Testosterone Therapy and Cardiovascular Risk

28 Pages·2015·0.46 MB·English

SPECIAL ARTICLE
Testosterone Therapy and Cardiovascular Risk:
Advances and Controversies
Abraham Morgentaler, MD; Martin M. Miner, MD; Monica Caliber, MSc;
y
Andre T. Guay, MD ; Mohit Khera, MD; and Abdulmaged M. Traish, PhD
Abstract
Tworecentstudiesraisednewconcernsregardingcardiovascular(CV)riskswithtestosterone(T)therapy.
ThisarticlereviewsthosestudiesaswellastheextensiveliteratureonTandCVrisks.AMEDLINEsearch
was performed for the years 1940 to August 2014 using the following key words: testosterone, androgens,
human, male, cardiovascular, stroke, cerebrovascular accident, myocardial infarction, heart attack, death, and
mortality.Theweightanddirectionofevidencewasevaluatedandlevelofevidence(LOE)assigned.Only
4 articles were identified that suggested increased CV risks with T prescriptions: 2 retrospective analyses
withseriousmethodologicallimitations,1placebo-controlledtrialwithfewmajoradversecardiacevents,
and1meta-analysisthatincludedquestionablestudiesandevents.Incontrast,severaldozenstudieshave
reportedabeneficial effectofnormalTlevelsonCVrisks andmortality.Mortality andincident coronary
artery disease are inversely associated with serum T concentrations (LOE IIa), as is severity of coronary
artery disease (LOE IIa). Testosterone therapy is associated with reduced obesity, fat mass, and waist
circumference (LOE Ib) and also improves glycemic control (LOE IIa). Mortality was reduced with T
therapy in 2 retrospective studies. Several RCTs in men with coronary artery disease or heart failure re-
ported improved function in men who received T compared with placebo. The largest meta-analysis to
date revealed no increase in CV risks in men who received T and reduced CV risk among those with
metabolicdisease.Insummary,thereisnoconvincingevidenceofincreasedCVriskswithTtherapy.On
thecontrary,thereappearstobeastrongbeneficialrelationshipbetweennormalTandCVhealththathas
not yet been widely appreciated.
ª2015MayoFoundationforMedicalEducationandResearch n MayoClinProc.2015;90(2):224-251
I
n November 2013 and January 2014, 2 cautioningagainsttheuseofTtherapyinolder
studieswerepublishedreportingincreased menandinmenwithahistoryofcoronaryar-
Foreditorial cardiovascular (CV) risks in men who tery disease (CAD).6 The US Food and Drug
comment,seepage received testosterone (T) prescriptions.1,2 Administration (FDA) announced plans to re-
163 Thesearticlesgainedwidemediaattention.Me- view the CV safety of T products 2 days after
dia coverage of these studies was frequently publicationofthesecondarticle.7Plaintiffattor-
FromMen’sHealthBostonand combined with data indicating rapidly in- neysbegananationwidecampaignseekingcases
HarvardMedicalSchool,
creasedsalesofTproducts,3,4raisingconcerns of myocardial infarctions (MIs) and strokes in
ChestnutHill,MA(A.M.);
DepartmentofFamilyMedicine that the pharmaceutical industry was promot- menwhohadusedTproductsforaclassaction
andDepartmentofUrology, ingatreatmentassociatedwithimportantrisks. lawsuit. These concerns thrust T therapy into
BrownUniversity,MiriamHos-
pitalMen’sHealthCenter, This view was captured best by a New York thenews,wherethereportedCVrisksanchored
Providence,RI(M.M.M.);Fort Times editorial entitled “Overselling Testos- a variety ofunrelated concerns regarding other
Lauderdale,FL(M.C.);Depart- terone,Dangerously.”5 aspectsofTtherapy,suchasoveruseandabuse,
mentofEndocrinology,Lahey Theimpactofthesestudiesonpatientman- false claims byantiagingmedicine, profiteering
Clinic,Burlington,MA(A.T.G.);
ScottDepartmentofUrology, agement and the ensuing public attention has bylow-Tclinics,andthefailureofmentoaccept
BaylorCollegeofMedicine, been substantial. Men discontinued treatment, therigorsofnaturalaging.
Houston,TX(M.K.);and
occasionallycriticizingtheirphysiciansforput- It is beyond the scope of this article to
DepartmentofBiochemistry
andDepartmentofUrology, tingtheirhealthatrisk;somephysiciansstopped address these various issues. Instead, we wish
BostonUniversitySchoolof prescribing T products, and others warned to address the key scientificquestion, namely,
yMDeedciceiansee,dB.oston,MA(A.M.T.). against treatment of T deficiency (TD) (also whetherTtherapyisassociatedwithincreased
called hypogonadism or, more casually, low T). CVrisks.Thisreviewencompassesananalysis
The Endocrine Society issued a statement of the literaturepreviously submitted byus to
224 MayoClinProc. n February2015;90(2):224-251 n http://dx.doi.org/10.1016/j.mayocp.2014.10.011
www.mayoclinicproceedings.org n ª2015MayoFoundationforMedicalEducationandResearch
TESTOSTERONE AND CARDIOVASCULAR RISK
the FDA and to the European Medicines negative changes in body composition as well
Agency to assist with their own investigations asincreasedriskofincidentdiabetesmellitus.11
of this topic. This article provides in-depth The goal of T therapy is to alleviate symptoms
analysis of studies suggesting increased CV and signs by restoring T concentrations to
risks with T therapy, a historical perspective, optimallevelswithinthephysiologicrange.
