Macrolide Resistance Among Invasive Streptococcus pneumoniae Isolates FREE

Macrolides are recommended as first-line therapy for adults in the United States and Canada with community-acquired pneumonia.^1- 3 Newer macrolides (eg, azithromycin, clarithromycin) are popular because of convenient dosing schedules and fewer adverse effects. Unfortunately, resistance to macrolides among Streptococcus pneumoniae, the most common cause of community-acquired pneumonia,⁴ has been increasing in the United States.^5- 8

Pneumococcal macrolide resistance is usually expressed as 1 of 2 phenotypes. The first, known as M phenotype, is an efflux pump associated with the mefE gene that results in efflux of macrolides from the cell.⁹ These organisms exhibit resistance to 14- and 15-member macrolide antibiotics, including erythromycin, clarithromycin, and azithromycin. M phenotype isolates typically have moderate levels of macrolide resistance (ie, minimum inhibitory concentrations [MICs] in the range of 1-32 µg/mL) and are almost always susceptible to clindamycin.¹⁰ A second phenotype, MLS_B, results from methylation of adenine residues in the 23S ribosomal RNA (rRNA) peptidyl transferase center, domain 5, and is the result of the ermAM gene encoding the erythromycin-ribosomal methylase.^11- 12 The methylation blocks the binding of macrolide, lincosamide (eg, clindamycin), and streptogramin B (eg, quinupristin/dalfopristin) antibiotics. The MLS_B phenotype is associated with very high macrolide MICs (>64 µg/mL) and resistance to clindamycin. This phenotype is often associated with resistance to tetracyclines and chloramphenicol, as genes encoding this resistance phenotype can be clustered on conjugative and composite transposons.¹⁰ Other reported mechanisms of erythromycin resistance are rare^10,13; therefore, the resistance phenotypes described above correlate well with genetic subtype.¹⁴

Resistance phenotypes may have relevance for clinicians; some have postulated that MICs of M phenotype strains are in a range where pneumonia patients may respond to newer macrolides,¹⁵ whereas MICs of MLS_B phenotype strains clearly are in a range where treatment failures could occur. Gay et al¹⁶ reported that increases in the proportion of pneumococci with M phenotype explained the increase in macrolide resistance. The study also noted that the macrolide MICs of M phenotype isolates were increasing. We examined trends of macrolide resistance among invasive S pneumoniae across the United States, trends in macrolide use, the prevalence of strains with M and MLS_B phenotypes, factors associated with macrolide resistance, and MICs of macrolide-resistant isolates.

From 1995 to 1999, the Active Bacterial Core surveillance (ABCs) system conducted continuous active, laboratory-based surveillance for invasive S pneumoniae in 30 counties (15.4 million people, estimated population in 1999) throughout Georgia, California, Minnesota, Oregon, Tennessee, and Maryland as well as the entire state of Connecticut.⁸ A 7-county New York area joined ABCs in 1998. The state of New York is included in the 1999 surveillance results, but is not included in the overall trend analysis due to lack of continuous data. ABCs defines a case of invasive pneumococcal disease as the isolation of S pneumoniae from a normally sterile site. ABCs methods are defined elsewhere.^8,17

Isolates were defined as susceptible, intermediate, or resistant to agents tested according to National Committee for Clinicial Laboratory Standards definitions.¹⁸ We defined an isolate as being macrolide-resistant if it was intermediate or resistant to a particular agent. Erythromycin-resistant isolates (MIC >0.5 µg/mL) were further classified as having either M or MLS_B phenotype. Isolates that were erythromycin-resistant and clindamycin-susceptible were classified as M phenotype, and those that were both erythromycin- and clindamycin-resistant were classified as MLS_B phenotype.¹⁴

Statistical analyses were conducted using SAS version 6 (SAS Institute Inc, Cary, NC) statistical software. We used the χ² test to compare proportions and χ² test for trend with temporal analyses. P<.05 was considered statistically significant. Cumulative incidence rates for 1999 were calculated using 1999 population estimates from the US Census Bureau. Data for macrolide prescriptions in the United States were obtained from the National Ambulatory Medical Care Survey/National Hospital Ambulatory Medical Care Survey conducted by the National Center for Health Statistics at the Centers for Disease Control and Prevention.^19- 21

We determined odds ratios for groups at risk for macrolide-resistant infections in 1999 using multivariable logistic regression modeling. We constructed 2 logistic regression models: (1) erythromycin-susceptible vs M phenotype and (2) erythromycin-susceptible vs MLS_B phenotype. The dependent (outcome) variable was erythromycin susceptibility and the independent (predictor) variables included ABCs site, age group, race, and penicillin susceptibility. We assessed colinearity and 2-way interactions for all variables in the final multivariable model.

