Vasectomy and Risk of Prostate Cancer

ORIGINAL CONTRIBUTION
Vasectomy and Risk of Prostate Cancer
Brian Cox, MB, PhD
Context Vasectomy is a common method of contraception, but concern exists about
Mary J. Sneyd, MB, PhD
a reported association with risk of prostate cancer.
Charlotte Paul, MB, PhD
Objective To examine whether vasectomy increases risk of prostate cancer.
Brett Delahunt, MD, FRCPA
Design, Setting, and Participants National population-based case-control study
of 923 new cases of prostate cancer among men aged 40 to 74 years from the New
David C. G. Skegg, MB, DPhil
Zealand Cancer Registry who were on the general electoral roll. Controls (n=1224)
were randomly selected from the general electoral roll, with frequency matching to
VASECTOMYISTHESAFEST casesin5-yearagegroups.Cases(3-15monthsafterdiagnosis)andcontrolswere
method of surgical steriliza-
interviewed by telephone between January 1997 and November 1999.
tion.1 It has become an in-
Main Outcome Measures Relative risk (RR) of prostate cancer for men who had
creasingly common proce-
had a vasectomy vs those who had not.
dure in many countries since the early
Results There was no association between prostate cancer and vasectomy (RR, 0.92;
1970s. Worldwide about 42 to 60 mil-
95% confidence interval [CI], 0.75-1.14) nor with time since vasectomy (RR, 0.92;
lion couples rely on vasectomy for con-
95% CI, 0.68-1.23 for
25 years since vasectomy). Adjustment for social class, geo-
traception.2 Concerns about the safety
graphic region, religious affiliation, and a family history of prostate cancer did not affect
of vasectomy arose following reports in
these RRs.
1990 of an increased risk of prostate can-
Conclusions Vasectomy does not increase the risk of prostate cancer, even after 25
cer from 2 hospital-based data systems
years or more.
analyzed using case-control meth-
JAMA. 2002;287:3110-3115
www.jama.com
ods.3,4 The analysis of a further set of
cases and controls from 1 of these hos-
cancer would be of great importance.
approach the patients was sought from
pital-based surveillance systems found
New Zealand, a country with the high-
each patient’s consultant and general
no significant association, suggesting
est vasectomy prevalence in the world12
practitioner. Consultants were also
that the earlier finding was due to
and statutory notification of cancer, was
asked for the clinical TNM stage at di-
chance.5 Reviews of these and other early
considered ideal for a study to exam-
agnosis.13 Approval for the study was
studies concluded that vasectomy was
ine this association.
given by all regional ethics commit-
probably not a risk factor for prostate
tees in New Zealand, and men were in-
cancer.6,7 Moreover, a biological expla-
METHODS
vited to give consent for interview.
nation of any association seemed un-
The source population of cases com-
likely.8 In 1993, however, 2 large co-
prised all New Zealand men newly di-
Cases
hort studies showed a significantly
agnosed with prostate cancer. To maxi-
The National Cancer Registry for-
increased risk, particularly 20 or more
mize the response rate and avoid
warded histology reports of prostate
years after vasectomy (when the rela-
confounding by ethnicity, the study was
cancer14 for all men diagnosed be-
tive risk [RR] was 1.9 in both studies).9,10
restricted to men of European de-
tween April 1, 1996, and December 31,
Although the evidence was inconsis-
scent. In New Zealand, Maori people
1998. There were 3186 men aged 40 to
tent, sufficient concern arose for many
choose whether to be on the general
74 years with prostate cancer reported
urologists to screen vasectomized men
electoral roll or a separate Maori elec-
early for prostate cancer and to discour-
toral roll. To be eligible for inclusion
Author Affiliations: Department of Preventive and So-
cial Medicine, Dunedin School of Medicine (Drs Cox,
age vasectomy in men with a strong fam-
in this national population-based study,
Sneyd, Paul, and Skegg), and Department of Pathol-
ily history of prostate cancer.11
case and control subjects had to be
ogy, Wellington School of Medicine (Dr Delahunt),
University of Otago, Dunedin, New Zealand.
In view of the widespread use of va-
listed on the general electoral roll, have
Corresponding Author and Reprints: Brian Cox, MB,
sectomy and the relatively common oc-
a traceable telephone number, have
PhD, Hugh Adam Cancer Epidemiology Unit, Depart-
currence of prostate cancer, any asso-
been married, and be aged 40 to 74
ment of Preventive and Social Medicine, University of
Otago Medical School, PO Box 913, Dunedin, New
ciation of vasectomy with prostate
years. Approval for the study team to
Zealand (e-mail: brian.cox@stonebow.otago.ac.nz).