and a systematic literature review. Because of EstablishedbenefitsofTtherapyinhypo-
the large number of studies reviewed, much gonadal men include improved sexual desire
of the information is presented in tables, with and function,12-15 improved energy, mood,
textlimitedtosummariesofdata. and vitality,15-19 increased lean mass,14,19-22
There are no large, long-term, placebo- decreased waist circumference,23-27 reduced
controlled randomized clinical trials in the total body fat mass,19-22 and increased bone
fieldofTtherapytoprovidedefinitiveconclu- mineral density.28-31 Promising new data
sions about CV risk. Nonetheless, there exists reveal that T therapy improves insulin sensi-
a rich literature spanning many decades that tivity32-34 and reduces blood glucose23,25,35
provides valuable information. As described and hemoglobin A (HbA )23,25,27,35 levels
1c 1c
in more detail subsequently, the 2 recent arti- inmenwithtype2diabetesorobesity.
cles contradict this literature, and on careful Biochemical confirmation of TD has tradi-
review, neither provides credible evidence of tionally been made on the basis of low serum
increased CV risks. Only 2 additional studies concentrationsoftotalT(TT).Althoughnospe-
are generally cited as support for that view. cific value reliably distinguishes men who will
Incontrast,manydozensofstudies,including respondto treatment fromthose who will not,
a modest number of randomized controlled recommended thresholds for low TT range
trials (RCTs), indicate that low serum T con- from 300 ng/dL (10.4 nmol/L)9 to 400 ng/dL
centrations are associated with increased CV (13.9 nmol/L).36 Because a majority of circu-
risk and mortality and that T therapy may lating T is rendered biologically unavailable
have clinically relevant CV benefits. This last due to tight binding to sex hormoneebinding
point will be new to many readers. A recently globulin (SHBG), the unbound fraction called
published meta-analysis of 75 placebo- freeTand/ortheportionofTweaklyboundto
controlled studies, the largest to date, found albumin may be more indicative of a man’s
no evidence of increased CV risk with T ther- true androgen status.37,38 These 2 fractions
apy and clear evidence of improved metabolic together represent bioavailable T. Men with
profiles.8 Given the personal suffering of men high-normal or elevated SHBG concentrations
withTDaswellasthepublichealthburdenof mayhaveTTconcentrationswithinthenormal
TD, the recent controversy regarding T and range yet may still have TD due to reduced
CV diseasepresents animportantopportunity freeTconcentrations.Thisissuemaybepartic-
tounderstandthescienceunderlyingthiscrit- ularly relevant for older men because SHBG
ical medical issue. levels increase with age.39 Levels below 65 to
100 pg/mL (<174-288 pmol/L) for calculated
BACKGROUND freeTand0.8to1.5ng/dL(27-52pmol/L)for
Testosterone deficiency is a clinical syndrome directlymeasuredfreeThavebeenusedclini-
characterized by a set of signs and symptoms cally to identify men who are candidates for
in combination with low serum T concentra- treatment.10,39,40 However, laboratory-
tions.9,10 Symptoms include decreased libido, provided reference ranges are problematic
erectiledysfunction,difficultyachievingorgasm, because they are not clinically based and
reduced intensity of orgasm, fatigue, decreased vary widely between assays, and even among
energy, depressed mood, irritability, and laboratoriesusingthesameassays.41
decreased sense of well-being. Objective signs The prevalence of symptomatic TD ranges
include anemia, decreased bone density, from 2.1% to 12.8% in middle-aged to older
reduced muscle strength and mass, increased men, with an incidence of 12 new cases per
body fat mass (both visceral and total), and 1000 person-years in the United States and
weightgain.9,10 Androgen deprivationtherapy, Europe.42 Populations at high risk for low
usedinthetreatmentofadvancedprostatecan- serum T levels include men with type 2 dia-
cer,causesprofoundTDandisassociatedwith betes, obesity, chronic obstructive pulmonary
MayoClinProc. n February2015;90(2):224-251 n http://dx.doi.org/10.1016/j.mayocp.2014.10.011 225
www.mayoclinicproceedings.org
MAYO CLINIC PROCEEDINGS
disease, infection with human immunodefi- from34to77yearsofagewithanginapectoris
ciency virus, and chronic opioid use, all with whoreceivedTtherapywithfollow-upranging
prevalenceratesgreaterthan30%.42 fromafewmonthsto5years.Improvementwas
Prescription rates for T products have noted in 91%, with no appreciable improve-
increased substantially over the past mentinpatientstreatedwithplacebo.59
decade.3,43,44Thisincreasehasledtoconcerns Awealthofmoderndataaccumulatedover
that T products are overprescribed because of thepast2decadeshasgenerallyrevealedthata
aggressive marketing. However, as recently as low serum T level is associated with increased
2007, the FDA indicated that as few as 5% risks of atherosclerosis, CV risk factors, and
of the hypogonadal population was treated,45 mortality and that T therapy has beneficial ef-
with other studies reporting similarly low fectsonmultipleriskfactorsandriskbiomarkers
treatment rates.46 The increase in T prescrip- relatedtotheseclinicalconditions.Notably,TD
tions seems to have resulted from increased hasbeenprojectedtobeinvolvedinthedevel-
awarenessofTDandthebenefitsofTtherapy opmentofapproximately1.3millionnewcases
