The New England Journal o f Me di c i ne

A PROSPECTIVE STUDY OF RISK FACTORS FOR SYMPTOMATIC URINARY TRACT

INFECTION IN YOUNG WOMEN

THOMAS M. HOOTON, M.D., DELIA SCHOLES, PH.D., JAMES P. HUGHES, PH.D., CAROL WINTER, A.R.N.P.,
PACITA L. ROBERTS, M.S., ANN E. STAPLETON, M.D., ANDY STERGACHIS, PH.D., AND WALTER E. STAMM, M.D.

ABSTRACT diaphragm with spermicide,4-9 delayed postcoital mic-

Background Although acute urinary tract in- turition,4,7,10,11 and the ABO-blood-group nonsecre-
fections are common in young women, the asso- tor phenotype.12-14 However, these factors have been
ciated risk factors have not been defined prospec- identified primarily in small case–control studies re-
tively. porting widely varying risk estimates. Furthermore,
Methods We recruited sexually active young questions about the comparability of the groups
women who were starting a new method of con- used in these studies and the potential for recall bias
traception at a university health center or a health and measurement errors are important limitations of
maintenance organization (HMO) and monitored retrospective studies. In addition, all of these recent
them for six months for symptomatic urinary tract
infections. Daily diaries and serial interviews were
studies have been conducted in college students,
used to collect data on potential risk factors. and the findings in these women may not apply to
Results Among 796 women, the incidence of other young women. As yet, there have been no
urinary tract infections per person-year was 0.7 in prospective studies large enough to address inde-
the university cohort (mean age, 23 years; n  348) pendent associations between these and other po-
and 0.5 in the HMO cohort (mean age, 29; n  448). tentially modifiable risk factors and urinary tract in-
In both cohorts, there were strong dose–response fection. We therefore conducted a prospective study
relations between the risk of infection and both using daily diaries to avoid recall bias and to seek
recent use of a diaphragm with spermicide (re- more conclusive evidence of risk factors that predis-
spective relative risks for one, three, and five days pose young women to urinary tract infection. Our
of use in the past week, 1.42, 2.83, and 5.68 in
main objectives were to determine the incidence of
the university cohort, P0.001; and 1.29, 2.14, and
3.54 in the HMO cohort, P  0.04) and recent sexual urinary tract infection in two populations of young
intercourse (respective relative risks for one, three, women and to identify factors independently asso-
and five days with intercourse in the past week, 1.37, ciated with an increased risk of urinary tract infec-
2.56, and 4.81 in the university cohort, P0.001; tion in these women.
and 1.24, 1.91, and 2.96 in the HMO cohort,
P  0.002). The risk of acute infection was also as- METHODS
sociated with a history of recurrent infection (rela-
Study Design
tive risk, 5.58 in the university cohort and 2.10 in the
HMO cohort) but not with cervical-cap use, ABO- This study was conducted at the University of Washington Stu-
blood-group nonsecretor phenotype, or delayed dent Health Center and the Group Health Cooperative of Puget
postcoital voiding. Sound. The University of Washington Student Health Center
Conclusions Among sexually active young wom- provides primary care for approximately 85 percent of the 34,000
en the incidence of symptomatic urinary tract infec- students enrolled in the university. The Group Health Coopera-
tive is a staff-model health maintenance organization (HMO) lo-
tion is high, and the risk is strongly and independ-
cated in western Washington. The age, sex, and racial distribu-
ently associated with recent sexual intercourse, tions of the approximately 370,000 enrollees are similar to those
recent use of a diaphragm with spermicide, and a of the surrounding community.15
history of recurrent urinary tract infections. (N Engl Subjects were eligible for inclusion in the study if they were
J Med 1996;335:468-74.) healthy women between 18 and 40 years old, were about to start
©1996, Massachusetts Medical Society. or had started within 6 weeks of enrollment a new method of
contraception, had not used this method in the previous 3 months,
and had had no more than one urinary tract infection in the pre-
vious 12 months. Subjects were ineligible if they were pregnant
or planned to become pregnant within the next 6 months, if they

A
N estimated 7 million episodes of acute had a chronic illness requiring medical supervision, if they had
cystitis occur annually in the United
States,1 and the annual costs of caring for
these infections in young women are
thought to exceed $1 billion.2 Thus, improved means From the Departments of Medicine (T.M.H., C.W., P.L.R., A.E.S.,
of preventing acute cystitis could lead to important W.E.S.), Biostatistics (J.P.H.), Epidemiology (A.S.), and Pharmacy (A.S.),
reductions in both morbidity and health care costs. University of Washington School of Medicine; and the Center for Health
Studies, Group Health Cooperative of Puget Sound (D.S.) — both in Se-
Factors that may influence the risk of urinary tract in- attle. Address reprint requests to Dr. Hooton at Harborview Medical Cen-
fection include recent sexual intercourse,3-7 use of a ter Madison Clinic, 1001 Broadway (Suite 206), Seattle, WA 98122.