3110 JAMA, June 19, 2002—Vol 287, No. 23 (Reprinted)
©2002 American Medical Association. All rights reserved.
by guest on March 15, 2011
jama.ama-assn.org
Downloaded from

---

VASECTOMY AND RISK OF PROSTATE CANCER
during this time. The reports were re-
after diagnosis, 89 men were ex-
Table 1. Potentially Eligible Cases
viewed by 2 authors (M.J.S. and B.C.),
cluded, and an additional 39 were ex-
and Controls
and 31 men without confirmed adeno-
cluded because they had never been
Cases
Controls
carcinoma of the prostate, or with can-
married. The outcome for the 1082 po-
Total potentially eligible
1082
1582
cer found incidentally at cystoprosta-
tentially eligible cases is shown in
men
tectomy for bladder cancer, were
TABLE 1.
Not interviewed
125
321
excluded. Because a project in the
The pathological specimens of a ran-
Died before interview
31
42
completed
Auckland area was being completed at
dom sample of 100 cases were inde-
Doctor refused
13
0
the beginning of this study, 446 men
pendently reviewed by one of the au-
Subject refused
39
196
aged 40 to 74 years, diagnosed before
thors (B.D.), and all 100 original
Too ill
34
22
April 1, 1997, and living in the Auck-
diagnoses were confirmed.
Not traced or living
3
41
land region, were excluded. Control
overseas
subjects living in this area were ap-
Controls
Language difficulties
5
20
proached only after April 1, 1997. Be-
Electoral registration is compulsory for
Interviewed
957
1261
cause we were notified of more pros-
all people 18 years or older in New Zeal-
Excluded after interview
4
1
(ineligible)
tate cancer cases than expected, during
and, and the information is available in
Interviewed and eligible
953
1260
the last 6 months of the study a ran-
electronic form for research purposes.
dom selection of approximately 50% of
About 95% of adults are listed on the
prostate cancer cases (stage T2 or
electoral roll. Control subjects were ran-
to the length of time elapsed since va-
higher) was included, giving a total of
domly selected from the general elec-
sectomy, only 49 records were located.
2132 potentially eligible men with pros-
toral roll and frequency matched in
For those that were traced, all self-
tate cancer of all stages.
5-year age groups with the cases of pros-
reports of vasectomy were confirmed.
Electoral roll listings (Maori or gen-
tate cancer; 1819 men in all. Tele-
Interviewers did not know whether
eral electoral roll) were found for 2027
phone numbers were found for 1689 of
they were interviewing a case or a con-
of the 2118 men (96%) sought (14 were
these men (93%), to whom a letter seek-
trol until partway through the tele-
excluded due to death). Telephone
ing participation in a study of health
phone interview, after information
numbers were found for 1862 of the
problems was sent. If no response was
about vasectomy had been collected.
1883 men (99%) on the general elec-
received, consent for interview was
The study hypothesis was not dis-
toral roll for whom numbers were
sought by telephone and an interview
closed to subjects.
sought. The rest (144) were either on
date arranged.
the Maori roll, dead, reported too late,
All control subjects were interviewed
Definition of Terms
known to have stage T1 cancers be-
between January 1997 and November
A positive family history of prostate can-
fore the telephone number was sought,
1999. The 24 men selected as controls
cer was defined as having a brother or
or were excluded because the consul-
who had a previous diagnosis of pros-
father reported to have prostate can-
tant (medical specialist, usually a urolo-
tate cancer were excluded, as were the
cer. A New Zealand scale based on oc-
gist) did not participate. One consul-
83 men who had never been married.
cupation was used to assign 6 levels of
tant replied to requests for staging
The outcome for the remaining 1582
social class from the longest-held oc-
information only after the study com-
control subjects is shown in Table 1.
cupation.15 Due to the small numbers
menced. Altogether the TNM stage was
in the lowest 2 categories, these were
known for 1554 men but unknown for
Data Collection
combined to create 5 levels of social
308 men. Initially the design of the
Information was collected using iden-
class for analysis. The regional bound-
study did not include men with stage
tical methods for cases and controls. In-
aries of 4 recent health purchasing au-
T1 prostate cancer. However, it was
terviews were conducted by telephone,
thorities defined geographic regions of
later decided to include a random
took about 30 minutes, and sought in-
similar population size from the north
sample of about 200 men with stage T1
formation about previous illnesses, va-
to the south of the country.
prostate cancer, diagnosed between July
sectomy, smoking and alcohol con-
Self-reported average weekly con-
1, 1996, and November 30, 1997. All
sumption, prostate-specific antigen
sumption of beer, wine, and spirits was
other T1 cases (344) were excluded,
(PSA) testing and digital rectal exami-
combined using estimated alcohol con-
leaving 1210 men with prostate can-
nation (DRE), previous urological symp-
tents of 3.5%, 12.5%, and 37.5%, re-
cer of all stages.
toms and operations, any family his-
spectively. For liqueurs and fortified
Cases had to be interviewed be-
tory of cancer, and sociodemographic
wine, estimates of alcoholic content of
tween 3 and 15 months after their di-
characteristics. All interviews were con-
40% and 18%, respectively, were used.
agnosis. Cases were interviewed by tele-
ducted by 3 trained interviewers. Vali-
Total weekly alcohol consumption over
phone from January 1997 to November
dation of the history of vasectomy for a
the previous 5 years was categorized into
1999. Due to delays beyond 15 months
sample of 103 men was attempted.12 Due
quartiles with cut points of 52, 129, and
©2002 American Medical Association. All rights reserved.
