REVIEW

Health-Related Lifestyle Factors and Sexual Dysfunction: A

Meta-Analysis of Population-Based Research
Mark S. Allen, PhD, and Emma E. Walter, PhD

ABSTRACT

Background: Sexual dysfunction is a common problem among men and women and is associated with negative
individual functioning, relationship difficulties, and lower quality of life.
Aim: To determine the magnitude of associations between 6 health-related lifestyle factors (cigarette smoking,
alcohol intake, physical activity, diet, caffeine, and cannabis use) and 3 common sexual dysfunctions (erectile
dysfunction, premature ejaculation, and female sexual dysfunction).
Methods: A comprehensive literature search of 10 electronic databases identified 89 studies that met the in-
clusion criteria (452 effect sizes; N ¼ 348,865). Pooled mean effects (for univariate, age-adjusted, and
multivariable-adjusted estimates) were computed using inverse-variance weighted random-effects meta-analysis
and moderation by study and population characteristics were tested using random-effects meta-regression.
Results: Mean effect sizes from 92 separate meta-analyses provided evidence that health-related lifestyle factors
are important for sexual dysfunction. Cigarette smoking (past and current), alcohol intake, and physical activity
had dose-dependent associations with erectile dysfunction. Risk of erectile dysfunction increased with greater
cigarette smoking and decreased with greater physical activity. Alcohol had a curvilinear association such that
moderate intake was associated with a lower risk of erectile dysfunction. Participation in physical activity was
associated with a lower risk of female sexual dysfunction. There was some evidence that a healthy diet was related
to a lower risk of erectile dysfunction and female sexual dysfunction, and caffeine intake was unrelated to erectile
dysfunction. Publication bias appeared minimal and findings were similar for clinical and non-clinical samples.
Clinical Translation: Modification of lifestyle factors would appear to be a useful low-risk approach to decreasing
the risk of erectile dysfunction and female sexual dysfunction.
Strengths and Limitations: Strengths include the testing of age-adjusted and multivariable-adjusted models
and tests of potential moderators using meta-regression. Limitations include low statistical power in models
testing diet, caffeine, and cannabis use as risk factors.
Conclusion: Results provide compelling evidence that cigarette smoking, alcohol, and physical activity are
important for sexual dysfunction. Insufficient research was available to draw conclusions regarding risk factors for
premature ejaculation or for cannabis use as a risk factor. These findings should be of interest to clinicians
treating men and women with complaints relating to symptoms of sexual dysfunction. Allen MS, Walter EE.
Health-Related Lifestyle Factors and Sexual Dysfunction: A Meta-Analysis of Population-Based Research. J
Sex Med 2018;XX:XXXeXXX.
Copyright
2018, International Society for Sexual Medicine. Published by Elsevier Inc. All rights reserved.
Key Words: Alcohol; Caffeine; Cannabis; Cigarette Smoking; Diet; Erectile Dysfunction; Exercise; Female
Sexual Dysfunction; Physical Activity; Premature Ejaculation

INTRODUCTION frequent and persistent problems within normal sexual func-

Improving sexual and reproductive health in poor and high- tioning and is associated with lower quality of life for sufferers
income nations is a leading public health priority of the World and their partners and families.2,3 Normal sexual function has
Health Organization.1 Sexual dysfunction broadly refers to been described as a biopsychosocial process that involves an
interaction of psychological, endocrine, vascular, and neurologic
Received December 17, 2017. Accepted February 11, 2018.
systems,4e7 and these systems are susceptible to disruption from
University of Wollongong, Wollongong, NSW, Australia
health-related lifestyle choices.8e13 The focus of this meta-
analysis is on the relation between health-related lifestyle
Copyright ª 2018, International Society for Sexual Medicine. Published by
Elsevier Inc. All rights reserved. choices (cigarette smoking, alcohol, diet, physical activity,
https://doi.org/10.1016/j.jsxm.2018.02.008 caffeine, and cannabis use) and reported incidence of sexual

J Sex Med 2018;-:1e18 1

2 Allen and Walter

dysfunction. The findings of this research synthesis might be and other common sexual dysfunctions in men and women. In
used to improve prognostic capabilities that could be extremely addition, we explore non-adjusted, age-adjusted, and
valuable to health care professionals treating men and women multivariable-adjusted associations. This is important because
with complaints relating to symptoms of sexual dysfunction. risk factors might change across the lifespan33 and health
The most common sexual complaint in male sexual medicine behaviors are known to be confounded.38,39 For instance, people
is erectile dysfunction.14 Erectile dysfunction is defined as a who exercise more tend to be more social and often report a
consistent inability to obtain or maintain an erection that is greater alcohol intake.39 The connection between alcohol and
sufficient for sexual activity.15 Erectile dysfunction primarily lower risk of erectile dysfunction might be a manifestation of
affects men older than 40 years (with prevalence rates increasing other (positive) health outcomes that accompany a moderate
across the adult lifespan)6 and the worldwide prevalence of alcohol intake. Multivariable-adjusted models provide a better
erectile dysfunction is predicted to reach 322 million cases by indication of the independent contributions of the health
2025.16 Premature ejaculation refers to ejaculation that occurs behavior of interest and can provide important information on
within approximately 1 minute of vaginal penetration, the whether multimodal interventions might be more effective than
inability to delay ejaculation during most vaginal penetrations, unimodal interventions in treating sexual dysfunction. The
and the negative personal consequences including distress.15 present research aimed to investigate these associations through
Premature ejaculation is unrelated to age and affects approxi- quantitative analysis of published data.
mately 20% to 30% of men worldwide.6,17 Female sexual
dysfunction is the name given to the collection of sexual METHODS
difficulties experienced by women. These include a loss of
interest or desire for sexual activity, sexual arousal disorders This research synthesis was prepared in accordance with the
(eg, minimal vaginal lubrication from sexual stimulation; pain Preferred Reporting Items for Systematic Reviews and Meta-
from vaginal penetration), and an absence of feelings of sexual Analyses (PRISMA) statement for the reporting of systematic
arousal (for major categories of female sexual dysfunction, see18). reviews and meta-analyses40 (Supplementary File S1).
The worldwide prevalence of female sexual dysfunction in
premenopausal women is estimated to be 41%.19 Eligibility Criteria
To date, at least 15 narrative (non-systematic) reviews have Studies assessing the relation between sexual dysfunction and
concluded that lifestyle factors, including cigarette smoking, health-related lifestyle factors were included if the following
alcohol, physical inactivity, poor diet, and cannabis use, are criteria were met: (i) the study was published in a peer-reviewed
associated with an increased risk of sexual dysfunction.6,15,20e32 scientific journal; (ii) the sample was representative of the
Research synthesis has been less common but lifestyle factors population under study—representativeness was established
have featured in previous meta-analytic reviews.33e37 A meta- using standard criteria41 and was considered met if recruitment
analysis of alcohol consumption and erectile dysfunction in 11 procedures were outlined and characteristics of the study
cross-sectional studies33 found that regular consumption of population were sufficiently described—and excluded conve-
alcohol was associated with a lower risk of erectile dysfunction nience samples (eg, undergraduate students); (iii) the study
(odds ratio [OR] ¼ 0.79, 99% CI ¼ 0.67e0.92), and a meta- included a measure of sexual dysfunction; (iv) the study
analysis of 7 cross-sectional studies34 found that moderate included a measure of cigarette smoking, alcohol, physical
(k ¼ 4, OR ¼ 0.63, 95% CI ¼ 0.43e0.93) and high (k ¼ 4, activity, diet, caffeine, or cannabis use; (v) the study did not
OR ¼ 0.42, 95% CI ¼ 0.22e0.82) levels of physical activity involve experimental manipulation of independent variables (ie,
were associated with a lower risk of erectile dysfunction. 3 meta- the study was observational in nature); and (vi) the study used
analyses also explored cigarette smoking and erectile dysfunction. human participants.
The 1st, a meta-analysis of 19 studies,35 found that smokers had
an increased risk of erectile dysfunction compared with non- Search Strategy
smokers (relative risk ¼ 12.4%, 95% CI ¼ 10.8e13.9). The A systematic search of 10 electronic databases from 2000 to
2nd, a meta-analysis of 3 to 4 prospective cohort studies,36 found the search date was conducted in April 2017. The databases
that, compared with non-smokers, past smokers (OR ¼ 1.20, searched were PubMed; Science Direct; Scopus; SPORTdiscus;
95% CI ¼ 1.11e1.30) and current smokers (OR ¼ 1.51, 95% Proquest; Web of Science; and psycINFO, psycARTICLES,
CI ¼ 1.34e1.71) had an increased risk of erectile dysfunction. MEDLINE, and CINAHL through EBSCO. The search terms
The 3rd, a dose-response meta-analysis of 10 studies,37 found used were sexual dysfunction [or sexual function*/ or erectile
that the number of cigarettes smoked per day was associated with dysfunction/ or erectile function*/ or erectile/ or erection/ or
an increased risk of erectile dysfunction (OR for 10 cigarettes ¼ impotence/ or impoten*/ or orgasm*/ or sex* drive*] and
1.14, 95% CI ¼ 1.09e1.18). alcohol* [or exercis*/ or physical activity/ or cigarette/ or
The present meta-analysis builds on this foundation of smoking/ or smok*/ or cannabis/ or drug abus*/ or drug use/ or
research evidence and extends the focus to other health behaviors marijuana/ or nutrition/ or caffeine*/ or diet*/ or fruit/ or

