You are on page 1of 8

Vol. 161, No.

6
Printed in U.S.A.
DOI: 10.1093/aje/kwi078

American Journal of Epidemiology


Copyright 2005 by the Johns Hopkins Bloomberg School of Public Health
All rights reserved

Risk of Urinary Tract Infection and Asymptomatic Bacteriuria among Diabetic and
Nondiabetic Postmenopausal Women

Edward J. Boyko1,2, Stephan D. Fihn2,3, Delia Scholes4,5, Linn Abraham5, and Barbara Monsey5
1

Epidemiologic Research and Information Center, VA Puget Sound Health Care System, Seattle, WA.
Department of Medicine, School of Medicine, University of Washington, Seattle, WA.
3
Northwest Health Services Research and Development Program, VA Puget Sound Health Care System, Seattle, WA.
4
Department of Epidemiology, School of Public Health and Community Medicine, University of Washington, Seattle, WA.
5
Center for Health Studies, Group Health Cooperative of Puget Sound, Seattle, WA.
2

Received for publication February 2, 2004; accepted for publication October 26, 2004.

diabetes mellitus; postmenopause; prospective studies; risk factors; urinary tract infection; women

Abbreviations: GHC, Group Health Cooperative of Puget Sound; UTI, urinary tract infection.

Urinary tract infection (UTI) accounts for considerable


morbidity among adult women; it is responsible for 3.6
million office visits annually in the United States and direct
costs of $1.6 billion (1, 2). The belief that diabetes mellitus,
which is estimated to affect 16 million persons in the United
States, predisposes to UTI is widespread (3, 4). Diabetes
causes several abnormalities of the host defense system that
might result in a higher risk of certain infections, including
UTI (5). These include immunologic impairments, such as
impaired migration, intracellular killing, phagocytosis, and
chemotaxis of polymorphonuclear leukocytes from diabetic
patients (6), and neuropathic complications, such as impaired bladder emptying (7). In addition, a higher glucose

concentration in the urine may create a culture medium for


pathogenic microorganisms.
Although the relation between diabetes and asymptomatic bacteriuria has been the subject of several controlled
studies, the association between diabetes and UTI risk has
not been examined until recently. A case-control study of
postmenopausal women demonstrated 2.2-fold relative odds
of microbiologically confirmed UTI in relation to diabetes
mellitus. This increase in risk was confined mainly to
women undergoing pharmacologic treatment for diabetes
(8). A randomized controlled trial of the effects of hormone
therapy on coronary heart disease events among 2,763
postmenopausal women aged 4479 years with established

Correspondence to Dr. Edward J. Boyko, VA Puget Sound Health Care System (S-152E), 1660 South Columbian Way, Seattle, WA 98108
(e-mail: eboyko@u.washington.edu).

557

Am J Epidemiol 2005;161:557564

Downloaded from http://aje.oxfordjournals.org/ by guest on April 2, 2015

No prospective data exist on the risk of microbiologically conrmed urinary tract infection (UTI) and
asymptomatic bacteriuria (AB) in relation to diabetes mellitus and its characteristics. The authors prospectively
(19982002) followed 218 diabetic and 799 nondiabetic Washington State women aged 5575 years for UTI and
AB. The baseline examination and two annual follow-up examinations included urine culture, measurement of
hemoglobin A1c and postvoid residual bladder volume, and a survey of diabetes and other characteristics.
Surveillance for UTI included self-reports conrmed by microbiologic culture and medical record review. UTI
incidence per 100 person-years was 12.2 for diabetic women and 6.7 for nondiabetic women (relative risk (RR)
1.8, 95% condence interval (CI): 1.2, 2.7). AB incidence per 100 person-years was 6.7 for diabetic women and
3.0 for nondiabetic women (RR 2.3, 95% CI: 1.3, 3.9). In Cox models adjusted for multiple covariates, the
increased UTI risk occurred mainly in women taking insulin (RR 3.7, 95% CI: 1.8, 7.3) and women with a longer
diabetes duration (10 years; RR 2.6, 95% CI: 1.3, 5.1) compared with nondiabetic women. No clear linear
trend between hemoglobin A1c and UTI or AB risk was seen. Postmenopausal women with diabetes have higher
risks of UTI and AB in relation to diabetes duration and severity but not to recent glucose control.

