Defining the risk factors for S. aureus infection in patients with HIV infection is needed so that high-risk patients who may benefit from preventive strategies can be identified. We report a prospective, multihospital cohort study of HIV-infected patients that assessed the risk factors for S. aureus infection, with special emphasis on nasal carriage.

Methods

Top Methods Results Discussion Author & Article Info References

We enrolled all HIV-infected patients seen in the outpatient clinics of three acute care Veterans Affairs Medical Centers (Ann Arbor, Michigan; Boston, Massachusetts; and Pittsburgh, Pennsylvania) who agreed to participate. Patients were followed every 3 months for a minimum of 2 years or until death. Cultures of the anterior nares for S. aureus were obtained every 3 months and within 48 hours of hospital admission.

For patients who developed clinical S. aureus infection, all available clinical and nasal surveillance isolates were strain typed by using pulsed-field gel electrophoresis (10). Isolates were considered to represent the same strains if their profiles on pulsed-field gel electrophoresis did not differ by more than two restriction fragments (11).

Definitions

Patients were considered evaluable only if at least three nasal cultures had been obtained. Nasal carriage was classified as follows: 1) persistent if patients had two or more positive cultures separated by one or no negative culture, 2) intermittent if patients had two or more positive cultures with more than one negative culture, or 3) transient if only one of three or more cultures was positive (9). A noncarrier was defined as a patient in whom none of at least three cultures was positive.

Patients were considered to have community-acquired S. aureus infection if they acquired the infection as an outpatient and if they had not been hospitalized in the 4 weeks preceding the infection.

Statistical Analysis

We performed univariate analysis of contingency data by using the chi-square or Fisher exact test (Prophet, version 5.0, BBN, Cambridge, Massachusetts). Factors determined to be significant for S. aureus infection by univariate analysis (P < 0.1) were further assessed one by one in separate logistic regression models (Prophet) to search for possible confounding risks. Both odds ratios and attributable risks were calculated for univariate analyses; only odds ratios were calculated for multivariate analyses. Attributable risk, odds ratios, and 95% CIs were calculated by using True Epistat (Epistat Services, Richardson, Texas) (12). A Cox proportional-hazards model (True Epistat) was constructed to incorporate timing into the analysis. Kaplan-Meier probability plots were constructed with S. aureus infection as the end point. The curves were stratified by nasal carriage and CD4 cell count and were compared by using the Mantel-Cox rank test (Prophet). Kaplan-Meier plots were also used to evaluate mortality.

Results

Top Methods Results Discussion Author & Article Info References

A total of 231 enrolled patients had 1134 nares cultures, of which 26% (296 of 1134) were positive for S. aureus. Forty-nine percent of patients (114 of 231) had at least one culture positive for S. aureus. Thirteen percent of patients (30 of 231) were excluded from the analysis of nasal carriage because fewer than 3 nares cultures were performed. Thirty-four percent of patients (69 of 201) were nasal carriers; of these patients, 38% (26 of 69) were persistent carriers and 62% (43 of 69) were intermittent carriers. Twenty percent of patients (40 of 201) carried S. aureus transiently and were not considered to be nasal carriers. Ninety-two patients (46%) never had S. aureus isolated from their nares. Methicillin-susceptible S. aureus was recovered from the nares of 83% of patients who were nasal carriers (91 of 109).

Nasal carriage of S. aureus was significantly more common in patients with a dermatologic condition (52% [16 of 31]) than in those without such a condition (9% [15 of 170]) (P = 0.04; attributable risk, 0.09 [95% CI, 0.002 to 0.182]). Nasal carriage was also significantly more common in patients who were not receiving trimethoprim-sulfamethoxazole prophylaxis (79% [77 of 98]) than in those receiving this type of prophylaxis (25% [26 of 103]) (P = 0.008; attributable risk, 0.26 [CI, 0.07 to 0.44]). No association was seen between nasal carriage and CD4 count or neutropenia (leukocyte count < 1000 cells/mm³).