468  Aug u s t 1 5 , 1 9 9 6

The New England Journal of Medicine

Downloaded from nejm.org on December 2, 2022. For personal use only. No other uses without permission.
Copyright © 1996 Massachusetts Medical Society. All rights reserved.
 R I S K FAC TO R S FO R SY M P TO M AT I C U R I N A RY T R AC T I N F E C T I O N I N YO U N G WO M E N

used systemic antimicrobial agents within the previous 14 days, or ture-confirmed and probable urinary tract infections were com-
if they had a known anatomical or functional abnormality of the bined for analyses. Contraceptive use was based on data reported
urinary tract. in the diaries, except oral-contraceptive use, which was assumed
At the University of Washington, potentially eligible women to be used daily as prescribed. Because of the differences between
were referred to the study nurse by the family-planning counselor. subjects at the two study sites, all analyses were conducted sepa-
Most such women had not started their new method of contra- rately for women at each site.
ception at the time of study enrollment and began it at the initial Univariate analyses were performed for age, marital status, race
visit with the study nurse. At the Group Health Cooperative, po- or ethnic group, history of urinary tract infections, ABO-blood-
tentially eligible women were identified through a review of an group secretor status, and sexual and contraceptive practices. The
automated pharmacy data base.16 Eligible subjects were sent let- incidence rates of urinary tract infection were determined by di-
ters of invitation, contacted by telephone, and if they agreed to viding the total number of urinary tract infections during follow-
participate, scheduled for a clinic visit. Many of these women had up by the total number of person-years at risk.
begun their new method of contraception at the time of their en- Multivariate analyses of the risk of urinary tract infection were
rollment in the study. For both groups of women, the starting based on data reported in the diaries for the frequency of both
date of the study was considered to be the date of enrollment. contraceptive use and intercourse. The Cox proportional-hazards
The study was approved by the University of Washington and model20 was used to determine the relative contribution of vari-
Group Health Cooperative human-subjects review committees, ous factors to the risk of urinary tract infection. We used the
and all subjects gave written informed consent. counting-process formulation of the Cox model21 in order to in-
clude multiple urinary tract infections per woman, and we used
Evaluation of Subjects the robust variance correction of Lin22 to control for possible cor-
relations within subjects. Time zero for the Cox model was taken
At the initial visit, all subjects were interviewed with a stand- to be the beginning of the study, and real calendar time was used
ardized questionnaire and submitted a midstream urine specimen as the time scale for all analyses. A woman was considered not to
for evaluation of bacteriuria and pyuria as well as blood and saliva be at risk for a new urinary tract infection for five days after the
for blood typing and determination of ABO-blood-group secre- diagnosis of a urinary tract infection.
tor status. At both sites, the subjects were provided with a daily
diary and instructions to indicate in the diary the days on which
the following occurred: sexual intercourse; use of contraception,
with the type indicated; postcoital voiding within one hour of co-
itus; and vaginal and urinary symptoms. The university subjects
were asked to return to the clinic for a directed history taking,
diary review, and urine culture each month or whenever urinary TABLE 1. CHARACTERISTICS OF THE STUDY
symptoms developed during follow-up. The HMO subjects were SUBJECTS WHO MADE AT LEAST ONE
seen every two months for a history taking and midstream urine FOLLOW-UP VISIT.*
collection. At the initial visit, the HMO subjects were also in-
structed in the use and handling of Oxoid dipslides (Unipath,
Ogdensburg, N.Y.). At each clinic visit, the diary was examined UNIVERSITY HMO
by the research nurse to encourage compliance. In alternate SUBJECTS SUBJECTS
months, each subject was interviewed by telephone to collect in- CHARACTERISTIC AT ENROLLMENT (N  348) (N  448)
formation about episodes of urinary tract infection since the last percent
visit, and she was asked to return a urine dipslide to the clinic for
the determination of bacteriuria. The HMO subjects were asked Age
to see their primary care providers for evaluation and treatment 18–22 yr 62 11
if they had urinary symptoms. Subjects at both sites were followed 23–30 yr 31 49
for six months. 31–40 yr 7 40
Marital status
Laboratory Studies Married 10 61
Never married 85 30
Midstream urine specimens for cultures of aerobic bacteria Other 4 9
were collected from the subjects according to previously de- Race or ethnic group
scribed methods.17 The microbiologic procedures used to isolate White 79 88
Black 2 6
and identify organisms from urine cultures have also been de-
Asian or Pacific Islander 10 3
scribed previously.17 ABO-blood-group secretor status was deter- Other† 9 3
mined in a blinded fashion according to standard methods.18
History of any urinary tract 28 58
infection
Definition of Urinary Tract Infection
History of 2 urinary tract 14 38
Subjects were considered to have a culture-confirmed urinary infections
tract infection if they had dysuria, frequency, or urgency (or all ABO-blood-group secretor 75 76
three) and 102 colony-forming units of a uropathogen per mil- Contraceptive method started
liliter of midstream urine.19 Subjects were considered to have a at study enrollment
probable urinary tract infection if they had dysuria, frequency, or Oral contraceptive 52 29
urgency (or all three) and they had been given a diagnosis of a Diaphragm and spermicide 30 46
Cervical cap 15 15
urinary tract infection by their provider in the absence of a urine
Spermicide alone‡ 3 9
culture or, in the absence of provider documentation of a urinary
tract infection, there was documentation of treatment with a uri- *Because of rounding not all categories total 100
nary antimicrobial agent. percent.
†This category includes Native Americans, His-
Statistical Analysis panics, and other groups.
Only subjects who returned to the clinic for at least one follow- ‡Spermicide alone refers to the use of spermicide
up visit were included in the analyses. Except where specified, cul- in the absence of a diaphragm or cervical cap.