among both physicians and patients, coupled ofCVdisease,1.1millionnewcasesofdiabetes,
with reduced concern regarding prostate can- and over 600,000 osteoporosis-related frac-
cer risk.47 Although it has been asserted that tures.60 Over a 20-year period, TD has been
direct-to-consumer marketing is responsible estimated to be directly responsible for ap-
for this growth, the evidence would suggest proximately $190 to $525 billion in inflation-
otherwise, as 2010 industry data failed to adjusted US health care expenditures.60 In
include any T products within the 25 most- addition,longitudinalmodelspredictincreased
marketed drugs in the United States.48 outpatient visits and costs related to low base-
Whereas it was once believed that TD line serum T levels independent of socioeco-
only occurred in association with several nomic and lifestyle factors; even when
rare disorders, such as pituitary tumors, Kli- controllingforage,menaged20to79yearsat
nefelter syndrome, or mumps orchitis, it is baseline with low serum T levels had 29%
now understoodthatTDiscommonandoften more outpatient visits and 38% higher outpa-
idiopathic. Symptoms and signs result directly tientcostsat5-yearfollow-up.61
from reduced serum concentrations of T, Numerousinterventionstudieshaveconsis-
regardlessofetiology,andcanbeinducedexper- tently found improvements in CV risk factors
imentallyinhealthyvolunteersofallagessimply such as fat mass, obesity, waist circumference,
byreducingserumTconcentrations.20Therela- blood pressure, and glycemic control. These
tivelyhighprevalenceofTDinmiddle-agedand importantfindingsprovideareasonablebiolog-
oldermenhasgivenrisetotheconceptofmale ical mechanism to explain the frequently
menopause,orandropause.Althoughtheseterms observedoutcomeofincreasedmortalityamong
havesomeconceptualappeal,TDdiffersimpor- men with the lowest quartiles or quintiles of
tantlyfrommenopauseinthatmostmenareun- serum T or with frank TD.62-68 Importantly,
affected, and the onset is gradual over an TD in older men is associated with increased
extended time course.49 In addition, the age- riskofdeathduringthe20yearsafterdiagnosis,
related decline in serum T is relatively small,39 independent of multiple traditional risk factors
andmuchofthedeclinecanbeattributedtoco- andseveralpreexistinghealthconditions.63,65
morbidconditionssuchasobesity.39,50,51Atthis Small randomized, placebo-controlled T
time, there are insufficient data to support T trials have documented reduction in carotid
therapyfordiseasepreventionorantiaging. intima-media thickness with T therapy, raising
Testosterone therapy has been in use for the possibility that normalizing serum T may
more than 70 years for the treatment of TD,52 actuallycausereversalofatherosclerosisincrit-
and several of the earliest reports documented ical vascular beds. Moreover, 2 studies pub-
benefits specifically for CV disease. From the lished within the past 2 years, one in a
late 1930s into the 1950s, several studies re- Veterans Administration population69 and the
portedmarkedbenefitsofTtherapyinpatients other in diabetic men,70 found mortality
withperipheralvasculardisease53,54andangina reduced by half in men with TD who received
pectoris.55-59 Lesser59 described 100 consecu- T prescriptions compared with similar men
tive patients (92 men and 8 women) ranging who did not, It was therefore surprising that
226 MayoClinProc. n February2015;90(2):224-251 n http://dx.doi.org/10.1016/j.mayocp.2014.10.011
www.mayoclinicproceedings.org
TESTOSTERONE AND CARDIOVASCULAR RISK
publication of 2 retrospective studies reporting interpretation of relevant studies in those
increased risks of CV adverse events would areas, and level of evidence supporting that
cause such great concern. In one of these conclusion was adjudicated by 2 authors
studies,Vigenetal1reportedincreasedratesof (M.M., A.T.).
MIs, strokes, and deaths in men who received
T prescriptions compared with untreated men; ANALYSIS
thisstudyusedunvalidatedstatisticalmethodol- In contrast to many dozens of studies docu-
ogy that reversed the raw data, which actually menting the beneficial CV effects of T therapy
revealed that the percentage of adverse events in humans, there appear to be only 4 articles
in T-treated men was lower by half compared thatsuggestincreasedCVrisk.These4articles
withuntreatedmen.71Largedataerrorsrevealed werealsoidentifiedbytheFDAanalysis.These
postpublicationledtoacallforretractionbyno articles are the 2 retrospective large dataset
less than 29 medical societies.72 The second analyses of Vigen et al1 and Finkle et al,2 a
study by Finkle et al2 reported an increased meta-analysis by Xu et al,75 and a report of
rate of nonfatal MI up to 90 days after receipt incidental CV adverse events in a placebo-
ofaTprescriptioncomparedwiththeprevious controlled T gel study designed to assess
12months.However,MIratespostprescription muscular and functional benefits in elderly,
werelow,therewasnocontrolgroup,andmeth- frail men with mobility limitations by Basaria
odological issues again rendered the study re- etal.76Althoughfewinnumber,thesestudies
sultshighlyquestionable.73 have garnered considerable attention in the
In response to the publication of these 2 medical literature and lay press, and therefore
studies and their attendant publicity, the merit individual analysis here.
