Limiting the Spread of Resistant Pneumococci: Biological and Epidemiologic Evidence for the Effectiveness of Alternative Interventions

Ecologic Niche of S. pneumoniaeThe human nasopharynx is the primary reservoir for S. pneumoniae and the main source of person-to-person transmission. Pneumococcal infections are preceded by bacterial colonization of the nasopharyngeal mucosa (44, 96), where the bacteria can persist as part of the commensal flora without causing disease. The nasopharyngeal flora is complex, consisting of normally nonpathogenic bacteria such as aerobic α-hemolytic streptococci, anaerobic streptococci, and Prevotella melaninogenica in addition to more commonly pathogenic organisms such as S. pneumoniae,Haemophilus influenzae, and Moraxella catarrhalis(11).

Nasopharyngeal carriage of S. pneumoniae is more common in young children than adults and varies by geographic region. Variation in the prevalence of nasopharyngeal carriage may be due to genetic differences in the host that affect the likelihood of nasopharyngeal colonization and also to socioeconomic conditions such as crowding, sanitation, family size, and day care contact (92). The duration of carriage also varies, depending on the host's age and on the serotype of the colonizing strain, and typically ranges between 1 and 17 months (44).

The factors responsible for the transition from carriage to disease remain poorly understood. In a longitudinal study of infants in Alabama, 15% of colonization events resulted in infection (44). Disruption of natural barriers, for example by damaged bronchial epithelium following influenza virus infection, can lead to invasive pneumococcal infections (75). Some viruses and cytokines also enhance bacterial adherence in vitro and may have a similar effect in vivo (75). Additionally, some strains ofS. pneumoniae have virulence determinants, for example, different capsular polysaccharide types, that enhance their likelihood of causing disease.

Genetic Basis of Pneumococcal ResistanceThe genetic basis of resistance plays a key role in determining how resistance develops and spreads. A number of biological features distinguish pneumococci from many other pathogens with acquired drug resistance, such as mycobacteria and Escherichia coli. First, both sensitive and resistant pneumococci can be carried asymptomatically in the nasopharynx, which is also the primary source of person-to-person transmission (46, 86). Second, for the majority of commonly used antibiotics, de novo pneumococcal resistance is rarely the result of single point mutations or plasmid carriage (86). Trimethoprim-sulfamethoxazole resistance and low-level fluoroquinolone resistance are notable exceptions in which resistance conferred by single point mutations is an important mechanism in clinical pneumococcal isolates (Table1). In contrast, resistance to most major classes of antibiotics (Table 1) is acquired either by transformation (the uptake of free DNA from closely related strains or species, which is then incorporated by homologous recombination) or transfer via conjugative transposons (large segments of DNA encoding antibiotic resistance and self-transfer functions).

Additionally, resistant strains can vary widely in their degree of resistance to certain drugs, a graded phenomenon measured as the MIC of antibiotic required to inhibit growth. Different MICs are associated with different genetic alterations. For the β-lactam antibiotics, high-level resistance requires sequential transformation events which may then be followed by point mutations, whereas intermediate resistance can be conferred by a single transformation event (Table 1). Recently, some clinical pneumococcal isolates were found to survive in the presence of vancomycin, although they could not grow in the presence of the drug and thus were considered tolerant rather than resistant (73). This vancomycin-tolerant phenotype was associated with tolerance to a range of antibiotics in addition to vancomycin (73).

These biological characteristics directly influence the population dynamics of pneumococcal resistance. In contrast to diseases that do not have a carrier state, asymptomatic carriers of pneumococci contribute more to pneumococcal transmission than infected individuals. Thus, reducing the number of infected individuals in a population does not necessarily reduce the potential for transmission of resistant strains within the community. Moreover, because the preconditions for acquiring resistance by transformation and conjugative transposons are stringent and de novo resistance does not result from single point mutations for the majority of antibiotic classes, de novo evolution of resistant pneumococci from an initially sensitive strain occurs only rarely. However, evolution of increased MICs in individuals initially infected with intermediate-resistant strains may be possible, particularly for drug classes such as fluoroquinolones, for which increases in MIC can be achieved by single point mutations. For most of the major antibiotic classes, resistant pneumococci spread primarily by clonal amplification (48), and resistant pneumococcal infections result primarily from the acquisition of resistant strains from other nasopharyngeal carriers in the community.

Association between Resistance Profile and SerogroupThere are 90 serogroups of S. pneumoniae, classified on the basis of antigenic differences in the capsular polysaccharide. Currently, most clinical isolates with high-level resistance to β-lactam antibiotics belong to serogroups 6, 9, 14, 19, and 23 (56). In particular, resistance is associated with serotypes 6B, 9V, 9A, 14, 19F, and 23F. Although each serotype encompasses genetic diversity, resistant isolates often belong to prevalent, well-documented clonal groups (41, 48). For reasons not well understood, a majority of these clonal groups have also acquired resistance to other drugs, including erythromycin, chloramphenicol, trimethoprim-sulfamethoxazole, and tetracycline.