Staphylococcus aureus Infection

Thirteen patients developed 21 episodes of S. aureus infection (Table 1), consisting of bacteremia (10 episodes, including 2 with endocarditis), pneumonia (1 episode), and localized cutaneous or subcutaneous infections (10 episodes). Two patients had both bacteremia and localized subcutaneous infections. Seventy percent of the cases of bacteremia and 100% of the subcutaneous and cutaneous infections were community acquired.

Ten episodes of bacteremia occurred in nine patients; four of these episodes were due to methicillin-resistant S. aureus. The median CD4 cell count within the 2 months before onset of bacteremia was 50 cells/mm³ (range, 2 to 140 cells/mm³). Three of the nine bacteremic patients were neutropenic (absolute neutrophil count, 500 to 1000 cells/mm³ for two patients and <500 cells/mm³ for one patient) because of chemotherapy for lymphoma (one patient), ganciclovir therapy for cytomegalovirus retinitis (one patient), or end-stage AIDS (one patient). Of the five bacteremic patients for whom staphylococcal nasal carriage was evaluable, four were persistent or intermittent carriers. Three bacteremic patients had multiple episodes of S. aureus infections; all were nasal carriers with CD4 counts of 60 cells/mm³ or less (Table 1).

Factors Associated with Staphylococcus aureus Infection

Univariate Analysis

Patients who were nasal carriers of S. aureus (10% [7 of 69]) were significantly more likely than those who were not nasal carriers (3% [4 of 132]) to develop S. aureus infection (P = 0.04; odds ratio, 3.6 [95% CI, 0.9 to 15.4]; attributable risk, 0.44 [CI, 0.02 to 0.87]). Likewise, patients with vascular catheters (17% [5 of 29]) were significantly more likely than those without a vascular catheter (4% [8 of 202]) to develop S. aureus infection (P = 0.01; odds ratio, 5.1 [CI, 1.3 to 19.0]; attributable risk, 0.30 [CI, 0.004 to 0.588]). Patients with neutropenia were also more likely to develop S. aureus infection (27% [3 of 11]) than those without neutropenia (4.5% [10 of 220]) (P = 0.02; odds ratio, 7.9 [CI, 1.4 to 40.8]; attributable risk, 0.19 [CI, –0.04 to 0.43]).

Patients with lower CD4 counts were more likely to develop S. aureus infection. The median CD4 count for patients with S. aureus infection was 40 cells/mm³, whereas the median CD4 count for patients without infection was 160 cells/mm³ (P = 0.048). Patients with diabetes mellitus (22%) and dermatoses (13%) were more likely than patients without these conditions to have S. aureus infection (5% and 4%, respectively); these results were not statistically significant (P = 0.08 for both conditions; odds ratio, 5.5 [CI, 1.2 to 34.8] and 2.1 [CI, 0.75 to 12.3], respectively; and attributable risk, 0.12 [CI, –0.08 to 0.31] and 0.20 [CI, –0.079 to 0.48], respectively).

Multivariate Analyses

In a logistic regression model that included S. aureus nasal carriage and CD4 count, presence of a vascular catheter, or neutropenia, only nasal carriage, CD4 count (P = 0.04; odds ratio, 5.1 [CI, 1.04 to 24.7]), and presence of a vascular catheter (P = 0.04; odds ratio, 5.1 [CI, 1.04 to 24.7]) were significantly associated with S. aureus infections by multivariate analysis (Table 2). When nasal carriage, CD4 count, and presence of a vascular catheter were included in the model, only nasal carriage (P = 0.02; odds ratio, 5.1 [CI, 1.3 to 20]) and presence of a vascular catheter (P = 0.03; odds ratio, 4.9 [CI, 1.1 to 21.9]) were significantly associated with S. aureus infection. In a Cox proportional-hazards model, the same factors were found to be associated with S. aureus infection (Table 2). When we adjusted for different risk factors, the odds ratio and hazards ratio estimates for nasal carriage became unstable as the hazard ratio ranged from a nonsignificant 3.0 to a significant 5.8. Nonetheless, S. aureus infection and nasal carriage seem to be associated, although it is difficult to estimate the magnitude of the association. Adjusting by center did not substantially change the results.

