quent meta-analyses showed variable results and number than those in all previous randomized
most, including a recent patient-level meta-analy- trials of testosterone combined.
sis,15 did not confirm these findings. Cardiac The incidence of pulmonary embolism was
events in those trials were not uniformly adjudi- higher with testosterone than with placebo. Al-
cated, definitions were broad and inconsistent, though most reported cases of thrombosis as-
and the trials were not adequately powered for sociated with testosterone therapy have been in
cardiovascular outcomes. Retrospective observa- men with underlying thrombophilia,16 a meta-
tional studies, which were limited by potential analysis of randomized trials did not show an

association between venous thromboembolic of testosterone20,21 as well as other therapies for

events and testosterone use in wider populations.15 symptomatic conditions such as menopause,22,23
Our findings support current guidelines that obesity,24,25 and chronic pain.26 These findings
testosterone should be used with caution in men reflect the challenge of retaining patients who
who have had previous thromboembolic events.17 have persistent symptoms and choose to switch
There were more cases of nonfatal arrhythmias medications or discontinue trial participation. The
warranting intervention, atrial fibrillation, and consistency of the results derived from the inten-
acute kidney injury among patients who received tion-to-treat and on-treatment sensitivity analy-
testosterone than among those who received ses suggests that the noninferiority conclusions
placebo; these adverse events were not expected. were unlikely to have been influenced by nonad-
A cohort study suggested that normalization of herence. The effect of nonretention, however, is
low testosterone levels by testosterone-replacement more difficult to quantify, and the possibility of
therapy was associated with a decrease in the in- bias due to informative censoring cannot be ruled
cidence of atrial fibrillation.18 In the current trial, out despite a similar incidence of nonretention in
the small increase in blood pressure observed in the two trial groups. It is somewhat reassuring
the testosterone group was similar to that re- that of the patients who discontinued trial partici-
ported previously with other testosterone formu- pation, nearly half did so after end-of-trial visits
lations.19 began, and outcome data were available for more
Our trial design specified that noninferiority than 80% of the potential person-time follow-up.
testing for cardiovascular safety be performed in Controlled trials have shown that the use of
the population of all patients who had under- testosterone in older men improves sexual func-
gone randomization and who had received at tion,27-29 increases volumetric and areal bone min-
least one dose of testosterone or placebo. This eral density,30 corrects unexplained anemia,31,32 and
modified intention-to-treat analysis preserves the moderately reduces depressive symptoms.33 How-
integrity of randomization but may attenuate dif- ever, because testosterone deficiency is not a life-
ferences between trial groups and bias toward threatening condition, uncertainty about cardio-
noninferiority when patients are not adherent to vascular outcomes has weighed on treatment
or discontinue their assigned testosterone or decisions by clinicians and patients.17 Our find-
placebo. We therefore planned three on-treat- ings regarding the cardiovascular safety of tes-
ment sensitivity analyses and specified that the tosterone may facilitate a more informed consid-
trial would not conclude until accrual of at least eration of the potential benefits and risks of
256 primary end-point events that met the crite- testosterone therapy among middle-aged and older
ria for the principal sensitivity analysis. By con- men with hypogonadism.
sidering only those end-point events that occurred Among men with hypogonadism and estab-
while patients were receiving testosterone or pla- lished cardiovascular disease or multiple risk
cebo or within a time frame of physiologic effect factors for incident cardiac events, testosterone-
after discontinuation of testosterone or placebo, replacement therapy was noninferior to placebo
on-treatment analyses can strengthen safety sig- with respect to the occurrence of major adverse
nals by providing more specific accounting of cardiac events during a mean 22-month follow-
outcomes related to testosterone therapy. These up, and the overall incidence of adverse events
sensitivity analyses provided support for the non- was low.
inferiority conclusions. Supported by AbbVie, Acerus Pharmaceuticals, Endo Pharma-
The incidence of adherence and of retention ceuticals, and Upsher–Smith Laboratories.
was lower in this trial than in most cardiovascu- Disclosure forms provided by the authors are available with
the full text of this article at NEJM.org.
lar outcome studies. Similar levels of adherence A data sharing statement provided by the authors is available
and retention have been reported in other trials with the full text of this article at NEJM.org.
Appendix
The authors’ full names and academic degrees are as follows: A. Michael Lincoff, M.D., Shalender Bhasin, M.B., B.S., Panagiotis Fleva-
ris, M.D., Ph.D., Lisa M. Mitchell, R.N., B.S.N., Shehzad Basaria, M.D., William E. Boden, M.D., Glenn R. Cunningham, M.D., Chris-
topher B. Granger, M.D., Mohit Khera, M.D., M.P.H., Ian M. Thompson, Jr., M.D., Qiuqing Wang, M.S., Kathy Wolski, M.P.H., Debo-
rah Davey, R.N., Vidyasagar Kalahasti, M.D., Nader Khan, M.D., Michael G. Miller, Pharm.D., Michael C. Snabes, M.D., Ph.D., Anna
Chan, Pharm.D., Elena Dubcenco, M.D., Xue Li, Ph.D., Tingting Yi, Ph.D., Bidan Huang, Ph.D., Karol M. Pencina, Ph.D., Thomas G.
Travison, Ph.D., and Steven E. Nissen, M.D.