It is not yet clear why these particular serotypes have a higher probability of containing antibiotic resistance determinants. One possibility is that these serotypes, which are commonly isolated from children, may be carried for longer durations and thus exposed to increased antibiotic pressure. It is also possible that barriers to transformation play a role in preventing some serotypes from acquiring horizontally spread resistance determinants. The proportion of clinical isolates that are transformable in the laboratory has not been determined. Competence for genetic transformation is a transient property, optimal competence conditions differ between strains, and various capsular types may reduce or totally inhibit transformation. Differences in the competence regulon may limit the acquisition of resistance to particular clonal groups (76, 93).

Despite evidence of a strong association between resistance patterns and serotype, the genes encoding capsular serogroup can also be exchanged between strains by transformation (18). Thus, it is possible that highly resistant clones may become members of serogroups that are currently not associated with multidrug resistance. The potential for capsular switching between strains with different resistance profiles was recently demonstrated in laboratory mouse experiments in which a multidrug-resistant serotype 23F strain acquired a serotype 3 capsule and also acquired increased virulence (72). Capsular switching has particular implications for the conjugate pneumococcal vaccine, which confers protection against a limited number of common serotypes.

Transport Medium for Nasopharyngeal SpecimensBecause the nasopharynx is the site of person-to-person transmission of pneumococci, characterizing the susceptibility profile of nasopharyngeal pneumococci is essential to assessing interventions to limit the spread of pneumococcal resistance. A recently developed transport medium for nasopharyngeal specimens enables nasopharyngeal specimens to be preserved for future pneumococcal isolation rather than requiring immediate processing (M. A. Bronsdon, K. L. O'Brien, P. Yagupsky, et al., 38th Intersci. Conf. Antimicrob. Agents Chemother. 1998, abstr. D-5 and D-25). The medium consists of skim milk, tryptone, glucose, and glycerol and has been shown to permit high rates of pneumococcal recovery when maintained at cold temperatures. Use of this transport medium facilitates field investigations and large-scale studies of pneumococcal carriage. A key advantage of this medium compared to direct plating methods is that the specimen collected on the swab is preserved in the transport medium and can be used for detailed analyses, such as the identification of minority clones.

Rapid Methods for Detecting ResistanceCurrently, isolating pneumococci from clinical specimens and determining penicillin susceptibility takes 2 to 3 days. More rapid approaches are being developed that will allow identification of pneumococcal isolates and resistance profiles in shorter times. Some approaches still require a primary bacterial isolation step, whereas others do not. Unique features of these more rapid methods under development are summarized in Table 2.

Detecting Carriage of Multiple StrainsEarly investigations of colonization by multiple pneumococcal strains attempted to amplify potential minority clones by inoculating specimens into mice. A current standard approach to identifying individuals who carry multiple pneumococcal strains with different resistance profiles is to characterize multiple colonies from a single nasopharyngeal specimen. However, if one strain predominates and other strains are present at very low densities, representing, for example, less than 1% of the total pneumococcal population, the number of colonies that need to be characterized from a single nasopharyngeal swab becomes unwieldy by conventional procedures. A pioneering study of multiple colonization by S. pneumoniae and H. influenzae in children in Papua New Guinea attempted to circumvent this problem by plating nasal specimens directly onto large (11-cm diameter) plates and serotyping 50 suspect pneumococcal colonies per specimen (43).

To assess whether an individual carries a minority resistant pneumococcal population, an antibiotic selective plating approach can be a useful alternative. At the initial pneumococcal isolation step, in addition to plating the nasopharyngeal specimen on a medium favorable for pneumococcal growth (typically a blood agar medium supplemented with gentamicin to prevent growth of gram-negative bacteria), the specimen can also be plated on a medium additionally supplemented with the antibiotic of interest, for example, penicillin. Individuals are identified as carrying a minority resistant strain if the isolates obtained from the nonselective plate are susceptible to penicillin and the isolate from the selective plate is confirmed by the MIC as being penicillin resistant. If the nasopharyngeal specimen is initially stored in an appropriate transport medium, selective plating can be reserved for individuals identified as carrying a dominant sensitive population.

A powerful approach would be to combine replica plating with colony DNA hybridization to detect minority populations containing specific resistance genes. In particular, this holds promise for gene targets such as ermAM, tetM, mefE, andcat, which are not found in sensitive pneumococcal populations.

Association between Recent Antibiotic Use and Carriage of Resistant PneumococciAlthough the genetic and environmental mechanisms that result in pneumococcal resistance are complex, the primary risk factor for the carriage and spread of resistant pneumococci is clear. Epidemiologic studies have repeatedly identified recent antibiotic use as the strongest risk factor for the carriage and spread of resistant pneumococci, at both the community (2, 7, 57) and individual (30, 35, 62) levels.