(Reprinted) JAMA, June 19, 2002—Vol 287, No. 23 3111
by guest on March 15, 2011
jama.ama-assn.org
Downloaded from

---

VASECTOMY AND RISK OF PROSTATE CANCER
250 mL per week. Cigarette smoking
sion 6.1 (SPSS Inc, Chicago, Ill) was
ent religious affiliation or geographic
was defined as smoking more than 1
used for all analyses. Adjustment for age
region of residence (data not shown).
cigarette a day for more than a year.
was made by using age as a continu-
A small but statistically insignificant el-
ous variable in the regression models.
evation in RR was present for men of
Statistical Analyses
social classes 3 and 4 relative to the
Most respondents (97%) were of Eu-
RESULTS
highest social class, 1, but no signifi-
ropean descent, and the analysis was
The mean ages for cases and controls
cant heterogeneity in risk of prostate
confined to the 923 cases and 1224 con-
were 66.3 years and 65.1 years, respec-
cancer by social class was observed.
trols in this ethnic group. For each con-
tively. The age-adjusted RRs for se-
There was no association with a his-
trol a reference date was calculated by
lected characteristics of case and con-
tory of sexually transmitted disease. A
subtracting 340 days (11.2 months)
trol subjects are shown in TABLE 2. Men
history of smoking (current smokers
from the date of interview to corre-
who reported having a father or brother
and ex-smokers), consumption of al-
spond with the mean time between di-
with a diagnosis of prostate cancer had
cohol in the past 5 years (each quar-
agnosis and interview among the cases.
an increased risk of prostate cancer (RR,
tile compared with nondrinkers), and
After interview, 1 control was ex-
2.63; 95% confidence interval [CI], 1.88-
number of children were not related to
cluded because he was younger than 40
3.68). Inclusion of uncles with a diag-
risk of prostate cancer (data not shown).
years on his reference date.
nosis of prostate cancer to define a posi-
Self-report of a PSA test at least 18
Standard methods for analysis of case-
tive family history did not appreciably
months before interview was more
control studies were used. Relative risks
alter the estimated RR.
common among cases (RR, 1.36; 95%
were estimated by calculating odds ra-
The RR of prostate cancer did not dif-
CI, 1.08-1.71) and most pronounced for
tios using logistic regression. SPSS ver-
fer significantly among men of differ-
stage T1 and T2 cases (TABLE 3). Con-
trols were significantly less likely than
cases to know whether they had had a
Table 2. Characteristics of Cases and Controls*
PSA test (RR, 0.56; 95% CI, 0.43-0.72).
No. (%)
Cases were significantly less likely to
Relative Risk
have had a DRE at least 18 months prior
Cases
Controls
of Prostate Cancer
Characteristic
(N = 923)
(N = 1224)
(95% Confidence Interval)
to interview than controls (RR, 0.79;
Family history of prostate cancer
95% CI, 0.66-0.94), an effect confined
No
819 (89)
1165 (95)
1.00
to stage T3 and T4 cases. Cases were
Yes
104 (11)
59 (5)
2.63 (1.88-3.68)
significantly less likely to know whether
Religious affiliation†
they had had a DRE (RR, 2.45; 95% CI,
Catholic
124 (13)
159 (13)
1.00
1.21-4.94), an effect confined to stage
Protestant
583 (63)
803 (66)
0.95 (0.73-1.23)
T1 and T2 cases.
Other
52 (6)
74 (6)
0.93 (0.60-1.42)
None
161 (18)
186 (15)
1.17 (0.85-1.61)
The age-adjusted RRs of prostate can-
Social class level†
cer from vasectomy overall and for se-
1 (High)
102 (11)
149 (12)
1.00
lected subgroups are shown in TABLE 4.
