A Reassessment of the Importance of Low-Count Bacteriuria in Young Women with Acute Urinary Symptoms

Methods

Patients

The study was done at the gynecology clinic at The Ohio State University student health center between April 1991 and May 1992. The study protocol was reviewed and approved by the institutional review board at The Ohio State University. All nonpregnant women who attended the clinic for any reason were eligible to participate, except for those who had taken an oral or injectable antimicrobial drug in the previous week. During a 2- to 3-hour period each day, the clinic nurse consecutively selected each eligible patient who was able to provide a urine specimen at the time of the clinic visit. Oral informed consent was obtained. Each patient was interviewed using a standardized questionnaire that included demographic information; reason for the visit; current symptoms; recent use of antimicrobial drugs; history of symptoms of cystitis; contraceptive use; date of most recent intercourse; and pregnancies. Refusals were less than 5%.

The patients were divided into four categories according to current symptoms. These consisted of those with: 1) urinary symptoms (urinary frequency, urgency, small amount of urine, hematuria or dysuria without fever, flank pain or loin pain); 2) vaginal symptoms (vaginal discharge, itching, swelling, redness, soreness or vulvar burning); 3) urinary and vaginal symptoms (some combination of urinary and vaginal symptoms as described above); and 4) asymptomatic (no urinary or vaginal symptoms).

Laboratory Methods

Clean-catch midstream urine specimens were obtained using a standardized procedure [4]. Patients were asked to cleanse the inner vulval area with a green soap towelette, to rinse twice by wiping from front to back with moistened gauze pads, and to collect a mid-stream urine specimen while keeping the labia spread. The specimens were cultured as soon as they were obtained by streaking the surface of MacConkey and blood agar plates with 0.01 mL and 0.001 mL loops. Reagent strips (Chemstrip, Boehringer Mannheim, Indianapolis, Indiana) were then dipped into the urine. The specimens were refrigerated and transported to the laboratory within 3 hours of collection. Specific gravity was determined with a refractometer (Hand Protometer, National Instrument Co., Baltimore, Maryland). Creatinine levels were measured at the clinical laboratory of The Ohio State University Hospital. A drop of Kova stain (Hycor Biomedical, Garden Grove, California) was added to an aliquot of urine, and the leukocyte count was determined by hemocytometry. The plates were incubated for 48 hours at 37 C. Counts were done for each morphologically distinct colony. The species were identified using standard microbiologic methods.

The following procedure was done to determine whether urine, obtained from each patient, could support growth of low numbers of E. coli that might be present in the specimen. An aliquot, obtained from each of 331 consecutive specimens of urine, was incubated at 37 C. The urine was subcultured on MacConkey agar at 0, 2, 4, and 20 hours of incubation using 0.01 mL and 0.001 mL loops. The plates were incubated for 24 hours at 37 C. Colonies were counted and identified by standard methods.

Statistical Analysis

The data were processed using the EpiInfo, Version 5.0, software program (Centers for Disease Control and Prevention, Atlanta, Georgia). EpiInfo and EpiStat software programs were used to do chi-square tests and most Fisher exact test calculations. The StatXact software program (Cytel Software Corporation, Cambridge, Massachusetts) was used to do the Fisher exact test for those comparisons in which one or both variables had more than two levels. Regression equations and correlation coefficients were calculated using the EpiInfo program. Analysis for linear trends in proportions was done by the method of Schlesselman [16].

Results

Characteristics of the Patients

The patients consisted of 639 women among whom 388 were asymptomatic; 83 had urinary symptoms only; 53 had urinary and vaginal symptoms; and 115 had vaginal symptoms only. The demographic characteristics of the four groups of women were similar in regard to age; marital status; pregnancies; use of contraceptives; and time of recent sexual intercourse except for less recent intercourse among women with vaginal symptoms compared with asymptomatic women (P = 0.016). Women with urinary symptoms were slightly younger than asymptomatic control women (22.3 3.7 [SD] years versus 23.6 4.3 years, respectively; P = 0.012) and were more often single (96.4% versus 86.1%; P = 0.009) and, although not statistically different, fewer women had a history of pregnancy (10.8% versus 14.9%, respectively; P > 0.2). They were more likely to previously have had urinary tract infections (73.5% versus 44.8%, respectively; P < 0.001) and to have received antibiotics within 6 weeks before the clinic visit (22.9% versus 9.3%, respectively; P < 0.001). No statistical differences occurred among the groups in contraceptive practices. About two thirds of the 512 sexually active women used oral contraceptives, about one third used condoms, and fewer than 2% used the diaphragm for contraception.