Moreover, among patients with invasive disease, recent antibiotic use correlates with an increased risk of infection with nonsusceptible (intermediate and resistant) pneumococci (35). This suggests not only that recent antimicrobial use increases the risk that an individual will carry, and therefore potentially transmit, resistant pneumococci, but that among infected individuals, it also increases the risk that individuals will develop invasive pneumococcal illness caused by resistant strains.

Additional risk factors for resistant pneumococcal carriage identified by observational studies include young age (with highest carriage among individuals <1 year old) and attendance at day care centers (3, 30, 33, 49, 58). Human immunodeficiency virus (HIV) infection may also be a risk factor in some populations, for example, urban areas of South Africa (54).

Mechanisms Responsible for Association between Antibiotic Use and CarriageThe biological mechanisms resulting in the association between recent antibiotic use and carriage of resistant S. pneumoniae have not been well studied. Possible mechanisms fall into three broad categories: the unmasking of a resistant pneumococcal clone that was present at a low density in the nasopharynx prior to antimicrobial therapy; replacement of sensitive pneumococci with resistant pneumococci acquired from the community during or after antimicrobial therapy; and within-host evolution of a resistant pneumococcal population in response to the selective pressure of antimicrobial therapy. Distinguishing features of these alternative mechanisms are noted in more detail in Table3. A better understanding of which of these mechanisms generate the observed association will provide an important framework for developing interventions to prevent the spread of resistance. For example, if replacement of sensitive strains with resistant strains from the community is the primary mechanism, then antibiotics that clear the nasopharynx may increase the risk of resistant carriage more than antibiotics that do not. In contrast, if unmasking is the primary mechanism, antibiotics that clear the nasopharynx of both sensitive and resistant strains may be the most effective treatment. The type of treatment least likely to select for resistance may vary with the prevalence of resistant strains in the community. This prevalence influences both the risk of carrying multiple pneumococcal strains and the risk of acquiring resistant strains during and after treatment.

Longitudinal studies of nasopharyngeal carriage before and after the initiation of antibiotic treatment provide data that can help to distinguish between alternative mechanisms for the increased risk of carrying resistant pneumococci following antimicrobial treatment (Table3). Two recent studies of nasopharyngeal carriage among children undergoing antibiotic therapy suggest that initially sensitive strains very rarely acquire resistance during antimicrobial therapy (19, 28). There is some evidence, however, that selection for genetic alterations increasing resistance may increase the MICs for initially nonsusceptible strains. The most common mechanism behind the association of antibiotic use and carriage of resistance, however, appeared to be either unmasking of a minority subpopulation of resistant strains or acquisition of new resistant strains during or after antimicrobial therapy (19, 28). A study that directly evaluated evidence for the unmasking hypothesis found an elevated rate of transition from carriage of susceptible to nonsusceptible strains in children undergoing antibiotic treatment but not in a control group, but the rate of transition from no pneumococci to carriage of nonsusceptible pneumococci was not elevated (39). This suggests that unmasking of resistant strains played a role in generating the association between recent antibiotic use and carriage of resistant strains. Although the sensitivity of the screening method for minority resistant clones in this study was low, 3.6% of children with sensitive pneumococci were found to carry minority resistant populations before treatment, and in one case DNA fingerprinting was able to document that the minority clone present at baseline was the resistant clone that predominated in the nasopharynx posttreatment (39).

RationaleThe identification of recent antibiotic use as the primary risk factor for carriage of resistant pneumococci has led to the development of control and prevention programs that promote more judicious antibiotic use. Inappropriate prescription of antibiotics contributes to a large proportion of unnecessary antibiotic use in both developed and developing countries. For example, studies of antibiotic prescribing in the United States have documented that antibiotics are often prescribed for the treatment of viral upper respiratory infections (42). There is evidence from several countries that rates of antibiotic use correlate directly with the prevalence of resistant pneumococcal strains (2, 7, 57). For example, a cross-sectional study of antimicrobial consumption and the carriage of penicillin-resistant pneumococci in children in five different communities in Iceland found that children in communities with higher levels of antimicrobial consumption were at higher risk for nasopharyngeal carriage of resistant pneumococci (2). Similarly, a longitudinal analysis of sentinel surveillance for invasive pneumococcal disease in the United States found that increases in the prevalence of penicillin-resistant pneumococci in the early 1990s correlated with increased prescription rates for certain β-lactam agents, while a fairly constant prevalence of tetracycline-resistant pneumococci was associated with slight decreases in the prescription rate for tetracycline (12).

Will a decline in antibiotic use resulting from judicious antibiotic use campaigns necessarily lead to declines in the prevalence of resistant pneumococci? If resistant strains are rare enough in the population that they may be lost through stochastic changes or if resistant strains have a fitness disadvantage relative to sensitive strains in the absence of antibiotics, reducing antibiotic selective pressure is expected to lead to a reduced prevalence of resistant strains. Few studies have directly addressed fitness disadvantages (for example, growth rate or competition disadvantages) associated with resistance in pneumococci. For the case of penicillin-resistant pneumococci, which result from alterations in essential cell wall synthesis enzymes called PBPs, it would seem likely that alterations in these enzymes result in growth deficiencies. However, resistant clinical isolates with PBP alterations not only maintain their resistance during extensive laboratory passaging in the absence of drug selection, but appear to be unaffected in general physiology (51). To address these and other issues of pneumococcal resistance, empiric observations of the impact of decreasing antibiotic selection pressure on the prevalence of resistant strains are necessary.