FDA7 announced its intention to review CV
risks with T products 2 days after publication Vigen et al,JAMA2013
ofthestudybyFinkleetal.2Althoughanadvi- Vigenetal1conductedaretrospectiveanalysis
sory committee meeting was scheduled for of men who had undergone coronary angiog-
September 17, 2014, the FDA had already raphy within the Veterans Administration
made public its own analysis of the literature health care system. The authors reported that
in its denial of a petition by the group Public the overall rate of MI, stroke, and death in
Citizen to add black box warnings and other men with serum T levels less than 300 ng/dL
restrictions to T products.74 The FDA’s com- (to convert to nmol/L, multiply by 0.0347)
ments are included where appropriate in the was higher inmen who receivedaT prescrip-
analysis that follows. tion compared with untreated men. Although
no statistically significant differences were
METHODS noted at years 1, 2, or 3, the overall rate of
A MEDLINE search was performed for the events over the course of the study was
years1940toAugust2014usingthefollowing reported to be significantly higher (29%) in
key words: testosterone, androgens, human, T-treated men.
male, cardiovascular, stroke, cerebrovascular ac- Strangely, the actual rate of adverse events
cident,myocardialinfarction,heartattack,death, was only half as great in the T group (123
and mortality. Additional studies were sought events in 1223 men at risk, 10.1%) as in the
by examining publications with their own untreated group (1587 events in 7486 men,
literaturereviews.Tableswerecreatedwithre- 21.2%) (Figure).77 The authors failed to
sultsprovidedbyabstractsorfromthefulltext acknowledgethisfactandcametoanopposite
of the article, depending on the adequacy of interpretation of the data based on complex
abstracted information. statistics that included adjustment for more
A review of the literature was performed than 50 variables. The methodology in this
for 9specifictopicsrelated toTand CVrisks: study, ie, stabilized inverse propensity treat-
mortality; incident CAD; severity of CAD; ment weighting applied to Kaplan-Meier
ischemic stroke; carotid intima-media thick- curves with time as a covariate, was described
ness; obesity/fat mass; lipid profiles; glycemic in an article by senior author Michael P. Ho
control; and inflammatory markers. A sum- just a year earlier (2012) as follows: “Clearly,
mary statement was made regarding the assessing and confirming adequate covariate
MayoClinProc. n February2015;90(2):224-251 n http://dx.doi.org/10.1016/j.mayocp.2014.10.011 227
www.mayoclinicproceedings.org
MAYO CLINIC PROCEEDINGS
found to include nearly 10% women. In
10 25
responsetotheseerrors,29internationalmed-
Untreated group 21.2%
8 T-treated group 20 ical societies and more than 160 physician-
scientists from 32 countries petitioned JAMA
nts (%) 6 15 10.1% tmoismreatnraacgtemthenist aarntidcle,cocnittianmgin“agtiroons”s dthataat
Eve 4 10 rendered the study “no longer credible.”72
2 5 The medical societies urging retraction are
listed in Table 1.
0 0 In summary, this study used an unvali-
Death MI Stroke All events
dated methodology to reverse the results of
FIGURE. Actual percentage of individuals who experienced an adverse raw data, which revealed a lower percentage
cardiovascular event in the testosterone (T)-treated and untreated groups of adverse events in the T-treated group
inthestudybyVigenetal.1Theauthorsreportedahigherrateofadverse compared with untreated men. Putting aside
eventsin theT-treated groupusinginverse stabilized propensity weighting themultipledisturbingdataerrorsthatunder-
inwhichaneventwascountedasmorethan1eventintheT-treatedgroup minetheintegrityofthestudy,thislowerper-
and less than 1 event in the untreated group. MI ¼ myocardial infarction.
centage of adverse CV events in the T-treated
FromJSexMed,77withpermission.ª2014InternationalSocietyforSexual
group is consistent with the results of 2 prior
Medicine.
studiesthatreportedmortalityreducedbyhalf
in T-treated men compared with untreated
men.69,70 Those studies are reviewed in
balance in IPTW time-varying models is chal- further detail in the “Testosterone Prescrip-
lenging and needs further study. Further tions and CV Outcomes” section.
work with simulations and contrasts to other TheFDA’sconcludingcommentregarding
methods and other study applications would Vigen et al1 was as follows: “Given the
help elucidate the advantages and disadvan- described limitations of the study by Vigen
tages of this approach.”78 The authors fail to etalitisdifficulttoattributethereportedfind-
provideanycitationsorotherevidenceintheir ings to testosterone treatment.”74
2013 article by Vigen et al1 that these major
methodological concerns have been resolved Finkle et al,PLoSOne2014
and that the methodology itself is accurate or The study by Finkle et al2 was a retrospective
accepted by other investigators. study of a health insurance database that re-
This article has already undergone 2 offi- ported rates of nonfatal MI in the period up
cial corrections. The first, published January to90daysafteraTprescriptionandcompared
15, 2014, was for misreporting of primary re- these rates to MI rates in the previous 12
sults as “absolute risk,” a term that suggests months. The postprescription period was the
the results were based on raw data.79 In time to first prescription refill, which for an
response to criticism following publication, unspecified number of men would have been
the article was corrected to replace the term 30 days rather than 90 days. The authors re-
absoluteriskwithKaplan-Meierestimatedcumu- ported the rate ratio of MI postprescription
lative percentages with events, a term that more to preprescription was 1.36, and the rate in
accurately reflects the highly statistical nature menolderthan65yearswas2.19.Incompar-
of the results. On March 5, 2014, JAMA pub- ison,noincreaseinMIratewasnotedformen
lished a second correction.80 In response to a who received a prescription for a phosphodi-
letter challenging the exclusion of 1132 men esterase type 5 inhibitor (PDE5i).
