Effect of Testosterone Treatment on Volumetric Bone Density
and Strength in Older Men With Low Testosterone
A Controlled Clinical Trial
Peter J. Snyder, MD; David L. Kopperdahl, PhD; Alisa J. Stephens-Shields, PhD; Susan S. Ellenberg, PhD;
Jane A. Cauley, DrPH; Kristine E. Ensrud, MD, MPH; Cora E. Lewis, MD, MPH; Elizabeth Barrett-Connor, MD;
Ann V. Schwartz, PhD, MPH; David C. Lee, PhD; Shalender Bhasin, MD; Glenn R. Cunningham, MD;
Thomas M. Gill, MD; Alvin M. Matsumoto, MD; Ronald S. Swerdloff, MD; Shehzad Basaria, MD;
Susan J. Diem, MD, MPH; Christina Wang, MD; Xiaoling Hou, MS; Denise Cifelli, MS; Darlene Dougar, MPH;
Bret Zeldow, MS; Douglas C. Bauer, MD; Tony M. Keaveny, PhD
IMPORTANCE
As men age, they experience decreased serum testosterone concentrations,
decreased bone mineral density (BMD), and increased risk of fracture.
OBJECTIVE To determine whether testosterone treatment of older men with low
testosterone increases volumetric BMD (vBMD) and estimated bone strength.
DESIGN, SETTING, AND PARTICIPANTS Placebo-controlled, double-blind trial with treatment
allocation by minimization at 9 US academic medical centers of men 65 years or older with 2
testosterone concentrations averaging less than 275 ng/L participating in the Testosterone
Trials from December 2011 to June 2014. The analysis was a modified intent-to-treat
comparison of treatment groups by multivariable linear regression adjusted for balancing
factors as required by minimization.
INTERVENTIONS Testosterone gel, adjusted to maintain the testosterone level within the
normal range for young men, or placebo gel for 1 year.
MAIN OUTCOMES AND MEASURES Spine and hip vBMD was determined by quantitative
computed tomography at baseline and 12 months. Bone strength was estimated by finite
element analysis of quantitative computed tomography data. Areal BMD was assessed by
dual energy x-ray absorptiometry at baseline and 12 months.
RESULTS There were 211 participants (mean [SD] age, 72.3 [5.9] years; 86% white; mean [SD]
body mass index, 31.2 [3.4]). Testosterone treatment was associated with significantly
greater increases than placebo in mean spine trabecular vBMD (7.5%; 95% CI, 4.8% to 10.3%
vs 0.8%; 95% CI, −1.9% to 3.4%; treatment effect, 6.8%; 95% CI, 4.8%-8.7%; P < .001),
spine peripheral vBMD, hip trabecular and peripheral vBMD, and mean estimated strength of
spine trabecular bone (10.8%; 95% CI, 7.4% to 14.3% vs 2.4%; 95% CI, −1.0% to 5.7%;
treatment effect, 8.5%; 95% CI, 6.0%-10.9%; P< .001), spine peripheral bone, and hip
trabecular and peripheral bone. The estimated strength increase s were greater in trabecular
than peripheral bone and greater in the spine than hip. Testosterone treatment increased
spine areal BMD but less than vBMD.
CONCLUSIONS AND RELEVANCE Testosterone treatment for 1 year of older men with low
testosterone significantly increased vBMD and estimated bone strength, more in trabecular
than peripheral bone and more in the spine than hip. A larger, longer trial could determine
whether this treatment also reduces fracture risk.
TRIAL REGISTRATION clinicaltrials.gov Identifier: NCT00799617
JAMA Intern Med. 2017;177(4):471-479. doi:10.1001/jamainternmed.2016.9539
Published online February 21, 2017. Last corrected on March 4, 2019.
Editorial pages 459 and 461
Author Video Interview and
JAMA Report Video
Related articles pages 480
and 491
Supplemental content
Related articles at jama.com
Author Affiliations: Author
affiliations are listed at the end of this
article.
Corresponding Author: Peter J.
Snyder, MD, Perelman School of
Medicine at the University of
Pennsylvania, 12-135 Translational
Research Bldg, 3400 Civic Center
Blvd, Philadelphia, PA 19104-5160
(pjs@mail.med.upenn.edu).