Impact on ResistanceCountrywide initiatives to reduce antibiotic resistance by reducing antibiotic use were attempted in Finland for the case of macrolide-resistant group A streptococci (GAS) and in Iceland for the case of penicillin-resistant S. pneumoniae. GAS have some features in common with S. pneumoniae, including a carriage state and a similar genetic mechanism of macrolide resistance. In Finland, in response to a dramatic rise in the prevalence of erythromycin-resistant GAS from 5% in 1988 to 1989 to 13% in 1990, national guidelines recommending reductions in the outpatient use of macrolide antibiotics were issued in 1991. Total macrolide consumption in outpatient therapy declined from 2.4 daily doses per 1,000 inhabitants in 1991 to a level of 1.4 in 1992; the daily doses per 1,000 inhabitants remained between 1.3 and 1.7 during the next 5 years (82). From 1992 to 1996 the prevalence of erythromycin-resistant GAS isolates also decreased, from 16.5% in 1992 to 8.6% in 1996. This decline was evident after 1993, approximately 2 years after the release of national guidelines. Thus, a reduction in antibiotic use was temporally associated with a subsequent decline in the prevalence of resistant GAS strains. Although the prevalence of erythromycin-resistant GAS declined, it remained close to 9%; moreover, there was an increase in the consumption of azithromycin and roxithromycin that correlated with the decline in erythromycin use.

In Iceland, the incidence of penicillin-nonsusceptible pneumococci among patients with invasive pneumococcal disease increased from 0% in 1988 to 16.3% in 1992 (K. G. Kristinsson, M. A. Hualmarsdottir, and T. Gudnason, 35th Intersci. Conf. Antimicrob. Agents Chemother. 1995, abstr. C9). A majority of penicillin-nonsusceptible strains were also resistant to macrolide antibiotics and to trimethoprim-sulfamethoxazole. The authorities responded to this rapid increase in nonsusceptible pneumococci by launching a public health campaign against the overuse of antibiotics that was aimed both at the public and at physicians. Additionally, children undergoing antibiotic therapy and known carriers of nonsusceptible pneumococci with colds or coughs were discouraged from going to day care centers. The number of daily doses of antibiotics in Iceland decreased from 23.6 per 1,000 inhabitants in 1990 to 21.4 in 1994, due largely to decreased use of macrolides and trimethoprim-sulfamethoxazole (Kristinsson et al., abstr.). Furthermore, in contrast to the steady increases of previous years, the incidence of penicillin-nonsusceptible invasive pneumococcal isolates declined from a peak of 19.8% in 1993 to 18.6% in 1994; among pneumococcal carriers, the proportion of healthy children attending day care who were colonized by a nonsusceptible strain also decreased from 20% in 1992 to 16% in 1995 (Kristinsson et al., abstr.). Thus, these data suggest that decreased antibiotic use reversed the trend of an increasing incidence of nonsusceptible strains in Iceland, although the reductions in both antibiotic consumption and the incidence of resistant pneumococci are not as striking as those observed for GAS in Finland.

Two recent controlled studies assessed the impact of well-defined judicious antibiotic use interventions on antibiotic prescribing practices and the incidence of resistant pneumococcal strains. The impact on antibiotic prescription rates of an intervention targeting four adult primary care practices in Denver, Colorado, was evaluated using a prospective, nonrandomized controlled trial design (42). The intervention, which was aimed at reducing antibiotic prescribing for outpatient adults diagnosed with uncomplicated acute bronchitis, consisted of household and office-based patient educational materials, clinician education, practice profiling, and academic detailing. During the 1-year study period, practices that received the full intervention showed a significant decline in antibiotic prescription rates, from 74 to 48%, whereas no significant decline was detected for control sites (42). Practices which received only a limited intervention, which consisted solely of office-based educational materials, also did not show significant declines. While this study did not address the impact of reduced prescription rates on the incidence of resistant pneumococci, it does suggest that active education of clinicians about the importance of judicious antibiotic use can lead to reduced rates of antibiotic prescription for specified conditions.