who had suffered adverse events in the non- Because the source was an insurance
T group, the authors revealed they had made claims database, available information was
a series of errors. The number of men in this limited to diagnosis codes, procedure codes,
excluded group was changed from 1132 to and prescriptions. There was no information
128 men, a difference of greater than 1000 available regarding several standard CV risk
men.Thevalue forasecondgroupwasfound factors,suchasdiabetes,hypertension,hyper-
to be incorrect by more than 900 men. Most lipidemia,smokinghistory,orobesity,andno
astonishingly, the all-male study group was information concerninganyblood testresults,
228 MayoClinProc. n February2015;90(2):224-251 n http://dx.doi.org/10.1016/j.mayocp.2014.10.011
www.mayoclinicproceedings.org
TESTOSTERONE AND CARDIOVASCULAR RISK
including serum T, or lipid profiles. The
TABLE1. ProfessionalSocietiesCallingforRetractionofVigenetal1,a
weakness of the dataset as an investigative
tool for assessment of CV risk is underscored AmericanSocietyforMen’sHealth(ASMH)
by the fact that the end point in the study, BrazilianSocietyofEndocrinologyandMetabolism
CanadianMen’sSexualHealthCouncil
nonfatal MI, was determined solely by the
CanadianSocietyfortheStudyofMen’sHealth(CSSAM)
use of an insurance diagnosis code without
EuropeanSocietyfortheStudyoftheAgingMale(ESSAM)
verification that MI occurred and without
EuropeanSocietyforSexualMedicine(ESSM)
measures to increase the likelihood of
GermanSocietyforMenandHealth
capturingatrueevent,suchaslimitingevents IndonesianAndrologistAssociation
to the primary diagnosis code at hospital InternationalSocietyofMen’sHealth(ISMH)
discharge.81 An error rate as high as 12% InternationalSocietyforSexualMedicine(ISSM)
has been reported when such measures are InternationalSocietyfortheStudyoftheAgingMale(ISSAM)
not followed.82 IrishAssociationofSexualMedicine
Methodologically, it was inappropriate to ItalianSocietyofAndrology
ItalianSocietyofAndrologyandSexualMedicine
compare posttreatment rates of MI to pretreat-
JapanASEANCouncilforMen’sHealthandAging
mentratesbecausetheseratesmeasuredifferent
things.73 The posteT prescription MI rate JapaneseSocietyforMen’sHealth
KoreanSocietyforSexualMedicineandAndrology
reasonably reflects the naturally occurring MI MalaysianMen’sHealthInitiative
rateinthispopulation,albeitwiththeaforemen-
MalaysianSocietyofAndrologyandtheStudyoftheAgingMale
tioned caveats regarding the accuracy of this Men’sHealthInitiativeofBritishColumbia(Canada)
type of database investigation. However, MexicanAssociationofBoneandMineralMetabolism
because men were included in the study on MiddleEastSocietyforSexualMedicine
thebasisofreal-worldreceiptofaTprescrip- RussianSocietyforMen’sHealth
tion,thepreeTprescriptionMIreflectedhow SouthAsianSocietyforSexualMedicine
often health care providers were willing to SexualMedicineSocietyofNorthAmerica
SociedadeLatinoamericanadeMedicinaSexual(LatinAmericanSocietyforSexual
prescribe T to men with a recent (within 12
Medicine)
months) MI. Any reluctance to prescribe T
SocietyforMen’sHealth,Singapore
to men with recent MI would result in a
SocietyfortheStudyofAndrogenDeficiency
reduced preprescription MI rate. The rates
SocietyfortheStudyofAndrologyandSexology,Singapore(SSASS)
of MI in the preprescription and postpre-
scription periods thus measure different aCurrentlistcanbefoundathttp://www.androgenstudygroup.org/co-signers/list-of-co-signersmedical-
organizations.
things, and the comparison is therefore
meaningless.
Moreover, the reported MI rates posteT lower, or unchanged in association with a T
prescriptionwerealllow,overallandforallsub- prescription.
groups.Forexample,theoverallpostprescription Finally, the comparison with men who
MIratereportedbyFinkleetal2was4.75events receivedaPDE5iprescriptionisuninterpret-
per1000person-years.Thisresultcompareswith able. These were 2 dissimilar groups treated
arateof13expectedMIsper1000person-years with dissimilar medications for dissimilar
using the National Institutes of Health heart indications. Phosphodiesterase type 5 inhib-
attack risk calculator, inputting age 54 years itorsareknowntohavevasodilatoryproper-
(the mean age of the study participants) and ties, and one medication in this class,
favorable risk parameters (nonsmoker; total sildenafil, is approved for treatment of pul-
cholesterol, 230 mg/dL [to convert to mmol/L, monary hypertension, confounding any
multiply by 0.0259]; high-density lipoprotein comparison because of the possibility that
[HDL] cholesterol, 40 mg/dL [to convert to PDE5i’s may have beneficial effects on CV
mmol/L, multiply by 0.0259]; systolic blood risk. This is a classic case of comparing
pressure, 140 mm Hg).83TheobservedMIrate apples and oranges.