J Sex Med 2018;-:1e18

Health-Related Behavior and Sexual Dysfunction 3

Figure 1. Flow diagram for database search and record screening.

vegetable/ or Mediterranean diet/ or risk factor] (Supplementary which multiple studies had used the same dataset, we selected the
File S2). A single researcher screened the titles, key words, and study with the better score on the risk-of-bias assessment. 89
abstracts of each study for eligibility (Supplementary File S3 studies were included in the research synthesis (Supplementary
presents details of search engine hits). If a study appeared to File S4 provides the full reference list for included studies).
meet the eligibility criteria or if the relevance of the study was
uncertain, full texts were obtained. Introduction sections and
reference lists of identified studies were manually searched for Data Extraction and Risk of Bias
further relevant articles by 2 researchers (using a snowball search Data extraction was performed by 2 researchers. Information
strategy). Full texts of all identified studies were independently extracted from each study included the sample size, nation where
assessed for inclusion by 2 researchers. the study was completed, age and sex of participants, effect size
estimates, and information used to assess risk of bias. Risk of bias
(study quality) was assessed using the AXIS tool.41 The scale is
Study Selection designed for non-experimental studies and includes 20 items that
Figure 1 presents the screening procedure. 1,685 records were measure aspects of study quality including justification of sample
identified through electronic databases and another 83 were size, representativeness of the sample, a description of non-
located through manual searches. After title, abstract, and key responders, use of validated measures, description of statistical
word screening, the full texts of 391 studies were assessed for methods, discussion of non-response bias, and reporting of
eligibility. The main reasons for exclusion were experimental funding and conflicts of interest (Supplementary File S5 presents
manipulation of independent variables, incongruent or no the risk-of-bias computation table). Each study was assigned a
measure of health behavior, non-representative sample, and score from 0 to 20 to indicate risk of bias, with higher scores
duplicate data from a sample already included. In instances in reflecting lower risk (higher study quality).

J Sex Med 2018;-:1e18

 4
Table 1. Characteristics of included studies*
Sex Age (y), Sexual Risk of
Study Data N Study sample Nation (% male) mean (SD) dysfunction Health behavior bias

Ahmed et al New 1,139 PHC patients Qatar Male 30 ED Cigarette smoking 16
(2011) (current, non),
alcohol (yes, no),
physical activity
(yes, no)
Ahn et al New 1,570 General Korea Male 40e79 ED Cigarette smoking 18
(2007) population (past, current),
alcohol (yes, no),
caffeine (yes, no)
Akkus et al TEDPSG 1982 General Turkey Male 40 ED Cigarette smoking 14
(2002) population (ever use), alcohol
(yes, no), caffeine
(yes, no), physical
activity (moderate,
active)
Al-Hunayan New 323 Patients with Kuwait Male 41.7 (0.6) ED Cigarette smoking 20
et al (2008) T2DM (yes, no)
Allen and ELSA 6916 General UK Male þ 66.2 (9.1) ED, female Alcohol (past 12 mo), 19
Desille (2017) population female sexual physical activity
(44.7) dysfunction (mild, moderate,
vigorous), diet
(fruit, vegetables
[portions/d])
Al-Naimi et al New 1,052 Hospital Qatar Male 41.9 (14.4) ED Cigarette smoking 16
(2014) outpatients (past, current)
Andersen et al EPISONO 467 General Brazil Male 20e80 ED Physical activity 19
(2010) population (episodes/wk)
Anyfanti et al New 557 Rheumatologic Greece Male þ 54.1 (14.1) ED, female Cigarette smoking 17
(2013) patients female sexual (current, non),
(80.6) dysfunction alcohol
Apostolo et al New 110 Patients with Italy Male 59.2 (8.9) ED Physical activity 16
(2009) heart failure (peak VO2)
Asboe et al New 711 Patients with Europe Male ED Cigarette smoking 17
(2007) HIV (yes, no), alcohol
(yes, no)
J Sex Med 2018;-:1e18

(continued)

Allen and Walter

J Sex Med 2018;-:1e18

Health-Related Behavior and Sexual Dysfunction

Table 1. Continued
Sex Age (y), Sexual Risk of
Study Data N Study sample Nation (% male) mean (SD) dysfunction Health behavior bias
Austoni et al New 16,724 General Italy Male ED Physical activity 16
(2005) population (yes, no), cigarette
smoking (past,
current), alcohol
(drinks/d)
Bacon et al HPFS 22,086 Health USA Male 40e75 ED Physical activity 16
(2006) practitioners (weekly METs),
cigarette smoking
(past, current, non),
alcohol (g/d)
Bener et al New 605 Patients with Qatar Male 35e75 ED Cigarette smoking 14
(2008) stroke (past, current, non)
Blanker et al Krimpen 1,661 General Netherlands Male 50e78 ED, premature Cigarette smoking 19
(2001) study population ejaculation (current, non),
alcohol (U)
Bohm et al ONTARGET 1,357 Patients with Germany þ Male 64.9 (6.4) ED Cigarette smoking 17
(2007) CVD Canada (past, current, non)
Borgquist MPP-RES 830 General Sweden Male 61.8 (8.0) ED Cigarette smoking 16
et al (2008) population (yes, no), alcohol
(yes, no)
Cabral et al New 370 PHC patients Brazil Female 49.8 (8.1) Female sexual Physical activity 19
(2014) dysfunction (active, moderate)
Cassidy et al HPFS 25,096 General USA Male 40e75 ED Diet (flavonoid) 18
(2016) population
Cayan et al New 1,217 General Turkey Female 38.3 (12.7) Female sexual Cigarette smoking, 17
(2016) population dysfunction alcohol (yes/no),
physical activity
(yes/no)
Cheng and Ng New 923 PHC patients Hong Kong Male 26e70 ED Physical activity 18
(2007) (frequency, type,
duration), cigarette
smoking (past,
current, non),
alcohol (yes, no)
(continued)

5
 6
Table 1. Continued
Sex Age (y), Sexual Risk of
Study Data N Study sample Nation (% male) mean (SD) dysfunction Health behavior bias
Chen et al New 2,584 Outpatients with China Male 65.8 (7.7) ED Cigarette smoking 18
(2016) LUTS/BPH (past, current, non),
alcohol (past,
current, non), diet
(vegetables)
Chew et al WAMHS 1,580 General Australia Male 57.9 ED Cigarette smoking 17
(2009a) population (past, current, non)
Chew et al WAMHS 1,580 General Australia Male 57.9 ED Alcohol (past, current, 18
(2009b) population non, frequency)
Cho et al New 3,501 PHC patients Korea Male 40.6 (11.5) ED Cigarette smoking 14
(2003) (pack-years),
alcohol (non to
light, moderate,
heavy), physical
activity (yes, no)
Choi et al New 900 PHC patients Korea Female 30e48 Female sexual Cigarette smoking 16
(2015) dysfunction (past, current,
non), alcohol (past,
current, non)
Corona et al New 755 Outpatients Italy Male Premature Cigarette smoking 14
(2004) with sexual ejaculation (past, current, non),
dysfunction alcohol (drinks/d)
Dombek et al New 111 Patients with Brazil Female 55.9 (4.8) Female sexual Cigarette smoking 20
(2016) postmenopausal dysfunction (past, current, non),
MS alcohol (yes, no)
Elbendary et al New 434/272 Patients with Egypt Male 32.7 (3.0)/33.6 ED Cigarette smoking 16
(2009) ED/matched (5.9) (<100, 100e200,
controls >200), cannabis
Esposito et al CAPRI 595 patients with Italy Female 57.9 (6.9) Female sexual Cigarette smoking 16
(2010) T2DM dysfunction (past, current, non),
physical activity
(METs)
Ettala et al New 1,000 General Finland Male 45e70 ED Cigarette smoking 20
(2014) population (yes, no), alcohol
(frequency), physical
J Sex Med 2018;-:1e18

activity (intensity)