An evaluation of judicious antibiotic use interventions in Alaskan villages (K. Petersen, L. Chiou, J. Kokesh, L. Carlson, and S. Stool, Seventh Int. Symp. Recent Adv. Otitis Media, abstr. 67) was designed to evaluate both antibiotic prescribing rates and the prevalence of nonsusceptible pneumococci in villages receiving the intervention versus control villages. The intervention focused on otitis media and consisted of an educational program for village health aides and their consultants, workshops on the appropriate diagnosis of otitis media, and information sheets to providers and village residents. Preliminary results suggest that respiratory health care visits for children <5 years of age decreased by 23% and antimicrobial prescriptions per respiratory visit decreased by 22% in the intervention region, while no declines were evident in control villages that did not receive the intervention (78). Additionally, although the overall prevalence of pneumococcal carriers did not decrease in the intervention area, the proportion of all pneumococcal isolates that were nonsusceptible to penicillin (MIC, ≥0.12 μg/ml) declined by 28% in the intervention region but did not decline significantly in the control regions (78).

Thus, studies from the village to country level demonstrate that judicious antibiotic use interventions can result in reduced antibiotic prescriptions and also lead to decreases in pneumococcal resistance. Declines in penicillin-resistant pneumococci in Iceland and in Alaskan villages receiving interventions suggest that judicious-use campaigns may have the potential to limit the spread of resistant pneumococci. Further controlled studies of interventions designed to look at the prevalence of resistant pneumococci as an outcome are needed to address the effectiveness of this class of interventions. Questions of importance include the time required for clinically relevant declines in resistance and whether declines in resistance are unique to a particular set of circumstances (e.g., small communities with centralized health care such as in Alaska) or can be expected as a general result when antibiotic use is decreased.

Shorter Duration of TherapyFor pathogens without a carrier state, the goal of antibiotic therapy is to eradicate the pathogen. Thus, treatment must last long enough to ensure that the pathogen is eradicated. In contrast, for pathogens such as S. pneumoniae which are commonly carried asymptomatically, eradication of the pathogen from the site of carriage is not always possible. Consequently, shorter durations of therapy have been proposed as a means of resolving clinical symptoms and at the same time reducing an individual's exposure to antibiotics. Moreover, shorter courses may reduce the total amount of time that pathogens are exposed to MICs below levels that inhibit resistant strains but above levels that inhibit sensitive strains. Such conditions exert strong selective pressure for resistant strains.

A number of recent studies have demonstrated that shorter-course antimicrobial therapy for acute otitis media is as effective clinically as the standard 10-day course of amoxicillin for many patients (79). The World Health Organization currently recommends 5 days of treatment with trimethoprim-sulfamethoxazole or amoxicillin for treatment of acute respiratory infections in developing countries. Additionally, the recent introduction of drugs such as azithromycin and ceftriaxone, which can be administered in short courses, has led to an increasing popularity of short-course treatments for pneumococcal infections.

In some cases, short-course therapies reduce a patient's exposure to antibiotics. This may reduce the risk of carrying resistant pneumococci following antibiotic treatment, depending on how quickly unmasking or replacement occurs. However, although duration may be shorter with some treatments, for example, a single dose of ceftriaxone, antibiotic concentrations may still remain above the MIC for long periods following treatment (24). In such cases, a shorter treatment duration does not result in reduced exposure to antibiotics.

Consistent with the hypothesis that shorter treatment durations decrease the probability of carrying a dominant resistant population, an observational study of 941 children 3 to 6 years old in France identified a low daily dose of β-lactam antibiotic combined with a long duration of treatment (>5 days) as risk factors for pharyngeal carriage of resistant strains (45). However, a recent multicenter trial of 5 versus 10 days of amoxicillin-clavulanate therapy for acute otitis media for young children (20) found that the overall risk of a child's carrying penicillin-nonsusceptibleS. pneumoniae did not increase after antibiotic treatment in either treatment group. Because this study was a randomized, double-blind trial with sufficient power to detect other differences between treatment groups, these results may reflect the impact of treatment duration more clearly than those from the observational, retrospective study. Moreover, a study investigating changes in the nasopharyngeal flora 3 to 4 days after the initiation of antibiotic therapy found that substantial changes in the S. pneumoniaepopulation had already occurred by that time (28). Together, these observations suggest that shorter-course therapy may not reduce the risk of carrying resistant pneumococci and that significant changes in pneumococcal populations occur within the first days of antimicrobial therapy. Further controlled studies designed to assess the risk of carrying resistant pneumococci after different durations of antimicrobial therapy are needed to directly address this question. The challenge in such studies is obtaining a sufficient sample size of children posttreatment who still carry pneumococci.

Higher-Dose TherapyHigher-dose therapy has been suggested to reduce the risk of acquiring a dominant resistant pathogen population, regardless of whether a pathogen has a carrier state (Table 4). The theory behind this modification is to achieve high enough drug concentrations at the site of infection to eradicate both sensitive and resistant strains. In vivo studies of respiratory pathogens using animal models (22, 23) have identified specific pharmacokinetic and pharmacodynamic parameters that maximize the bactericidal activity of antibiotics. For β-lactam drugs, macrolides, and trimethoprim-sulfamethoxazole, the duration of time that serum levels remain above the MIC has been identified as the important determinant of bacteriologic efficacy. For such drugs, bactericidal efficacies of 85 to 100% were generally achieved if the drug concentration exceeded the MIC for at least 40% of the dosing interval (23). Studies of bacteriologic efficacy in children with acute otitis media have confirmed that bacteriologic cure is associated with dosing regimens in which β-lactam concentrations exceeded the MIC for 40% or more of the dosing interval (23). In contrast, for quinolones, aminoglycosides, azalides, and azithromycin, the area under the serum concentration-versus-time curve in relationship to the MIC is the most important parameter (23, 25). For these drugs, bactericidal efficacy increases with drug concentration.

