Risk Factors for Group B Streptococcal Disease in Adults

Methods

Identification of Cases

Active surveillance for invasive group B streptococcal infection was done in 1991 and 1992 in an aggregate population of 6.6 million persons using previously described methods [1]. The population included all residents of three counties in California (Alameda, Contra Costa, and San Francisco); eight counties in Georgia (Cobb, Clayton, Dekalb, Douglas, Fulton, Gwinnett, Newton, and Rockdale); and Baltimore City and Baltimore County, Maryland. Briefly, regional surveillance staff made biweekly calls to infection-control practitioners or designated contacts in the microbiology laboratories serving all acute-care hospitals in the surveillance areas for reports of sterile-site (for example, blood or cerebrospinal fluid) isolates of group B streptococcus. A case report form requesting information on demographic characteristics, site of isolation, and clinical syndrome was completed for each identified case. To ensure complete reporting, periodic audits of all laboratories were done.

A case of invasive adult group B streptococcal disease was defined as isolation of group B streptococcus from a normally sterile site in a resident of a surveillance area who was 18 years of age or older and who was not pregnant or postpartum. If group B streptococcus was first isolated from a specimen obtained 2 or more days after hospital admission, the case was defined as nosocomial. Cases with positive cultures obtained from 1 June 1991 to 30 June 1992 in California and Georgia and from 1 November 1991 to 31 October 1992 in Maryland were eligible for inclusion.

Selection of Controls

For each case-patient, a listing of all patients hospitalized on the same day and of the hospital service to which they were admitted was obtained and reviewed. Controls were selected by identifying the three preceding patients admitted to the same general service (for example, medicine instead of oncology or surgery instead of orthopedic surgery) as the case-patient. If three appropriate controls could not be identified, admission lists for previous days were sequentially reviewed until three controls for each case-patient were enrolled.

Data Collection

We used standardized forms to abstract medical records for case-patients and controls. Information on demographic variables, outcome, outpatient medications, presenting signs and symptoms, underlying conditions, and the clinical syndrome associated with group B streptococcal infection was collected. For all cases, isolation of group B streptococcus from a sterile site was documented and information on isolation of other organisms from blood was obtained. For nosocomial cases, additional information was obtained on medications administered and procedures done in the hospital before the first positive group B streptococcal specimen was collected. For controls matched to case-patients with nosocomial infection, this additional information was abstracted for the same number of days after admission as for the matched case-patient.

Statistical Analysis

We used the chi-square test to compare proportions of categorical variables among case-patients. We did matched univariate and multiple conditional logistic regression analyses [17] using the Statistical Analysis System procedure PHREG [18] to estimate the risk for group B streptococcal disease associated with individual underlying conditions and exposures. Potential interaction among the variables associated with disease was evaluated by including two-way interaction and main-effect terms in multivariate models. We used stepwise selection, incorporating variables that were significant (P < equals 0.05) in the univariate analysis, to identify a final model for nosocomial disease. To ensure that the risk associated with the underlying conditions was assessed after we controlled for age, we forced a categorical variable that coded for age into the community-acquired model and then used stepwise selection to determine the other variables in the final model.

Results

Chart abstractions were completed for 219 of the 290 (76%) case-patients identified by surveillance (97 from Baltimore, 72 from the San Francisco metropolitan area, and 50 from the Atlanta metropolitan area) and for 645 matched controls from 54 hospitals (2.9 controls per case-patient). Seventy-one case-patients identified by surveillance were not included in the study for the following reasons: the hospital chart was not available for review (50 case-patients), the hospital did not participate in the study (9 case-patients), a sterile-site group B streptococcal culture could not be documented (5 case-patients), the patient was not admitted to the hospital (3 case-patients), the hospital closed (2 case-patients), or the patient died in the emergency department (2 case-patients). Information from the surveillance case report form indicates that the age and sex distribution and mortality rate of the cases not included did not significantly differ from those of the case-patients included in the study.