Allen and Walter

(continued)
J Sex Med 2018;-:1e18

Health-Related Behavior and Sexual Dysfunction

Table 1. Continued
Sex Age (y), Sexual Risk of
Study Data N Study sample Nation (% male) mean (SD) dysfunction Health behavior bias
Fedele et al Fedele et al 9,756 Patients with Italy Male 20e69 ED Cigarette smoking 13
(2000) (1998) T1DM or T2DM (past, current,
non), alcohol
(consumption/wk)
Feldman et al MMAS 513 General USA Male 40e70 ED Cigarette smoking 20
(2000) population (current, passive),
alcohol (drinks/d),
physical activity
(moderate to heavy),
diet (saturated and
unsaturated fat)
Furukawa et al DOGO 340 Patients with Japan Male 57 (10) ED Alcohol 17
(2016) T2DM (consumption/wk)
Gades et al OCSUSHS 1,329 General USA Male 40e79 ED Cigarette smoking 16
(2005) population (current, former,
per day)
Ghalayini et al New 905 General Jordan Male 43.6 ED Cigarette smoking 16
(2010) population (current, past, non),
physical activity
(active, inactive)
Giugliano et al New 529 Patients with Italy Female 35e70 Female sexual Diet (Mediterranean) 18
(2010) T2DM dysfunction
Green et al New 2002 PHC patients UK Male 55e70 ED Cigarette smoking 12
(2001) (past, current, total
years), alcohol
Hart et al MACS 1,340 Homosexual USA Male 50 ED Cigarette smoking 18
(2012) adults (cumulative
pack-years)
He et al (2007) InterASIA 4,763 General China Male 35e74 ED Cigarette smoking 17
population (current), alcohol,
physical activity
Huang et al LHID2000 5,763/ Patients with Taiwan Male 55.5/55.7 ED Alcohol (abuse) 18
(2012) 17,289 ED/matched
controls
Ibrahim et al New 509 Gynecologic Egypt Female 39.5 Female sexual Cigarette smoking 17
(2013) patients dysfunction (yes, no)
Jaafarpour et al New 400 General Iran Female 28.2 (2.3) Female sexual Cigarette smoking 20
(2013) population dysfunction (yes, no)
(continued)

7
 8
Table 1. Continued
Sex Age (y), Sexual Risk of
Study Data N Study sample Nation (% male) mean (SD) dysfunction Health behavior bias
Janiszewki et al NHANES 3,941 General USA Male 44.9 (15.8) ED Cigarette smoking 18
(2009) population (past, current, non),
alcohol (non, light,
heavy), physical
activity (inactive,
moderate, active),
diet (fat consumption)
Johnson et al ECA 3,004 General USA Male þ 18e96 Sexual Cannabis 16
(2004) population female dysfunction (lifetime use)
(40%) (any DSM-III)
Junozovic et al New 243 Patients with Bosnia Male 25e65 ED Cigarette smoking 13
(2010) T2Dm (current, past, non)
Kang et al New 1,165 PHC patients Korea Male 54.6 (9.6) ED Cigarette smoking 19
(2016) (past, current, non),
alcohol (drinks/wk),
physical activity
(inactive, minimal,
active)
Klein et al WESDR 365 Patients with USA Male 34.4 (8.4) ED Cigarette smoking 15
(2005) T1DM (current, past, pack
years), alcohol
(history, current)
Korneyev et al New 1,083 PHC patients Russia Male 42.8 (14.1) ED Alcohol (mL/wk) 16
(2016)
Kratzik et al AVMstudy 674 General Austria Male 45e60 ED Cigarette smoking 17
(2009) population (number/d), alcohol
(drinks/d), physical
activity (kcal/wk)
Kupelian et al BACH 2,301 PHC patients USA Male 47.6 ED Cigarette smoking 20
(2008) (pack-years), alcohol
(drinks/d), physical
activity (low,
moderate, high)
Lau et al New 1,178 General China Female 20e40 Female sexual Cigarette smoking 16
(2005) population dysfunction (yes, no), alcohol
J Sex Med 2018;-:1e18

(yes. no)
40

Allen and Walter

Laumann et al MARSH 2,173 General USA Male ED Cigarette smoking 16
(2007) population (packs/d), alcohol
(drinks/d), physical
activity
(continued)
J Sex Med 2018;-:1e18

Health-Related Behavior and Sexual Dysfunction

Table 1. Continued
Sex Age (y), Sexual Risk of
Study Data N Study sample Nation (% male) mean (SD) dysfunction Health behavior bias
Lee et al (2010) MHS 816 General China Male 31e60 ED Cigarette smoking 17
population (current, never,
>20), alcohol
(drinks/d)
Lianjun et al New 1,457 PHC patients China Female 33.4 (8.58) Female sexual Alcohol (yes, no) 18
(2011) dysfunction
Lopez et al NHANES 3,724 General USA Male 20 ED Caffeine (mg/d) 20
(2015) population
Lyngdorf and New 2,210 General Denmark Male 40e79 ED Cigarette smoking 15
Hemmingsen population (past, current, non)
(2004)
Mak et al New 799 General Belgium Male 40e69 ED Cigarette smoking 16
(2002) population (past, current, non),
alcohol (g/d), physical
activity (high, low)
Malavige et al OSDS 510 PHC patients UK Male 56.9 (9.7) ED Physical activity 19
(2015) (frequency), cigarette
smoking (current)
Martin-Morales New 2,476 General Spain Male 25e70 ED Cigarette smoking 16
et al (2001) population (>40/d), alcohol
(abuse)
Millett et al ASHR 8,367 General Australia Male 37.0 ED cigarette smoking 17
(2006) population (past, current, per
day), alcohol
(yes, no, per day)
Mirone et al New 2,010 PHC patients Italy Male 18 ED Cigarette smoking 18
(2002) (past, current, non)
Mofid et al New 700 Patients with Iran Male 20e69 ED Cigarette smoking 16
(2009) T1DM and (current, non)
T2DM
Moriera et al CSED 602 General Brazil Male 51.0 ED Cigarette smoking 17
(2002) population (past, current), alcohol
(yes, no), caffeine
(per week), physical
activity (low, average,
high)
Moriera et al GSSAB 1,200 General Korea Male þ 40e80 Female sexual Cigarette smoking 15
(2006) population female dysfunction, (past, current),
(50%) ED, premature physical activity
ejaculation (low, average, high)
(continued)

9
 10
Table 1. Continued
Sex Age (y), Sexual Risk of
Study Data N Study sample Nation (% male) mean (SD) dysfunction Health behavior bias
Nazarpour et al New 405 Postmenopausal Iran Female 52.8 (3.7) Female sexual Physical activity (type, 20
(2016) adults dysfunction frequency, duration)
Ngai et al New 950 PHC patients Hong Kong Male ED Cigarette smoking 18
(2013) (past, current, per day)
Nicolosi et al CNSEDEIC 2,412 General Brazil, Italy, Male 40e70 ED Tobacco (past, current, 19
(2003) population Japan, non, per day), alcohol
Malaysia (per week), physical
activity (low, average,
high)
Okulate et al New 829 Military Nigeria Male 36.7 (9.9) ED Alcohol (problem use) 17
(2003) personnel
Oksuz and New 2,288 General Turkey Male 28.2 (9.3) Male sexual Diet 16
Malhan population dysfunction (presence, absence)
(2005)
Olugbenga-Bello New 600 General Nigeria Male 33.6 (13.4) ED Cigarette smoking 19
et al (2013) population (past, current, non),
alcohol (past, current,
non, drinks/d)
Oyelade et al New 241 General Nigeria Male 46.7 (13.7) ED Cigarette smoking 14
(2016) population (yes, no), alcohol
(yes, no), physical
activity (yes, no)
Paiva et al New 216 Survivors of Brazil Female 51.9 (9.2) Female sexual Physical activity (mild or 19
(2016) breast cancer dysfunction moderate, none)
Polsky et al New 101/234 ED/matched Canada Male 63.9 (6.4)/64.0 ED Cigarette smoking (past, 18
(2005) controls (7.0) current, non), alcohol
(drinks/wk)
Ponholzer et al New 2,869 PHC patients Austria Male 45.8 (12.5) ED Physical activity 17
(2005) (<1 or >1/wk)
Ramirez et al New 440 Patients with Spain Male 53.7 (9.6) ED Cigarette smoking 20
(2016) lipid and (past, current, non),
vascular risks alcohol (U/d), physical
activity (hours of
exercise), diet
(Mediterranean diet)
J Sex Med 2018;-:1e18