This pharmacological framework suggests that if serum antibiotic concentrations can be maintained above the MIC for at least 40% of the dosing interval, even strains with high-level resistance to β-lactam or macrolide antibiotics will be eradicated from the site of infection. This has led to the hypothesis that high-dose therapy may also eradicate resistant pneumococci from the nasopharynx, thus reducing the risk of resistant pneumococcal carriage following therapy. Direct investigations of antibiotic concentrations required to eradicate pneumococci from the nasopharynx have not been performed, although there is evidence that carriage rates are reduced significantly following antibiotic treatment.

Even if high-dose therapy is more likely to eliminate nasopharyngeal carriage of S. pneumoniae resistant to the drug used for treatment, this will not prevent acquisition of resistant strains from the community. In fact, if, as some studies suggest, there is a “niche” for pneumococci in the nasopharynx that opens when pneumococci or other members of the nasopharyngeal flora are cleared by antimicrobial therapy (11), high-dose therapy may increase the likelihood of recolonization by a resistant strain in communities where the prevalence of resistance is high. Again, these important questions can be best resolved by clinical trials in which participants are randomized to receive either low or high doses of a given antibiotic and nasopharyngeal carriage is carefully and sequentially monitored.

Choosing Drugs That Minimize Selection for ResistanceAnother proposed treatment modification to limit the spread of resistance is to use drugs that minimize selection for resistance. This modification was initially proposed for pathogens for which within-host evolution of resistance in response to treatment is a primary concern. In contrast, for pneumococci, the principal concern is not to avoid within-host evolution of the pathogen but to minimize the opportunities for unmasking of minority resistant strains or acquisition of resistant strains from the community.

While transformation is the predominant mechanism for resistance to β-lactam antibiotics in clinical isolates of S. pneumoniae, antibiotic-resistant point mutants are readily selectable in the laboratory. There is also laboratory evidence that drugs differ in their activity against S. pneumoniae and in their potential to select for resistance and cross-resistance. For example, an in vitro study of S. pneumoniae (14) found that aminopenicillins selected for resistance to themselves and to cephalosporins, while cephalosporins, with the exception of cefixime, were shown to select only for resistance to themselves. This is consistent with observations that cephalosporin resistance is conferred by alterations to the PBP genes pbp1a andpbp2x, while high-level penicillin resistance also requires alterations in those two genes and in the pbp2b gene (47, 71). The relevance of these laboratory findings to clinical situations is unclear, since point mutations do not appear to play a similar role in in vivo bacterial populations.

In the clinical context, studies of nasopharyngeal carriage of S. pneumoniae before and after antibiotic treatment show clearly that some drugs eliminate S. pneumoniae from the nasopharynx more effectively than others. A study of the effects of cefpodoxime-proxetil and amoxicillin-clavulanate treatment on nasopharyngeal carriage ofS. pneumoniae in children in France (19) found that the amoxicillin-clavulanate treatment led to a significantly greater decline in the number of children carrying S. pneumoniae posttreatment (a 64% decline in the amoxicillin-clavulanate group versus a 41% decline in the cefpodoxime-proxetil group). Similarly, a study comparing the effects of cefuroxime-axetil and cefaclor on nasopharyngeal carriage during the first days following treatment initiation found that the decline in the prevalence of S. pneumoniae carriage in the cefuroxime-axetil group at day 4 or 5 following treatment initiation was three times greater than in the cefaclor group (28).

These studies, however, were not designed to assess whether the drugs with lower activity against S. pneumoniae were also associated with a decreased risk of carrying a resistant strain posttreatment. Recent investigations of the impact of azithromycin on streptococcal carriage have shown that azithromycin is extremely effective at reducing nasopharyngeal carriage of S. pneumoniae. However, two studies of mass treatment with azithromycin found that this resulted in an increased prevalence of macrolide-resistant S. pneumoniae 1 month after treatment (63, 70).

Thus, available evidence, along with the fact that pneumococcal resistance has been observed for all of the major drug classes used to treat pneumococcal infections, suggests that trying to minimize selection for resistant pneumococci by choosing alternative drugs for treatment is likely to play only a minor role in minimizing the spread of resistant pneumococci.