558 Boyko et al.

coronary heart disease showed a significant 1.4- to 1.8-fold


increase in the risk of physician-diagnosed UTI in association with self-reported diabetes (9).
Given that a uniform screening process was not used in
either of those studies, the possibility exists that a higher
level of UTI detection occurred in women with recognized
diabetes which led to an apparently higher risk of UTI in
relation to this metabolic condition. Other important
unanswered questions include the association between
glycemic control and risk and the role of asymptomatic
bacteriuria as a precursor to UTI in diabetic women.
To address these issues, we conducted a prospective study
of postmenopausal women to investigate the relation of
clinically confirmed diabetes to the risks of UTI and
asymptomatic bacteriuria.
MATERIALS AND METHODS

The study was conducted at the Group Health Cooperative of Puget Sound (GHC), a staff-model nonprofit health
maintenance organization with approximately 450,000
enrollees. Women aged 5575 years were eligible to
participate if they resided in Pierce County, King County,
or Snohomish County in Washington State, had been
enrolled in GHC for at least 1 year, had not had a natural
menstrual cycle in the preceding 12 months, and had not had
a UTI in the preceding 90 days.
Women were randomly selected from the GHC enrollment file. We also selected additional women from the GHC
Diabetes Registry who were frequency-matched, by age, to
the main cohort. Exclusion criteria included residential
nursing care, wheelchair dependency, dementia, severe
psychiatric disorder, urinary catheterization, renal dialysis,
active nonskin cancer, and chronic antibiotic use. After
checking GHC databases for indicators of ineligibility, we
contacted potential participants by letter and telephone to
determine their eligibility and willingness to participate. We
made up to 10 attempts to reach each woman by phone. The
study protocol was approved by the human subjects
committees of the University of Washington and GHC.
Signed informed consent was obtained from all participants.
Women who participated were followed prospectively for
2 years. Exposure and outcome data were collected from
multiple sources: interviews and research clinic visits
conducted at baseline and at 12 and 24 months and GHC
laboratory, hospital, and pharmacy data (for identification
and verification of UTI and diabetes).
Primary exposures of interest included diabetes presence
and characteristics, demographic features, sexual activity,
and bladder function. The primary outcome of interest was
acute UTI, as defined by a midstream urine specimen yielding 100,000 colony-forming units per ml of a uropathogenic organism (any aerobic gram-negative rod, Lancefield
group B or D streptococci, or Staphylococcus saprophyticus), accompanied by dysuria, urgency, or frequency for 2
weeks or less.
We collected interview data at baseline and at 12 and 24
months between 1998 and 2002 to assess history of UTI,

Symptomatic UTI surveillance

Participants were instructed to contact research personnel


if any new symptom occurred that suggested UTI. A
standardized interview was then conducted, and, if infection
was thought to be likely (presence of dysuria, urgency, or
frequency), the patient was asked to provide an Oxoid dipslide (Unipath, Ogdensburg, New York) urine specimen via
mail, which was processed to identify and quantify uropathogens. The subject was also referred to her primary
physician for evaluation.
In addition, surveillance of GHC clinical records was
performed on a monthly basis to capture episodes of
symptomatic UTI not reported to research staff. GHC
diagnosis codes generated by outpatient clinical encounters
and laboratory databases were searched for urinalysis results
suggestive of infection (trace blood or any bacteria) or urine
culture results indicative of infection; if such results were
present, these women were interviewed by telephone for
the presence of UTI symptoms. If the women were
Am J Epidemiol 2005;161:557564

Downloaded from http://aje.oxfordjournals.org/ by guest on April 2, 2015

Study setting and participants

diabetes presence and characteristics, and sexual activity.


Women who reported having diabetes were invited to
complete a supplemental questionnaire that obtained data
on age, weight, and height at onset of diabetes, history of
ketoacidosis, and type of diabetes.
During research clinic visits, women underwent measurement of postvoid residual bladder volume using a portable ultrasound device (BladderScan BVI 2500; Diagnostic
Ultrasound Corporation, Seattle, Washington) and submitted clean-catch midstream urine specimens for microbiologic culture. The presence of diabetes was ascertained by
self-report of a physicians diagnosis. In addition, women
who did not self-report diabetes but were included in the
GHC Diabetes Registry were considered to have diabetes.
This registry captures information from GHC laboratory,
pharmacy, and hospital discharge summary databases at
monthly intervals to continuously update the diabetes status
of enrollees. Subjects are entered into this registry for
specified time periods (indicated in parentheses) if they
meet the following criteria: 1) a hospital discharge diagnosis
of diabetes at any time (indefinite); 2) a random plasma
glucose level greater than 200 mg/dl (12 months); 3) a
fasting plasma glucose level greater than 125 mg/dl (12
months); 4) a hemoglobin A1c concentration greater than
7.0 percent (12 months); and 5) receipt of a prescription for
insulin or an oral hypoglycemic agent (3 years). The
presence of diabetes in women from the GHC Diabetes
Registry was confirmed by record review. For identification
of previously undiagnosed diabetes, fasting plasma glucose
was measured at baseline, and a level greater than 125 mg/dl
was defined as diabetes. Women meeting diabetes criteria
underwent high-pressure liquid chromatography testing of
hemoglobin A1c at baseline and at 12 and 24 months
(Variant II hemoglobin A1c; Bio-Rad, Hercules, California).
Women who reported an age of diabetes onset of less than
30 years, one or more episodes of ketoacidosis, and
continuous use of insulin since diagnosis were considered
to have type 1 diabetes.