2
163 (18)
223 (18)
1.06 (0.77-1.47)
For both cases and controls, the mean
3
366 (40)
460 (38)
1.14 (0.85-1.52)
age at vasectomy was 41.3 years. The
4
227 (25)
280 (23)
1.17 (0.86-1.60)
prevalence of vasectomy in controls, ad-
5 (Low)
62 (7)
110 (9)
0.81 (0.54-1.21)
justed to the age distribution of New
Sexually transmitted diseases‡
Zealand European men aged 40 to 74
No
900 (98)
1189 (97)
1.00
years, was 44%.12 A history of vasec-
Yes
22 (2)
34 (3)
0.97 (0.56-1.68)
tomy was less common among cases
PSA test at least 18 months
before interview§
than controls, although this was not sta-
No
632 (69)
822 (67)
1.00
tistically significant (RR, 0.92; 95% CI,
Yes
188 (20)
175 (14)
1.36 (1.08-1.71)
0.75-1.14). Adjustment for social class,
Don’t know
102 (11)
227 (19)
0.56 (0.43-0.72)
geographic region, religious affilia-
DRE at least 18 months
tion, and a family history of prostate
before interview
No
487 (53)
594 (49)
1.00
cancer had little effect on the RR of
Yes
411 (45)
618 (50)
0.79 (0.66-0.94)
prostate cancer from vasectomy (RR,
Don’t know
25 (3)
12 (1)
2.45 (1.21-4.94)
0.92; 95% CI, 0.74-1.13).
*PSA indicates prostate-specific antigen; DRE, digital rectal examination. Relative risk was adjusted by individual year
Similar proportions of cases (9%) and
of age in logistic regression.
controls (10%) had a vasectomy 25 years
†Three cases and 2 controls were excluded because of incomplete information.
‡One case and 1 control were excluded because they declined to respond.
or more previously, representing 38% of
§One case was excluded because response could not be interpreted.
all men with a vasectomy. The RR of pros-
3112 JAMA, June 19, 2002—Vol 287, No. 23 (Reprinted)
©2002 American Medical Association. All rights reserved.
by guest on March 15, 2011
jama.ama-assn.org
Downloaded from

---

VASECTOMY AND RISK OF PROSTATE CANCER
Table 3. Relative Risk (95% Confidence Interval) of Prostate Cancer for at Least 1 PSA Test or DRE 18 Months or Longer Before Interview,
Compared With None During That Period, by Stage of Prostate Cancer*
Stage
Test
T1
T2
T3
T4
All Stages
PSA
No
1.00
1.00
1.00
1.00
1.00
Yes
2.30 (1.61-3.29)
1.48 (1.12-1.95)
0.62 (0.39-1.01)
0.47 (0.14-1.55)
1.36 (1.08-1.71)
Don’t know
0.93 (0.61-1.43)
0.49 (0.35-0.69)
0.48 (0.30-0.77)
0.35 (0.10-1.16)
0.56 (0.43-0.72)
DRE
No
1.00
1.00
1.00
1.00
1.00
Yes
1.21 (0.89-1.64)
0.83 (0.67-1.03)
0.50 (0.37-0.69)
0.36 (0.17-0.75)
0.79 (0.66-0.94)
Don’t know
3.42 (1.25-9.36)
3.01 (1.40-6.49)
1.00 (0.28-3.64)
. . .
2.45 (1.21-4.94)
*PSA indicates prostate-specific antigen; DRE, digital rectal examination. Relative risk was adjusted by individual year of age in logistic regression. Ellipses indicate too few subjects
for estimation of relative risk.
tate cancer was not elevated for men
whose vasectomy was at least 25 years
Table 4. Prostate Cancer After Vasectomy According to Time Since Vasectomy, Age at
Vasectomy, Age, Family History, and Stage of Cancer*
previous (RR, 0.92; 95% CI, 0.68-1.23).
No. (%)
Also, no trend in RR with years since
vasectomy was apparent. Adjustment for
Cases
Controls
Relative Risk
Characteristic
(N = 923)
(N = 1224)
(95% Confidence Interval)
social class, geographic region, reli-
Vasectomy
gious affiliation, and a family history of
No
707 (77)
891 (73)
1.00
prostate cancer did not appreciably alter
Yes
216 (23)
333 (27)
0.92 (0.75-1.14)
the estimates of RR for each period since
Years since vasectomy†
vasectomy. The RR of prostate cancer
No vasectomy
707 (77)
891 (73)
1.00
after vasectomy was not increased for any
0-14
25 (3)
55 (5)
0.76 (0.46-1.26)
age at vasectomy. There was also no sta-
15-19
42 (5)
55 (5)
1.16 (0.76-1.78)
tistically significant variation in RR by age
20-24
61 (7)
96 (8)
0.92 (0.66-1.30)
at diagnosis. For the few men aged 40 to
25
86 (9)
120 (10)
0.92 (0.68-1.23)
49 years at diagnosis, an increased RR of
Age at vasectomy†
No vasectomy
707 (77)
891 (73)
1.00
prostate cancer after vasectomy was
34
32 (3)
46 (4)
1.23 (0.76-2.01)
observed, but this was not statistically sig-
35-39
50 (5)
96 (8)
0.81 (0.56-1.17)
nificant. The RR of prostate cancer from
40-44
74 (8)
88 (7)
1.15 (0.83-1.59)
vasectomy was not appreciably differ-
45-49
35 (4)
58 (5)
0.78 (0.50-1.20)
ent among men with or without a fam-
50
23 (2)
38 (3)
0.76 (0.45-1.30)
ily history of prostate cancer.