among men who received a T prescription was A key concept not addressed by Finkle
thus approximatelyone-thirdtheexpectedrate. et al2 is that TD itself has been identified as
In the absence of a control group of men who a risk factor for CV events (reviewed in the
were untreated, it is impossible to determine “Testosterone and Mortality section). Given
whether these reported MI rates were higher, the short T exposure time of 30 to 90 days,
MayoClinProc. n February2015;90(2):224-251 n http://dx.doi.org/10.1016/j.mayocp.2014.10.011 229
www.mayoclinicproceedings.org
MAYO CLINIC PROCEEDINGS
one unexplored possibility is that any in frail elderly men performed in the United
observed increased risk of MI was due to the Kingdom, there were 2 MACE (1 death, 1
underlying condition rather than from the MI), both occurring in the placebo group.19
treatment. For this reason, we agree with the As Basaria et al76 concluded, “The lack of a
FDAanalysis,whichconcludedthat“itisdiffi- consistentpatternintheseeventsandthesmall
culttoattributetheincreasedriskfornon-fatal numberofoveralleventssuggestthepossibility
MI seen in the Finkle study to testosterone that the differences detected between the two
alone and not consider that the study partici- trial groups may have been due to chance
pants might have remained hypogonadic and alone.” A subsequent analysis suggested that
thus at higher risk for non-fatal MI.”74 eventswereassociatedwithhigherserumcon-
centrationsofTachievedwithtreatmentusing
Basaria et al,NewEnglandJournalof higher than approved doses of T, contrary to
Medicine2010 therecommendationsoftheEndocrineSociety
Basaria et al76 conducted a prospective ran- guidelines.9 It is impossible to conclude from
domized trial designed to investigate whether this study that T prescriptions confer an
Tgelprovidedgreatermuscularandfunctional increased risk of CV events. The FDA
benefits than placebo in a population of frail concluded, “The Basaria study does appear to
elderly men treated for 6 months. The study showanempiricaldose-dependentassociation
did indeed find a benefit of T treatment over between testosterone and cardiovascular risk,
placeboformuscularandfunctionalresponses butitwasnon-conclusivebecauseofthesmall
butwasterminatedearlybecauseoftheobser- sample size and small number of events
vation of increased adverse events categorized reportedinthestudy,aswellasthelimitations
as“cardiovascular”inthetreatmentarm.There withrespecttoconfirmingtheevents.Theau-
were 23 of theseevents in the T arm and 5 in thorsofthisstudyhaveexplicitlyindicatedthat
theplaceboarm. the differences between the groups in cardio-
However, this study was not designed to vascular adverse events might have been due
investigateCVevents,andnoneoftheseevents tochancealone.”74
wereprimaryorsecondaryendpoints.Mostof
the reported “events” were incidentally noted, Xu et al,BMCMedicine2013
subjective, or of questionable clinical impor- Inameta-analysisofCVeventsin27placebo-
tance, such as palpitations, pedal edema, and controlled T studies of 12 weeks’ duration or
premature ventricular contractions. None of longer, Xu et al75 reported that more CV
theseadverseeventsweredefinedbeforestudy events occurred in men who received T
enrollment, and there was no attempt to sys- compared with those who received placebo.
tematically investigate all participants for the This study is the only one of several meta-
presence of any of these events. The most analyses and systematic reviews to suggest
frequently reported adverse event was pedal any increased risk with T therapy. As with
edema, with 5 cases in the T group and none allmeta-analyses,resultsaregreatlyinfluenced
in the placebo group. Given that the placebo bythedefinitionsofendpointsofinterestand
group was composed of more than 100 frail the selection of studies. In this case, the au-
elderly men with multiple comorbidities, it thors specifically included only studies in
seems unlikely thatnone ofthem had any de- which one or more CV events were reported,
gree of pedal edema. This flaw underscores meaning that studies without any CV events
concerns about the interpretation of these wereexcluded.Thisselectionprocessexagger-
data by Basaria et al76 as indicativeof true CV ates the apparent rate of events and distorts
riskratesbecauseitviolatedfoundationalcon- absolute differences in event rates between
cepts in clinical trials, ie, defined end points groups. In addition, just 2 of the 27 studies
andsystematicdataacquisition. contributed35%ofallCVeventsintheTarm.
Fourmajoradversecardiacevents(MACE) The disproportionate influence of these 2
occurred: 1 death, 2 MIs, and 1 stroke, all of studies on the outcome of the meta-analysis
whichoccurredintheTgroup.Althoughtrou- merits closer scrutiny. One is the study by
bling, this asymmetry is not uncommon with Basaria et al76 discussed previously, in which
rare events in clinical trials. In a similar study 18 of 23 events (incorrectly reported as 25
230 MayoClinProc. n February2015;90(2):224-251 n http://dx.doi.org/10.1016/j.mayocp.2014.10.011
www.mayoclinicproceedings.org
TESTOSTERONE AND CARDIOVASCULAR RISK
events by Xu et al75) would not normally Finkleetal2couldeacharguablybedescribed
qualify as CV events. The other was the asdocumentingprotectiveeffectsofTtherapy
1986 Copenhagen study84 in which a nonap- on CV risk because the percentage of events
proved oral formulation of micronized T was was lower by half in the former and overall
administered at a remarkably high dose of MIrateswereonlyafractionofexpectedrates
600 mg/d to men with cirrhosis of the liver, inthelatter.EventsreportedbyBasariaetal76
resultinginserumTconcentrationsexceeding werenotsystematicallyinvestigatedinallindi-
4000 ng/dL (approximately 140 nmol/L) in a viduals, and most events were of questionable
quarter of the T group and levels as high as clinical importance. Finally, the results of
21,000 ng/dL (745 nmol/L), a value approxi- Xuetal75appeartobeduetotheinclusionof
mately20times the upper limit ofthe normal questionable events and studies, and their
range. Because these oral forms of T are conclusions are contradicted by several other
knowntocauselivertoxicityviaafirst-passef- meta-analyses.
fect, it should be no surprise that markedly
supraphysiologic T doses in a hepatically REVIEW OF EXISTING LITERATURE
compromised population would prove harm- Any objective assessment of the literature
ful.Moreover,theauthorsappeartohavecate- regarding T and CV effects must recognize a
gorizedanybleedingeventas“cardiovascular,” broad, rich literature in which numerous
including the most frequently observed cause studies reveal increased CV concerns with TD
of death in this study, bleeding from esopha- andimprovementinavarietyofCVriskfactors
geal varices. The authors listed 21 events as and some CV outcomes with T therapy. That
CV, yet only 1 (MI) could reasonably be literature has been summarized and tabulated
consideredasCV.Thistrialhaslittlerelevance and is included here in Tables 2-9. Summary
to the question as to whether the clinical use statements and levels of supporting evidence
of T therapy increases CV risks. Its inclusion areprovidedinTable10.
and misreporting led to distorted results.
