Multicenter national surveillance of S. pneumoniae was performed from November 1994 to April 1995 (16) and again from November 1997 to April 1998 (17). All isolates during these two surveillance studies were recovered from consecutive nonhospitalized patients from either the lower respiratory tract or a sterile site (blood or cerebrospinal fluid). Briefly, isolates were transported from study centers to a central laboratory, where they were confirmed as S. pneumoniae by conventional identification methods (16). Susceptibility testing was performed by the reference broth microdilution method recommended by the National Committee for Clinical Laboratory Standards (NCCLS) (18). Susceptibility was determined by using the established NCCLS breakpoints (19). For penicillin, breakpoints of 0.1 to 1.0 µg/mL for intermediate and > 2 µgg/mL for resistant were used; for this analysis, both intermediate and resistant categories were considered resistant.

Twenty-four medical centers were surveyed during both study periods. For 23 of these centers, data for outpatient antimicrobial-drug use were obtained for the surrounding metropolitan statistical area. These data were expressed in terms of number of prescriptions written per 100,000 population per month during the 48-month period that included the two surveillance studies (20). This period (May 1994 through April 1998) included four consecutive respiratory virus seasons. We divided the 23 medical centers into high-, intermediate-, and low-use centers for each antimicrobial-drug class. With the change in penicillin resistance as the dependent variable of interest, we used one-way ANOVA to compare mean change in resistance to penicillin between high-, intermediate-, and low-use centers. We then analyzed covariance models to evaluate the relationship between antimicrobial-drug use categories and changes in penicillin resistance. Alpha was set at 0.05, and all p-values were two-tailed.

We compiled the penicillin and erythromycin susceptibility test results for S. pneumoniae isolates collected in 1994-95 and 1997-98 from all 23 centers (Table 1). Overall, penicillin nonsusceptibility (MIC > 0.1 µg/mL) increased by 8.9%. In 1994-95, 269 (22.2%) of the 1,211 S. pneumoniae isolates were intermediate or fully resistant to penicillin, while in 1997-98, 337 (31.1%) of the 1,083 isolates were in these categories. When the change in percent penicillin susceptibility at each center was considered, the overall mean increase in penicillin resistance was 8.3% (-14.6% to 39.2%) among the 23 centers that participated in both surveys (Table 1).

Antimicrobial-drug use data for beta-lactams, tetracyclines, quinolones, and macrolides were calculated for the high-, intermediate-, and low-use tertiles in our analysis (Table 2). The mean increase in penicillin resistance was compared among high-, intermediate-, and low-use centers for the major antibiotic classes (Table 3). The beta-lactams were most strongly associated with an increase in penicillin resistance (2.8%, 8.8%, and 13.3% increases in low-, intermediate-, and high-use tertiles, respectively, p=0.20).

Univariate analysis of covariance was performed, with change in penicillin resistance as the dependent variable and the antimicrobial-drug use category for each antimicrobial-drug class as independent variables. When all classes for which data were available (beta-lactams, tetracyclines, macrolides, and quinolones) were entered into a model, only the macrolides and beta-lactams were statistically significant (p< 0.1) as explanatory variables and were therefore included in the final model (Table 4). Higher beta-lactam use was strongly associated with increased resistance to penicillin (F=8.7, p=0.008). Conversely, higher macrolide use was associated with decreased resistance to penicillin (F=5.4, p=0.031). The overall model explained a significant amount of the variance in penicillin resistance at these 23 centers (F=4.8, p=0.02).

A separate analysis showed no significant association between beta-lactam, macrolide, quinolone, or tetracycline use and change in the percentage of erythromycin resistance. However, an overall increase in erythromycin resistance was observed (Table 1).

Numerous studies have associated antimicrobial-drug use patterns in hospitals with the emergence of resistance among nosocomial pathogens (21-25). However, S. pneumoniae is usually acquired outside the hospital environment; therefore, establishing a relationship between antimicrobial-drug use and resistance requires outpatient data, as well as susceptibility test results. A large-scale study of this type is costly and difficult to perform in the United States, given the problems inherent in collecting accurate data from multiple outpatient settings. To generate hypotheses and support the planning of such a study, we used data collected for other purposes to explore the relationship between outpatient antimicrobial-drug use and resistance among S. pneumoniae isolates.

We found an association between the outpatient use of beta-lactam antimicrobial drugs in metropolitan areas and changes in the penicillin susceptibility of S. pneumoniae isolates sampled from tertiary care centers in those metropolitan areas. Determining whether this association is spurious or causal requires further investigation, given the limitations of our study design. Since information about each patient's previous antimicrobial-drug use was not available, we were unable to make a direct connection between patient use and risk for resistance. Furthermore, since antimicrobial-drug use data are presented as the total number of prescriptions per month in the population, the data may not accurately reflect use. Patient compliance, dosage prescribed, and duration of antibiotic use may differ from region to region. In addition, the data are for large populations, and the S. pneumoniae isolates represent a small sample from one study center in each metropolitan statistical area. These samples may not accurately reflect the true prevalence of resistance in the study population. For this reason, we grouped the study centers into tertiles on the basis of use, to decrease the impact of a small number of resistant isolates at a single study center. Finally, this analysis was retrospective. These surveillance surveys were not designed to evaluate the association between antimicrobial-drug use and changing resistance patterns among S. pneumoniae.

However, the use of antimicrobial agents in a population would be expected to contribute to the emergence and spread of resistance within that population, and our data support this hypothesis for beta-lactam use and penicillin resistance. The fact that beta-lactam use was associated with increased penicillin, but not erythromycin, resistance among pneumococcal isolates in our study suggests a specific association. Furthermore, this positive association with penicillin resistance was not seen for antimicrobial-drug classes other than the beta-lactams; neither was resistance associated with the total number of antimicrobial-drug prescriptions.

Erythromycin resistance also increased during our study. The lack of a strong association between use and erythromycin resistance may reflect the fact that beta-lactams were the most commonly prescribed in the metropolitan statistical areas we studied, and the impact of these drugs was therefore greater and easier to detect. In addition, a relationship between resistance to penicillin and resistance to virtually all other oral antimicrobial-drug classes has been described (2,16-17), making colinearity a potential problem in evaluating the impact of specific classes on resistance to a single antimicrobial agent or class. If the relationship between penicillin resistance and resistance to other antimicrobial-drug classes is due to the clonal spread of already multidrug-resistant strains (rather than emergence of resistance under antimicrobial pressure), the impact of a specific class of agent on the spread of a specific resistance in S. pneumoniae might vary by region, depending on the coresistance pattern of the predominant PRSP clones in that area.

Other investigators have reported an association between prescriptions for outpatients and rates of resistance in Western Europe (26), Hungary (27), and Iceland (28). In these studies, lower use of antimicrobial drugs in general and beta-lactams in particular is associated with lower rates of isolation of resistant strains. Our study supports this association and underscores the importance of implementing measures to decrease the inappropriate use of antibiotics in the outpatient setting (29).

Despite several limitations, our data support the hypothesis generated in previous studies that outpatient antimicrobial-drug use plays an important role in the development and spread of resistance. In future epidemiologic studies, antimicrobial-drug use should be carefully matched with resistance in well-defined populations and should include prospective evaluation of interventions to reduce the use of certain classes of antimicrobial agents for outpatients.