Age at diagnosis (or reference data)
When the case series comprised only
40-49
11 (1)
33 (3)
4.47 (0.75-26.72)
men with stage T1 prostate cancer, va-
50-54
30 (3)
56 (5)
0.83 (0.34-2.06)
sectomy was again less common among
55-59
78 (8)
122 (10)
0.81 (0.46-1.44)
men with prostate cancer, although not
60-64
189 (20)
253 (21)
1.00 (0.66-1.51)
significantly so (RR, 0.87; 95% CI,
65-69
287 (31)
422 (34)
1.00 (0.70-1.43)
0.61-1.25). When only stage T4 cases
70-74
328 (36)
338 (28)
0.69 (0.43-1.11)
were included, the RR was reduced fur-
Family history of prostate cancer
ther (RR, 0.53; 95% CI, 0.20-1.41) but
No
819 (89)
1165 (95)
1.04 (0.93-1.16)
was still not significant. The RR was
Yes
104 (11)
59 (5)
0.95 (0.64-1.40)
Stage
highest for stage T2 cases (RR, 1.02;
T1
202 (22)
1224 (100)
0.87 (0.61-1.25)
95% CI, 0.80-1.32). Adjustment for a
T2
481 (52)
1224 (100)
1.02 (0.80-1.32)
family history of prostate cancer, and
T3
205 (22)
1224 (100)
0.81 (0.56-1.18)
reporting at least 1 PSA test or at least
T4
35 (4)
1224 (100)
0.53 (0.20-1.41)
1 DRE 18 months or longer before in-
T2
721 (78)
1224 (100)
0.94 (0.75-1.17)
terview, made no appreciable differ-
*Relative risk was adjusted by individual year of age in logistic regression.
ence to the results obtained for each
†Two cases and 7 controls were excluded because year of vasectomy was unknown.
stage.
When the analysis was confined to
ter vasectomy was found (RR, 0.82; 95%
tomy was found among those with can-
stage T1 cases, no elevation in the RR
CI, 0.48-1.40). Similarly, no elevation
cer at stage T2 or higher (RR, 0.94; 95%
of prostate cancer 25 years or more af-
in RR 25 years or more after vasec-
CI, 0.69-1.29).
©2002 American Medical Association. All rights reserved.
(Reprinted) JAMA, June 19, 2002—Vol 287, No. 23 3113
by guest on March 15, 2011
jama.ama-assn.org
Downloaded from

---

VASECTOMY AND RISK OF PROSTATE CANCER
COMMENT
tomy and prostate cancer in our study,
tomy, there was little confounding by
This large national population-based
as risk was not altered by adjustment for
dietary factors9; associations between
case-control study provides strong evi-
a prior PSA test or DRE. Furthermore,
prostate cancer and dietary nutrients (in-
dence that vasectomy does not in-
no association between prostate cancer
cluding some vegetable fats and anti-
crease the risk of prostate cancer. No
and vasectomy was found for any stage
oxidants) appear weak.18-21 While sele-
elevation in risk was found 25 years or
of disease, suggesting that misclassifi-
nium intake may affect the risk of
more after vasectomy, even though 38%
cation of stage of disease at diagnosis was
prostate cancer,22 confounding of any
of men with vasectomy were in this
unlikely to have influenced the results.
association with vasectomy seems un-
group. Adjustment for a family his-
The variation in RRs for different
likely. Alcohol consumption23 and ciga-
tory of prostate cancer or other pos-
stages of prostate cancer associated with
rette smoking24 have not been consis-
sible explanatory variables did not affect
a PSA test or DRE at least 18 months be-
tently associated with an increased risk
these results.
fore interview is consistent with the pre-
of prostate cancer, which was con-
Although many men are diagnosed
sentation of men with prostate cancer.
firmed in our study. We also examined
and treated for prostate cancer as out-
For example, men with more advanced
religious affiliation and geographic varia-
patients, the cases for this study came
cancer are less likely to have a DRE be-
tion because these factors could be as-
from the entire population through
fore their date of diagnosis. Men whose
sociated with both the use of vasec-
statutory notification of cancers by all
cancer was diagnosed through investi-
tomy and the incidence of prostate
pathology laboratories. Response rates
gation for persistent or rising PSA lev-
cancer. In fact, they were found not to
for both cases and controls were high,
els would be more likely to have stage
be confounders in our study.