WithoutthecontributionsoftheCopenha- Testosterone Prescriptions and CV
genstudy84andthenon-MACEinthestudyby Outcomes
Basariaetal,76theratesofadverseCVeventsin ThreeadditionalstudiesbeyondthoseofVigen
the Tand placebo groupswere similar,with a etal1andFinkleetal2haveinvestigatedmortal-
slightly lower rate in the T group (78 events ity or MI rates in association with T ther-
in 1599 men [4.88%] vs 60 events in 1174 apy.69,70,156 Shores et al69 studied 1031 men
men[5.1%],respectively).Itshouldbeunder- intheVeteransAdministrationhealthcaresys-
scoredthatthisistheonlyoneofseveralprevi- temwithserumTlevelsoflessthan250ng/dL.
ously published meta-analyses and systematic Themeanagewas62years,andmeanfollow-
reviews to report increased risk with T treat- up was 40.5 months. Mortality in T-treated
ment.85-87 Recently, Corona et al8 reported a men was reduced by half in treated men
meta-analysis of 75 studies, compared with compared with untreated men, at 10.3% vs
the 27 analyzed by Xu et al.75 The meta- 20.7%, respectively (P¼.0001). Multivariate
analysis by Corona et al8 included 3016 men adjustment for age, body mass index, T level,
treated with T and 2446 treated with placebo medical morbidity, diabetes, and coronary
for a mean treatment duration of 34 weeks. heart disease yielded a hazard ratio (HR) of
Theyfoundnosignificantassociationbetween 0.61 (95% CI, 0.42-0.88; P¼.008), indicating
TtherapyandCVevents,eitherassingleevents significant reduction in mortality with T
orascombinedCVendpoints.Studiesinmen therapy.
with metabolic derangements revealed a pro- Muraleedharan et al70 investigated a group
tectiveeffectofTtreatment.88 of 581 diabetic men followed for a mean of
5.8 years. Men with low levels of serum T,
Summary defined as a serum T level of less than 10.4
None of the 4 studies cited as evidence sup- nmol/L (300 ng/dL), had increased mortality
portinganincreasedCVriskwithTadministra- compared with men with T values above this
tion provide compelling evidence of increased threshold.AdjustedmortalityinthelowTgroup
risk. Indeed, the articles by Vigen et al1 and was17.2%comparedwith9.0%inthenormalT
MayoClinProc. n February2015;90(2):224-251 n http://dx.doi.org/10.1016/j.mayocp.2014.10.011 231
www.mayoclinicproceedings.org
2
32 TABLE2. ObservationalStudiesofSerumTestosteroneConcentrationsandMortalitya
Subfractionof
Reference,year testosterone Agerange/mean Meanfollow-
(study) usedforanalysis Samplesize age(y) up(y) Population Majorfinding Remarks
Studiesdocumentinganegative(inverse)associationbetweenendogenoustestosteroneandmortality
Pyeetal,892014(European TT,FT 2599 40-79 4.3 Community-based, Testosteronedeficiencyisassociated HRoflowTT<8nmol/L(irrespectiveofsymptoms)for
MaleAgingStudy) Europe withsubstantiallyhigherrisksofall- all-causemortality,2.3(95%CI,1.2-4.2)
causeandcardiovascularmortality, HRoflowTT<8nmol/LandFT<220pmol/Lfor
towhichboththelevelof all-causemortality,5.5(95%CI,2.7-11.4)
testosteroneandthepresenceof HRwith3sexualsymptoms(irrespectiveofserum
sexualsymptomscontribute testosteronelevel;comparedwithasymptomatic
independently men),3.2(95%CI,1.8-5.8)
Similarrisksobservedforcardiovascularmortality
Haringetal,902013 TT 254 75.5 5and10 Community-based, HigherbaselineTTconcentrations HR(perquartileincrement)ofhighFTforall-cause
(FraminghamHeart UnitedStates wereassociatedwithlower mortality,0.74(95%CI,0.56-0.98)
Study) mortalityriskat5y
Correctionformultiplestatistical
testing(P<.005)renderedthis
M associationstatistically
ay nonsignificant
o
C Repeatedanalysesat10-yfollow-up
lin revealednosignificantassociation
P
ro betweensexsteroids,
c
.n gonadotropinsortheirtrajectories
F andmortality
e
bru Muraleedharanetal,702013 581 5.8 Type2diabetes Lowtestosteronelevelspredictan Mortalityincreasedinthelowtestosteronegroup(17.2%)
ary increaseinall-causemortality comparedwiththenormaltestosteronegroup(9%;
20 duringlong-termfollow-up. P¼.003)whencontrolledforcovariates
1
5;9 Testosteronereplacementmay InCoxregressionmodel,multivariate-adjustedHRfor
0(2 improvesurvivalinhypogonadal decreasedsurvivalwas2.02(95%CI,1.2-3.4;P¼.009).