Proportions of case-patients and controls with selected demographic characteristics and underlying medical conditions are listed in Table 1. The age range for case-patients was 22 to 99 years (median age, 68 years), and the age range for controls was 18 to 99 years (median age, 60 years). Case-patients who died during the hospitalization that was studied died a median of 5 days after collection of the specimen from which group B streptococcus was isolated (range, 0 to 187 days).

Characteristics of Invasive Group B Streptococcal Disease

Site of Isolation

Group B streptococcus was isolated from blood in 201 cases (92%). Other sites of isolation included synovial fluid (6 cases), peritoneal fluid (6 cases), pleural fluid (4 cases), and cerebrospinal fluid (2 cases). It was also isolated from an abscess, bone, bone marrow, bronchial washing, intervertebral disc-space fluid, gall bladder, and liver tissue (1 case each). More than one site of isolation could be reported for each case.

Polymicrobial Bacteremia

Polymicrobial bacteremia, defined as the isolation of other bacterial species from blood cultures collected on the same day as a positive blood culture for group B streptococcus, was identified in 53 of 201 case-patients (26%) with group B streptococcal bacteremia. Staphylococcus aureus was isolated from 24 case-patients and was the only additional organism isolated from 17 case-patients. The age distribution, mortality rate, and proportion of nosocomial cases were similar between case-patients with polymicrobial bacteremia and those with bacteremia caused only by group B streptococcus.

Clinical Syndromes

The clinical syndromes associated with group B streptococcal infection in case-patients are listed in Table 2. Seven of the 19 case-patients with cellulitis were women with a history of breast cancer in whom cellulitis was associated with invasive group B streptococcal infection. Five of these 7 case-patients had cellulitis of the arm or chest wall on the side of a mastectomy done 3 months to 12 years before admission, 1 had cellulitis of the chest wall opposite the site of the previous mastectomy, and 1 had cellulitis of the arm after axillary node dissection.

Characteristics Associated with In-Hospital Death

Age was associated with in-hospital mortality. Case-patients 65 years of age or older were significantly more likely to die of group B streptococcal infection than were younger case-patients (in-hospital mortality rates, 28% and 13%, respectively; P = 0.008). Although the overall mortality rate for patients with a clinical diagnosis of pneumonia was higher than the rate for other patients (41% and 18%, respectively; P = 0.009) and the mortality rate for patients with cellulitis was lower (0% and 23%, respectively; P = 0.02), it was not significantly associated with clinical diagnosis after we controlled for the effect of age.

Among case-patients with community-acquired bacteremia, those who had hypothermia on admission (temperature less than equals 36.1 °C) were significantly more likely to die during the hospitalization than case-patients who presented with fever (temperature more than equals 38.3 °C) or intermediate temperatures (mortality rates of 42%, 12%, and 21%, respectively; P = 0.01). This association remained after we controlled for age.

Risk Factors for Disease

Most patients with invasive group B streptococcal infection had chronic underlying medical conditions (Table 1). Ninety-three percent of case-patients and 77% of controls had at least one of the following conditions: diabetes mellitus, cancer, cardiovascular disease, chronic renal insufficiency, alcoholism, injecting drug use, HIV infection, neurologic disease, or cirrhosis.

In univariate analysis, advanced age was associated with disease; age 65 years or older was associated with a 1.8 times greater risk for disease (P < 0.001). Sex and race were not significantly associated with disease. Case-patients were not more likely than controls to have known HIV infection or anatomical asplenia or to have received an immunosuppressive medication in the 2 weeks before hospital admission.

Community-Acquired Infection

Table 3 lists the conditions found in univariate analysis to be significantly more common among case-patients with community-acquired infection than among matched controls. Age and variables significantly associated with disease in univariate analysis were included in a multivariate model, the results of which are also shown in Table 3. The association of age and group B streptococcal disease was of borderline statistical significance after we controlled for the effect of other underlying conditions in the multivariate model.

Sixty-three percent of case-patients (108 of 171) with community-acquired infection had at least one of the six underlying conditions that were significantly associated with disease in multivariate analysis, and 58% (100 of 181) had diabetes mellitus, previous stroke, cirrhosis, or breast cancer.