Association between Bacterial Species and the Presence of Urinary Symptoms and Pyuria

The microorganisms isolated by quantitative culture of the urine of asymptomatic women and those with urinary symptoms only are shown in Table 1. Only E. coli and S. saprophyticus were found more often in patients with urinary tract symptoms (P < 0.001). Other gram-negative enteric bacteria (Klebsiella pneumoniae, Proteus mirabilis, and Enterobacter aerogenes) were found more frequently among women with urinary symptoms, but the numbers were small (P > 0.02). Escherichia coli and S. saprophyticus were more frequently associated with pyuria (>20 leukocytes/mm³) than any of the other microorganisms (P < 0.001) (data not shown). Most microorganisms were isolated in pure culture. Mixed cultures of E. coli with other enteric gram-negative bacteria were observed in five patients and a mixture of E. coli and S. saprophyticus in one patient. All the other bacterial species listed in the Table were recovered more frequently from asymptomatic women. Accordingly, in the remainder of the report, only bacterial counts for E. coli, other enteric bacteria, and S. saprophyticus are shown.

Association between Bacterial Counts and the Presence of Symptoms of Urinary and Vaginal Infection

The distribution of patients according to bacterial counts and symptoms is shown in Table 2. Among the asymptomatic control patients, the frequency of significant bacteriuria (>10⁵ CFU/mL) was 3.1%; low-count bacteriuria (>10² to 10⁴ CFU/mL) was present in 10.2%. In contrast, about one third (32.5%) of women with urinary symptoms had significant bacteriuria, and about one half (45.8%) had low-count bacteriuria. Women with combined urinary and vaginal symptoms were similar to those with urinary symptoms only but had a lower rate of significant and low-count bacteriuria. The distribution of bacterial counts among women with vaginal symptoms did not differ from asymptomatic control women (P > 0.2). The odds ratio of an association between bacteriuria and urinary symptoms increased as the bacterial count increased (P < 0.001) (Table 2). The same association was also noted for women who had urinary and vaginal symptoms (P < 0.001).

Association of Urinary and Vaginal Symptoms with Quantitative Counts of Leukocytes in the Urine

The distribution of leukocytes in the urine, according to symptoms at the time of presentation, is shown in Figure 1. A cutoff of 20 leukocytes/mm³ appeared to most clearly separate the groups. Using this criterion, pyuria was present in 19.6% of asymptomatic women; in 37.4% of women with vaginal symptoms; in 56.6% with combined vaginal and urinary symptoms; and in 77.1% with urinary symptoms. Most women (61.4%) with urinary symptoms were found to have more than 100 leukocytes/mm³. This degree of pyuria was rarely present in asymptomatic women or those with only vaginal symptoms unless the bacterial count was more than 10⁵ CFU/mL.

Relation of Bacterial Counts in the Urine to Pyuria

The extent of the inflammatory response, as determined by the number of leukocytes in the urine, was examined in relation to the quantitative bacterial count (Figure 2). A stepwise increase occurred in the frequency and magnitude of pyuria as the bacterial count increased. Most women (70.8%) with counts of more than 10⁵ CFU/mL had more than 100 leukocytes/mm³. This degree of pyuria was rarely present in women with counts of 10² or less than 10² CFU/mL.

Relation between Urinary Concentration and Bacterial Counts

To determine whether low counts of bacteria and leukocytes found in urine of women with urinary symptoms were due to dilution by high fluid intake and frequent voiding, the urine specific gravity and concentration of creatinine were compared with the bacterial count of each voided specimen. No statistical differences occurred in the mean specific gravity (P > 0.2) or concentration of creatinine (P > 0.2) between bacterial counts of 10² to 10⁴ and more than 10⁵ CFU/mL. Further, no change occurred in the slope of the regression line for specific gravity or creatinine level across the entire range of the bacterial counts (regression equation, Y = 1.0175 0.00001 X, R = 0.07 for specific gravity; Y = 167.81 0.00011 X, R = 0.03 for urinary creatinine level). The specific gravity and creatinine level were highly correlated (r = 0.84).

