Use of Spermicide-Coated Condoms and Other Risk Factors for Urinary Tract Infection Caused by Staphylococcus saprophyticus FREE

STAPHYLOCOCCUS saprophyticus is responsible for 7% to 26% of urinary tract infections (UTIs) in young sexually active women.^1- 6 Because this bacterium is not ordinarily present in the rectovaginal flora, the manner in which it causes UTI has remained uncertain.⁷ While a variety of risk factors for acquiring UTI caused by coliform organisms have been documented, predisposition to infection by S saprophyticus has not been thoroughly evaluated. In addition to younger age and sexual activity, the other widely acknowledged risk factor for UTI caused by this organism is the season of the year, since these infections occur more frequently in the summer and fall.^2,6,8- 10

One well-known risk factor for UTIs caused by coliform organisms is exposure to vaginal spermicides. Several studies have confirmed a higher risk of UTIs among women using a diaphragm with spermicides compared with sexually active women using other types of contraceptives.^11- 17 This appears to result from the toxic effect of spermicides on the vaginal flora.¹⁸ Users of a diaphragm plus spermicide have increased levels of introital and periurethral colonization with coliform organisms,^19- 21 possibly because of enhanced adherence of the bacteria to the vaginal mucosa.²² Bacteriuria with Escherichia coli is more frequent after intercourse in women who use spermicidal foam or diaphragms with spermicides.²³ Women who are exposed to spermicide-coated condoms have also recently been shown to have a significantly elevated risk of UTI caused by E coli.²⁴ To our knowledge, no studies to date have assessed whether spermicide use predisposes to UTI caused by S saprophyticus .

In conjunction with a larger study of the risk factors for UTI,²⁴ we performed a population-based, case-control study to assess host factors that might predispose to acute UTI caused by S saprophyticus, including sexual activity and contraceptive practices. We were especially interested in whether women who had experienced a recent UTI caused by S saprophyticus were more likely than randomly selected control patients to have been exposed to spermicides, including condoms coated with nonoxynol 9, the most commonly used vaginal spermicide.

The study was conducted at Group Health Cooperative (GHC) of Puget Sound, a staff-model health maintenance organization in Seattle, Wash, that has approximately 500000 contract enrollees and is the oldest cooperative health maintenance organization in the United States. Eligibility was limited to women receiving care in Pierce, King, Thurston, and Snohomish Counties in western Washington State. There were approximately 75600 female GHC enrollees between the ages of 18 and 40 years who received care in the 4 counties during the period of study, 1990 to 1993.

Case patients were women between the ages of 18 and 40 years who had been members of GHC for at least 1 year and had an acute UTI caused by S saprophyticus or E coli diagnosed within the prior month. We reviewed the computerized laboratory files monthly to identify all women with a urine culture yielding more than 10⁵ colonies of S saprophyticus or E coli per milliliter during the preceding month. We excluded women whose urine culture was obtained during treatment for a previously diagnosed UTI, who had asymptomatic bacteriuria, or who had been pregnant within the previous year. This study focuses exclusively on the cases caused by S saprophyticus. Risk factors for E coli are the subject of another article.²⁴

We reviewed the medical records of potential case patients for documentation of an acute, symptomatic UTI, defined as the presence of dysuria, frequency, or urgency for 2 weeks or less. Absence of symptoms led to exclusion. The remaining women were eligible for interview. After securing permission from their primary care physician, we sent a letter describing the study, advising we would call to schedule an interview, and explaining how to decline participation.

From GHC enrollment files, we randomly selected women to serve as control patients. Like case patients, control patients had to have been GHC enrollees for the preceding 12 months. Control patients were frequency matched according to date of birth (within 2 years) to women who were selected as case patients for the larger study.

We also reviewed the hospital discharge and laboratory files to eliminate women who had a clinical record of having a UTI, a urine culture positive for more than 10⁵ uropathogenic organisms per milliliter within the prior month, or who had been pregnant within the prior year. After receiving permission from their primary care physician, potentially eligible control patients were contacted by mail in the same manner as women in the case group.