minimizing the possible effects of selec-
T1 or T2 cancer. Some diagnoses of stage
Three studies found no association be-
tion bias. Pathological review of a sample
T1 cancer are incidental findings from
tween serum testosterone, dihydrotes-
of diagnostic specimens confirmed the
transurethral resection of the prostate for
tosterone, their ratio, or 3- , 17- an-
diagnosis of prostate cancer.
urological symptoms. Detection of pros-
drostanediol glucuronide and the risk of
Possible interviewer bias was avoided
tate cancer through PSA testing and the
prostate cancer.25-27 However, serum lev-
by collecting information about vasec-
possible inclusion of postsurgical stag-
els may not reflect differences in pros-
tomy before interviewers were aware of
ing information could lead to detection
tatic concentrations of these hor-
the case or control status of each sub-
bias among stage T1 and T2 cases.
mones. Because no major causative
ject. Vasectomy status was obtained by
What are possible confounders of the
factor for prostate cancer has been de-
self-report, and validation of vasec-
association between prostate cancer and
termined, there is potential for con-
tomy proved difficult, but in the small
vasectomy? Intercountry variation in
founding by some unknown factor to in-
sample in which validation was sought
prostate cancer incidence and mortal-
fluence associations with vasectomy.
and was possible, all reports of vasec-
ity, time trends, and the general ab-
Hence, small elevated RRs obtained from
tomy were confirmed. Unlike many ex-
sence of a social class gradient do not
case-control or cohort studies require
posures measured in epidemiologic in-
support a sexually transmitted disease
cautious interpretation.
vestigations, few men are in any doubt
etiology for prostate cancer. However,
Several studies have reported an as-
about whether they have undergone va-
a recent nested case-control study that
sociation between vasectomy and pros-
sectomy; where vasectomy history has
found an association with positive hu-
tate cancer, with RRs as high as 6.7 in 1
been confirmed, self-report of age at va-
man papillomavirus (HPV)-16 and
case-control study and 1.9 in some co-
sectomy is highly correlated with phy-
HPV-18 serology has rekindled inter-
hort studies.3,4,9,10,28 Most case-control
sician-reported age.16
est in this hypothesis.17 If vasectomy
studies reporting an association have
More frequent contact with urolo-
were associated with increased num-
used hospital-based cases and controls,
gists among men with previous vasec-
bers of partners and a sexually trans-
which may be more susceptible to se-
tomy, particularly in the United States,
mitted infection were a risk factor for
lection bias than population-based stud-
might have produced detection bias in
prostate cancer, this would be a poten-
ies. In a case-control study in China, the
some studies. In New Zealand, vasec-
tial confounder of the association be-
RR varied from 2.0 to 6.7, depending on
tomy was initially performed by urolo-
tween vasectomy and prostate cancer.
the control series used in the analysis, in-
gists, but soon many family doctors of-
While no association between a history
dicating the degree to which selection
fered this service. Therefore, any
of a sexually transmitted disease and
bias may influence results.28 Our re-
detection bias from close surveillance
prostate cancer was found in our study,
sults are consistent with several recent
by urologists is unlikely. Also, annual
the low prevalence of previous sexu-
studies that have not found a signifi-
medical checks have not been a major
ally transmitted disease suggests that
cant association between vasectomy and
feature of the New Zealand health care
considerable underreporting occurred.
prostate cancer.29-34 In particular, sev-
service.
We did not collect dietary informa-
eral cohort studies using record linkage
Detection bias does not appear to have
tion. In a previous study reporting an as-
have not found any increased risk of
affected the relationship between vasec-
sociation of prostate cancer with vasec-
prostate cancer among men with vasec-
3114 JAMA, June 19, 2002—Vol 287, No. 23 (Reprinted)
©2002 American Medical Association. All rights reserved.
by guest on March 15, 2011
jama.ama-assn.org
Downloaded from

---

VASECTOMY AND RISK OF PROSTATE CANCER
tomy.35-37 A systematic review of 9 case-
Statistical expertise: Cox.
bric 185. Geneva, Switzerland: World Health Orga-
Obtained funding: Cox, Skegg.
nization; 1977.
control and 5 cohort studies reported
Administrative, technical, or material support: Cox,
15. Elley W, Irving J. The Elley-Irving socio-
wide variation in estimated RRs, which
Sneyd, Delahunt.
economic index. N Z J Educ Stud. 1985;20:115-128.
were sensitive to differences in the study
Study supervision: Cox, Paul, Skegg.
16. Lesko SM, Louik C, Vezina R, et al. Vasectomy
Financial Disclosure: Dr Cox receives funding from
and prostate cancer. J Urol. 1999;161:1848-1853.
population, study design, presence of de-
the Director’s Cancer Research Trust.