The risk for S. aureus infection determined by Kaplan-Meier analysis was approximately 10% for every 6 months in patients who were nasal carriers of S. aureus and had CD4 counts less than 100 cells/mm³. These patients had a significantly higher rate of infection than did other patients (P = 0.02).

Outcome

The overall mortality rate was 35% (79 of 231 patients). Although the mortality rate did not differ between patients who were nasal carriers and those who were not nasal carriers (P > 0.2, Mantel-Cox test), it was significantly higher among those with clinically apparent S. aureus infection (62% [8 of 13]) than among those without such infection (32% [71 of 218]; P = 0.04, Mantel-Cox test).

Strain Typing

Both infecting and preceding surveillance isolates were available from seven patients with infection. Each patient was infected with a unique strain. For six of these seven patients, the infecting and surveillance isolates had indistinguishable SmaI digests on pulsed-field gel electrophoresis and were considered to represent the same strain.

Discussion

Top Methods Results Discussion Author & Article Info References

In our study, the prevalence of clinical S. aureus infection was 6% (13 of 231 patients), with 21 episodes of infection occurring in 13 patients. Eleven episodes were invasive, and 10 were local cutaneous or subcutaneous infections. Two of the invasive episodes were complicated by endocarditis with multiple pulmonary emboli (Table 1). Staphylococcus aureus bacteremia resulted in a mortality rate of 67% (6 of 9 patients). Patients with S. aureus bacteremia were twice as likely to die as other patients.

In addition to high morbidity and mortality, S. aureus infections were associated with a high recurrence rate, as noted by others (2). Three of seven patients who survived an initial S. aureus infection developed recurrent S. aureus infection. Recurrent S. aureus infection in patients with HIV infection has been attributed to phagocytic dysfunction or impaired intracellular killing by monocytes and granulocytes (13-15) but may also be related to persistent nasal carriage of staphylococci. In our patients, S. aureus carriage was clearly associated with S. aureus infection: Ten percent (7 of 69) of nasal carriers ultimately became infected with S. aureus compared with only 3% (4 of 132) of noncarriers (P = 0.04).

Studies by other investigators have also demonstrated an increase in frequency of S. aureus nasal carriage among HIV-infected patients compared with those without HIV infection (16, 17). Of our 231 patients, 49% (114 of 231) had at least one positive nares culture, a prevalence similar to that reported by others (27% to 55%) (16, 18).

Strain typing showed that S. aureus from the infected site was almost always the same strain as that previously cultured from the nares. This is consistent with the conclusion that S. aureus in the anterior nares serves as the reservoir for S. aureus colonization at other sites. Risk factors for staphylococcal infection were staphylococcal nasal carriage, presence of a vascular catheter, and CD4 count less than 100 cells/mm³. The risk for S. aureus infection over time was approximately 10% for every 6 months in patients who were nasal carriers of S. aureus and had CD4 counts less than 100 cells/mm³. This subgroup of patients with a high rate of infection might benefit from intervention directed toward clearing nasal carriage of staphylococci.

Study limitations are that we included only U.S. veterans, and our sample of patients with staphylococcal infection was small. However, given the biological plausibility of our findings, other investigators may want to confirm them with a larger sample and a more heterogeneous population of patients.

Prophylactic administration of antimicrobial agents that are effective against nasal carriage of S. aureus, such as rifampin or topical mupirocin, have been shown to decrease nasal carriage and, in turn, decrease the rate of clinical infection in patients undergoing hemodialysis (6, 19) and patients receiving long-term ambulatory peritoneal dialysis (7). Our study therefore provides the foundation for controlled trials of prophylaxis directed against nasal carriage of S. aureus in HIV-infected patients at greatest risk for S. aureus infection, such as those with low CD4 counts.