Ricci et al New 2010 PHC patients Italy Male 18 ED Cigarette smoking 18

Allen and Walter

(2003) (past, current, non)
(continued)
J Sex Med 2018;-:1e18

Health-Related Behavior and Sexual Dysfunction

Table 1. Continued
Sex Age (y), Sexual Risk of
Study Data N Study sample Nation (% male) mean (SD) dysfunction Health behavior bias
Rosen et al LookAHEAD 373 Patients with USA Male 45e75 ED cigarette smoking 17
(2009) T2DM (past, current, non),
physical activity
(METs)
Safarinejad New 2,444 General population Iran Male 41.1 ED Cigarette smoking 17
(2003) (past, current, non)
Shaeer et al New 1,814 PHC patients Egypt, Male 25e70 ED Tobacco use (ever, 17
(2003) Nigeria, current), alcohol,
Pakistan physical activity
(inactive, moderate,
active), caffeine
(cups of coffee/d)
Shiri et al TAMUS 1,130 General population Finland Male 50e70 ED Cigarette smoking 17
(2004) (past, current, non),
alcohol (g/wk), caffeine
(cups of coffee/d)
Shiri et al New 312 Patients with Iran Male 55.2 (13.4) ED Tobacco smoking 16
(2006) T1DM and (past, current, non),
T2DM diet (fruit intake)
Simon et al New 295 Urologic patients USA Male 62 ED Cigarette smoking 17
(2015) (past, current, non),
physical activity
(MET hours/wk)
Teles et al New 3,548 PHC patients Portugal Male 40e69 ED Physical activity (never, 18
(2008) some, regular)
Vallejo-Medina New 356 General population Spain Male ED, premature Alcohol (severity 20
and Sierra ejaculation of use), cannabis
(2013) (severity of use)
Wang et al SLCDC 1,466 Patients with DM Canada Male 65.1 ED Cigarette smoking 18
(2013) (past, current, non),
alcohol (yes, no),
physical activity
(active, moderate,
inactive), diet (fruit þ
vegetable
consumption/d)
(continued)

11
 12
Table 1. Continued
Sex Age (y), Sexual Risk of
Study Data N Study sample Nation (% male) mean (SD) dysfunction Health behavior bias
Weber et al 45up 108,477 General population Australia Male 45 ED Cigarette smoking 20
(2013) (past, current, per
day), alcohol (drinks/
wk), physical activity
(sessions in past
week)
Wu et al (2012) FAMHES 2,686 PHC patients China Male 20e79 ED Cigarette smoking 20
(past, current, non),
alcohol (yes, no),
physical activity
(active, inactive)
Zedan et al New 658/821 Patients with ED/ Egypt Male 46.3 (23.5)/44.5 ED Cigarette smoking 16
(2010) matched (20.2) (current, non)
controls
Zhang et al New 25,446 General population China Female 20e70 Female sexual Alcohol (yes, no), 19
(2017) dysfunction cigarette smoking
(yes, no)
45up = 45 and Up Study; ASHR = Australian Study of Health and Relationships; AVMstudy = Androx Vienna Municipality study; BACH = Boston Area Community Health; BPH = benign prostatic hyperplasia;
CAPRI = Campanian post-PrandIal hyperglycemia group; CNSEDEIC = Cross National Study on the Epidemiology of Erectile Dysfunction and its Correlates; CSED = Cross-National Study of Erectile
Dysfunction; CVD = cardiovascular disease; DOGO = Dogo study; DSM-III = Diagnostic and Statistical Manual of Mental Disorders, 3rd Edition; ECA = Epidemiological Catchment Area Project; ED = erectile
dysfunction; ELSA = English Longitudinal Study of Ageing; EPISONO = The Epidemiologic Sleep Study; FAMHES = Fangchenggang Area Male Health and Examination Survey; GSSAB = Global Study of Sexual
Attitudes and Behaviors; HPFS = Health Professionals Follow-Up Study; InterASIA = International Collaborative Study of Cardiovascular Disease in Asia; Krimpen = Krimpen Study; LHID2000 = Taiwan
Longitudinal Health Insurance Database 2000; LookAHEAD = Action for Health in Diabetes study; LUTS = lower urinary tract symptoms; MACS = Multicenter AIDS Cohort Study; MARSH = Male Attitudes
Regarding Sexual Health study; MET = metabolic equivalent; MHS = Hong Kong Men's Health Survey; MMAS = Massachusetts Male Aging Study; MPP-RES = Malmo Primary Prevention study;
MS = multiple sclerosis; NHANES = National Health and Nutrition Examination Survey; OCSUSHS = Olmsted County Study of Urinary Symptoms and Health Status; ONTARGET = Ongoing Telmisartan Alone
and in Combination with Ramipril Global Endpoint Trial; OSDS = Oxford Sexual Dysfunction Study; PHC = primary health care; SLCDC = Survey on Living with Chronic Diseases in Canada; T1DM = type 1
diabetes mellitus; T2DM = type 2 diabetes mellitus; TAMUS = Tampere Aging Male Urological Study; TEDPSG = Turkish Erectile Dysfunction Prevelance Study Group; VO2 = maximal oxygen consumption;
WAMHS = Western Australia Men's Health Study; WESDR = Wisconsin Epidemiologic Study of Diabetic Retinopathy.
*Please see Supplementary File S4 for complete reference list.
J Sex Med 2018;-:1e18

Allen and Walter

Health-Related Behavior and Sexual Dysfunction 13

Table 2. Random-effects models for health-related lifestyle factors and risk of sexual dysfunction*
Multivariable-adjusted
Non-adjusted estimates Age-adjusted estimates estimates

k OR (95% CI) k OR (95% CI) k OR (95% CI)