Diabetes and Urinary Tract Infection Risk

symptomatic, they were asked to send a dip-slide urine


specimen to us for microbiologic culture.
Statistical methods

RESULTS

A total of 799 nondiabetic women and 218 diabetic


women agreed to participate, representing 34 percent and
26 percent of the total number of eligible nondiabetic
and diabetic women, respectively. The random subject
sampling method yielded 65 diabetic women from the
GHC enrollment file, to whom an additional 153 women
were added by random selection from the diabetes registry.
A high proportion of participants completed the 12- and
24-month follow-up visits (87 percent and 81 percent,
respectively).
Diabetic women were more likely to be non-White,
slightly older, less educated, of lower income, and less
sexually active and more likely to have a prior UTI
than nondiabetic women (table 1). Diabetic women had
a greater mean residual bladder volume (49.1 ml vs. 33.1
ml; p 0.022) and were more likely to have a residual
bladder volume greater than 200 ml (7.4 percent vs. 2.8
percent; p 0.003). Forty-six women had asymptomatic
bacteriuria at the baseline examination; frequencies of this
condition were similar by diabetes status (diabetes absent,
4.0 percent; diabetes present, 6.5 percent; p 0.138).
Among nondiabetic women, 71 developed UTI during
follow-up, with 55 (6.9 percent) having one episode, nine
(1.1 percent) having two episodes, and seven (0.9 percent)
having three or more episodes. Among diabetic women,
26 developed UTI, with 16 (7.3 percent) having one
episode, seven (3.2 percent) having two episodes, and three
Am J Epidemiol 2005;161:557564

(1.4 percent) having three or more episodes. Regarding


diabetes characteristics, 30.1 percent of diabetic women had
had the disease for 10 years or more, 1.4 percent had type 1
disease, 48.6 percent took oral hypoglycemic medication,
and 19.0 percent used insulin.
Diabetic women had significantly higher risks of overall
UTI and a first episode of asymptomatic bacteriuria (table 2).
No distinct linear trend was seen between categorized
hemoglobin A1c levels using two different sets of cutpoints
and risk of either asymptomatic bacteriuria or UTI. Higher
risks of UTI and asymptomatic bacteriuria were seen in
diabetic women treated with medication, but statistical
significance was observed only in the insulin-treated
women. Significantly higher risks of asymptomatic bacteriuria and UTI were seen among women who had had
diabetes for 10 or more years.
Multivariable failure-time analyses were performed to
adjust for age, history of previous UTI, sexual activity,
baseline asymptomatic bacteriuria, postvoid residual bladder volume, and ethnicity (table 3). All episodes of UTI and
the first episode of asymptomatic bacteriuria were considered as outcomes. In unadjusted analyses, diabetes was
related to a significantly higher risk of both outcomes, but
this difference diminished for UTI after adjustment. The
significant association between diabetes and higher risk of
asymptomatic bacteriuria remained largely unchanged after
additional adjustment for postvoid residual bladder volume
and ethnicity.
A weak trend was seen between hemoglobin A1c and risk
of asymptomatic bacteriuria or UTI in adjusted models
(table 3). However, no category of hemoglobin A1c was
associated with a significantly elevated risk of UTI (except
for >7.9 percent), while the lowest and highest categories
were associated with a significantly higher risk of asymptomatic bacteriuria in one scheme and the middle and
highest categories were associated in the other scheme. A
significantly higher risk of UTI was seen among insulintreated women. A significantly higher risk of asymptomatic
bacteriuria was seen with insulin treatment but also with
oral medication use in a model adjusted for all covariates.
When oral medication treatment was set as the reference
category, insulin therapy was associated with a significantly
higher risk of UTI compared with oral medication in
a model adjusted for all covariates (relative risk 2.9, 95
percent confidence interval: 1.2, 6.9). Higher risks of UTI
and asymptomatic bacteriuria were seen only among
women with a longer duration of diabetes.
Uropathogens observed in UTI did not differ by diabetes
status (diabetic women vs. nondiabetic women: Escherichia
coli, 74.4 percent vs. 75.8 percent; Klebsiella sp., 7.0
percent vs. 6.3 percent; Proteus sp., 7.0 percent vs. 5.3
percent; group B Streptococcus, 2.3 percent vs. 3.2 percent;
Enterococcus sp., 0 percent vs. 5.3 percent; other organisms,
9.3 percent vs. 9.3 percent (p 0.863)). Bacterial pathogens
other than E. coli were more often responsible for the first
observed episode of asymptomatic bacteriuria in diabetic
women (diabetic women vs. nondiabetic women: E. coli,
56.5 percent vs. 78.0 percent; Klebsiella sp., 21.7 percent vs.
9.8 percent; Proteus sp., 0 percent vs. 0 percent; group B
Streptococcus, 4.3 percent vs. 4.9 percent; Enterococcus sp.,