):22 menwithtype2diabetes TRT(meanduration,41.6(cid:2)20.7mo;n¼64)associated
4
-2 withareducedmortalityof8.4%comparedwith
51 19.2%(P¼.002)intheuntreatedgroup(n¼174)
n
Multivariate-adjustedHRfordecreasedsurvivalinthe
h
ttp untreatedgroupwas2.3(95%CI,1.3-3.9;P¼.004)
://d Hydeetal,912012(Health TT,FT 4249 70-88/77 5.1 Population-based, Lowerfreetestosterone(andhigher HRoflowFTforall-causemortality,1.62(95%CI,1.20- M
x
.d inMenStudy) Australia SHBGandLHlevels)were 2.19)for100vs280pmol/L A
www.mayoclinicproceedingsoi.org/10.1016/j.mayocp.2014.10 Incatphmeasirgussoehotsdroeetciasc-riatsltietpetSeyrdeHodcnBCifiweGVcit(Dhapannraemadldlly-oihcscreiatgteusah,dlseietloryn,wmoLwHenorh-r)CitflaerVelitDey HRfoorfl1o0w0FvTsf2o8r0CpVmDoml/Lortality,1C.7o1nt(i9nu5e%dCoIn,1n.e1x2t-2p.a6g2e) YOCLINICPROCEEDIN
.o.0 G
rg11 S
wM T
ww.mayoclinicproceedings.orgayoClinProc.February2015n STAtuLBedrLiceEhsRbe2adfu.eom(rcCsetunouemcdtneyat,e)iln,yn9e2utaie2nr0dg1a2negaTutTiSsvte,ueedbFsTtf(foroaisncrttveaiorenonranslyeoesf)isassSoacmia2pt0lieo69snizebetwAegeenarageneng(edy/o)mgeeannouMseteaunspt7foo(.7ysll)toewr-onCeoaarnongndPiaoormgpyruaolparthtioyanlity,contAinuc2wmeo5diemt(hnObfriHanetef)aeDdlreMrdveisaeedjfinosicftgrosinerfiiinnfincccdoayairnnlooagtrnflgyaFeraTycsosaohncodiartteodf MMuu(Clltt9iiIvv5,aa%1rr.ii2aaC3tteeI-,--2aa1.5dd.65jjuu0)ss-tt2ee.dd79HH)RRRffeoomrraaCrllkV-scDaumseomrtaolirttya,li1ty.7,72.1(915% ESTOSTERONEANDCA
;90(2 angiography Multivariate-adjustedHRfornon-CVDmortality,2.33 RD
):2 (95%CI,1.45-3.47) IO
24-2 Haringetal,932011 TT 9.9 Kidneydysfunction MeasuredTTlevelsmayhelpdetect HRforall-causemortalityinmenwithkidneydysfunction, VA
51 high-riskindividualsforpotential 1.4(95%CI,1.02-1.92) SC
n therapeuticinterventionsand HRforall-causemortalityinmenwithkidneydysfunction U
h L
ttp://d iamndproouvetcommoertalityriskassessment andlowTT,2.52(95%CI,1.08-5.85) ARR
x.do Kyriazisetal,942011 TT 111 27mo HDpatients TestosteronedeficiencyinmaleHD PatientswithtestosteronedeficiencyhadincreasedCVD ISK
i.o patientsisassociatedwithincreased andall-causemortality,evenafteradjustmentforage,
rg/1 CVDandall-causemortality bodymassindex,serumalbuminandC-reactive
0.1 Increasedarterialstiffnessmaybea proteinlevels,prevalentCVD,andHDduration
0
1 possiblemechanismexplainingthis TestosteronelevelswereinverselyrelatedtoPWV
6
/j.m association (r¼(cid:3)0.441;P<.001)
ay TheassociationoftestosteronewithCVDmortality,but
o
cp notwithall-causemortality,waslostafteradjustingfor
.2
0 PWV,anindexofarterialstiffness
1
4.1 Carreroetal,952011 TT 260 48-67/59 3 ESRD,Sweden Testosteronedeficiencyis ORoflowTTforcardiovascularcomorbiditywas2.51
0
.0 independentlyassociatedwith (95%CI,1.32-4.76)andfordeathwas2.0(95%CI,
1
1
cardiovascularcomorbidityand 1.01-3.97)
deathinlogisticregressionanalyses
Testosteronedeficiencyisacommon
findingamongmalepatientswith
ESRD,anditisindependently
associatedwithinflammation,
cardiovascularcomorbidityand
outcome
Haringetal,652010(Study TT 1954 20-79/58.7 7.2 Population-based, LowTTisassociatedwithincreased HRoflowTTforall-causemortality,1.92(95%CI,1.18-
ofHealthinPomerania) Germany riskofmortalityfromallcausesand 3.14;P<.001)
CVdisease HRoflowTTforCVmortality,2.56(95%CI,1.15-6.52;
P<.05)
Malkinetal,962010 TT,BT 930 Notreported 6.9 CHD(positive LowBTisinverselyassociatedwith HRoflowBTforall-causemortality,2.2(95%CI,1.4-3.6;
angiographic timetoall-causeandvascular P<.0001)
findings) mortality HRoflowBTforvascularmortality,2.2(95%CI,1.2-3.9;
P¼.007)
Continuedonnextpage
2
3
3