Research
JAMA Internal Medicine | Original Investigation
(Reprinted) 471
© 2017 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 08/26/2022
A
s men age, they experience decreases in serum testos-
terone concentration.
1,2
They also experience de-
creases in areal bone mineral density (aBMD),
3-5
volu-
metric bone mineral density (vBMD),
6
and estimated strength
6
and an increase in fractures.
7
When men of any age develop
severely low testosterone due to known disease, their BMD
decreases
8-11
and fractures increase.
12,13
In men who are frankly
hypogonadal, testosterone treatment improves BMD,
14-16
trabecular architecture,
17
and mechanical properties.
18
Prior studies of the effect of testosterone treatment on
bone in older men, however, have not been conclusive.
19-22
In 1 placebo-controlled study, testosterone treatment did not
improve spine BMD overall, but in a regression model, lower
serum testosterone predicted a significantly greater effect of
testosterone treatment on spine BMD.
19
Another study dem-
onstrated a significant increase in spine and hip BMD in
testosterone-treated men, but supraphysiologic doses of
testosterone were used.
21
We report here the results of the Bone Trial of the Testos-
terone Trials (T-Trials), a group of 7 coordinated trials of the
effects of testosterone treatment of older men with low tes-
tosterone concentrations.
23,24
The purpose of the Bone Trial
was to determine whether testosterone treatment would im-
prove vBMD and estimated bone strength.
Methods
Study Design
The T-Trialswere conducted at 12 US sites; 9 of them participated
in the Bone Trial. The study design has been described.
23
To en-
roll in the T-Trials overall, participants had to qualify for at least
1 of the 3 main trials.
24
If they qualified, they could participate
in the Bone Trial. Participants were randomly assigned to re-
ceive testosterone or placebo gel double-blindly for 1 year. This
report describes the efficacy results for the Bone Trial.
The protocol (Supplement 1) was approved by the institu-
tional review boards of all participating institutions. All men
provided written, informed consent. A data safety monitor-
ing board approved the protocol and monitored unblinded
safety data.
Participants
Participants were recruited and screened as described.
25
Re-
spondents were screened first by telephone interview and then
during 2 clinic visits. To be included in the T-Trials, men had to
be at least 65 years old, have subjective and objective evidence
of impaired sexual or physical function or reduced vitality, and
have a serum testosterone concentration on 2 morning speci-
mens that averaged less than 275 ng/dL (to convert to nano-
moles per liter, multiply by 0.0347). Potential participants were
excluded if they were at increased risk of conditions that tes-
tosterone treatment might exacerbate. Potentialparticipants for
the Bone Trial were also excluded if they were taking a medi-
cation known to affect bone, except for calcium and over-the-
counter vitamin D preparations; if they did not have at least 1
evaluable lumbar vertebra; or if they had a dual-energy x-ray
absorptiometry (DXA) T-score at any site of less than −3.0.
Treatment
We allocated participants to receive testosterone or placebo gel
by minimization.
26,27
Balancing variables included participa-
tion in each of the main trials, clinical site, testosterone con-
centration greater than or less than 200 ng/dL, and age older
or younger than 75 years. The testosterone preparation was
AndroGel 1% in a pump bottle (AbbVie). Placebo gel was simi-
lar. The initial dose was 5 g daily. Serum testosterone concen-
tration was measured at months 1, 2, 3, 6, and 9 in a central
laboratory (Quest Clinical Trials), and the dose of testoste-
rone gel was adjusted after each measurement to attempt to
keep the concentration within the normal range for young men.
To maintain blinding when the dose was adjusted in a partici-
pant taking testosterone, the dose was changed simultane-
ously in a participant taking placebo by a staff person in the
Data Coordinating Center according to a prespecified algo-
rithm; no site personnel knew the treatment allocation.
All participants were given and instructed to take 1 tablet
containing 600 g of elemental calcium and 400 units of vita-
min D
3
twice a day with meals.
Assessments
At the end of the trials, the serum concentrations of testoste-
rone and estradiol were measured by liquid chromatography
and tandem mass spectroscopy and free testosterone by equi-
librium dialysis in the Brigham Research Assay Core Labora-
tory, Boston, Massachusetts.
24
Samples from baseline and
months 3, 6, 9, and 12 from each participant were measured
in the same assay run.