Downloaded from http://aje.oxfordjournals.org/ by guest on April 2, 2015

The primary outcomes of interest were episodes of


asymptomatic bacteriuria and UTI during the follow-up
period. We assessed the statistical significance of differences in means or frequencies of baseline characteristics by
diabetes status using the t test or the chi-squared test (10).
We computed incidence, incidence density ratios, and 95
percent confidence limits for first UTI according to risk
factors of interest using standard methods (11). Cox proportional hazards modeling was used to estimate hazard
ratios for UTI and asymptomatic bacteriuria while adjusting
for covariates. Since multiple occurrences of UTI were
included in these models, we used the Anderson-Gill model
with a correction to the variance for multiple observations
per woman (12, 13). The assumption of proportional
hazards was tested in all Cox models and found to be valid
(14). Women with asymptomatic bacteriuria at baseline
were excluded from analyses of asymptomatic bacteriuria
risk during the 2-year follow-up. We planned the study with
a target sample size of 800 nondiabetic women and 200
diabetic women to achieve 80 percent power to detect
a relative risk of 2.2 in relation to diabetes, assuming an
annual UTI incidence of 3 percent and a two-sided alpha
value of 0.05.

559

560 Boyko et al.

TABLE 1. Characteristics of women with and without diabetes mellitus, Group Health Cooperative of Puget Sound, 19982002
Characteristic

Without diabetes (n 799)


No.

5559

262

32.8

6064

166

20.8

6569

168

7075

203

With diabetes (n 218)


No.

Total (n 1,017)

No.

60

27.5

322

31.7

47

21.6

213

20.9

21.0

47

21.6

215

21.1

25.4

64

29.4

267

26.3

Age (years)

Annual income*
<$10,000

1.1

13

7.2

20

2.4

$10,000<$35,000

245

37.1

77

42.5

322

38.2

$35,000<$50,000

145

21.9

45

24.9

190

22.6

$50,000<$75,000

141

21.3

25

13.8

166

19.7

$75,000

123

18.6

21

11.6

144

17.1

Highest level of education*


High school graduate

23

2.9

15

7.0

38

3.8

340

42.7

125

58.1

465

46.0

College graduate

247

31.0

35

16.3

282

27.9

Graduate or professional degree

186

23.4

40

18.6

226

22.4

Ethnicity*
White

721

90.5

167

77.0

888

87.6

African-American

22

2.8

24

11.1

46

4.5

Asian or Pacic Islander

33

4.1

17

7.8

50

4.9

Other

21

2.6

4.1

30

3.0

501

63.0

134

61.5

635

62.7

99

12.5

32

14.7

131

12.9

160

20.1

46

21.1

206

20.3

35

4.4

2.8

41

4.0

364

45.7

77

35.3

441

43.4

Never

443

56.6

146

67.9

589

59.0

15

258

33.0

58

27.0

316

31.7

6

82

10.5

11

5.1

93

9.3

276

35.0

67

30.9

343

34.1

15

370

46.9

87

40.1

457

45.4

610

83

10.5

29

13.4

112

11.1

1120

34

4.3

16

7.4

50

5.0

21

26

3.3

18

8.3

44

4.4

Marital status
Married or living as married
Widowed
Divorced or separated
Never married
Sexual historyy
Sexual intercourse during past year
Frequency of intercourse per month

History of UTIy,z (no. of past UTIs in lifetime)

UTI during follow-up


No UTI during follow-up

684

85.6

166

76.1

850

83.6

Symptomatic UTI during follow-up

71

8.9

26

11.9

97

9.5

Asymptomatic UTI only

44

5.5

23

10.6

67

6.6

1.4

0.3

UTI, but symptoms unknown


* Nondiabetic women vs. diabetic women: p < 0.001 (chi-squared test).
y Nondiabetic women vs. diabetic women: p < 0.005 (chi-squared test).
z UTI, urinary tract infection.

Am J Epidemiol 2005;161:557564

Downloaded from http://aje.oxfordjournals.org/ by guest on April 2, 2015

11 years

Diabetes and Urinary Tract Infection Risk

561

TABLE 2. Incidence of and incidence ratios for all episodes of symptomatic urinary tract infection and a rst episode of
asymptomatic bacteriuria, by diabetes status and characteristics at the time of study entry, Group Health Cooperative of Puget Sound,
19982002
All symptomatic urinary tract infections
Characteristic

Incidence
per 100 PY

First episode of asymptomatic bacteriuria


Incidence
per 100 PY

No. of
events

PY

41

1,388.1

3.0

1.0y

1.2, 2.7

23

342.8

6.7

2.3

1.7

1.1, 2.7

15

221.2

6.8

2.3

1.2, 4.2

2.3

1.1, 4.4

62.4

8.0

2.7

0.8, 6.9

0.8, 4.6

49.1

4.1

1.4

0.2, 5.3

1.6

1.0, 2.7

11

177.0

6.2

2.1

1.0, 4.2

1.7

0.7, 3.3

77.9

10.3

3.5

1.4, 7.5

1.4, 4.7

77.8

3.9

1.3

0.3, 4.1

No. of
events

PY*

Nondiabetic
(n 799)