From 1995 to 1999, 18430 cases of invasive S pneumoniae infection were identified in 7 ABCs areas (excluding New York) ranging from 2505 in 1995 to 3419 in 1999. Isolates from 15 481 (84%) cases underwent antimicrobial susceptibility testing; 2273 (14.7%) had decreased susceptibility to macrolides; 18 (0.1%) were intermediate and 2255 (14.6%) fully resistant. Macrolide resistance increased from 10.6% of S pneumoniae isolates in 1995 to 20.4% in 1999. Increases in the proportion of isolates with macrolide resistance were seen in each individual surveillance site; increases were significant in all states except California (P = .20).

The proportion of isolates with M phenotype increased from 7.4% in 1995 to 16.5% in 1999 (P<.001) and accounted for the overall increase seen in macrolide resistance. An increase in the proportion of isolates with M phenotype was seen for children younger than 5 years and persons 5 years or older (Figure 1A and Figure 1B). The erythromycin MIC₅₀ increased in M phenotype isolates from 4 µg/mL in 1995 to 8 µg/mL in 1999. During the 5-year period, the proportion of isolates with MLS_B phenotypes remained stable (3.4% in 1995 and 3.7% in 1999).

The number of macrolide prescriptions increased significantly (13%) from approximately 17.7 million prescriptions (69 prescriptions/1000 persons) in 1993 to approximately 21.2 million prescriptions (78 prescriptions/1000 persons) in 1999 (P = .01). The most dramatic increases were among children younger than 5 years, in whom macrolide prescriptions increased 320% from 0.7 million prescriptions (37 prescriptions/1000 persons) in 1993 to approximately 3.1 million prescriptions (164 prescriptions/1000 persons) in 1999 (Figure 1A). In children younger than 5 years, azithromycin has comprised the majority of recent macrolide prescriptions with nearly 0.8 million prescriptions (41 prescriptions/1000 persons) in 1996 increasing to nearly 2.3 million prescriptions (120 prescriptions/1000 persons) in 1999. Total macrolide prescription rates among individuals 5 years and older was unchanged with 16.9 prescriptions (71 prescriptions/1000 persons) in 1993 and 18.1 million prescriptions (71 prescriptions/1000 persons) in 1999 (Figure 1B). In this group, azithromycin accounted for 3 million prescriptions (12 prescriptions/1000 persons) in 1996 and 7.8 million prescriptions (37 prescriptions/1000 persons) in 1999. Prescriptions for all antibiotics decreased 15% (535 to 450 prescriptions/1000 persons from 1993 to 1999, respectively) with a 20% decrease among those younger than 5 years (1408 to 1130 prescriptions/1000 persons from 1993 to 1999, respectively).

We identified 4132 invasive pneumococcal cases in the ABCs areas in 1999, of which 3620 (87.6%) isolates underwent antimicrobial susceptibility testing. S pneumoniae was isolated from patients with bacteremic pneumonia (n = 1957, 54.1%), bacteremia and no focal diagnosis (n = 1342, 37.1%), meningitis (n = 192, 5.3%), and other diagnoses (n = 129, 3.6%). Children younger than 5 years were the source for 913 (25.2%) isolates. Isolates collected in each surveillance region ranged from 199 in California to 726 in Georgia.

Overall, 709 (19.6%) strains were fully resistant (MIC ≥1.0 µg/mL) and 2 (0.1%) strains were intermediate (MIC = 0.5 µg/mL) to erythromycin. M phenotype isolates accounted for 82% and MLS_B phenotype accounted for 18% of macrolide-resistant isolates. Macrolide resistance was more common in infections occurring in children younger than 5 years (279/913, 30.6%) than in individuals 5 years and older (431/2698, 16.0%; P = .001). The proportion of macrolide-resistant isolates was higher among whites (439/1914, 22.9%) than blacks (179/1246, 14.4%; P = .001). Macrolide resistance was most common in Georgia (31.5%) and Tennessee (29.1%) and least common in New York (9.0%) and California (8.5%).