Efficacy outcomes in the Bone Trial were assessed at base-
line and after 12 months of treatment. The primary efficacy out-
come was percent change from baseline in vBMD of trabecu-
lar bone in the lumbar spine, as assessed by means of
quantitative computed tomography (QCT). Volumetric BMD
was chosen as the main method of assessment rather than
aBMD by DXA because it is not artifactually influenced by os-
teophytes and aortic calcification
4,28
and because it can dis-
tinguish between trabecular bone, which testosterone af-
fects primarily, and cortical bone.
18
Secondary outcomes were
vBMD of peripheral bone and whole bone of the lumbar spine
and trabecular, peripheral, and whole bone of the hip; esti-
mated strength of the same sites by finite element analysis
(FEA) from computed tomographic (CT) data; and aBMD
of the spine and hip by DXA.
Key Points
Question Will testosterone treatment of older men with low
testosterone improve their bone density and strength?
Findings Testosterone treatment of older men with low
testosterone increased volumetric trabecular bone mineral density
of the lumbar spine and estimated bone strength significantly
compared with placebo.
Meaning These results suggest that a larger and longer trial to
determine whether testosterone treatment decreases fracture risk
in this population is warranted.
Research Original Investigation Testosterone Treatment and Volumetric Bone Density and Strength
472 JAMA Internal Medicine April 2017 Volume 177, Number 4 (Reprinted) jamainternalmedicine.com
© 2017 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 08/26/2022
All T-Trials participants were asked about fractures every
3 months during treatment and at 6 and 12 months after-
wards. An independent adjudicator reviewed radiographic re-
ports of all reported fractures and adjudicated without knowl-
edge of treatment allocation.
Computed tomographic scans of the lumbar spine and
the hip were performed at baseline and month 12. The QCT
reading center trained the technicians at each of the 9 clini-
cal sites to ensure a consistent imaging technique. The spine
scan extended from mid-T12 to mid-L4; L1 and L2 measure-
ments were used preferentially, but if not assessable, L3 was
used; the values of 2 vertebrae were averaged. The hip scan
extended from 1 cm above the femoral heads to 2 cm below
the lesser trochanters; both hips were used if assessable and
the results averaged. Each image included an external bone
mineral phantom (Mindways Software) beneath the partici-
pant for calibration. A second phantom (Mindways) was
scanned monthly to detect any field nonuniformity or scan-
ner drift. The mean (range) coefficient of variation for all
scanners was 0.23% (0.13%-0.29%). This phantom was also
used for cross-calibration for the 12 participants at 2 sites
whose scans were acquired using different scanners at the
baseline and 12-month visits. A third phantom (European
Spine Phantom, QRM GmbH) was used to verify cross-
calibration.
Image processing, vBMD measurements, and finite ele-
ment strength analyses were performed at a central site (O.N.
Diagnostics), blinded to treatment group, by analysis of the CT
scans using VirtuOst software. O.N. Diagnostics also main-
tained quality control of the CT data collection. Construction
of the finite element models has been described.
29-31
For the
vertebrae, trabecular vBMD was measured using an elliptical
region of interest in the trabecular centrum (eFigure 1 in
Supplement 2). Whole bone and peripheral vBMD were de-
fined, respectively, as the mean vBMD for the whole verte-
bral body, and the outer 2 mm of bone, which included the cor-
tex and neighboring trabecular bone. To measure vertebral
strength, uniform axial compression was applied virtually to
the finite element model through a layer of bone cement
(eFigure 1 in Supplement 2); the whole bone strength was de-
fined as the force at 2% deformation. Trabecular strength was
similarly measured after removing the outer 2 mm of bone, and
peripheral strength was calculated as whole-bone strength
minus trabecular strength.
For the femur, whole-bone vBMD was measured as the
mean density of the entire model. Each model was then di-
vided into a trabecular compartment (all bone with an appar-
ent density less than 1 mg/cm
3
and more than 3 mm from the
periosteum) and a peripheral compartment (all bone not in the
defined trabecular compartment containing the cortex and
some adjacent trabecular bone) (eFigure 1 in Supplement 2).
Trabecular and peripheral vBMD were measured as the mean
vBMD of their respective compartments. Femoral strength was
measured by simulation of a sideways fall. Trabecular strength
was similarly measured after assigning 2 reference densities
to the peripheral compartment; and peripheral strength was
measured after assigning a single reference density to the tra-
becular compartment.