Vol ume 335 Numbe r 7  469

The New England Journal of Medicine

Downloaded from nejm.org on December 2, 2022. For personal use only. No other uses without permission.
Copyright © 1996 Massachusetts Medical Society. All rights reserved.
 The New England Journal o f Me di c i ne

TABLE 2. INCIDENCE OF URINARY TRACT INFECTION ACCORDING TO SELECTED

CHARACTERISTICS.*

CHARACTERISTIC UNIVERSITY SUBJECTS (N  348) HMO SUBJECTS (N  448)

NO. OF INCIDENCE NO. OF INCIDENCE
NO. OF URINARY PER PERSON- NO. OF URINARY PER PERSON-
PERSON- TRACT YR OF PERSON- TRACT YR OF
YR INFECTIONS EXPOSURE YR INFECTIONS EXPOSURE

Age
18–22 yr 92 61 0.7 20 10 0.5
23–30 yr 45 35 0.8 86 44 0.5
31–40 yr 9 2 0.2 70 28 0.4
Marital status
Married 15 7 0.5 109 36 0.3
Never married 126 88 0.7 52 32 0.6
Other 5 3 0.6 15 14 0.9
Race
White 116 84 0.7 156 68 0.4
Nonwhite 31 14 0.5 21 14 0.7
History of urinary tract
infection
No previous urinary tract 107 53 0.5 74 21 0.3
infections
1 urinary tract infection in 20 12 0.6 36 16 0.4
lifetime
2 urinary tract infections 20 33 1.6 64 42 0.7
in lifetime
ABO-blood-group secretor
Yes 109 74 0.7 131 59 0.4
No 37 24 0.6 40 22 0.6
Oral-contraceptive use† 75 33 0.4 49 18 0.4
No. of days with intercourse
in past 7 days†
0 40 10 0.2 54 6 0.1
1 30 14 0.5 42 18 0.4
2 26 17 0.6 27 19 0.7
3–7 36 45 1.2 26 25 1.0
No. of days diaphragm and
spermicide used in past
7 days†
0 108 56 0.5 116 44 0.4
1 12 5 0.4 18 12 0.7
2 7 7 1.0 9 7 0.8
3–7 6 18 3.0 6 5 0.8
No. of days cervical cap used
in past 7 days†
0 121 74 0.6 136 59 0.4
1 4 5 1.2 5 3 0.6
2 4 6 1.5 3 0 0.0
3–7 4 1 0.2 3 6 2.0
No. of days spermicide alone
used in past 7 days†
0 124 79 0.6 136 53 0.4
1–7 9 7 0.8 12 15 1.2

*For comparison, the incidence among women who were taking oral contraceptives; did not have
intercourse or use a diaphragm and spermicide, a cervical cap, or spermicide alone in the past 7 days;
and had a history of one or fewer urinary tract infections was 0.05 (20.8 person-years) in the univer-
sity cohort and 0.09 (11.8 person-years) in the HMO cohort.
†This information was obtained from the subjects’ diaries. The categories marked with a dagger
are not mutually exclusive.