Potential case and control patients who had not notified us that they wished to be excluded were contacted by telephone to make a final determination of their eligibility and willingness to participate. We made up to 10 attempts to reach each woman by telephone. When contacted, we asked potential participants whether they had received materials about the study and confirmed that they had been active members of GHC for the past year and had not been pregnant during this time. We eliminated any women who did not meet these eligibility criteria, as well as those who no longer lived in the area or who told us that they had not been sexually active during the past month. We also eliminated women with certain medical conditions, including neurologic problems that might interfere with voiding, known anatomical abnormalities of the urinary tract, an indwelling urinary catheter, or diabetes mellitus.

We then described the study procedures, requested verbal consent, and scheduled an interview with those who consented. All procedures used in the study were approved by the human subjects committees of the University of Washington, Seattle, and GHC.

We conducted all interviews by telephone with the aid of computer software that used branching logic and was programmed to perform automatic range checking and examine internal consistency (CiII, Sawtooth Software, Ketchum, Idaho). The interview was extensively tested and revised prior to the study. During the first and last months of the study, 10% of interviews were monitored by a second interviewer who simultaneously recorded responses to assess reliability. Interrater reliability exceeded 98% for all items.

For each monthly sample, the 15th of the month was programmed into the interviewing software as the point of reference for all questions relating to events during the month prior to the UTI, for example, "In the month prior to July 15th did you use a diaphragm?" Because matching was not 1-to-1, this enabled us to assign a comparable date to case and control patients and ask identical questions. Women in the case and control groups were interviewed a median of 75 days (mean, 77 days) after the reference date, respectively.

We interviewed patients about their general medical, obstetric, and gynecologic history, recent frequency of sexual activity, and number of sexual partners. We made detailed inquiries about contraceptive practices during the previous month and year, including frequency and manner of use and specific brands of products used. To classify the type of condom used, we inquired whether the product was coated with a lubricant or a spermicide (and specifically, nonoxynol 9). To aid patients with product identification, interviewers used a notebook listing 107 commercially available condoms²⁵ and photographs of the packaging for 58 major brands. To minimize possible bias in ascertaining exposure status, interviewers were kept unaware whether of they were speaking with a case or a control patient until the final set of questions, which dealt with the patient's history of previous UTIs.

Data were analyzed using statistical software (EGRET, SERC, Seattle, and SAS, SAS Institute Inc, Cary, NC). Bivariate comparisons of characteristics by case status were tested for significance with the t test for continuous variables and the χ² test for categorical variables. Odds ratios (ORs) and 95% confidence intervals (CIs) were computed by standard methods.²⁶

To adjust for potential confounding, we constructed several logistic models using case or control status as the dependent variable.²⁶ In these models, sexual activity was treated as a continuous variable, coded as the frequency per week during the previous month. Condom exposure was treated as either a dichotomous variable (yes or no) or an ordinal or continuous variable representing frequency of use. To test for the presence of a dose-response relationship between frequency of condom use and UTI, we created a set of dummy variables representing different levels of condom use.

Using computerized laboratory and administrative files, 186 women were selected as potential case patients. Review of their medical records eliminated 18 women as ineligible and the records of 8 others were unavailable. One woman's physician declined to allow her to participate. Exclusion of these 27 yielded 159 women whom we attempted to contact as case patients, 14 of whom were deemed ineligible on the basis of information gathered during the screening interview, leaving a total of 145 eligible case patients (Table 1). We initially identified 1543 potential control patients, of whom 389 were found to be ineligible after screening, leaving 1154 eligible control patients.

Of the 159 potential case patients we attempted to contact, 23 (14%) could not be contacted, 26 (16%) refused to be interviewed, and 14 (9%) were found to be ineligible after interview (Table 1). Of the 1543 potential control patients we attempted to contact, 184 (12%) could not be reached, 341 (22%) refused to be interviewed, and 389 (25%) were ineligible. Many of those who could not be contacted were no longer receiving care from GHC and/or had moved away but administrative records had not yet been updated. The final analysis included 96 case and 629 control patients.