Dual-energy x-ray absorptiometry scans of the lumbar
spine and hip were obtained at the baseline and 12-month
visits using Hologic densitometers. Quality control of DXA
was centrally monitored by the University of California
San Francisco Coordinating Center, DXA QA Group. The DXA
operators at each of the 9 sites were certified at the beginning
of the trial. Scans were analyzed locally, using the same
software version at baseline and follow-up, and sent to the
Coordinating Center for incorporation into a central data-
base. Flagged scans and a random sample of scans were re-
viewed for quality. Longitudinal performance of densitom-
eters was monitored with regular scanning of a spine phantom.
Statistical Analyses
Sample size was based on a prior study in hypogonadal men
that showed a mean (SD) increase in trabecular vBMD of 14%
(3%) over 18 months of testosterone treatment.
14
We posited
a 9% improvement over 12 months, assuming no change in the
placebo group and the same standard deviation. To achieve
90% power with a 2-sided significance level of .05, we re-
quired 172 men; we targeted 200 men to compensate for
nonadherence and dropout.
Analyses followed the intention-to-treat principle; men al-
located to testosterone were compared with men allocated to
placebo, regardless of adherence or T level achieved. All par-
ticipants who had baseline and month 12 scans were in-
cluded in the analyses. Each outcome reported here was pre-
specified. The effect of testosterone compared with placebo
on percent change in bone outcomes was evaluated by mul-
tivariable linear regression, adjusted for balancing factors as
required for the analysis of interventions allocated by mini-
mization. Multiple imputation was used to assess the influ-
ence of missing month 12 scans on the primary outcome analy-
sis. Imputation models included demographic and clinical
variables listed in Table 1. The Markov chain Monte Carlo
method was used to impute missing values. All analyses were
conducted at a 2-sided significance level of .05.
Multivariable linear regression models with interactions
of treatment and baseline factors were used to examine
whether the magnitude of the effect of testosterone treat-
ment differed according to baseline vBMD, total serum tes-
tosterone level, or estradiol level. Unadjusted linear regres-
sion was used to determine, in men in the testosterone arm,
the association of the percent change in trabecular vBMD of
the lumbar spine from baseline to month 12 with absolute
change in total testosterone and estradiol from baseline to
month 12.
Analyses did not adjust for multiple comparisons be-
cause the bone outcomes were likely highly correlated, mak-
ing such adjustments overly conservative.
Results
Participants and Treatment
Recruitment began in December 2011. Targeted enrollment was
completed in June 2013, and treatment was completed in June
2014. Of the 295 men who enrolled in one of the main T-Trials,
Testosterone Treatment and Volumetric Bone Density and Strength Original Investigation Research
jamainternalmedicine.com (Reprinted) JAMA Internal Medicine April 2017 Volume 177, Number 4 473
© 2017 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 08/26/2022
at the 9 clinical trial sites from the inception of the Bone Trial,
211 met Bone Trial entry criteria and enrolled (Figure 1). Allo-
cation to testosterone or placebo treatment was the same for
each participant as in the T-Trials overall. One hundred eighty-
nine participants (90%) completed 12 months of treatment and
had analyzable baseline and 12-month scans. Noncompletion
was more frequent among men in the placebo arm (16 [15.8%]
placebo, 6 [5.5%] testosterone; P = .01); demographic charac-
teristics, baseline hormone levels, and baseline bone strength
and density measures did not differ between completers and
noncompleters.
At baseline, the participants had low serum testosterone
concentrations for young men (Table 1). Baseline characteris-
tics in the 2 treatment arms were similar, including total and
free testosterone levels and aBMD. The mean T scores for the
spine and hip were not low (Table 1). Mean body mass index,
alcohol consumption, and serum estradiol level were slightly
higher in the placebo-treated men.
Treatment with testosterone increased the median se-
rum concentrations of total testosterone, free testosterone, and
estradiol to within the normal ranges for young men (Figure 2).