Nosocomial Disease

Forty-eight of the 219 case-patients (22%) had nosocomial infection and had been hospitalized for a median of 4 days (range, 2 to 366 days) before collection of the specimen from which group B streptococcus was isolated. Table 3 lists the variables significantly associated with disease in the univariate analysis and in the multivariate analysis that controlled for both age and the variables found to be significant in the univariate analysis.

Discussion

Our results indicate that several chronic underlying conditions are independently associated with invasive group B streptococcal infection in nonpregnant adults. Diabetes mellitus was strongly associated with both community-acquired and nosocomial group B streptococcal disease. In addition, breast cancer, cirrhosis, neurogenic bladder, decubitus ulcer, and previous stroke were associated with community-acquired infection. Age was also associated with an increased risk for community-acquired and nosocomial disease but was of borderline statistical significance in the multivariate model of community-acquired disease.

These results expand on the findings of previous case series [5-11] and population-based studies [2, 12], which indicated that diabetes mellitus, cancer, liver disease, alcoholism, and neurologic disease are common among adults with group B streptococcal disease. The most recent population-based study also compared the estimated incidence of group B streptococcal disease in persons with diabetes mellitus with that in the total population and found a significantly higher risk for disease among adults with diabetes across all age groups. However, the magnitude of risk was highest for diabetic patients younger than 45 years of age [2]. An elevated risk for disease was also noted for persons with HIV infection and young adults with cancer.

We did not identify a significant association between group B streptococcal disease and the presence of malignant disorders (other than breast cancer) or HIV infection in any age group. This may be partly attributable to our study design, which would tend to bias against finding a true association if it exists, because our control group of hospitalized persons probably had a higher prevalence of underlying conditions than the general population. Therefore, our failure to detect a significant association does not mean that a condition is not a risk factor for disease. Conversely, a significant association between a condition and group B streptococcal disease in our study suggests that an even stronger association might be found if case-patients were compared with members of the general population. In addition, although we included 219 cases in our study, this sample size may not have been sufficient to detect an association for rare conditions. Asplenia in particular is a rare condition that has been documented in several patients with group B streptococcal sepsis [11, 19-21].

The absence of an association with malignant disorders in general may also be a result of the heterogeneous nature of the conditions in this category. As previously suggested [22], it is possible that only a subset of patients with cancer are at increased risk for group B streptococcal infection. We collected information on type of cancer, presence of metastases, and previous receipt of chemotherapy or radiation treatment and did not find an association between solid tumors, hematologic malignancy, metastatic cancer, or previous cancer treatment and either community-acquired or nosocomial group B streptococcal infection.

Reports of population-based surveillance for invasive group B streptococcal disease have also indicated elevated rates of disease among men and black persons [1, 2, 12]. The absence of an association between these variables and disease in our analysis suggests that sex and race are not in themselves risk factors for disease but may be surrogate markers for the presence of other conditions.

We identified case-patients through laboratory-based surveillance for sterile-site isolates of group B streptococcus. This system would not identify persons with group B streptococcal infections from whom sterile-site cultures were not obtained. This may have introduced a detection bias by preferentially including patients who were more likely to seek medical attention or who were more likely to be suspected of having a potentially treatable infection by their health care provider. However, because our surveillance system was population-based, we believe that our case-patients were more likely to be representative of the population of persons with invasive group B streptococcal infection than patients identified by surveillance at a single hospital or group of hospitals.

We could not include 71 of the 290 patients identified by the surveillance system in the study, and this may also have introduced bias. However, available data on the patients who were not included suggest that these patients were not significantly different from patients enrolled in the study. Because blood or other sterile-site cultures may not have been obtained from controls, some patients with unrecognized group B streptococcal infection may have been misclassified as controls. Although we cannot exclude this possibility, we believe that the number of misclassified patients, if any, is small and that this type of bias would be unlikely to significantly affect our findings.