Treatment with Multiple DrugsTreatment of infections with multiple drugs has been a cornerstone of minimizing the spread of resistance in both HIV andMycobacterium tuberculosis. For people infected with these pathogens, treatment durations are often long (many years for HIV and 6 to 12 months for M. tuberculosis) and within-host evolution of resistance is common. Because the probability that a strain will acquire resistance to two or more drugs is the product of the probabilities that a strain will acquire resistance to each individual drug, simultaneous treatment with multiple drugs is effective at preventing infections with resistant strains. Although treatment of pneumococcal infections with multiple drugs has been suggested to improve clinical efficacy in some circumstances (16, 32), there is currently no evidence to suggest that it is an effective intervention to limit the development of resistant pneumococci. Because treatment durations for persons infected with pneumococci are typically much shorter than for persons with tuberculosis and HIV, and because for most antibiotic classes acquisition of resistance is rarely due to single point mutations, treatment with multiple drugs is unlikely to be a useful approach to prevent the acquisition of pneumococcal resistance. If the unmasking of minority resistant strains is a primary mechanism of acquiring dominant resistant populations, treatment with multiple drugs might in some cases reduce the potential for unmasking to occur. However, if acquisition of resistant strains from the community is the primary mode of acquiring a dominant resistant pneumococcal population, treatment with multiple drugs may have the adverse effect of exerting selection pressure for acquisition of strains with multiple resistance. The risk of this will depend on the prevalence of multidrug resistance in the community.

Cycling of DrugsMany hospitals have implemented formulary control of drugs administered and have scheduled cyclical changes of antibiotics at the hospital level in an effort to reduce the incidence of multiresistant nosocomial infections (68). While formulary control of antibiotics can lead to more judicious antibiotic use and can successfully limit the use of antibiotics such as vancomycin (83), evidence that “cycling” of antibiotics (periodically substituting a new member of a drug family or a new drug family for one currently in use) reduces the incidence of resistant infections remains unclear (68). Moreover, it is possible that some forms of cycling can increase the risk of acquiring a multiresistant infection (9). Multicenter studies that control for confounding factors will prove useful in assessing the impact of such strategies (68).

While formulary control may help limit the spread of resistant nosocomial infections, the majority of pneumococcal infections are community acquired, and except in small, isolated communities served by a limited number of health facilities, such control of antibiotics is not feasible. If cycling is found to be useful for preventing resistance in common nosocomial infections, it is possible that a similar strategy could minimize the development of multiresistant pneumococcal disease in patients in long-term care facilities, for example, the elderly and persons with AIDS.

Improved Compliance with Treatment RegimensDirectly observed therapy for individuals with tuberculosis has been an important intervention to minimize the risk of developing a resistant infection. Would a similar emphasis on improved compliance with treatments for pneumococcal infections help minimize the spread of resistant pneumococci? Increased compliance may improve treatment outcome by increasing the time that antibiotic concentrations are maintained above the MIC (23). Whether improved patient compliance also reduces the patient's risk of carrying resistant pneumococci is less clear. Missing antibiotic doses may increase the time during which the concentration of antibiotic in the nasopharynx is in the dangerous zone between the MICs for sensitive and resistant strains, increasing the selective pressure for resistance. Depending on the pattern of how doses are missed, however, the concentration of antibiotic may also drop below the MIC, reducing the selective pressure for either unmasking or replacement. Moreover, because the etiologic agent causing respiratory infections is rarely identified and antibiotics are often prescribed for nonbacterial infections, a general public health campaign to increase compliance with antibiotics prescribed for respiratory symptoms might paradoxically contribute to the inappropriate use of antibiotics. Thus, for the case of suspected pneumococcal infections, it is unclear how the traditional message that patients should finish their entire course of medication affects pneumococcal resistance.

Noncompliance may be more important in developing countries, where antibiotics are often available without a prescription. In some countries this may contribute to misuse and overuse of antibiotics; for example, surveillance in a rural area of Bangladesh showed that 48% of the antibiotics purchased for adults were purchased in quantities that were often less than a single day's dose (50). However, the availability of antibiotics without a prescription does not always result in antibiotic use without a physician's advice. Data from a periurban community in Mexico suggest that the majority of antibiotics were prescribed by a physician (13); a study of antibiotic use in Manila, the Philippines, suggests that while two thirds of antimicrobial agents were purchased without a prescription, over half of these purchases were made upon the recommendation of a physician (60).

Pneumococcal VaccinesA polysaccharide vaccine that is currently available can protect individuals >2 years of age against invasive pneumococcal infections caused by 23 common S. pneumoniae serotypes. The efficacy of this vaccine varies depending on a person's age and immune status; in case-control studies, the effectiveness of the vaccine against invasive disease has ranged from 56 to 81% (15). Strongest recommendations for vaccination are for persons over 65 years of age and those aged 2 to 64 years with the following risk factors: chronic cardiovascular disease, chronic pulmonary disease, diabetes mellitus, or functional asplenia (15).