Efficacy
Testosterone treatment increased mean lumbar spine tra-
becular vBMD (primary outcome) by 7.5% (95% CI, 4.8% to
10.3%), compared with 0.8% (95% CI, −1.9% to 3.4%) by pla-
cebo (Figure 3A and Table 2), a difference of 6.8% (95% CI,
4.8% to 8.7%; P < .001; r
2
= 0.26). The mean difference was
somewhat less (4.0%; 95% CI, 3.0% to 5.0%) in sensitivity
analyses for missing month 12 scans but still significant
(P < .001).
The magnitude of the treatment effect on trabecular vBMD
of the spine did not vary significantly by baseline total testos-
terone, estradiol, or vBMD. The magnitude of the percent in-
crease in spine trabecular vBMD from baseline to month 12 in
testosterone-treated men, however, was significantly associ-
ated with changes in total testosterone (β = 0.01, ρ = 0.25,
P = .01) and estradiol (β = 0.17, ρ = 0.37, P < .001) (eAppendix
1 and eFigure 2 in Supplement 2). A 200 ng/dL increase in tes-
tosterone was associated with a 6.1% increase in trabecular
vBMD, and a 15 pg/mL increase in estradiol was associated with
a 6.3% increase.
Testosterone treatment also increased peripheral and
whole-bone vBMD of the spine and trabecular, peripheral, and
whole-bone vBMD of the hip (Figure 3A and Table 2). The mag-
nitudes of the increases were less in the hip than in the spine
but still statistically significant.
Based on FEA of QCT data, testosterone treatment also
increased estimated bone strength. Testosterone treatment
increased estimated strength of spine trabecular bone by
10.8% (95% CI, 7.4% to 14.3%), compared with 2.4% (95% CI,
−1.0% to 5.7%) in placebo-treated men (Figure 3B and
Table 2). The difference was 8.5% (95% CI, 6.0% to 10.9%;
P < .001). Testosterone treatment also significantly
increased estimated strength of peripheral and whole bone
Table 1. Baseline Characteristics of Participants in the Bone Trial
Characteristic
Testosterone
(n = 110)
Placebo
(n = 101)
Age, mean (SD), y 72.3 (6.3) 72.4 (5.5)
Race, No. (%)
White 93 (84.5) 88 (87.1)
African American 6 (5.5) 4 (4.0)
Other 11 (10.0) 9 (8.9)
Concomitant conditions, mean (SD)
BMI, mean (SD) 30.7 (3.7)
a
31.8 (3.1)
Alcohol use, mean (SD), No. drinks/wk 2.5 (3.5)
a
4.0 (5.3)
Smoking, No. (%)
Current smoker 6 (5.5) 7 (6.9)
Ever smoker 70 (63.6) 72 (71.3)
Diabetes 43 (39.1) 40 (39.6)
Serum steroid hormone, mean (SD)
Total testosterone, ng/dL 229.6 (65.3) 238.8 (64.0)
Free testosterone, pg/mL 61.2 (20.0) 64.5 (21.1)
Estradiol, pg/mL 20.5 (6.7)
a
22.4 (6.4)
DXA areal BMD, mean (SD), g/cm
2
Lumbar spine 1.2 (0.2) 1.2 (0.2)
Total hip 1.0 (0.2) 1.0 (0.1)
Femoral neck 0.8 (0.1) 0.8 (0.1)
DXA BMD T-score,
b
mean (SD)
Lumbar spine 1.3 (1.8) 1.2 (1.8)
Total hip 0.7 (1.2) 0.6 (1.2)
Femoral neck −0.3 (1.1) −0.3 (1.2)
Abbreviations: BMD, bone mineral density; BMI, body mass index (calculated as
weight in kilograms divided by height in meters squared); DXA, dual-energy
x-ray absorptiometry.
SI conversion factors: To convert testosterone to nanomoles per liter, multiply
by 0.0347; to convert free testosterone to picomoles per liter, multiply by 3.47;
to convert estradiol to picomoles per liter, multiply by 3.67.
a
P < .05 compared with placebo (t test).
b
Calculated from young female referent database.