95

1,420.8

6.7

Diabetic
(n 215)

43

352.1

12.2

7.5% (n 137)

26

227.4

11.4

7.6%8.5% (n 42)

10

65.5

15.3

49.1

14.3

2.1

20

182.2

11.0

81.1

11.1

14

78.7

17.8

2.7

IR*

95% CI*

IR

95% CI

Diabetes status
1.0y
1.8

1.3, 3.9

Hemoglobin A1c
concentration

>8.5% (n 30)
Hemoglobin A1c
concentration

>7.9% (n 49)
Duration of
diabetes
<10 years (n 149)

19

239.1

7.9

1.2

0.7, 2.0

13

230.6

5.6

1.9

0.9, 3.6

10 years (n 64)

24

112.1

21.4

3.2

2.0, 5.1

10

111.3

9.0

3.0

1.4, 6.2

115.9

4.3

0.7

0.2, 1.6

111.0

6.3

2.1

0.8, 4.8

Oral medications
(n 104)

17

167.7

10.1

1.5

0.9, 2.6

10

164.0

6.1

2.1

0.9, 4.2

Insulin treatment
(n 40)

21

67.6

31.1

4.7

2.8, 7.5

66.9

9.0

3.0

1.1, 7.2

Diabetes treatment
No medications
(n 69)

* PY, person-years; IR, incidence ratio; CI, condence interval.


y Referent.

13.0 percent vs. 4.9 percent; other organisms, 4.3 percent vs.
2.4 percent (p 0.071)).
DISCUSSION

We observed a higher risk of acute symptomatic UTI and


asymptomatic bacteriuria among postmenopausal women
with diabetes. Women taking insulin were mainly those at
higher risk, possibly because of more severe diabetes, since
the use of insulin may be a marker for disease severity (15).
Risk of UTI was higher with increasing duration of diabetes.
The risk of incident asymptomatic bacteriuria was significantly increased in relation to insulin use in all multivariable
models and to oral medication use in one model. Consistent
with one other study, we found no significant association
between glycemic control as assessed by hemoglobin A1c
level and risk of UTI, while the pattern of risk elevation for
the relation of asymptomatic bacteriuria to hemoglobin A1c
was inconsistent with a dose-response effect (16). The high
proportion of diabetic woman-years (67 percent) in the
lowest hemoglobin A1c category (<7.5 percent) may have
Am J Epidemiol 2005;161:557564

impaired our ability to detect this association in this


analysis. Recategorization of hemoglobin A1c values
(tables 2 and 3) to redistribute more woman-years to higher
exposure categories did not produce a clearer dose-response
effect.
The association between diabetes and risks of UTI and
asymptomatic bacteriuria could not be explained by confounding due to frequency of sexual intercourse (a major
risk factor for UTI (17)), a history of UTI, postvoid residual
bladder volume, or ethnicity. The presence of asymptomatic
bacteriuria at baseline did not confound the association
between diabetes characteristics and UTI risk. Although
diabetic persons may be more susceptible to infection by
uncommon organisms, we found most of their infections to
be due to typical uropathogens, which suggests that diabetes
facilitates the same route of infection as that for UTI in
nondiabetic persons (i.e., ascending infection from the
urethra). The finding that asymptomatic bacteriuria more
often involved Klebsiella and Enterococcus in diabetic
women suggests that defenses against these organisms
may be reduced. Geerlings et al. (18) found similar

Downloaded from http://aje.oxfordjournals.org/ by guest on April 2, 2015

7.0% (n 107)
7.0%7.9% (n 53)

562 Boyko et al.

TABLE 3. Hazard ratios for asymptomatic bacteriuria and symptomatic urinary tract infection according to diabetes status and
characteristics, Group Health Cooperative of Puget Sound, 19982002*
First asymptomatic bacteriuria
Model