Ability of Indigenous Escherichia coli To Grow in the Patients' Urine

To examine the possibility that low-count bacteriuria might result from poor growth of the microorganisms in urine, each patient's urine was incubated at 37 C and was cultured quantitatively at periodic intervals (Figure 3). Most (70.9%) urine specimens initially containing less than 10² CFU/mL were sterile at 20 hours. For urine specimens initially containing more than 10² to less than 10³ CFU/mL, growth began by 4 hours with full growth (>10⁵ CFU/mL) by 20 hours. For specimens initially containing more than 10³ to 10⁴ CFU/mL, most grew by 2 and 4 hours and all achieved full growth by 20 hours.

Comparison of the Leukocyte Esterase Test with the Chamber Count as Tests for Pyuria

Using a hemocytometer count of 10 leukocytes/mm³ as the criterion for pyuria for all 639 patients, the sensitivity of the leukocyte esterase test was 79.7%; the specificity was 96.2%; the positive predictive value was 94.7%; and the negative predictive value was 84.8%. Using a criterion of 20 leukocytes/mm³ for the same group of patients, the sensitivity was 92.5%; the specificity was 88%; the positive predictive value was 79.2%; and the negative predictive value was 98%.

Sources of Pyuria Other Than the Urinary Tract

Pyuria (>20 leukocytes/mm³) was present in 19.6% of asymptomatic women and in 37.4% of women with vaginal symptoms. To determine whether the source of the pyuria might have been from the vagina, 47 asymptomatic women and 20 women with vaginitis who had a routine gynecologic examination were examined in the following manner. Shortly after a clean-voided specimen of urine was obtained, the patient was placed on an examining table, and a cotton swab was inserted deep in the vagina. The moistened portion of the swab was then inserted into a tube containing 1 mL of buffered saline and was mixed thoroughly. The fluid was serially diluted in buffered saline and was examined by the leukocyte esterase test. The test was confirmed by counting the leukocytes by hemocytometer at the highest dilution at which the leukocyte esterase test was positive.

The urine of 9 of the 47 asymptomatic women (19.1%) was positive by the leukocyte esterase test and contained more than 20 leukocytes/mm³. The vaginal fluid of 8 of the 9 women with pyuria (88.9%) also gave a positive leukocyte esterase test and contained more than 20 leukocytes/mm³ at a dilution of more than 1:16 compared with 15 of 38 (39.5%) of those without pyuria (P = 0.009). The urine of 9 of the 20 women with vaginal symptoms (45%) was positive by the same tests. The vaginal fluid of all 9 (100%) was positive by these tests at dilution of more than 1:16 compared with 7 of 11 (63.6%) of those without pyuria (P = 0.068).

Discussion

Our study confirms the observations of Stamm and colleagues [12, 13] that the urine of many women with urinary symptoms and pyuria contains E. coli, S. saprophyticus, and enteric gram-negative bacteria at bacterial counts of less than 10⁵ CFU/mL. Unlike Maskell [17], we did not find that fastidious microorganisms are causal agents of the urethral syndrome and agree with Brumfitt and others [18, 19] in this regard. It is possible that lactobacilli and other organisms may have a protective role, because they were recovered statistically more often from asymptomatic women. It was also not possible to divide patients into those with vaginitis and lower urinary tract complaints, as recommended by Komaroff [20]. An additional category of combined urinary and vaginal symptoms had to be added. This group was found to be at greater risk for both low-count and significant bacteriuria (>10⁵ CFU/mL) than were patients who were asymptomatic or those with vaginal symptoms only.