17. Dillner J, Knekt P, Boman J, et al. Sero-
tection bias, and inadequate control se-
Funding/Support and Disclaimers: This work was sup-
epidemiological association between human-
ported by the National Institute of Child Health and
papillomavirus infection and risk of prostate cancer.
lection.38 Also, publication bias was
Human Development/National Institutes of Health
Int J Cancer. 1998;75:564-567.
likely.38 The low prevalence of vasec-
through Interagency Agreement Y1-HD-7053 with the
18. Ghadirian P, Lacroix A, Maisonneuve P, et al. Nu-
United States Agency for International Development
tomy among control subjects in some
tritional factors and prostate cancer. Cancer Causes
(USAID). The content of this publication does not nec-
Control. 1996;7:428-436.
studies raises the possibility of bias in the
essarily reflect the views or policies of the Depart-
19. Willett WC. Specific fatty acids and risks of breast
collection of vasectomy information from
ment of Health and Human Services, nor does men-
and prostate cancer. Am J Clin Nutr. 1997;66(suppl):
tion of trade names, commercial products, or
1557S-1563S.
cases and controls.39 If prostate cancer
organizations imply endorsement by the US govern-
20. Norrish AE, Jackson RT, Sharpe SJ, Skeaff CM.
takes some time to develop after expo-
ment. This investigation also received financial sup-
Prostate cancer and dietary carotenoids. Am J Epide-
port from the Special Programme of Research, De-
miol. 2000;151:119-123.
sure to a causative agent, an elevation in
velopment, and Research Training in Human
21. Nomura AMY, Stemmermann GN, Lee J, Craft NE.
the RR within a short time after the op-
Reproduction, World Health Organization. Technical
Serum micronutrients and prostate cancer in Japa-
support for this study was provided by Family Health
eration is likely to be due to detection bias
nese Americans in Hawaii. Cancer Epidemiol Biomar-
International (FHI) and USAID (AID/CCP-3079-A-
kers Prev. 1997;6:487-491.
or some form of misclassification of dis-
00-5022-00, subagreement 6900), although the views
22. Yoshizawa K, Willet WC, Morris SJ, et al. Study
expressed in this article do not necessarily reflect those
ease or exposure.
of prediagnostic selenium level in toenails and the risk
of FHI or USAID.
of advanced prostate cancer. J Natl Cancer Inst. 1998;
Another possible explanation for the
Acknowledgment: We are grateful for advice from
90:1219-1224.
inconsistency in the results is that, in ar-
Judith Fortney, PhD, Olav Meirik, MD, and Tim Far-
23. Breslow RA, Weed DL. Review of epidemiologic
ley, PhD, about the preparation and conduct of the
eas of relatively low vasectomy preva-
studies of alcohol and prostate cancer: 1971-1996. Nutr
study. We thank Christina Benfell, Anthea Joel, BA,
Cancer. 1998;30:1-13.
lence, men seeking vasectomy may be
Andrew Joel, Ruth Shaw, and Sally Wood, RN, who
24. Rohan TE, Hislop TG, Howe GR, et al. Cigarette
assisted in the conduct of this investigation. Stuart
more likely to possess other factors that
smoking and risk of prostate cancer. Eur J Cancer Prev.
Gowland, FRACS, helped in the planning of the re-
1997;6:382-388.
increase their risk of prostate cancer,
search. We are grateful to the many urologists, gen-
25. Vatten LJ, Ursin G, Ross RK, et al. Androgens in se-
eral practitioners, and especially to the men who par-
whereas in areas of high vasectomy
rum and the risk of prostate cancer. Cancer Epidemiol
ticipated in this study.
Biomarkers Prev. 1997;6:967-969.
prevalence such as New Zealand, this
26. Guess HA, Friedman GD, Sadler MC, et al. 5 -
effect would be diluted by the addition
REFERENCES
reductase activity and prostate cancer. Cancer Epide-
of many men with vasectomy without
1. Schwingl P, Guess HA. Safety and effectiveness of
miol Biomarkers Prev. 1997;6:21-24.
vasectomy. Fertil Steril. 2000;73:923-936.
27. Dorgan JF, Albanes D, Virtamo J, et al. Relation-
such characteristics. However, in a study
2. Population Information Program. Vasectomy: New
ships of serum androgens and estrogens to prostate can-
that covered several geographic areas of
Opportunities. Baltimore, Md: Center for Communi-
cer risk. Cancer Epidemiol Biomarkers Prev. 1998;7:
1069-1074.
the United States with more than 2-fold
cation Programs, The Johns Hopkins University; 1992:
17-21.
28. Hsing AW, Wang R-T, Gu F-L, et al. Vasectomy
variation in vasectomy prevalence, an
3. Rosenberg L, Palmer JR, Zauber AG, et al. Vasec-
and prostate cancer risk in China. Cancer Epidemiol
Biomarkers Prev. 1994;3:285-288.
elevated risk of prostate cancer from
tomy and the risk of prostate cancer. Am J Epide-
miol. 1990;132:1051-1055.