All symptomatic UTIy

Covariates as measured at baseline


HRy

95% CIy

HR

95% CI

Diabetes status
Diabetes present

Age, sexual activity,z history of previous UTI

1.9

1.1, 3.2

1.5

0.9, 2.5

Diabetes present

Above covariates PRBV,y asymptomatic


bacteriuria

1.9

1.2, 3.3{

1.4

0.8, 2.2

Diabetes present

All of the above covariates ethnicity

2.1

1.2, 3.5{

1.4

0.9, 2.3

7.5%

1.9{

1.0, 3.5

1.2

0.7, 2.2

7.6%8.5%

1.9

0.7, 5.4

1.6

0.9, 3.0

>8.5%

2.5

1.0, 6.6

1.8

0.7, 4.7

HbA1cy concentration at
baseline and 1 year

HbA1c at baseline and 1 year

Age, sexual activity, previous UTI, PRBV,


asymptomatic bacteriuria

Above covariates ethnicity

{
2.0

1.0, 3.6

1.3

0.7, 2.3

7.6%8.5%

2.2

0.8, 6.1

1.8

0.9, 3.3

>8.5%

2.7

1.0, 7.2

1.9

0.7, 4.8

HbA1c concentration at
baseline and 1 year

Age, sexual activity, previous UTI, PRBV,


asymptomatic bacteriuria

7.0%

1.6

0.8, 3.3

1.3

0.7, 2.4

7.0%7.9%

2.3

1.0, 5.5

1.1

0.6, 2.1

>7.9%

2.4

1.1, 5.5

2.0

1.0, 4.1

HbA1c at baseline and 1 year

Above covariates ethnicity

7.0%

1.7

0.8, 3.5

1.3

0.7, 2.5

7.0%7.9%

2.5

1.0, 6.0

1.2

0.6, 2.3

>7.9%

2.6

1.1, 6.0

2.2

1.1, 4.3

Diabetes treatment being


administered

Age, sexual activity, previous UTI, PRBV,


asymptomatic bacteriuria

No medications

1.7

0.8, 3.9

0.5

0.2, 1.2

Oral medications

1.9

0.9, 3.8

1.2

0.6, 2.2

2.4

1.0, 5.9

3.2

1.6, 6.6

Insulin treatment
Diabetes treatment

Above covariates ethnicity

No medications

1.8

0.8, 3.9

0.5

0.2, 1.2

Oral medications

2.1

1.0, 4.3

1.3

0.7, 2.3

2.7

1.1, 6.6

3.7

1.8, 7.3

Insulin treatment
Duration of diabetes
(in years)

Age, sexual activity, previous UTI, PRBV,


asymptomatic bacteriuria

<10

1.7

0.9, 3.1

0.9

0.5, 1.6

10

2.5

1.2, 5.0

2.3

1.2, 4.6

<10

1.8

0.9, 3.3

0.9

0.5, 1.6

10

2.8

1.4, 5.7

2.6

1.3, 5.1

Diabetes duration (years)

Above covariates ethnicity

* The comparison group for all models was nondiabetic women.


y UTI, urinary tract infection; HR, hazard ratio; CI, condence interval; PRBV, postvoid residual bladder volume; HbA1c, hemoglobin A1c.
z Monthly average for sexual intercourse in the past year: 0, 15, 6.
Number of past UTIs: 0, 15, 6.
{ Women with asymptomatic bacteriuria at baseline were excluded from this model. No adjustment was performed for baseline asymptomatic
bacteriuria in models predicting asymptomatic bacteriuria as the outcome.

Am J Epidemiol 2005;161:557564

Downloaded from http://aje.oxfordjournals.org/ by guest on April 2, 2015

7.5%

Diabetes and Urinary Tract Infection Risk

Am J Epidemiol 2005;161:557564

diabetes argues in favor of causation as opposed to


surveillance bias as the explanation for these findings.
The hemoglobin A1c measurements used to assess glucose
control were obtained at baseline and the annual follow-up
visits. It is possible that these measurements did not reflect
glucose control throughout the entire year between follow-up
examinations, though this seems unlikely on the basis of our
data. We observed high correlations between year 1 and year
2 hemoglobin A1c measurements (correlation coefficient
0.64) and between year 2 and year 3 hemoglobin A1c
measurements (correlation coefficient 0.61) and small
mean differences and standard deviations (year 2 minus year
1: 0.4 percent (standard deviation, 1.3); year 3 minus year
2: 0.1 percent (standard deviation, 1.3)). However, the
possibility exists that degree of hyperglycemia was not
accurately captured for UTIs occurring more than 90 days
before the time of hemoglobin A1c measurement and that
more frequent performance of this test would have demonstrated an association between risk of this outcome and
glycemic control.
The exact onset of type 2 diabetes cannot be easily
determined, since symptoms may remain occult for years
(35). Error in estimating diabetes duration due to this
problem is likely to have been nondifferential, thereby
resulting in underestimation of effects associated with this
exposure. Given our screening protocol, it is unlikely that
diabetes was also present among women not reporting this
condition (36). Although our cohort was relatively large for
a study of UTI epidemiology, it is possible that it was of
insufficient size for detection of all clinically significant
associations. It is also possible that the increase in UTI risk
related to diabetes would be even greater in diabetic
populations with poorer glycemic control.
In conclusion, diabetes treated with insulin and diabetes
of longer duration were related to substantial increases in
the risks of UTI and asymptomatic bacteriuria among
postmenopausal women enrolled in the GHC health plan.
Predictors of UTI and asymptomatic bacteriuria were
similar, with the exception that diabetes was significantly
related to risk of asymptomatic bacteriuria but not UTI. The
higher risk of UTI was not explained by poorer glucose
control, higher postvoid residual volume, or asymptomatic
bacteriuria at baseline. UTI and asymptomatic bacteriuria
should be considered complications of diabetes mellitus in
postmenopausal women. These results also argue for diabetes severity and duration as the main determinants of
higher UTI and asymptomatic bacteriuria riska pattern
that resembles the relation between diabetes characteristics
and other complications, such as retinopathy or neuropathy
(36). Therefore, one would expect that improved diabetes
control might yield a reduction in incidence of these
complications, although this remains to be demonstrated
in properly conducted randomized controlled trials.