Several theories may explain the phenomenon of low-count bacteriuria in young women with acute urinary symptoms: 1) The patients had urinary tract infections, but the bacterial count was lowered by intake of high volumes of fluid and frequent voiding; 2) the organisms did not grow well in the urine because of the presence of an inhibitor[s]; 3) greater in vivo killing occurred by the bladder mucosal cells; 4) the infection was localized to the urethra, but the causative agents were diluted by the urinary stream; and 5) no bacterial infection was present, but colonization of the urethra with uropathogens was fostered by mechanical, irritative, or immunologic insults. Acute urinary symptoms may be caused by gonococci, chlamydia, herpes viruses, and Ureaplasma urealyticum, and low-count bacteriuria may not be etiologic [21, 22].

The virtual consistency of the urine-specific gravity and creatinine concentration, regardless of the bacterial count, makes dilution of urine an unlikely explanation. The relation between the bacterial count and the frequency of voiding or the residual volume of urine was not determined. All the symptomatic patients complained of urgency and frequent urination. The presence of urine inhibitors is also unlikely because the organisms grew well in the urine from which they were obtained. We were unable to examine the possibility of killing by the bladder wall, but this mechanism is limited to a small number of the microbes that adhere to the mucosal surface [23, 24]. We favor the more likely hypothesis that the infection had not become fully established in the bladder in women with very low count bacteriuria (>10² to 10⁴ CFU/mL) and may be an early phase of urinary tract infections possibly localized to the urethra. The finding of a step-wise increase in the frequency and magnitude of pyuria in relation to the bacterial count and the graded increase in the frequency of pyuria in symptomatic patients as the bacterial count increased suggest that this is a dynamic process. In contrast, urinary tract infections in asymptomatic women followed the more classic pattern in which pyuria (>100 leukocytes/mm³) was present only in those with very high counts (>10⁵ CFU/mL).

O'Grady and colleagues [25] have proposed that dysuria/pyuria or the urethral syndrome in women may be a transitional phase of urinary tract infections in which the urethra is the primary site of colonization and inflammation. According to this concept, bacteria may enter the bladder transiently but, because of urodynamic and other host-defense mechanisms, are not able to grow sufficiently to achieve the high densities that are observed in well-established urinary tract infections. This is consistent with the classic studies of Stamey and others [26-29], who showed that the urethra and vaginal introitus in women prone to recurrent urinary infections were much more likely to be colonized with enteric gram-negative bacteria than were healthy control women. It is also supported by our finding of prolonged, intermittent colonization of the urine by the same serotype of E. coli in women with or without a previous history of urinary tract infection [30]. Several authors [7, 10] have proposed that the urethral or the dysuria/pyuria syndrome might be a distinct clinical entity, possibly localized to the paraurethral glands, with transient colonization of the bladder urine. This concept was supported by the finding of more leukocytes in the first portion of partitioned voidings [31].

A single screening culture was used in our study as would be done in a practice setting. More strict criteria, including repeated cultures, would be required to define patients with asymptomatic bacteriuria or to enter patients into a clinical trial. The rate of asymptomatic bacteriuria of 3.1% (cutoff of >10⁵ CFU/mL) reported here would have been somewhat lower had repeated cultures been obtained. Voided urine in women is rarely sterile and may contain vaginal leukocytes. Thus, the finding of low bacterial counts of E. coli and other enteric bacteria and the low levels of pyuria in symptomatic patients in a single culture may be very difficult to distinguish from the background frequency of low-count bacteriuria of about 10% and of pyuria of about 20% observed in asymptomatic patients. It may be difficult to distinguish infection from periurethral contamination in women prone to recurrent infections. We share the concern expressed by Washington [32] that urine specimens are often carelessly collected and transported and may be difficult to interpret. We also doubt the practical utility of a criterion as low as 10² CFU/mL because this depends on finding one colony on a plate, is difficult to replicate, and has the least association with symptoms and pyuria.

Because of the difficulty in establishing a clear end point for diagnosis and response to therapy, and the possibility of spontaneous cure, we recommend that for the purposes of studies of natural history of urinary tract infections and clinical trials of anti-infective drugs, symptomatic patients with pyuria and low-count bacteriuria be considered separately from those with the traditionally accepted concept of significant bacteriuria (>10⁵ CFU/mL) [33].