The authors’ affiliations are as follows: the Cleveland Clinic Coordinating Center for Clinical Research, Department of Cardiovascular
Medicine, Cleveland Clinic, Cleveland (A.M.L., L.M.M., Q.W., K.W., D.D., V.K., S.E.N.); the Research Program in Men’s Health: Aging
and Metabolism, Brigham and Women’s Hospital, Harvard Medical School (S. Bhasin, S. Basaria, K.M.P.), Veterans Affairs Boston
Healthcare System and Massachusetts Veterans Epidemiology, Research, and Information Center, Boston University School of Medicine
(W.E.B.), and Marcus Institute for Aging Research, Beth Israel Deaconess Medical Center, Harvard Medical School (T.G.T.) — all in
Boston; AbbVie, North Chicago, IL (P.F., N.K., M.G.M., M.C.S., A.C., E.D., X.L., T.Y., B.H.); Baylor College of Medicine, Houston
(G.R.C., M.K.), and CHRISTUS Santa Rosa Health System and the University of Texas Health Science Center, San Antonio (I.M.T.) — all
in Texas; and Duke Clinical Research Institute, Durham, NC (C.B.G.).
References
1. Baillargeon J, Urban RJ, Kuo Y-F, et al. -communication-fda-cautions-about-using 23. Rossouw JE, Anderson GL, Prentice
Risk of myocardial infarction in older men -testosterone-products-low-testosterone RL, et al. Risks and benefits of estrogen
receiving testosterone therapy. Ann Phar- -due). plus progestin in healthy postmenopausal
macother 2014;48:1138-44. 12. Bhasin S, Lincoff AM, Basaria S, et al. women: principal results from the Wom-
2. Finkle WD, Greenland S, Ridgeway Effects of long-term testosterone treat- en’s Health Initiative randomized con-
GK, et al. Increased risk of non-fatal myo- ment on cardiovascular outcomes in men trolled trial. JAMA 2002;288:321-33.
cardial infarction following testosterone with hypogonadism: rationale and design 24. Bohula EA, Wiviott SD, McGuire DK, et
therapy prescription in men. PLoS One of the TRAVERSE study. Am Heart J 2022; al. Cardiovascular safety of lorcaserin in
2014;9(1):e85805. 245:41-50. overweight or obese patients. N Engl J Med
3. Muraleedharan V, Marsh H, Kapoor D, 13. Xue X, Agalliu I, Kim MY, et al. New 2018;379:1107-17.
Channer KS, Jones TH. Testosterone defi- methods for estimating follow-up rates in 25. Nissen SE, Wolski KE, Prcela L, et al.
ciency is associated with increased risk of cohort studies. BMC Med Res Methodol Effect of naltrexone-bupropion on major
mortality and testosterone replacement 2017;17:155. adverse cardiovascular events in over-
improves survival in men with type 2 dia- 14. Basaria S, Coviello AD, Travison TG, et weight and obese patients with cardiovas-
betes. Eur J Endocrinol 2013;169:725-33. al. Adverse events associated with testos- cular risk factors: a randomized clinical
4. Shores MM, Smith NL, Forsberg CW, terone administration. N Engl J Med 2010; trial. JAMA 2016;315:990-1004.
Anawalt BD, Matsumoto AM. Testosterone 363:109-22. 26. Nissen SE, Yeomans ND, Solomon DH,
treatment and mortality in men with low 15. Hudson J, Cruickshank M, Quinton R, et al. Cardiovascular safety of celecoxib,
testosterone levels. J Clin Endocrinol et al. Adverse cardiovascular events and naproxen, or ibuprofen for arthritis. N Engl
Metab 2012;97:2050-8. mortality in men during testosterone treat- J Med 2016;375:2519-29.
5. Vigen R, O’Donnell CI, Barón AE, et al. ment: an individual patient and aggregate 27. Brock G, Heiselman D, Maggi M, et al.
Association of testosterone therapy with data meta-analysis. Lancet Healthy Longev Effect of testosterone solution 2% on tes-
mortality, myocardial infarction, and stroke 2022;3(6):e381-e393. tosterone concentration, sex drive and en-
in men with low testosterone levels. JAMA 16. Glueck CJ, Wang P. Testosterone thera- ergy in hypogonadal men: results of a pla-
2013;310:1829-36. py, thrombosis, thrombophilia, cardiovas- cebo controlled study. J Urol 2016; 195:
6. Alexander GC, Iyer G, Lucas E, Lin D, cular events. Metabolism 2014;63:989-94. 699-705.
Singh S. Cardiovascular risks of exogenous 17. Bhasin S, Brito JP, Cunningham GR, et 28. Cunningham GR, Stephens-Shields AJ,
testosterone use among men: a systematic al. Testosterone therapy in men with hypo- Rosen RC, et al. Testosterone treatment
review and meta-analysis. Am J Med 2017; gonadism: an Endocrine Society clinical and sexual function in older men with
130:293-305. practice guideline. J Clin Endocrinol Metab low testosterone levels. J Clin Endocrinol
7. Corona G, Maseroli E, Rastrelli G, et 2018;103:1715-44. Metab 2016;101:3096-104.
al. Cardiovascular risk associated with 18. Sharma R, Oni OA, Gupta K, et al. 29. Steidle C, Schwartz S, Jacoby K, et al.
testosterone-boosting medications: a sys- Normalization of testosterone levels after AA2500 testosterone gel normalizes an-
tematic review and meta-analysis. Expert testosterone replacement therapy is associ- drogen levels in aging males with improve-
Opin Drug Saf 2014;13:1327-51. ated with decreased incidence of atrial fi- ments in body composition and sexual
8. Elliott J, Kelly SE, Millar AC, et al. Tes- brillation. J Am Heart Assoc 2017; 6(5): function. J Clin Endocrinol Metab 2003;88:
tosterone therapy in hypogonadal men: a e004880. 2673-81.
systematic review and network meta-anal- 19. White WB, Bernstein JS, Rittmaster R, 30. Snyder PJ, Kopperdahl DL, Stephens-
ysis. BMJ Open 2017;7(11):e015284. Dhingra O. Effects of the oral testosterone Shields AJ, et al. Effect of testosterone
9. Onasanya O, Iyer G, Lucas E, Lin D, undecanoate Kyzatrex™ on ambulatory treatment on volumetric bone density and
Singh S, Alexander GC. Association be- blood pressure in hypogonadal men. J Clin strength in older men with low testoster-
tween exogenous testosterone and cardio- Hypertens 2021;23:1420-30. one: a controlled clinical trial. JAMA In-
vascular events: an overview of systematic 20. Basaria S, Harman SM, Travison TG, et tern Med 2017;177:471-9.
reviews. Lancet Diabetes Endocrinol 2016; al. Effects of testosterone administration 31. Bachman E, Travison TG, Basaria S, et
4:943-56. for 3 years on subclinical atherosclerosis al. Testosterone induces erythrocytosis via
10. Xu L, Freeman G, Cowling BJ, School- progression in older men with low or low- increased erythropoietin and suppressed
ing CM. Testosterone therapy and cardio- normal testosterone levels: a randomized hepcidin: evidence for a new erythropoie-
vascular events among men: a systematic clinical trial. JAMA 2015;314:570-81. tin/hemoglobin set point. J Gerontol A Biol
review and meta-analysis of placebo-con- 21. Page ST, Amory JK, Bowman FD, et al. Sci Med Sci 2014;69:725-35.
trolled randomized trials. BMC Med 2013; Exogenous testosterone (T) alone or with 32. Roy CN, Snyder PJ, Stephens-Shields
11:108. finasteride increases physical perfor- AJ, et al. Association of testosterone levels
11. Food and Drug Administration. FDA mance, grip strength, and lean body mass with anemia in older men: a controlled
cautions about using testosterone products in older men with low serum T. J Clin En- clinical trial. JAMA Intern Med 2017;177:
for low testosterone due to aging: requires docrinol Metab 2005;90:1502-10. 480-90.
labeling change to inform of possible in- 22. Grady D, Herrington D, Bittner V, et al. 33. Snyder PJ, Bhasin S, Cunningham GR,
creased risk of heart attack and stroke Cardiovascular disease outcomes during et al. Effects of testosterone treatment in
with use. FDA drug safety communica- 6.8 years of hormone therapy: Heart and older men. N Engl J Med 2016;374:611-24.
tion. 2018 (https://www.fda.gov/drugs/drug Estrogen/progestin Replacement Study fol- Copyright © 2023 Massachusetts Medical Society.
-safety-and-availability/fda-drug-safety low-up (HERS II). JAMA 2002;288:49-57.