Cigarette smoking
Female sexual dysfunction 8 1.39 (0.92e2.10) — — 5 1.06 (0.51e2.21)
Past smoker 1 1.01 (0.48e2.12) — — 2 1.04 (0.53e2.03)
Current smoker 6 1.50 (0.92e2.44) — — 5 1.13 (0.91e1.41)
Premature ejaculation 3 1.30 (0.83e2.03) — — — —
Erectile dysfunction 43 1.41 (1.26e1.58) 20 1.30 (1.18e1.44) 31 1.43 (1.28e1.59)
Current smoker 40 1.34 (1.21e1.50) 17 1.41 (1.27e1.56) 28 1.47 (1.29e1.68)
Moderate 4 1.06 (0.89e1.26) 4 1.12 (0.88e1.42) 9 1.26 (1.10e1.44)
High 4 1.74 (1.28e2.35) 4 1.67 (1.18e2.36) 9 1.53 (1.31e1.80)
Past smoker 24 1.56 (1.30e1.89) 18 1.24 (1.11e1.39) 17 1.27 (1.15e1.41)
Moderate 2 1.87 (0.93e3.74) 3 1.13 (0.98e1.32) 6 1.17 (1.13e1.21)
High 2 2.53 (1.55e4.12) 3 1.54 (1.27e1.86) 6 1.64 (1.52e1.76)
General population 24 1.43 (1.25e1.64) 11 1.36 (1.11e1.66) 16 1.58 (1.35e1.85)
Clinical population 19 1.39 (1.14e1.69) 9 1.32 (1.21e1.44) 15 1.27 (1.09e1.47)
Alcohol intake
Female sexual dysfunction 5 1.05 (0.57e1.56) — — 5 0.76 (0.49e1.20)
Premature ejaculation 3 1.67 (0.53e5.22) — — — —
Erectile dysfunction 30 0.78 (0.68e0.89) 14 0.83 (0.70e0.98) 17 0.93 (0.82e1.05)
Alcohol drinker vs non-drinker 12 0.71 (0.53e0.95) 3 0.65 (0.33e1.34) 4 1.01 (0.93e1.10)
Moderate alcohol intake 11 0.77 (0.61e0.99) 9 0.84 (0.74e0.96) 10 0.77 (0.62e0.96)
High alcohol intake 15 1.19 (0.96e1.46) 10 1.21 (0.97e1.50) 10 1.02 (0.83e1.25)
General population 20 0.77 (0.63e0.95) 7 0.91 (0.59e1.41) 8 0.87 (0.69e1.11)
Clinical population 13 0.96 (0.79e1.15) 8 0.86 (0.71e1.04) 9 0.98 (0.86e1.12)
Physical activity
Female sexual dysfunction 6 0.67 (0.56e0.81) — — 1 0.91 (0.77e0.98)
Premature ejaculation 1 0.52 (0.28e0.97) — — — —
Erectile dysfunction 20 0.57 (0.48e0.68) 10 0.54 (0.44e0.67) 18 0.68 (0.58e0.79)
Exerciser vs non-exerciser 6 0.55 (0.37e0.81) 3 0.64 (0.44e0.93) 5 0.83 (0.74e0.93)
Moderate physical activity 7 0.71 (0.58e0.89) 4 0.45 (0.32e0.64) 8 0.58 (0.52e0.65)
High physical activity 7 0.45 (0.32e0.64) 4 0.32 (0.17e0.61) 8 0.56 (0.45e0.69)
General population 11 0.51 (0.39e0.67) 4 0.39 (0.26e0.59) 11 0.64 (0.49e0.82)
Clinical population 9 0.65 (0.52e0.80) 6 0.66 (0.53e0.81) 7 0.78 (0.68e0.90)
Diet
Female sexual dysfunction 2 0.83 (0.74e0.93) — — — —
Erectile dysfunction 7 0.86 (0.77e0.96) — — 4 0.93 (0.81e1.06)
Fruit ± vegetable intake 3 0.90 (0.73e1.11) — — 2 0.57 (0.21e1.52)
Fat consumption 1 0.75 (0.61e0.92) — — 1 1.05 (0.95e1.16)
Mediterranean 1 0.73 (0.48e1.14) — — — —
General population 4 0.87 (0.75e1.01) — — 2 1.00 (0.89e1.10)
Clinical population 3 0.72 (0.49e1.06) — — 2 0.57 (0.21e1.52)
Caffeine
Erectile dysfunction 2 0.52 (0.31e0.87) 5 1.08 (0.61e1.91) 2 1.24 (0.81e1.90)
Coffee drinker (yes vs no) 2 0.52 (0.31e0.87) 1 0.54 (0.29e1.00) — —
Frequency of caffeine — — 2 2.00 (1.43e2.79) 1 1.54 (1.01e2.36)
(per day or week)
Moderate intake — — 3 1.03 (0.51e2.08) 1 0.90 (0.59e1.38)
High intake — — 3 1.16 (0.56e2.41) 1 1.10 (0.71e1.71)
Cannabis
Mean of outcomes 3 1.81 (1.31e2.49) — — — —
Erectile dysfunction 2 1.75 (0.93e3.29) — — — —
(continued)

J Sex Med 2018;-:1e18

14 Allen and Walter

Table 2. Continued
Multivariable-adjusted
Non-adjusted estimates Age-adjusted estimates estimates

k OR (95% CI) k OR (95% CI) k OR (95% CI)

Orgasm dysfunction 2 2.09 (1.36e3.20) — — — —
Premature ejaculation 1 1.76 (1.48e5.17) — — — —
OR ¼ odds ratio.
*All findings are for the mean of outcomes associated with the outcome variable. Past and current smoking estimates are compared with never-smokers. All
premature ejaculation estimates are for non-clinical samples. Problem alcohol use is not included in estimates for overall erectile dysfunction. Problem alcohol
use is included in grouping of high levels of alcohol use. High and moderate alcohol intakes are compared with a referent group of no or very low levels of
alcohol use. Moderate and high levels of physical activity are contrasted with no or very low physical activity levels. Moderate and high levels of caffeine intake
are compared with no caffeine intake. For diet, the erectile dysfunction estimate includes the study by Oksuz and Malhan (2005) on male sexual dysfunction.
For cannabis use, the orgasm dysfunction estimate combines the premature ejaculation measure in Vallejo-Medina and Sierra (2013) with the inhibited
orgasm measure (assessed in men and women) in Johnson et al (2004).

Analytic Strategy RESULTS

Calculation of the pooled mean effect size (OR and 95% CI)
The characteristics of included studies are presented in
was computed using inverse-variance weighted random-effects
Table 1. The 89 studies provided 452 effect sizes and a total
meta-analysis. The inverse-variance method, in which each
sample size of 348,865 participants. The participants were from
included effect size is given a weight equal to the inverse of its
29 nations with samples representing Asia (n ¼ 30), Europe
variance, allows more weight to be given to more precise
(n ¼ 26), North America (n ¼ 15), Africa (n ¼ 6), South
studies.42 In cases in which effect size estimates were reported for
America (n ¼ 5), Australia (n ¼ 5), and multicontinental
separate components of female sexual dysfunction (eg, lubrica-
(n ¼ 3). The mean age was 48.8 years (SD ¼ 10.7 years) and
tion problems, pain, difficulty researching orgasm), effect sizes
risk-of-bias scores ranged from 12 to 20 (mean ¼ 17.2,
were averaged within studies, resulting in 1 effect size for each
SD ¼ 1.8). Table 2 presents findings from 92 separate meta-
study. Non-adjusted, age-adjusted, and multivariable-adjusted
analyses. All meta-analyses with a sample size (k) larger than
estimates (usually adjusted for age, sociodemographic factors,
10 showed significant heterogeneity (Q statistic), supporting the
and other health behaviors) were tested in separate meta-analyses
use of meta-regression to search for potential moderators.
to avoid confounding the pooled mean effect. We estimated the
Individual study (raw) effect size estimates are available in
magnitude of heterogeneity across studies using the I2 and
Supplementary File S6 and effect size estimates expressed as ORs
Q statistics.43 A statistically significant Q statistic indicates
are reported in forest plots in Supplementary File S7. A more
meaningful heterogeneity. Visual inspection of funnel plots and
detailed synthesis of findings can be found in Supplementary
rank correlation tests44 were used to search for evidence of
File S8.
publication bias.
The potential impact of publication bias was assessed using
the trim-and-fill procedure.45 To test for the impact of Cigarette Smoking
moderating variables, we used a protocol for random-effects 69 studies provided 223 usable effect sizes for cigarette
meta-regression42 in which the correlation between the health smoking. Female sexual dysfunction was unrelated to cigarette
behavior and the component of sexual dysfunction was set as smoking for non-adjusted (k ¼ 8, OR ¼ 1.39, 95% CI ¼
the criterion variable and the moderating variable was set as the 0.92e2.10) and multivariable-adjusted (k ¼ 5, OR ¼ 1.06,
predictor, with studies being weighted by their inverse-variance 95% CI ¼ 0.51e2.21) estimates. Premature ejaculation also was
weights. Moderator effects were explored for world region unrelated to cigarette smoking (k ¼ 3, OR ¼ 1.30, 95% CI ¼
(continent), clinical sample, risk of bias, and level of exposure 0.83e2.03). Current cigarette smoking was associated with an
(dose and response). Risk of bias was set as an integer variable increased risk of erectile dysfunction for non-adjusted (k ¼ 40,
and other moderators were set as categorical variables. OR ¼ 1.34, 95% CI ¼ 1.21e1.50), age-adjusted (k ¼ 17,
Regression models were tested using maximum likelihood OR ¼ 1.41, 95% CI ¼ 1.27e1.56), and multivariable-adjusted
estimation. 3 moderator terms (world region, clinical sample, (k ¼ 28, OR ¼ 1.47, 95% CI ¼ 1.29e1.68) estimates. There
and risk of bias) had no missing values and were explored in was no evidence of publication bias for age-adjusted estimates,
combination (forced-entry multiple meta-regression). Level of but the trim-and-fill procedure indicated that 7 effects be trim-
exposure was explored in a separate regression model owing to med for non-adjusted (pseudo OR ¼ 1.23, 95% CI ¼
missing values on most studies. Publication bias and moderator 1.11e1.36) and multivariable-adjusted (pseudo OR ¼ 1.29,
effects were explored for associations that included at least 10 95% CI ¼ 1.13e1.48) estimates, resulting in smaller pooled
effect sizes.42 mean effects. Past cigarette smoking also was associated with an