Case and control patients were similar with regard to race, self-reported health, and education (Table 2). Because most control patients were matched according to age with women who had infections with E coli, the control group was slightly older than the case (S saprophyticus) group. Although only 28 Asian American women were in the sample, they were twice as likely to be in the case group (OR, 2.31; 95% CI, 0.95-5.62). Case patients were significantly more likely than control patients to be unmarried (unadjusted OR, 2.78; 95% CI, 1.79-4.35), to have had multiple sexual partners during the past year (OR, 2.41; 95% CI, 1.33-4.38), and to have had frequent sexual intercourse (OR, 1.20 per weekly episode; 95% CI, 1.06-1.37) and prior UTIs (OR, 2.45; 95% CI, 1.46-4.12).

We did not observe infections with S saprophyticus more frequently in the summer and fall; only 46% of infections occurred during the months April through September.

Condoms were the most common method of contraception and/or protection from sexually transmitted disease in the study population. They were used during the past year and past month by 68 (71%) and 51 (53%) women in the case group, and 195 (31%) and 120 (19%) women in the control group, respectively. Of those using condoms during the past month, 58 case patients (60%) and 157 control patients (25%) reported they definitely used condoms coated with nonoxynol 9. For 5 (5%) of all case patients and 34 (5.4%) of all control patients, use of a diaphragm plus spermicide was their main method of contraception. Thirty-seven percent of the total study sample used oral contraceptives during the past year while 19% of women or their partners had undergone surgical sterilization, 13% practiced the rhythm method, and 38% used another method or no method at all.

The unadjusted OR for an acute UTI caused by S saprophyticus among women who had used any type of condom during the previous month was 1.72 (95% CI, 1.06-2.79) (Table 3). The OR for a UTI associated with condom use during the previous year was 1.97 (95% CI, 1.28-3.05). The risk of UTI increased in proportion to the frequency of condom use during either the previous month or year.

To determine whether faulty recall about type of condom used may have resulted in bias due to misclassification regarding exposure to spermicides, we repeated this analysis using responses from only the 35 case and 160 control patients about whom the interviewers were highly confident in classifying their exposure to condoms within the previous year. The OR for UTI among women exposed to condoms within that time was 1.89 (95% CI, 1.15-3.10), nearly identical to the OR of 1.97 when all women were included.

When the reference group was defined as women with no exposure to any type of condom, the unadjusted risk of UTI with S saprophyticus was significantly elevated for women who reported using condoms coated with nonoxynol 9 during the prior month (OR, 3.63; 95% CI, 1.86-7.08) or year (OR, 3.08; 95% CI, 1.73-5.48), and increased with greater frequency (Table 3). The OR for women who had used nonoxynol 9–coated condoms infrequently, ie, less than once weekly in the prior year, was 2.16 (95% CI, 0.98-4.75), a much higher point estimate than the OR of 1.14 for an equivalent level of exposure to any type of condom. For those exposed to nonoxynol 9–coated condoms more than twice weekly during the prior year, the OR was 6.05 (95% CI, 2.20-16.60). Insufficient numbers precluded calculating similar dose-response figures for exposure during the past month.

When the reference group was defined as women who had been exposed to condoms that were not coated with nonoxynol 9, the ORs for UTI caused by S saprophyticus among women who did use nonoxynol 9–coated condoms during the previous month or year were 3.80 (95% CI, 1.40-10.30) and 2.03 (95% CI, 0.96-4.33), respectively (data not displayed in tables).

In addition to questions about condoms coated with nonoxynol 9, we also inquired about regular lubrication of the condom. Use of a lubricated condom in the past month was associated with a modestly increased risk of UTI caused by S saprophyticus (OR, 1.68; 95% CI, 1.03-2.73).

The unadjusted OR for acquiring a UTI caused by S saprophyticus among women using a diaphragm during the prior month was 0.97 (95% CI, 0.37-2.55). The proportion of women using a diaphragm with a spermicide or other types of spermicides (eg, spermicidal foam) was 5% and decreased during the course of the study.

Others have suggested that swimming in open water (eg, lakes or oceans) is a risk factor for UTIs caused by S saprophyticus.²⁷ We found no such association (OR, 1.27; 95% CI, 0.35-4.55) for this type of exposure or for other types of water exposure, such as bathing, showering, douching, and swimming in pools and hot tubs.

We also detected no association between UTIs caused by S saprophyticus and a history of kidney stones or infection or history of bacterial or yeast vaginitis.