ACKNOWLEDGMENTS

Funding for this study was provided by National


Institutes of Health grant RO1 DK43134. Drs. Fihn and

Downloaded from http://aje.oxfordjournals.org/ by guest on April 2, 2015

virulence determinants in E. coli isolated from women with


and without diabetes.
Regarding potential mechanisms, we found that diabetic
women had a higher mean postvoid residual bladder
volume, an indicator of bladder dysfunction and possibly
autonomic neuropathy (7). Adjustment of the diabetes-UTI
(and asymptomatic bacteriuria) association by postvoid
residual bladder volume had little effect on these associations. Thus, the main effects of diabetes on risks of
asymptomatic bacteriuria and UTI are probably not mediated by postvoid residual bladder volume.
We found that baseline asymptomatic bacteriuria, a
known risk factor for UTI, did not explain the higher risk
of UTI among diabetic women (19, 20). Adjustment for
asymptomatic bacteriuria in regression models of UTI risk
and diabetes characteristics, including insulin treatment and
duration, had little effect on these hazard ratios.
With the exception of two previous studies that examined
UTI risk in relation to presence of diabetes, much of the
pertinent literature has focused solely on bacteriuria (8, 9).
A number of cross-sectional studies of bacteriuria and
diabetes have been published (21). Of 12 reviewed studies,
75 percent reported a two- to fourfold higher prevalence of
bacteriuria in diabetic subjects as compared with controls
(2233). Two of these studies demonstrated higher bacteriuria prevalence with increasing diabetes duration but no
association with glycosylated hemoglobin level (23, 31),
which is consistent with our findings.
We took precautions to minimize potential bias in this
study. To prevent misclassification of UTI, we used a case
definition that is more than 95 percent specific for this
diagnosis (34) and confirmed the diagnosis using data
from medical records, laboratory files, dip-slide culture,
and patient interviews. To avoid the biases associated
with selection of study subjects from clinical settings, we
randomly selected these women from the entire GHC
enrollment population in the targeted counties. Fewer than
half of the eligible subjects agreed to participate given the
requirement for ongoing surveillance and follow-up examinations. Since only subjects without chronic urinary voiding
conditions or UTI were eligible for recruitment, it is
unlikely that the participation rate caused biased associations between the main exposures and outcomes of interest,
although a possible factitious association between higher
UTI risk and diabetes might have occurred if diabetic
women with a history of UTI were more likely to participate. The proportion of women completing follow-up was
high. Given that the population was mainly Caucasian, the
results may not be generalizable to other ethnic groups.
A major advantage of this study is that it reflects the
influence of diabetes and other risk factors on risks of
UTI and asymptomatic bacteriuria in a population-based
sample of generally healthy, community-dwelling women.
Although similarly intense surveillance methods were employed for UTI detection, it is possible that clinical providers
had a lower threshold for the evaluation of possible UTI
symptoms, thereby leading to a higher rate of this outcome in
diabetic women. This same bias would not apply to the
detection of asymptomatic bacteriuria, and the higher risk of
asymptomatic bacteriuria that we observed in relation to

563

564 Boyko et al.

Boyko received salary support from the US Department of


Veterans Affairs.
The National Institutes of Health had no role in the study
design, data collection, analysis, or interpretation of results
or in the writing of this report.