J Sex Med 2018;-:1e18

Health-Related Behavior and Sexual Dysfunction 15

increased risk of erectile dysfunction for non-adjusted (k ¼ 24, For studies reporting level of exposure, the non-adjusted (k ¼
OR ¼ 1.56, 95% CI ¼ 1.30e1.89), age-adjusted (k ¼ 18, 26, c21 ¼ 4.78, P ¼ .029, R2 ¼ 0.19) and age-adjusted (k ¼ 20,
OR ¼ 1.24, 95% CI ¼ 1.11e1.39), and multivariable-adjusted c21 ¼ 9.53, P ¼ .002, R2 ¼ 0.51) meta-regression models were
(k ¼ 17, OR ¼ 1.27, 95% CI ¼ 1.15e1.41) estimates. For significant. The models showed that a moderate alcohol intake
multivariable-adjusted estimates, the trim-and-fill procedure (mode ¼ 1e3 drinks/day; mean z 7.8 drinks/week), but not a
indicated that 6 effects be trimmed, resulting in a smaller pooled high alcohol intake (mode > 3 drinks/day; mean z 23.1 drinks/
mean effect (pseudo OR ¼ 1.18, 95% CI ¼ 1.06e1.30). week), was associated with a lower risk of erectile dysfunction
For current cigarette smoking on erectile dysfunction, there (Table 2). There were no significant multivariate effects for the 3
was a significant dose-response effect (see Supplementary File S9 moderator terms on alcohol intake.
for coding). For multivariable-adjusted estimates (k ¼ 18,
c21 ¼ 4.65, P ¼ .031, R2 ¼ 0.28), high levels of current cigarette Physical Activity
smoking (mode > 20/day) were associated with a greater risk of 38 studies provided 79 usable effect sizes for physical activity.
erectile dysfunction than low levels of current cigarette smoking Involvement in physical activity was associated with a lower risk
(mean z 8.6/day; b ¼ 0.19, 95% CI ¼ 0.36 to 0.02). A of female sexual dysfunction (k ¼ 6, OR ¼ 0.67, 95% CI ¼
dose-response effect also emerged for past cigarette smoking on 0.56e0.81), with no evidence of publication bias. Physical ac-
erectile dysfunction. For multivariable-adjusted estimates tivity also was associated with a lower risk of erectile dysfunction
(k ¼ 12, c21 ¼ 66.43, P < .001, R2 ¼ 1.00), high levels of past for non-adjusted (k ¼ 20, OR ¼ 0.57, 95% CI ¼ 0.48e0.68),
cigarette smoking (mean z 23 pack-years; <5 years quit) were age-adjusted (k ¼ 10, OR ¼ 0.54, 95% CI ¼ 0.44e0.67), and
associated with greater risk of erectile dysfunction than low levels multivariable-adjusted (k ¼ 18, OR ¼ 0.68, 95% CI ¼
of past cigarette smoking (mean z 10 pack-years; >10 years 0.58e0.79) estimates, with no evidence of publication bias. For
quit; b ¼ 0.33, 95% CI ¼ 0.41 to 0.25; Table 2). studies reporting level of exposure, the meta-regression model
There was no significant multivariate effect for the 3 moder- was significant for non-adjusted estimates (k ¼ 14, c21 ¼ 4.44,
ator terms (continent, clinical sample, risk of bias) on current or P ¼ .035, R2 ¼ 0.28). Observation of the regression slope
past cigarette smoking for non-adjusted or age-adjusted showed that high levels of physical activity (eg, 30 minutes of
estimates. For multivariable-adjusted estimates, there was a sig- moderate- or high-intensity activity 5 times per week) were
nificant multivariate effect for the 3 moderator terms on current associated with a lower risk of erectile dysfunction compared
cigarette smoking (k ¼ 28, c27 ¼ 23.52, P ¼ .001, R2 ¼ 0.75), with moderate levels of physical activity (eg, 20 minutes of
with a significant regression coefficient for world region (c27 ¼ moderate-intensity activity 3 times per week; b ¼ 0.42, 95%
14.47, P ¼ .013). Observation of dummy-coded variables CI ¼ 0.03e0.82). The meta-regression was non-significant for
showed that estimates were higher for Africa-based studies (k ¼ age-adjusted and multivariable-adjusted models. There was some
2, OR ¼ 2.42, 95% CI ¼ 1.41e4.15) than for other continents evidence of moderation by risk of bias, clinical sample, and
(all pairwise comparisons significant at P < .01) except for North continent (Supplementary File S8).
America. There also was a significant multivariate effect for past
cigarette smoking (k ¼ 17, c26 ¼ 35.11, P < .001, R2 ¼ 1.00),
with a significant regression coefficient for world region (c24 ¼ Diet, Caffeine, and Cannabis
18.90, P < .001). Observation of dummy-coded variables Less research has explored diet (10 studies, 17 effect sizes),
showed that estimates were lower for North American samples caffeine (6 studies, 14 effect sizes), and cannabis use (3 studies, 5
than for Australian (b ¼ 0.18, 95% CI ¼ 0.07e0.29, P ¼ .001) effect sizes). There was some evidence that a healthier diet (eg,
and Asian (b ¼ 0.30, 95% CI ¼ 0.01e0.59, P ¼ .043) samples. higher fruit and vegetable intake) related to a lower risk of female
sexual dysfunction (k ¼ 2, OR ¼ 0.83, 95% CI ¼ 0.74e0.93)
and erectile dysfunction (k ¼ 7, OR ¼ 0.86, 95% CI ¼
Alcohol Intake 0.77e0.96). For erectile dysfunction, we also explored publica-
52 studies provided 114 usable effect sizes for alcohol intake. tion bias (despite only 7 effect sizes in the model). There was a
Female sexual dysfunction and premature ejaculation were un- significant rank correlation test result (s ¼ 0.62, P ¼ .050),
related to alcohol intake (Table 2). Alcohol intake was associated with the trim-and-fill procedure indicating that 3 effects be
with a decreased risk of erectile dysfunction for non-adjusted trimmed, attenuating the significant effect (k ¼ 7, pseudo OR ¼
(k ¼ 30, OR ¼ 0.78, 95% CI ¼ 0.68e0.89) and 0.91, 95% CI ¼ 0.81e1.02). This finding indicates that the
age-adjusted (k ¼ 14, OR ¼ 0.83, 95% CI ¼ 0.70e0.98) association between diet and erectile dysfunction might be an
estimates, but not for multivariable-adjusted estimates (k ¼ 17, artifact of publication bias. Most studies reported age-adjusted
OR ¼ 0.93, 95% CI ¼ 0.82e1.05). There was no evidence of estimates for caffeine, with findings showing no association
publication bias for age-adjusted and multivariable-adjusted (Table 2). Cannabis use has rarely been explored, but combined
estimates, but the trim-and-fill procedure indicated that 8 effects from 3 studies indicated that cannabis use could be a risk
effects be trimmed for non-adjusted estimates (pseudo OR ¼ factor for sexual dysfunction (k ¼ 3, OR ¼ 1.81, 95%
0.91, 95% CI ¼ 0.78e1.05), attenuating the significant effect. CI ¼ 1.31e2.49).