Figure 1. Screening and Retention of Participants
295 Men assessed for eligibility
211 Allocated by minimization
110 Allocated to testosterone
110 Had baseline scan
No month 12 scan
3
1
1
1
Withdrew prior to month 12
Did not withdraw but had no
final scan
Data on final scan lost
Final scan not analyzable
104 Included in the analysis
101 Allocated to placebo
98 Had baseline scan
No month 12 scan
9
3
1
Withdrew prior to month 12
Did not withdraw but had no
final scan
Final scan not analyzable
85 Included in the analysis
84 Excluded
35
6
8
35
Did not meet inclusion
criteria
Excluded for other
reasons
Declined to participate
Not assessed
Research Original Investigation Testosterone Treatment and Volumetric Bone Density and Strength
474 JAMA Internal Medicine April 2017 Volume 177, Number 4 (Reprinted) jamainternalmedicine.com
© 2017 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 08/26/2022
(Figure 3B and Table 2). The magnitudes of the effects of tes-
tosterone treatment on estimated hip strength were less
than those on the spine, but still significant (Figure 3B and
Table 2).
By DXA, testosterone treatment increased mean aBMD
(3.3%; 95% CI, 2.01% to 4.56%) more than placebo (2.1%;
95% CI, 0.87% to 3.36%; P = .01, r
2
= 0.12) (Table 2). In the
total hip, testosterone treatment was associated with a mean
increase of 1.2% (95% CI, 0.19% to 2.17%) compared with
0.5% (95% CI, −0.45% to 1.46%) for placebo (P = .052,
r
2
= 0.13). In the femoral neck, testosterone treatment was
associated with a mean increase of 1.5% (95% CI, 0.02% to
2.97%) and placebo of 0.9% (95% CI, −0.49% to 2.35%;
P =.27,r
2
= 0.06).
Adjusting the analyses for the variables in which the 2 arms
differed at baseline (body mass index, alcohol use, and estra-
diol level) did not change appreciably any of the QCT or DXA
results.
During the treatment year, 6 fractures were reported and
confirmed in each treatment arm (eTable 1 in Supplement 2).
During the subsequent year of observation, 3 fractures were
reported and confirmed in the testosterone arm and 4 in the
placebo arm.
Figure 2. Median Serum Concentrations of Total Testosterone,
Free Testosterone, and Estradiol From Months 0 to 12 in Men
Treated With Testosterone or Placebo
Free testosterone
B
Median Free Testosterone, pg/mL
Time, mo
50
300
250
150
200
100
0
03 1296
P
<
.001
Total testosterone
A
Median Total Testosterone, ng/dL
Time, mo
100
900
700
500
300
800
600
400
200
0
03 1296
P
<
.001
Testosterone
Placebo
Estradiol level
C
Median Estradiol, pg/mL
Time, mo
10
60
50
30
40
20
0
03 1296
P
<
.001
The P values indicate the significance of the difference in serum concentrations
in men in the testosterone arm compared with men in the placebo arm. The
shaded areas represent the normal ranges for healthy young men. Error bars
indicate interquartile ranges.
SI conversion factors: To convert testosterone to nanomoles per liter, multiply
by 0.0347; to convert free testosterone to picomoles per liter, multiply by 3.47;
to convert estradiol to picomoles per liter, multiply by 3.67.
Figure 3. Effects of Testosterone or Placebo Treatment for 12 Months
on Volumetric Bone Mineral Density and Estimated Bone Strength
of Trabecular, Peripheral, and Whole Bone of the Spine and Hip,
as Assessed by Quantitative Computed Tomography
12
10
6
8
4
2
0
−2
Trabecular Peripheral Whole
Change From Baseline, %
Spine
P < .001
P < .001
P < .001
Effect of testosterone on volumetric bone mineral density
A
Testosterone
Placebo
Trabecular Peripheral Whole
Hip
P < .001
P < .001
P < .001
16
14
12
10
6
8
4
2
0
−2
Trabecular Peripheral Whole
Change From Baseline, %
Spine
P < .001
P < .001
P < .001
Effect of testosterone on estimated bone strength
B
Trabecular Peripheral Whole
Hip
P = .005
P < .001
P < .001
Bars indicate means, and error bars, 95% CIs. The P values indicate the
significance of the difference in change in percent volumetric bone mineral
density or estimated strength from baseline to 12 months for men in the
testosterone arm compared with the placebo arm.
Testosterone Treatment and Volumetric Bone Density and Strength Original Investigation Research
jamainternalmedicine.com (Reprinted) JAMA Internal Medicine April 2017 Volume 177, Number 4 475
© 2017 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 08/26/2022