However, the polysaccharide vaccine suffers from the following limitations: it does not induce an effective immune response in children less than 2 years old, an age group with a high burden of disease; it does not protect against nonbacteremic pneumococcal disease or common upper respiratory diseases (e.g., acute otitis media and sinusitis); it does not protect against carriage of pneumococci; and declining antibody levels following vaccination suggest that vaccine-induced immunity may wane over a 5- to 10-year period. The fact that the vaccine is not indicated for children <2 years of age and is primarily recommended for the elderly prevents it from being an effective intervention against pneumococcal disease in most developing countries. Because children are a large reservoir of resistant pneumococci, and because the vaccine has no effect on pneumococcal carriage, the polysaccharide vaccine has little potential to control the spread of resistant pneumococci.

Pneumococcal conjugate vaccines, in which purified polysaccharides of the epidemiologically most important serotypes are conjugated to a carrier protein, have recently been developed and tested in clinical trials. A 7-valent formulation received Food and Drug Administration licensure in February 2000, and several other formulations are also undergoing clinical trials. Preliminary data suggest that these vaccines are highly effective for children <2 years of age. No significant safety problems associated with vaccination have occurred in phase I to III trials conducted since 1992 (38). Clinical trials show that the vaccine is more than 95% effective at preventing invasive disease due to serotypes included in the vaccine (61; S. Black, H. Shinefield, P. Ray, et al., 38th Intersci. Conf. Antimicrob. Agents Chemother. 1998, abstr. LB-9). Furthermore, the vaccine has some efficacy against noninvasive disease (38).

The vaccine's ability to prevent invasive pneumococcal disease may also present an opportunity to reduce empiric use of antibiotics for young children with fever of unknown origin. Concern about invasive pneumococcal disease among young children who present with fever of unknown source is a common reason for antibiotic prescription for this age group. The importance of such empiric therapy has been questioned in the United States now that the incidence of invasive disease caused by H. influenzae type b has declined dramatically (66). Such concerns may be even less well founded among children who also received appropriate vaccination with the pneumococcal conjugate vaccine, and thus antibiotic use in this age group may be reduced in settings with high vaccination coverage.

Preliminary data also suggest that the pneumococcal conjugate vaccine reduces carriage of vaccine serotypes (31, 74), similar to an unexpected benefit of the conjugate H. influenzae type b vaccine (69). This effect has been demonstrated in infants (31) as well as toddlers (30), and it raises the possibility that the conjugate vaccine may be effective at preventing colonization with vaccine-related serotypes, reducing the opportunity for transmission of these serotypes within communities. Because the serotypes in current formulations of the vaccine include pediatric serotypes that are commonly associated with antibiotic resistance, vaccination of infants may reduce the circulation of antibiotic-resistant strains (31). Additionally, if the conjugate vaccine is effective at preventing invasive disease in children, and in particular if it reduces the rate of otitis media, antimicrobial therapy for pneumococcal infections is likely to decline, thus reducing the selective pressure for resistance. There is the possibility, however, that the vaccine prevents carriage of vaccine-included serotypes only. Evidence from a vaccine trial in the Gambia (74) and a recent study of day care attendees in Israel (R. Dagan, N. Givon, P. Yagupsky, et al., 38th Intersci. Conf. Antimicrob. Agents Chemother., 1998, abstr. G-52) suggest that carriage of nonvaccine serotypes increases following vaccination, although a previous study in Israel did not find evidence of increased carriage of nonvaccine serotypes (29). The impact of mass vaccination on carriage of nonvaccine serotypes remains to be seen, as well as whether widespread use of the conjugate vaccine increases the selective pressure for serotype switching or the acquisition of antibiotic resistance in strains not covered by the vaccine.

Xylitol SugarXylitol is a sugar that has been used as a sweetening substitute for sucrose because it has a preventive effect against dental caries. Recent evidence suggests that in addition to inhibiting growth ofStreptococcus mutans, the primary cause of dental caries, this sugar can inhibit the growth of S. pneumoniae and possibly also of other bacterial colonizers of the nasopharynx such asH. influenzae (90, 91).

A controlled trial of healthy children attending day care centers in Finland randomized the children to control groups that received syrup, chewing gum, and lozenges sweetened with sucrose and a low concentration of xylitol or to treatment groups that received a high daily dose of xylitol in the form of syrup, chewing gum, or lozenges (91). The children were then monitored for a 3-month period, and information on respiratory infections, acute otitis media, and antibiotic therapy was collected. While the xylitol exposure did not reduce the overall number of respiratory infections in treatment groups relative to their respective control groups, children who regularly received xylitol in the form of syrup or chewing gum had a significant (30 to 40%) reduction in the occurrence of acute otitis media, and all treatment groups received fewer antibiotic prescriptions than did controls (91).

Prophylactic use of xylitol may thus protect against acute otitis media, particularly for children with high exposure to S. pneumoniae, such as those attending day care centers. Because acute otitis media is the primary reason for prescription of antibiotic therapy in developed countries, reducing the incidence of acute otitis media could greatly reduce antibiotic use in these settings. In the above study, however, xylitol in all groups was administered five times per day; such frequent dosing may be difficult to achieve, and studies of the minimum dosing required to get a significant protective effect would be useful.