Nejmoa 2215025

Uploaded by

Copyright:

Available Formats

Nejmoa 2215025

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Nejmoa 2215025

Uploaded by

Copyright:

Available Formats

The n e w e ng l a n d j o u r na l of m e dic i n e

Cardiovascular Safety of Testosterone-

from cardiovascular causes, nonfatal myocardial infarction, nonfatal stroke, or coro-

n engl j med﻿﻿ nejm.org﻿ 1

2 n engl j med﻿﻿ nejm.org﻿

The New England Journal of Medicine

n engl j med﻿﻿ nejm.org﻿ 3

Table 1. Baseline Characteristics of Patients in the Full-Analysis Population.*

Testosterone Group Placebo Group

4 n engl j med﻿﻿ nejm.org﻿

The New England Journal of Medicine

on the assumption of a testosterone or placebo

Median Level (ng/dl)

70% of the total enrollment or if the pooled pri- 350

mittee and sponsor discontinued enrollment of Months since Randomization

n engl j med﻿﻿ nejm.org﻿ 5

Cumulative Incidence (%)

Cumulative Incidence (%)

E Nonfatal Myocardial Infarction: Safety Population F Nonfatal Stroke: Safety Population

Cumulative Incidence (%)

6 n engl j med﻿﻿ nejm.org﻿

The New England Journal of Medicine

n engl j med﻿﻿ nejm.org﻿ 7

Table 2. Adjudicated Cardiovascular End Points in the Safety Population.*

Testosterone Group Placebo Group Hazard Ratio

number of patients (percent)

in the testosterone group and in 87 patients (3.3%)terone-replacement therapy in middle-aged and

8 n engl j med﻿﻿ nejm.org﻿

The New England Journal of Medicine

Table 3. Investigator-Reported Adverse Events.*

Testosterone Group Placebo Group

number of patients (percent)

n engl j med﻿﻿ nejm.org﻿ 9

association between venous thromboembolic of testosterone20,21 as well as other therapies for

10 n engl j med﻿﻿ nejm.org﻿

The New England Journal of Medicine

n engl j med﻿﻿ nejm.org﻿ 11

You might also like

n engl j med nejm.org 1

2 n engl j med nejm.org

n engl j med nejm.org 3

4 n engl j med nejm.org

n engl j med nejm.org 5

6 n engl j med nejm.org

n engl j med nejm.org 7

8 n engl j med nejm.org

n engl j med nejm.org 9

10 n engl j med nejm.org

n engl j med nejm.org 11