J Sex Med 2018;-:1e18

16 Allen and Walter

DISCUSSION dysfunction was dose dependent, with high levels of physical

activity having a lower risk compared with moderate levels of
The findings of this research synthesis confirm that health-
physical activity. The connection between physical activity and
related lifestyle factors relate to risk of sexual dysfunction.
sexual function might be explained by the increase in testos-
Cigarette smoking (past and current), alcohol intake, and
terone that accompanies physical activity participation (and
physical activity had dose-dependent associations with erectile
competitive sport in particular)48,49 or other pathophysiologic
dysfunction. Risk of erectile dysfunction increased with greater
effects associated with obesity that often accompanies low
cigarette smoking and decreased with greater physical activity,
physical activity.50,51 These sexual dysfunctions also have a
whereas alcohol intake had a curvilinear association such that
strong connection to diabetes,52,53 which could be an artifact of
moderate intake was associated with a lower risk of erectile
low physical activity, given the observed connection between
dysfunction. Participation in physical activity was associated with
physical inactivity and onset of diabetes.9 Further research should
a lower risk of female sexual dysfunction. There was some evi-
explore the potential mediating role of testosterone secretion in
dence that a healthy diet related to a lower risk of sexual
the association between physical activity and sexual function and
dysfunction, whereas caffeine intake was unrelated to sexual
the role of physical inactivity in the association between sexual
dysfunction. Publication bias appeared minimal and findings
function and diabetes.
were similar for clinical and non-clinical samples.
Less robust conclusions can be formulated for diet, caffeine,
The finding that past and current cigarette smoking had dose-
and cannabis use, but the available evidence indicated that a
response associations with erectile dysfunction is consistent with
healthy diet was associated with a lower risk of erectile
findings from a previous meta-analysis.37 The moderation by
dysfunction in men and female sexual dysfunction in women.
world region (in multivariable-adjusted models, controlling for
Changes in diet have been found to accompany changes in
study quality) is an intriguing result and might reflect differences
testosterone,54 which might explain the connection between a
in tobacco policies or norms in various world regions (eg, nico-
healthier diet and better sexual function. For caffeine, most
tine content of popular local brands). However, the number of
studies reported age-adjusted estimates with no consistent
pooled studies within regions was small and we cannot discount
pattern of results. Caffeine is one of the most widely consumed
the possibility that the moderation effect is an artifact of low
psychoactive compounds and is often marketed for its physical
statistical power in the meta-regression model and encourage
and cognitive performance benefits.55 It would appear that
further research that explores regional differences. Less robust
caffeine intake has no benefit to sexual function, but there was a
conclusions can be formulated for other types of sexual
small number of pooled mean effects, meaning further research is
dysfunction given the small number of pooled effects, but the
required. Like alcohol, cannabis is a central nervous system
available evidence indicates that cigarette smoking is unimpor-
depressant11 that has harmful physical and cognitive effects
tant for female sexual dysfunction and premature ejaculation.
associated with long-term use.11 The potential connection be-
The findings also extend those of a previous meta-analysis that tween cannabis use and sexual dysfunction also requires further
found alcohol intake was associated with a lower risk of erectile investigation, but the available studies are suggestive of cannabis
dysfunction.33 Our findings indicate a curvilinear association use as a potential risk factor.
such that moderate alcohol intake was associated with a lower
The present meta-analysis has some important limitations that
risk, whereas high alcohol intake was associated with no risk
readers must consider when interpreting study findings. (i) Few
change. That the positive effect for moderate alcohol intake
studies explored risk factors for female sexual dysfunction and
remained significant in multivariable-adjusted models suggests
premature ejaculation, and the corresponding low statistical
that alcohol itself might be beneficial for erectile dysfunction,
power limits the conclusions that can be drawn for these sexual
rather than alcohol being confounded with other (positive)
dysfunctions. (ii) Studies often use self-report assessments that
health behaviors such as physical activity. Although alcohol is a
are open to response distortion in the form of social desirability
known central nervous system depressant,46 moderate alcohol
bias and this might have attenuated some effects. (iii) The meta-
intake also can increase sexual desire and lessen sexual anxiety,47
analysis targeted non-experimental research and therefore does
which could benefit erectile function. Similar to findings for
not provide information on cause and effect. Experimental
cigarette smoking, less robust conclusions can be formulated for
studies have started to provide evidence that modification of
alcohol and other sexual dysfunctions, but the available evidence
lifestyle factors can benefit sexual function,54,56e58 and our
indicates that alcohol is unimportant for female sexual dysfunc-
finding that cigarette smoking and physical activity remained
tion and premature ejaculation.
significant risk factors in multivariable-adjusted models indicates
Results also support a previous meta-analysis that found higher that multimodal interventions (targeting multiple lifestyle
physical activity was associated with a lower risk of erectile factors) might be more effective in treating symptoms than
dysfunction.34 Given the observed effect sizes, a physically unimodal interventions. We recommend future research test the
inactive lifestyle was the most important risk factor for erectile effectiveness of multimodal interventions against those that target
dysfunction and female sexual dysfunction. The effect for erectile a single lifestyle change.

J Sex Med 2018;-:1e18

Health-Related Behavior and Sexual Dysfunction 17

CONCLUSION 5. Latif EZ, Diamond MP. Arriving at the diagnosis of female

sexual dysfunction. Fertil Steril 2013;100:898-904.
Health care professionals have often noted that “lifestyle changes
6. Shamloul R, Ghanem H. Erectile dysfunction. Lancet 2013;
and risk factor modification must precede or accompany ED
381:153-165.
[erectile dysfunction] treatment.”14 Before clinicians recommend
more invasive treatment options (oral pharmacotherapy, vacuum 7. Berman JR. Physiology of female sexual function and
dysfunction. Int J Impot Res 2005;17:S44-S51.
constriction devices, testosterone replacement), a 1st-line low-risk
option might be to incorporate lifestyle changes. That effect sizes 8. Biswas A, Oh PI, Faulkner GE, et al. Sedentary time and its
for physical activity were largest, and remained high after control of association with risk for disease incidence, mortality, and
hospitalization in adults: a systematic review and meta-anal-
other potential confounding factors (in all subgroup analyses),
ysis. Ann Intern Med 2015;162:123-132.
would indicate that physical activity should be the primary lifestyle
change recommended for people reporting symptoms of sexual 9. Lee IM, Shiroma EJ, Lobelo F, et al. Lancet Physical Activity
Series Working Group. Effect of physical inactivity on major
dysfunction. We recommend further research into the various
non-communicable diseases worldwide: an analysis of burden
lifestyle factors that might increase risk of premature ejaculation
of disease and life expectancy. Lancet 2012;380:219-229.
and female sexual dysfunction.
10. Rehm J, Mathers C, Popova S, et al. Global burden of disease
Corresponding Author: Mark S. Allen, University of and injury and economic cost attributable to alcohol use and
Wollongong, Northfields Avenue, Wollongong, NSW alcohol-use disorders. Lancet 2009;373:2223-2233.
2522, Australia. Tel: 61242981911; Fax: 61242215945; 11. Hall W, Degenhardt L. Adverse health effects of non-medical
E-mail: [email protected] cannabis use. Lancet 2009;374:1383-1391.
12. Stämpfli MR, Anderson GP. How cigarette smoke skews
Conflicts of Interest: The authors report no conflicts of interest.
immune responses to promote infection, lung disease and
Funding: None. cancer. Nat Rev Immunol 2009;9:377-384.
13. Lucas M, Mirzaei F, Pan A, et al. Coffee, caffeine, and risk of
depression among women. Arch Intern Med 2011;171:1571-1578.
STATEMENT OF AUTHORSHIP
14. Hatzimouratidis K, Amar E, Eardley I, et al. Guidelines on male
Category 1 sexual dysfunction: erectile dysfunction and premature ejac-
(a) Conception and Design ulation. Eur Urol 2010;57:804-814.
Mark S. Allen 15. Lewis RW, Fugl-Meyer KS, Corona G, et al. Definitions/epide-
(b) Acquisition of Data miology/risk factors for sexual dysfunction. J Sex Med 2010;
Emma E. Walter 7:1598-1607.
(c) Analysis and Interpretation of Data
16. Aytac IA, McKinlay JB, Krane RJ. The likely worldwide increase
Mark S. Allen
in erectile dysfunction between 1995 and 2025 and some
Category 2 possible policy consequences. BJU Int 1999;84:50-56.
(a) Drafting the Article 17. Laumann EO, Nicolosi A, Glasser DB, et al. Sexual problems
Mark S. Allen among women and men aged 40e80 y: prevalence and cor-
(b) Revising It for Intellectual Content relates identified in the Global Study of Sexual Attitudes and
Emma E. Walter Behaviors. Int J Impot Res 2005;17:39-57.
Category 3 18. Basson R. Women’s sexual dysfunction: revised and expanded
definitions. CMAJ 2005;172:1327-1333.
(a) Final Approval of the Completed Article
Mark S. Allen 19. McCool ME, Zuelke A, Theurich MA, et al. Prevalence of female
sexual dysfunction among premenopausal women: a sys-
tematic review and meta-analysis of observational studies.
REFERENCES Sex Med Rev 2016;4:197-212.
1. World Health Organization. Progress report 2016: prevent HIV, 20. Esposito K, Giugliano F, Ciotola M, et al. Obesity and sexual
test and treat all: WHO support for country impact. Geneva: dysfunction, male and female. Int J Impot Res 2008;
World Health Organization; 2016. 20:358-365.
2. Atlantis E, Sullivan T. Bidirectional association between 21. Esposito K, Giugliano F, Maiorino MI, et al. Dietary factors,
depression and sexual dysfunction: a systematic review and Mediterranean diet and erectile dysfunction. J Sex Med 2010;
meta-analysis. J Sex Med 2012;9:1497-1507. 7:2338-2345.
3. Nappi RE, Cucinella L, Martella S, et al. Female sexual 22. Glina S, Sharlip ID, Hellstrom WJ. Modifying risk factors to pre-
dysfunction (FSD): prevalence and impact on quality of life vent and treat erectile dysfunction. J Sex Med 2013;10:115-119.
(QoL). Maturitas 2016;94:87-91. 23. Hannan JL, Maio MT, Komolova M, et al. Beneficial impact of
4. De Tejada IS, Angulo J, Cellek S, et al. Pathophysiology of exercise and obesity interventions on erectile function and its
erectile dysfunction. J Sex Med 2005;2:26-39. risk factors. J Sex Med 2009;6(Suppl 3):254-261.