To determine the independent contributions of factors that were significant in bivariate comparisons, we constructed a series of approximately 20 logistic regression models using marital status, frequency of sexual activity, number of sexual partners, history of UTI, condom exposure, and age as independent variables. In these multivariate models, the ORs for UTI associated with the frequency of sexual intercourse per week, coded as a continuous variable, ranged between 1.17 and 1.21 with a lower limit for the 95% CI of 1.03 to 1.06 and an upper limit of 1.35 to 1.38. A history of UTI caused by any organism was also significantly associated with the risk of UTI caused by S saprophyticus, with an OR of 3.58 to 3.72. Although being unmarried was significant on a bivariate basis, it was not so in any of the multivariate models.

The risk of UTI associated with spermicide-coated condoms was highly significant (OR, 3.75; 95% CI, 1.76-7.99) (Table 4, multivariate model 1). Moreover, with more frequent use of spermicide-coated condoms during the prior month, the relative odds of experiencing a UTI rose dramatically (Table 4, multivariate model 2). The OR for use of a spermicide-coated condom 2 or more times per week during the past month compared with no condom use was higher than 10 (95% CI, 2.05-52.97).

In several multivariate models in which control patients not using any type of condom served as the reference group, the risk of UTI associated with exposure only to uncoated condoms was not statistically significant (P>.05). This again suggests that most of the excess risk of UTI among condom users was related to nonoxynol 9 exposure. Among study patients exposed to coated condoms during the previous month, 67.5% of UTIs caused by S saprophyticus were due to coated condoms (attributable risk percent).

In comparison with sexually active women not using condoms, there was not an increased risk of UTI associated with S saprophyticus among women who had used other contraceptive methods, including vaginal insertion of a spermicidal agent (ie, foam, film, cream, jelly, or suppository). The ORs for exposure to any of these products were not statistically significant (0.61 [95% CI, 0.18-2.03] for exposure during the prior month and 0.41 [95% CI, 0.05-1.3] in the prior year), but the number of women using them was very small.

In this population-based, case-control study, we identified several important risk factors for UTI caused by S saprophyticus, the second most common bacterial pathogen causing UTI in young women. In agreement with several prior investigations,^6,8- 9,27 we found that younger age and sexual activity were strongly associated with the risk of UTI caused by this organism. We also found that the risk for a UTI was more than 3-fold higher among women who reported a prior UTI caused by any organism compared with women with no history of UTI. Although our study design did not permit us to determine whether the preceding UTI was caused by S saprophyticus, this finding is consistent with that of Rupp and associates,⁷ who found that women who had vaginal colonization with S saprophyticus reported having a UTI within the past year nearly 3 times more often than those who were not colonized. Similarly, Latham and colleagues⁶ reported that 6 of 72 patients with UTIs caused by S saprophyticus had recurrences with the same organism.

The similarity of risk factors for UTI caused by S saprophyticus and E coli suggests that these infections occur by the same route. Thus, although S saprophyticus is not ordinarily a constituent of the fecal flora, colonization probably begins in the rectum and extends to the vaginal introitus and periurethral area before reaching the bladder.^6- 7,28

As is the case for E coli infection,²⁴ we observed that condom use was also a risk factor for UTI caused by S saprophyticus. Coupled with the findings that this infection was also more common in individuals with multiple sexual partners and more frequent intercourse, this raises the possibility of sexual transmission. This organism has been suggested as a cause of urethritis in men.²⁹ Arguing against this being a sexually transmitted agent is our finding that the increase in risk of UTI was restricted to women using condoms coated with a spermicidal agent. Use of uncoated condoms did not appear to increase the risk of infection. Of the women participating in this study, 33% had been exposed to a condom in the past year and, at a minimum, 12% of these exposures involved spermicide-coated condoms.