RESULTS 146 person-years of follow-up in the university co-

Characteristics of the Study Population
hort and 177 person-years in the HMO cohort. The
women enrolled in the HMO were older (mean age,
A total of 363 university women and 456 HMO 29 years, vs. 23 years in the university cohort), more
members were enrolled in the study, and of these, likely to be married, more likely to be white, more
348 (96 percent) and 448 (98 percent), respectively, likely to have had a previous urinary tract infection,
made at least one follow-up visit. Overall, there were and more likely to be starting contraception with a

470  Aug u s t 1 5 , 1 9 9 6

The New England Journal of Medicine

Downloaded from nejm.org on December 2, 2022. For personal use only. No other uses without permission.
Copyright © 1996 Massachusetts Medical Society. All rights reserved.
 R I S K FAC TO R S FO R SY M P TO M AT I C U R I N A RY T R AC T I N F E C T I O N I N YO U N G WO M E N

diaphragm and spermicide or with spermicide alone

at study enrollment (Table 1). Diary information was TABLE 3. RELATIVE RISK OF URINARY TRACT INFECTION,
AS DETERMINED BY FITTING SEPARATE COX MODELS
complete for 98 percent of the follow-up days in the
FOR EACH COHORT.*
university cohort and 97 percent of the follow-up
days in the HMO cohort. University women report-
ed a median of 6.4 episodes of sexual intercourse per COVARIATE RELATIVE RISK (95% CI) P VALUE
month, as compared with a median of 5.0 in the HMO cohort†
HMO cohort. Married 0.44 (0.25–0.75) 0.002
History of 2 urinary tract infections 2.10 (1.23–3.57) 0.006
Incidence of Urinary Tract Infections at study entry
Any use of spermicide alone in past 3.24 (1.63–6.47) 0.001
Ninety-eight urinary tract infections (range, 0 to 7 days
3 per subject) occurred in the university cohort, as University cohort‡
compared with 82 (range, 0 to 7 per subject) in the Married 0.48 (0.21–1.13) 0.09
HMO cohort. Eighty-eight (90 percent) of the uri- History of 2 urinary tract infections 5.58 (3.24–9.63) 0.001
at study entry
nary tract infections among the university cohort Any use of spermicide alone in past 0.77 (0.38–1.55) 0.47
were culture-confirmed, as compared with 49 (60 7 days
percent) among the HMO cohort. The incidence of
*Additional covariates in the model are given in Figure 1. CI denotes
urinary tract infections (culture-confirmed or prob- confidence interval.
able infections) was 0.7 per person-year of follow-up †The analysis included only the 66 urinary tract infections in women
in the university cohort and 0.5 in the HMO co- who used oral-contraceptive pills, a diaphragm with spermicide, a cervical
cap, or spermicide alone and who provided diary information.
hort. Escherichia coli was the only pathogen or a co-
‡The analysis included only the 86 urinary tract infections in women
existing pathogen in 127 of the culture-confirmed who used oral-contraceptive pills, a diaphragm with spermicide, a cervical
urinary tract infections (93 percent). The rates of cap, or spermicide alone and who provided diary information.
urinary tract infection according to selected charac-
teristics are shown in Table 2.
Multivariate Analyses
Qualitatively, the results of multivariate models fection was also strongly associated with the risk
for the university cohort and HMO cohort were re- of urinary tract infection (relative risk, 5.58 in the
markably similar. Thus, for ease of comparison, we university cohort and 2.10 in the HMO cohort;
fit the same model to each study population sepa- P0.001 and P  0.006, respectively). Increasing
rately and present the results for categorical variables age was associated with a reduced risk of urinary
in Table 3 and for continuous variables in Figure 1. tract infection in each cohort but was significantly
Only women with complete data on all the covari- related only in the university cohort (Fig. 1). Simi-
ates of interest were included in the multivariate larly, being married was associated with a reduced
models. As a result, 86 of 98 urinary tract infections and almost identical relative risk in both cohorts,
in the university cohort and 66 of 82 urinary tract but the reduction in risk was statistically significant
infections in the HMO cohort were included in the only in the HMO cohort (P 0.002) (Table 3), pri-
multivariate models. marily because so few of the university women were
Recent diaphragm and spermicide use, recent sex- married (Table 1). Use of a cervical cap did not have
ual intercourse, and a history of recurrent urinary a statistically significant association with the risk of
tract infection were independent risk factors for uri- urinary tract infection. Although there was a trend
nary tract infection among women at each study site in both cohorts toward an increased risk with fre-
(Table 3 and Fig. 1). In both cohorts, use of a dia- quent cap use (Fig. 1), this association approached
phragm with spermicide had a strong association statistical significance only in the HMO cohort
and dose–response relation with the risk of urinary (P0.08).
tract infection (relative risks for one, three, and five The only important difference between the two
days of use in the past week were, respectively, 1.42, study groups in terms of risk factors for urinary tract
2.83, and 5.68 in the university cohort, P0.001; infection involved the use of spermicide alone as the
and 1.29, 2.14, and 3.54 in the HMO cohort, contraceptive method of choice. In the university
P  0.04). Similarly, there was a strong association cohort, spermicide use in the past week had no sig-
and a dose–response relation between recent sexual nificant effect on the risk of urinary tract infection,
intercourse and the risk of urinary tract infection whereas in the HMO cohort, spermicide use in the
(relative risks for one, three, and five days with inter- past week significantly increased the risk of urinary
course in the past week were, respectively, 1.37, tract infection (P0.001) (Table 3). Because of the
2.56, and 4.81 in the university cohort, P0.001; small number of women in these cohorts who used
and 1.24, 1.91, and 2.96 in the HMO cohort, spermicide alone as their contraceptive, we were un-
P  0.002). A history of recurrent urinary tract in- able to estimate a dose–response effect of spermicide