29. Platz EA, Yeole BB, Cho E, et al. Vasectomy and
vasectomy was found in regions with
4. Mettlin C, Natarajan N, Huben R. Vasectomy and
prostate cancer. Int J Epidemiol. 1997;26:933-938.
both high and low prevalence of vasec-
prostate cancer risk. Am J Epidemiol. 1990;132:1056-
30. Zhu K, Stanford JL, Daling JR, et al. Vasectomy
1061.
and prostate cancer. Am J Epidemiol. 1996;144:717-
tomy.10
5. Rosenberg L, Palmer JR, Zauber AG, et al. The re-
722.
The high prevalence of vasectomy
lation of vasectomy to the risk of cancer. Am J Epi-
31. Stanford JL, Wicklund KG, McKnight B, et al. Va-
sectomy and risk of prostate cancer. Cancer Epide-
and the large size of our case-control
demiol. 1994:140:431-438.
6. Guess HA. Invited commentary: vasectomy and pros-
miol Biomarkers Prev. 1999;8:881-886.
study provided a 99% statistical power
tate cancer. Am J Epidemiol. 1990;132:1062-1065.
32. Newell GR, Fueger JJ, Spitz MR, Babaian RJ. A
to detect an RR of 1.5 or higher at the
7. Skegg DCG. Vasectomy and risk of cancers of the
case-control study of prostate cancer. Am J Epide-
prostate and testis. Eur J Cancer. 1993;29A:935-
miol. 1989;130:395-398.
5% level of significance. There was 80%
936.
33. John EM, Whittemore AS, Wu AH, et al. Vasec-
statistical power to detect such a risk
8. Howards SS. Possible biological mechanisms for a
tomy and prostate cancer. J Natl Cancer Inst. 1995;
relationship between vasectomy and prostate can-
87:662-669.
after 25 years or more had elapsed since
cer. Eur J Cancer. 1993;29A:1060-1062.
34. Hayes RB, Pottern LM, Greenberg R, et al. Va-
vasectomy. That no association be-
9. Giovannucci E, Ascherio A, Rimm EB, et al. A pro-
sectomy and prostate cancer in US blacks and whites.
spective cohort study of vasectomy and prostate can-
Am J Epidemiol. 1993;137:263-269.
tween prostate cancer and vasectomy
cer in US men. JAMA. 1993;269:873-877.
35. Moller H, Knudsen LB, Lynge E. Risk of testicular
was found, even 25 years or more af-
10. Giovannucci E, Tosteson TD, Speizer FE, et al. A
cancer after vasectomy. BMJ. 1994;309:295-299.
ter vasectomy, strongly suggests that
retrospective cohort study of vasectomy and pros-
36. Nienhuis H, Goldacre M, Seagroatt V, et al. In-
tate cancer in US men. JAMA. 1993;269:878-882.
cidence of disease after vasectomy. BMJ. 1992;304:
there is no increased risk of prostate
11. Sandlow JI, Kreder KJ. A change in practice. Fer-
743-746.
cancer after this procedure.
til Steril. 1996;66:281-284.
37. Sidney S, Quesenberry CP Jr, Sadler MC, et al.
12. Sneyd MJ, Cox B, Paul C, Skegg DCG. High preva-
Vasectomy and the risk of prostate cancer in a cohort
Author Contributions: Study concept and design: Cox,
lence of vasectomy in New Zealand. Contraception.
of multiphasic health-checkup examinees. Cancer
Paul, Skegg.
2001;64:155-159.
Causes Control. 1991;2:113-116.
Acquisition of data: Cox, Sneyd.
13. Schroder FH, Hermanek P, Denis L, et al. The TNM
38. Bernal-Delgado E, Latour-Perez J, Pradas-Arnal F,
Analysis and interpretation of data: Cox, Sneyd, Paul,
classification of prostate cancer. Prostate Suppl. 1992;
Gomez-Lopez LI. The association between vasectomy
Delahunt, Skegg.
4:129-138.
and prostate cancer. Fertil Steril. 1998;70:191-200.
Drafting of the manuscript: Cox, Sneyd.
14. International Classification of Diseases: Manual
39. Perlman JA, Spirtas R, Keleghan J, et al. Vasec-
Critical revision of the manuscript for important in-
of the International Statistical Classification of Dis-
tomy and the risk of prostate cancer [letter]. Am J Epi-
tellectual content: Cox, Sneyd, Paul, Delahunt, Skegg.
eases, Injuries, and Causes of Death. 9th revision, ru-
demiol. 1991;134:107-108.
©2002 American Medical Association. All rights reserved.
(Reprinted) JAMA, June 19, 2002—Vol 287, No. 23 3115
by guest on March 15, 2011
jama.ama-assn.org
Downloaded from

---