After adjustment for other risk factors (frequency of sexual activity, number of sexual partners, history of UTI, or use of other contraceptive methods) the odds of UTI among women exposed to condoms coated with nonoxynol 9 were more than 3 times higher than for sexually active women who did not use coated condoms. The association between risk of UTI and exposure to coated condoms was consistent in every analysis performed. We also observed a strong dose-response relationship between the frequency of using spermicide-coated condoms and the risk of UTI. We hypothesize that, as in the case of gram-negative bacteria, vaginal spermicides damage the normal vaginal flora in a manner that promotes colonization with S saprophyticus, possibly by enhancing mucosal adherence. Like E coli, S saprophyticus possesses several adhesins that hemagglutinate sheep red blood cells and mediate attachment to uroepithelium.^30- 32 Also like E coli, these adhesins may be related to blood group antigens.³³

Surprisingly, unlike most other investigators, we did not find a seasonal variation in the incidence of these infections. Whether this represents a unique feature of the population we studied or changing epidemiologic patterns of this infection is uncertain.

We took several precautions to minimize biases that can affect case-control studies. To prevent misclassification, our case definition, acute urinary symptoms plus a culture yielding more than 10⁵ bacteria per milliliter, is more than 95% specific for the presence of a UTI.³⁴ We confirmed that all case patients met this criterion based on data obtained from computerized laboratory files, medical records, and patient interviews. To avoid the biases introduced by using control patients drawn from individuals visiting a clinic or hospital, we randomly selected control patients from the entire GHC enrollment base in the 4 counties studied. Exclusion criteria applied identically to the case and control groups.

To minimize recall bias and misclassification of exposure, we interviewed case patients as soon as possible following diagnosis of their UTI. Patients were told only that the investigators were interested in the epidemiologic characteristics of UTI. In gathering information about spermicide use, we asked questions about use of products both "coated with nonoxynol 9" and those "coated with spermicides" and obtained nearly identical results with both wordings. In addition, to help patients identify types and brands of condoms, the interviewers used a notebook with photographs of the packaging and a description of almost every condom marketed in the United States.

Interview data were gathered using a computerized system that provided instantaneous checks for consistency between items and out-of-range responses. Reliability among our interviewers was very high. Furthermore, we attempted to keep our interviewers uninformed about a patient's case or control status until exposure data had been collected.

Although we made every effort to eliminate bias, there were limitations to this study. First, a considerable number of women declined to participate or could not be contacted, suggesting the possibility of response bias. However, many of these women may have been deemed ineligible if their medical records had been reviewed and they had been interviewed, but the possibility remains that participants may have differed from nonparticipants in terms of the exposures of interest. Second, we had no comprehensive source of external data with which to validate patients' reported exposures. Many women could not positively identify the type and/or brand of condom to which they might have been exposed. Approximately 30% of control patients who were exposed to condoms in the prior month did not know whether the product was coated with a spermicide. Random misclassification of exposure that may have resulted would have tended to reduce the size of any effect observed. Furthermore, examining just the group about which we were most certain yielded a nearly identical point estimate. Finally, the women we studied had health insurance and were mainly white, possibly limiting generalizability to other groups of women. This population, however, is more inclusive than that seen in student health centers and emergency departments where women with UTIs are most often studied.

Despite these potential limitations, our results suggest a strong association between exposure to spermicide-coated condoms and UTI caused by S saprophyticus. This association is supported by the consistency of these findings with earlier studies regarding spermicide exposure and UTI and by the strong dose-response relationship. Women who use coated condoms and experience recurrent UTIs, particularly those caused by S saprophyticus, may be advised to consider other methods of contraception and protection from sexually transmitted diseases. In the future other topically applied, nonsurfactant agents that protect against sexually transmitted pathogens but are not associated with UTI risk may become available.

Accepted for publication June 17, 1997.

Supported by grants RO1 DK431341-02 and RO1 DK431341-03 from the Epidemiology Program of the Kidney Branch of the National Institute of Diabetes, Digestive and Kidney Disease, Bethesda, Md (Dr Fihn).

We thank Jayne Grafton and Dick Rayray for programming assistance; Marcia Hunt for helping to develop our questionnaire; Allison McMorris and Lair Showalter for helping to perform interviews; and Walter Stamm, MD, and Anne Stapleton, MD, for their constructive critiques of the manuscript.

Reprints: Stephan D. Fihn, MD, MPH, Health Services Research Program, VA Puget Sound Health Care System (152), 1660 S Columbian Way, Seattle, WA 98108.