Vol ume 335 Numbe r 7  471

The New England Journal of Medicine

Downloaded from nejm.org on December 2, 2022. For personal use only. No other uses without permission.
Copyright © 1996 Massachusetts Medical Society. All rights reserved.
 The New England Journal of Me di c i ne

University cohort
10.0 10.0
HMO cohort
P0.001
P0.001
Relative Risk

Relative Risk
P  0.04
P  0.002
1.0 1.0

0.1 0.1
0 1 2 3 4 5 0 1 2 3 4 5

Days with Intercourse in Past 7 Days Days Diaphragm and Spermicide Used
in Past 7 Days

2.0
10.0
Relative Risk

Relative Risk
P  0.08
1.0
P  0.53
P  0.35
1.0

P  0.04

0.5

0.1
0 1 2 3 4 5 20 25 30

Days Cervical Cap Used in Past 7 Days Age (yr)

Figure 1. Relative Risk of Urinary Tract Infection, as Determined by Fitting Separate Cox Models for Each Study Site for the Con-
tinuous Variables of Age, Days of Diaphragm Use with Spermicide in the Past Seven Days, Days with Sexual Intercourse in the Past
Seven Days, and Days of Cervical-Cap Use in the Past Seven Days.
The bars indicate the 95 percent confidence intervals. Additional covariates in the model are given in Table 3.

as we did with the use of a diaphragm with spermi- fection decreased from 2.1 in the full analysis to 1.4
cide, cervical cap, and intercourse. in this subanalysis (P0.25).
Other potential risk factors, including blood-group
secretor status, race or ethnic group (white vs. oth- DISCUSSION
er), parity, time of voiding after intercourse, and The overall incidence of symptomatic urinary
having a new sexual partner since the last visit, were tract infections in the two populations studied was
not statistically significant when added to the model 0.5 per person-year in the HMO cohort and 0.7 per
shown in Table 3 and Figure 1 (P0.20 for each person-year in the university cohort. Although the
variable at each site). incidence among the young, largely unmarried uni-
versity students was higher than among the older,
Multivariate Analyses with Only Culture-Confirmed largely married HMO members, the difference was
Urinary Tract Infections
not dramatic and suggests that neither the incidence
Using the model shown in Table 3 and Figure 1, of nor risk factors for urinary tract infection seen in
we repeated the analyses using only the subgroup of university women are unique. We are not aware of
patients with culture-confirmed urinary tract infec- other prospective studies with which these incidence
tions. These analyses yielded essentially the same re- data can be compared. Although the HMO popula-
sults as those described above (data not shown), tion at the Group Health Cooperative is representa-
except that in the HMO cohort, the relative risk as- tive of all women in the Puget Sound region,15 our
sociated with a history of recurrent urinary tract in- study subjects were not randomly selected and had