REFERENCES

Am J Epidemiol 2005;161:557564

Downloaded from http://aje.oxfordjournals.org/ by guest on April 2, 2015

1. Foxman B. Recurring urinary tract infection: incidence and


risk factors. Am J Public Health 1990;80:3313.
2. Schappert SM. Ambulatory care visits to physician offices,
hospital outpatient departments, and emergency departments:
United States, 1997. Vital Health Stat 13 1999(143):iiv, 139.
3. Viberti GC, Walker JD, Pinto J. Diabetic nephropathy. In:
Alberti KG, DeFronzo RA, Keen H, et al, eds. International
textbook of diabetes mellitus. West Sussex, United Kingdom:
John Wiley and Sons Ltd, 1992:13012.
4. Patterson JE, Andriole VT. Bacterial urinary tract infections
in diabetes. Infect Dis Clin North Am 1997;11:73550.
5. Stapleton A. Urinary tract infections in patients with
diabetes. Am J Med 2002;113(suppl 1A):80S4S.
6. Valerius NH, Eff C, Hansen NE, et al. Neutrophil and
lymphocyte function in patients with diabetes mellitus. Acta
Med Scand 1982;211:4637.
7. Hosking DJ, Bennett T, Hampton JR. Diabetic autonomic
neuropathy. Diabetes Care 1978;27:104354.
8. Boyko EJ, Fihn SD, Scholes D, et al. Diabetes and the risk
of acute urinary tract infection among postmenopausal
women. Diabetes Care 2002;25:177883.
9. Brown JS, Vittinghoff E, Kanaya AM, et al. Urinary tract
infections in postmenopausal women: effect of hormone
therapy and risk factors. Obstet Gynecol 2001;98:104552.
10. Armitage P, Berry G. Statistical methods in medical research.
Oxford, United Kingdom: Blackwell Scientific Publications,
1987:1079, 1279, 41116.
11. Kleinbaum D, Kupper L, Morgenstern H. Epidemiologic
research: principles and quantitative methods. Belmont, CA:
Wadsworth Publishing Company, 1982.
12. Lin DY, Wei LJ. The robust inference for the Cox proportional hazards model. J Am Stat Assoc 1989;84:10748.
13. Andersen PK, Gill RD. Coxs regression model for counting
processes: a large sample study. Ann Stat 1982;10:111020.
14. Rothman KJ, Greenland S. Modern epidemiology. 2nd ed.
Philadelphia, PA: Lippincott-Raven, Publishers, 1998.
15. Turner RC, Cull CA, Frighi V, et al. Glycemic control with
diet, sulfonylurea, metformin, or insulin in patients with
type 2 diabetes mellitus: progressive requirement for multiple
therapies (UKPDS 49). UK Prospective Diabetes Study
(UKPDS) Group. JAMA 1999;281:200512.
16. Adverse events and their association with treatment
regimens in the Diabetes Control and Complications Trial.
Diabetes Care 1995;18:141527.

17. Strom BL, Collins M, West SL, et al. Sexual activity,


contraceptive use, and other risk factors for symptomatic
and asymptomatic bacteriuria: a case-control study. Ann
Intern Med 1987;107:81623.
18. Geerlings SE, Meiland R, Hoepelman AI. Pathogenesis of
bacteriuria in women with diabetes mellitus. Int J Antimicrob
Agents 2002;19:53945.
19. Geerlings SE, Stolk RP, Camps MJ, et al. Risk factors for
symptomatic urinary tract infection in women with diabetes.
Diabetes Care 2000;23:173741.
20. Hooton TM, Scholes D, Stapleton AE, et al. A prospective
study of asymptomatic bacteriuria in sexually active young
women. N Engl J Med 2000;343:9927.
21. Boyko E, Lipsky B. Diabetes and infection. In: Harris M,
Cowie C, Stern M, et al, eds. Diabetes in America. 2nd ed.
Bethesda, MD: National Institute of Diabetes and Digestive
and Kidney Diseases, 1995:48599.
22. Kass EH. Asymptomatic infections of the urinary tract.
Trans Assoc Am Phys 1956;69:5664.
23. Keane EM, Boyko EJ, Reller LB, et al. Prevalence of
asymptomatic bacteriuria in subjects with NIDDM in San
Luis Valley of Colorado. Diabetes Care 1988;11:70812.
24. Abu-Bakare A, Oyaide SM. Asymptomatic bacteriuria in
Nigerian diabetics. J Trop Med Hyg 1986;89:2932.
25. Bahl AL, Chugh RN, Sharma KB. Asymptomatic bacteriuria
in diabetics attending a diabetic clinic. Indian J Med Sci
1970;24:16.
26. Hansen RO. Bacteriuria in diabetic and non-diabetic outpatients. Acta Med Scand 1964;176:72130.
27. Jaspan JB, Mangera C, Krut LH. Bacteriuria in Black
diabetics. South Afr Med J 1977;51:3746.
28. Joffe BI, Seftel HC, Distiller LA. Asymptomatic bacteriuria
in diabetes mellitus. South Afr Med J 1974;48:13068.
29. Ooi BS, Chen BT, Yu M. Prevalence and site of bacteriuria
in diabetes mellitus. Postgrad Med J 1974;50:4979.
30. OSullivan DJ, Fitzgerald MG, Meynell MJ. Urinary tract
infection: comparative study in the diabetic and normal
population. BMJ 1961;1:7868.
31. Schmitt JK, Fawcett CJ, Gullickson G. Asymptomatic bacteriuria and hemoglobin A1. Diabetes Care 1986;9:51820.
32. Vejlsgaard R. Studies on urinary infection in diabetics. I.
Bacteriuria in patients with diabetes mellitus and in control
subjects. Acta Med Scand 1966;179:17382.
33. Vigg B, Rai V. Asymptomatic bacteriuria in diabetics.
J Assoc Phys India 1977;25:5761.
34. Stamm WE, Counts GW, Running KR, et al. Diagnosis of
coliform infection in acutely dysuric women. N Engl J Med
1982;307:4638.
35. Harris MI, Klein R, Welborn TA, et al. Onset of NIDDM
occurs at least 47 years before clinical diagnosis. Diabetes
Care 1992;15:81519.
36. Harris M. Summary. In: Harris M, Cowie C, Stern M, et al,
eds. Diabetes in America. 2nd ed. Bethesda, MD: National
Institute of Diabetes and Digestive and Kidney Diseases,
1995:113.