Our findings suggest potential areas for further investigation of the pathogenesis of invasive group B streptococcal infection in nonpregnant adults. Low titers of antibody to group B streptococcal polysaccharide capsular antigens have been seen in women whose neonates have group B streptococcal infection [23, 24], suggesting that low titers of passively transferred maternal antibody are associated with neonatal infection. Serum samples from persons with insulin-dependent diabetes have shown inefficient phagocytic activity against type II group B streptococcus in vitro. This deficiency was corrected by the addition of specific antibody [25]. To our knowledge, immunologic factors associated with group B streptococcal infection among persons with other underlying diseases have not been specifically addressed.

The increased risk for group B streptococcal infection associated with certain underlying conditions may be a result of factors other than or in addition to defects in humoral immunity. Group B streptococcus can colonize the skin [26], gastrointestinal tract [27, 28], and nasopharynx [26] and can cause urinary tract infections in adults [29]. Several of the conditions associated with group B streptococcal disease in our study, such as decubitus ulcers, neurogenic bladder, mastectomy, and stroke may predispose patients to infection by altering host defenses and allowing invasion of the organism from a site of colonization.

Identification of conditions associated with an increased risk for infection also has implications for the targeting of group B streptococcal polysaccharide-protein conjugate vaccines currently under development for prevention of neonatal infection. These vaccines have been shown to be more immunogenic than polysaccharide vaccines and have induced protection in animal models [16, 30-32]. The major serotypes associated with human disease, including serotype V, which only recently emerged as an important cause of group B streptococcal disease [33, 34], have been conjugated to carrier proteins [15, 16, 30-32, 35]. Although isolate collection was not a part of this study, researchers doing serotyping of invasive group B streptococcal isolates from patients identified by the laboratory-based surveillance system in Atlanta and Maryland during roughly the same period found that nearly all the adult group B streptococcal isolates tested were the serotypes included in candidate vaccines [33, 34]. The apparently rapid emergence of type V disease emphasizes the importance of continued surveillance as a complement to vaccine development.

If a vaccine that is protective in adults is developed, our results indicate that approximately 60% of adults with community-acquired infection could be targeted on the basis of underlying conditions associated with an increased risk for infection. In our analysis, age was an independent risk factor for infection and persons 65 years of age or older had a significantly higher in-hospital mortality rate than younger persons with group B streptococcal infection. An increased risk for death as a consequence of infection has been used as a criteria in recommendations for influenza and pneumococcal vaccines [36, 37] and suggests that this group may also benefit from preventive measures. Sixteen percent (28 of 171) of case-patients with community-acquired infection were younger than 65 years of age and did not have any of the underlying conditions that were associated with disease in multivariate analysis. Although several underlying conditions were also associated with nosocomial infection, preventive measures may be less effective in preventing nosocomial disease, especially if severely ill hospitalized patients acquire infection as a complication of invasive procedures.

We identified several underlying medical conditions that were significantly more common among patients with invasive group B streptococcal disease than other hospitalized persons. Further research on immunologic factors associated with adult group B streptococcal infection and demonstration of the immunogenicity and efficacy of future vaccines are necessary before recommendations for preventive strategies for adults can be developed. In addition, cost-effectiveness and cost–utility evaluations of adult immunization strategies would aid in setting priorities for preventive measures. Our results suggest, however, that if the group B streptococcal vaccines being developed for prevention of neonatal disease protect against adult disease, a targeted vaccination strategy for adults at increased risk could substantially reduce the morbidity and mortality of this serious bacterial infection.

Drs. Schuchat and Wenger and Ms. Hilsdon: Mailstop C-09, Centers for Disease Control and Prevention, Atlanta, GA 30333.

Dr. Farley: Infectious Diseases Service, Veterans Affairs Medical Center (151), 1670 Clairmont Road, Decatur, GA 30033.

Dr. Harrison: Department of International Health, Johns Hopkins University School of Hygiene and Public Health, 615 North Wolfe Street, Baltimore, MD 21205.

Dr. Reingold: Division of Public Health, Biology, and Epidemiology, University of California at Berkeley, 140 Warren Hall #7360, Berkeley, CA 94720.