472  Augus t 1 5 , 1 9 9 6

The New England Journal of Medicine

Downloaded from nejm.org on December 2, 2022. For personal use only. No other uses without permission.
Copyright © 1996 Massachusetts Medical Society. All rights reserved.
 R I S K FAC TO R S FO R SY M P TO M AT I C U R I N A RY T R AC T I N F E C T I O N I N YO U N G WO M E N

past seven days has a risk of urinary tract infection

TABLE 4. PREDICTED RELATIVE RISK OF URINARY that is 2.6 times greater than that of a similar stu-
TRACT INFECTION FOR A 24-YEAR-OLD UNMARRIED
WOMAN FROM THE UNIVERSITY COHORT BASED ON
dent who has not had intercourse in the previous
THE MODEL IN TABLE 3 AND FIGURE 1.* week; if she had had intercourse daily for the past
week, the risk of urinary tract infection would be
9 times greater (Table 4).
NO. OF DAYS WITH
INTERCOURSE IN THE NO. OF DAYS DIAPHRAGM AND SPERMICIDE The use of a diaphragm with spermicide has been
PAST 7 DAYS USED IN THE PAST 7 DAYS associated with an increased risk of urinary tract in-
0 1 2 3 4 fection in several retrospective case–control stud-
relative risk of urinary tract infection ies.4-9 Our results confirm these earlier findings by
demonstrating the independence of this association
0 1.0 1.4
and, as with sexual intercourse, by demonstrating a
1 1.4 1.9
2 1.9 2.6 3.8
strong dose–response relation between the use of a
3 2.6 5.1 7.3 diaphragm and spermicide and the risk of urinary
4 3.5 14.1 tract infection. Our finding that spermicide use alone
5 4.8 was associated with urinary tract infection in the
6 6.6 HMO cohort is consistent with the results of our
7 9.0 previous studies, which showed that exposure to
spermicides increased the risk of vaginal coloniza-
*Although the model can predict the risk for any arbitrary
number of days with intercourse and contraceptive use, only tion14 and bacteriuria25 with E. coli, but not to the
cells for which we have at least 365 person-days of data are degree seen with the use of a diaphragm and sperm-
filled in. A lifetime history of more than one urinary tract in- icide. Spermicides probably increase the risk of uri-
fection multiplies the relative risk by 5.6. A qualitatively sim-
ilar pattern is seen in the HMO cohort, although the rela- nary tract infection by altering the vaginal environ-
tive-risk estimates are somewhat smaller (Table 3 and Fig. 1). ment in favor of colonization with uropathogens,
thus predisposing women who use these products to
urinary tract infection.26
This study also addressed the relation of cervical-
to meet specific enrollment criteria. Therefore, our cap use to urinary tract infection. Although we did
incidence data cannot be readily generalized to the not find that use of the cervical cap (even though it
population at large. However, given that in 1990 ap- is usually used in conjunction with spermicidal jelly)
proximately 53 million adolescent girls and women was significantly associated with urinary tract infec-
15 to 44 years old reported being sexually experi- tion, there was a trend in this direction in both study
enced in the United States,23 our incidence data sug- cohorts and, with a larger sample, a significant effect
gest that urinary tract infections in young women may have been detected. The weaker association of
constitute a major source of morbidity and health urinary tract infection with cervical-cap use may re-
care costs that may have been considerably underes- flect the fact that less spermicide is used with the
timated by information based on office visits.1 cap than with the diaphragm and it is better con-
In previous prospective studies of the relation be- tained, which may result in less alteration of the vag-
tween sexual intercourse and bacteriuria, Nicolle et inal flora and, thus, a lower risk of urinary tract in-
al. found that 15 of 19 episodes of bacteriuria (79 fection.
percent) developed within 24 hours of intercourse,3 As in two case–control studies,4,5 we demonstrated
whereas Kunin et al. found no association between a relation between a history of urinary tract infection
coital frequency and the presence of bacteriuria.24 and current urinary tract infection. This strong asso-
However, both studies were small, neither controlled ciation could reflect a biologic predisposition to uri-
for other possible risk factors such as contraceptive nary tract infection among certain women due to
method, and both used bacteriuria rather than symp- their ABO-blood-group nonsecretor status,27 an im-
tomatic infection as an end point. Several case–con- mune hyporesponsiveness to E. coli antigens,28 or oth-
trol studies have also demonstrated an association er factors. Alternatively, persistent abnormalities in
between coital frequency and the risk of urinary the vaginal microflora, resulting from either a prior
tract infection.4,5,8 Our prospective study confirms infection or prior antimicrobial therapy,29 or behav-
the results of both Nicolle et al.3 and these case– ioral factors could underlie this association. Of these
control studies in that we found a strong independ- factors, our data suggest that the nonsecretor pheno-
ent association and a dose–response relation be- type is not independently associated with an in-
tween sexual intercourse and the risk of urinary tract creased overall risk of sporadic urinary tract infection
infection. For example, our model (Table 3 and Fig. in women. However, there were too few recurrent
1) predicts that an unmarried, 24-year-old female urinary tract infections in the study to allow us to de-
university student who has had sexual intercourse termine the association of recurrent infection with
without a diaphragm and spermicide on three of the nonsecretor status or other factors. Further studies