J Sex Med 2018;-:1e18

18 Allen and Walter

24. Kovac JR, Labbate C, Ramasamy R, et al. Effects of cigarette 44. Begg CB, Mazumdar M. Operating characteristics of a rank
smoking on erectile dysfunction. Andrologia 2015;47:1087-1092. correlation test for publication bias. Biometrics 1994;
25. Ledda A. Cigarette smoking, hypertension and erectile 50:1088-1101.
dysfunction. Curr Med Res Opin 2000;16(Suppl 1):13-16. 45. Duval S, Tweedie R. A nonparametric “trim and fill” method of
26. McKay A. Sexuality and substance use: the impact of tobacco, accounting for publication bias in meta-analysis. J Am Stat
alcohol, and selected recreational drugs on sexual function. Assoc 2000;95:89-98.
Can J Hum Sex 2005;14:47-56. 46. Eckardt MJ, File SE, Gessa GL, et al. Effects of moderate
27. McVary KT, Carrier S, Wessells H. Smoking and erectile alcohol consumption on the central nervous system. Alcohol
dysfunction: evidence based analysis. J Urol 2001; Clin Exp Res 1998;22:998-1040.
166:1624-1632. 47. Allen MS, Desille AE. Health-related lifestyle factors and sex-
28. Palha AP, Esteves M. Drugs of abuse and sexual functioning. ual functioning and behavior in older adults. Int J Sex Health
In: Sexual dysfunction, Balon R (ed). Basel: Karger; 2008. p. 2017;29:273-277.
131-149. 48. Cumming DC, Wheeler GD, McColl EM. The effects of exercise
29. Peugh J, Belenko S. Alcohol, drugs and sexual function: a on reproductive function in men. Sports Med 1989;7:1-7.
review. J Psychoactive Drugs 2001;33:223-232. 49. Geniole SN, Bird BM, Ruddick EL, et al. Effects of competition
30. Tostes RC, Carneiro FS, Lee AJ, et al. Cigarette smoking and outcome on testosterone concentrations in humans: an
erectile dysfunction: focus on NO bioavailability and ROS updated meta-analysis. Horm Behav 2017;92:37-50.
generation. J Sex Med 2008;5:1284-1295. 50. Traish AM, Feeley RJ, Guay A. Mechanisms of obesity and
31. Vrentzos GE, Paraskevas KI, Mikhailidis DP. Dyslipidemia as a related pathologies: androgen deficiency and endothelial
risk factor for erectile dysfunction. Curr Med Chem 2007; dysfunction may be the link between obesity and erectile
14:1765-1770. dysfunction. FEBS J 2009;276:5755-5767.
32. Wolf R, Shulmam A. Erectile dysfunction and fertility related to 51. Chitaley K, Kupelian V, Subak L, et al. Diabetes, obesity and
cigarette smoking. J Eur Acad Dermatol 1996;6:209-216. erectile dysfunction: field overview and research priorities.
33. Cheng JY, Ng EM, Chen RY, et al. Alcohol consumption and J Urol 2009;182:S45-S50.
erectile dysfunction: meta-analysis of population-based 52. Maiorino MI, Bellastella G, Esposito K. Diabetes and sexual
studies. Int J Impot Res 2007;19:343-352. dysfunction: current perspectives. Diabetes Metab Syndr
34. Cheng JY, Ng EM, Ko JS, et al. Physical activity and erectile Obes 2014;7:95-105.
dysfunction: meta-analysis of population-based studies. Int J 53. Pontiroli AE, Cortelazzi D, Morabito A. Female sexual
Impot Res 2007;19:245-252. dysfunction and diabetes: a systematic review and meta-
35. Tengs TO, Osgood ND. The link between smoking and impo- analysis. J Sex Med 2013;10:1044-1051.
tence: two decades of evidence. Prev Med 2001;32:447-452. 54. Moran LJ, Brinkworth GD, Martin S, et al. Long-term effects
36. Cao S, Yin X, Wang Y, et al. Smoking and risk of erectile of a randomised controlled trial comparing high protein or high
dysfunction: systematic review of observational studies with carbohydrate weight loss diets on testosterone, SHBG, erectile
meta-analysis. PLoS One 2013;8:e60443. and urinary function in overweight and obese men. PLoS One
2016;11:e0161297.
37. Cao S, Gan Y, Dong X, et al. Association of quantity and
duration of smoking with erectile dysfunction: a dose- 55. Beauchamp G, Amaducci A, Cook M. Caffeine toxicity: a brief
response meta-analysis. J Sex Med 2014;11:2376-2384. review and update. Clin Pediatr Emerg Med 2017;18:197-202.
38. Allen MS, Vella SA. Longitudinal determinants of walking, 56. Gupta BP, Murad MH, Clifton MM, et al. The effect of lifestyle
moderate, and vigorous physical activity in Australian adults. modification and cardiovascular risk factor reduction on
Prev Med 2015;78:101-104. erectile dysfunction: a systematic review and meta-analysis.
Arch Intern Med 2011;171:1797-1803.
39. Pate RR, Heath GW, Dowda M, et al. Associations between
physical activity and other health behaviors in a representa- 57. Maiorino MI, Bellastella G, Caputo M, et al. Effects of Medi-
tive sample of US adolescents. Am J Public Health 1996; terranean diet on sexual function in people with newly
86:1577-1581. diagnosed type 2 diabetes: the MÈDITA trial. J Diabetes
40. Moher D, Liberati A, Tetzlaff J, et al; PRISMA Group. Preferred Complicat 2016;30:1519-1524.
reporting items for systematic reviews and meta-analyses: the 58. Martin CK, Bhapkar M, Pittas AG, et al. Effect of calorie re-
PRISMA statement. PLoS Med 2009;6:e1000097. striction on mood, quality of life, sleep, and sexual function in
41. Downes MJ, Brennan ML, Williams HC, et al. Development of healthy nonobese adults: the CALERIE 2 randomized clinical
a critical appraisal tool to assess the quality of cross-sectional trial. JAMA Intern Med 2016;176:743-752.
studies (AXIS). BMJ Open 2016;6:e011458.
42. Borenstein M, Hedges LV, Higgins J, et al. Introduction to
meta-analysis. Chichester, UK: Wiley; 2009. SUPPLEMENTARY DATA
43. Higgins JP, Thompson SG, Deeks JJ, et al. Measuring incon- Supplementary data related to this article can be found at
sistency in meta-analyses. BMJ 2003;327:557. https://doi.org/10.1016/j.jsxm.2018.01.024.

J Sex Med 2018;-:1e18