Vol ume 335 Numbe r 7  473

The New England Journal of Medicine

Downloaded from nejm.org on December 2, 2022. For personal use only. No other uses without permission.
Copyright © 1996 Massachusetts Medical Society. All rights reserved.
 The New England Journal of Me di c i ne

are needed to ascertain the basis for recurrent uri- 9. Fihn SD, Latham RH, Roberts P, Running K, Stamm WE. Association
between diaphragm use and urinary tract infection. JAMA 1985;254:240-5.
nary tract infection in young women. 10. Adatto K, Doebele KG, Galland L, Granowetter L. Behavioral factors
The conclusions and generalizability of our find- and urinary tract infection. JAMA 1979;241:2525-6.
ings are strengthened by the prospective study de- 11. Foxman B, Frerichs RR. Epidemiology of urinary tract infection. II.
Diet, clothing, and urination habits. Am J Public Health 1985;75:1314-7.
sign,30 the daily recording of information of interest, 12. Kinane DF, Blackwell CC, Brettle RP, Weir DM, Winstanley FP, Elton
the high follow-up rate over the six-month period of RA. ABO blood group, secretor state, and susceptibility to recurrent uri-
nary tract infection in women. BMJ 1982;285:7-9.
observation, and the similarity of the findings in two 13. Sheinfeld J, Schaeffer AJ, Cordon-Cardo C, Rogatko A, Fair WR. As-
different groups of women, one of which is repre- sociation of the Lewis blood-group phenotype with recurrent urinary tract
sentative of the population in the Puget Sound re- infections in women. N Engl J Med 1989;320:773-7.
14. Hooton TM, Roberts PL, Stamm WE. Effects of recent sexual activity
gion. Further studies should address how sexual and and use of a diaphragm on the vaginal microflora. Clin Infect Dis 1994;
contraceptive practices might be modified in young 19:274-8.
women to decrease the associated risk of urinary 15. Pearson DC, Grothaus LC, Thompson RS, Wagner EH. Smokers and
drinkers in a health maintenance organization population: lifestyles and
tract infection. health status. Prev Med 1987;16:783-95.
16. Saunders KW, Stergachis A, Von Korff M. Group Health Cooperative
of Puget Sound. In: Strom BL, ed. Pharmacoepidemiology. 2nd ed. Chi-
Supported by grants (DK 40045 and DK 47549) from the National In- chester, England: John Wiley, 1994:171-85.
stitute of Diabetes and Digestive and Kidney Diseases (to Dr. Stamm). 17. Counts GW, Stamm WE, McKevitt M, Running K, Holmes KK, Turck
M. Treatment of cystitis in women with a single dose of trimethoprim-
We are indebted to Elaine Henley, M.D., and Elaine Jong, M.D., sulfamethoxazole. Rev Infect Dis 1982;4:484-90.
18. Walker RH. Technical manual. 10th ed. Arlington, Va.: American As-
medical directors, Kathleen Slettebak, Ingrid Helsel, R.N., and the sociation of Blood Banks, 1990.
staff at Hall Health Center for helping with patient enrollment; to 19. Stamm WE, Counts GW, Running KR, Fihn S, Turck M, Holmes KK.
Natalie DeShaw for assistance with patient care and data collection Diagnosis of coliform infection in acutely dysuric women. N Engl J Med
at Hall Health Center; and to Cathy Hutchison, Darlene White, 1982;307:463-8.
Sarah McElroy, Jane Grafton, Patty Karlen, Fae Newmann, Wendy 20. Cox DR. Regression models and life-tables. J R Stat Soc [B] 1972;34:
Bensussen-Walls, Sandy Howard, Lillie Stevens, Mary Fors, and Joyce 187-220.
Burgess for helping with patient enrollment and data collection at 21. Andersen PK, Gill RD. Cox’s regression model for counting processes:
the Group Health Cooperative. a large sample study. Ann Stat 1982;10:1100-20.
22. Lin DY. Cox regression analysis of multivariate failure time data: the
marginal approach. Stat Med 1994;13:2233-47.
REFERENCES 23. Bureau of the Census. Statistical abstracts of the United States, 1995.
115th ed. National data book. Washington, D.C.: Government Printing
1. Schappert SM. National ambulatory medical care survey: 1992 summa- Office, 1995:82.
ry. Advanced data from vital and health statistics. No. 253. Hyattsville, 24. Kunin CM, Polyak F, Postel E. Periurethral bacterial flora in women:
Md.: National Center for Health Statistics, 1994. (DHHS publication no. prolonged intermittent colonization with Escherichia coli. JAMA 1980;
(PHS) 94-1250.) 243:134-9.
2. Johnson JR, Stamm WE. Diagnosis and treatment of acute urinary tract 25. Hooton TM, Hillier S, Johnson C, Roberts PL, Stamm WE. Escheri-
infections. Infect Dis Clin North Am 1987;1:773-91. [Erratum, Infect Dis chia coli bacteriuria and contraceptive method. JAMA 1991;265:64-9.
Clin North Am 1990;4:xiia.] 26. Hooton TM, Fennell CL, Clark AM, Stamm WE. Nonoxynol-9: dif-
3. Nicolle LE, Harding GKM, Preiksaitis J, Ronald AR. The association of ferential antibacterial activity and enhancement of bacterial adherence to
urinary tract infection with sexual intercourse. J Infect Dis 1982;146:579- vaginal epithelial cells. J Infect Dis 1991;164:1216-9.
83. 27. Stapleton A, Nudelman E, Clausen H, Hakomori S, Stamm WE. Bind-
4. Strom BL, Collins M, West SL, Kreisberg J, Weller S. Sexual activity, ing of uropathogenic Escherichia coli R45 to glycolipids extracted from vag-
contraceptive use, and other risk factors for symptomatic and asymptomatic inal epithelial cells is dependent on histo-blood group secretor status.
bacteriuria: a case-control study. Ann Intern Med 1987;107:816-23. J Clin Invest 1992;90:965-72.
5. Remis RS, Gurwith MJ, Gurwith D, Hargrett-Bean NT, Layde PM. 28. Hopkins WJ, Xing Y, Dahmer LA, Balish E, Uehling DT. Western blot
Risk factors for urinary tract infection. Am J Epidemiol 1987;126:685-94. analysis of anti-Escherichia coli serum immunoglobulins in women suscep-
6. Foxman B, Frerichs RR. Epidemiology of urinary tract infection. I. Di- tible to recurrent urinary tract infections. J Infect Dis 1995;172:1612-6.
aphragm use and sexual intercourse. Am J Public Health 1985;75:1308-13. 29. Hooton TM, Stamm WE. The vaginal flora and urinary tract infec-
7. Foxman B, Chi J-W. Health behavior and urinary tract infection in col- tions. In: Mobley HLT, Warren JW, eds. Urinary tract infections: molecular
lege-aged women. J Clin Epidemiol 1990;43:329-37. pathogenesis and clinical management. Washington, D.C.: American Soci-
8. Foxman B, Geiger AM, Palin K, Gillespie B, Koopman JS. First-time ety for Microbiology, 1996:67-94.
urinary tract infection and sexual behavior. Epidemiology 1995;6:162-8. 30. Taubes G. Epidemiology faces its limits. Science 1995;269:164-9.

MASSACHUSETTS MEDICAL SOCIETY REGISTRY ON CONTINUING MEDICAL EDUCATION

To obtain information about continuing medical education courses in New England, call
between 9 a.m. and 12 noon, Monday through Friday, (617) 893-4610, or in Massachusetts,
1-800-322-2303, ext. 1342.

474  Augus t 1 5 , 1 9 9 6

The New England Journal of Medicine

Downloaded from nejm.org on December 2, 2022. For personal use only. No other uses without permission.
Copyright © 1996 Massachusetts Medical Society. All rights reserved.