Serotype information was available for 3617 (87.5%) isolates, 19.6% of which were macrolide-resistant. Serotype 14 accounted for 44.5% of resistant strains. Among macrolide-resistant isolates, 87% of strains from children younger than 5 years and 72% of strains from individuals 5 years and older were serotypes included in the 7-valent conjugate vaccine. For those individuals 5 years and older, 95% of macrolide-resistant isolates were serotypes represented in the 23-valent polysaccharide vaccine.

A multivariable analysis of independent factors associated with M phenotype found a significantly higher proportion of M phenotype isolates among children younger than 5 years than from individuals 5 years and older (P<.001), from whites than from blacks (P<.001), and from individuals living in Georgia (P = .002), Tennessee (P = .01), and Minnesota (P = .01) compared with those living in California (Table 1). Decreased susceptibility to penicillin was found to be an independent factor associated with M phenotype in the multivariable model (P<.001).

M phenotype isolates seldom had reduced susceptibility to chloramphenicol (7.9%) or tetracycline (16.1%), but commonly had reduced susceptibility to other antimicrobial agents including penicillin (81.1%), cefotaxime (60.5%), and trimethoprim-sulfamethoxazole (87.8%) (Table 2). Very few of the isolates were resistant to fluoroquinolones (0.3%). In 1999, the median erythromycin MIC₅₀ was 8 µg/mL in all sites except Georgia and Oregon, where erythromycin MIC₅₀ was 16 µg/ml or greater. Ten percent of M phenotype isolates had an erythromycin MIC₅₀ of 16 µg/mL or greater.

Twenty-three different serotypes were found among 3617 isolates (Table 3). Among the 591 isolates with M phenotype, serotypes 14 (284, 48.0%), 6A (65, 10.9%), 23F (44, 7.4%), 19F (40, 6.8%), and 6B (37, 6.3%) were the most common. These serotypes are reported as percentages of their respective phenotypes.

The MLS_B phenotype accounted for 128 (18%) of macrolide-resistant isolates. In the multivariable analysis, only decreased susceptibility to penicillin was found to be independently associated with the MLS_B phenotype (P<.001) (Table 1). The majority of MLS_B phenotype isolates displayed reduced susceptibility to chloramphenicol (56.2%), tetracycline (85.9%), penicillin (85.1%), cefotaxime (57.0%), and trimethoprim-sulfamethoxazole (87.5%) (Table 2). Four serotypes—6B (30.2%), 14 (20.6%), 23F (19.8%), and 19F (12.7%)—made up the majority of MLS_B phenotype isolates (Table 3). These serotypes are reported as percentages of their respective phenotypes.

Macrolide resistance is now common in the United States among invasive pneumococci. In the setting of increasing macrolide use, the prevalence of macrolide resistance among pneumococci doubled from 1995 to 1999 and M phenotype accounted for this increase. Although M phenotype isolates were previously thought to be less concerning because of relatively low erythromycin MICs, in our sample they had increasingly higher erythromycin MICs and were often resistant to other antibiotic agents. The proportion of all invasive pneumococcal isolates with MLS_B phenotype did not change during our 5-year period of observation. The M phenotype has become the dominant resistance mechanism likely due to a combination of successful spread of clones within the pediatric population in the United States¹⁶ and selective pressure of macrolide use in this population.

Other countries have reported the MLS_B phenotype as more common.^22- 25 Globally, the predominant phenotype for macrolide-resistant pneumococci varies regionally. The M phenotype is predominant in North America,^6,26 while several European countries have reported the dominance of MLS_B phenotype.^22,27 Differences in types of antibiotic used and prevalence of resistant clones probably account for this difference.

The observed increase in erythromycin MICs over this short time among M phenotype isolates is concerning. For maximum efficacy, an antibiotic should reach concentrations at the infection site higher than the specific antibiotic MIC in the infecting organism. It is unclear whether pneumococci with M phenotype are clinically significant, especially since newer macrolides (eg, clarithromycin and azithromycin) achieve higher concentrations intracellularly in tissue and in the epithelial lining fluid of the lung than concentrations achieved in blood.^28- 29 We found, however, that most M phenotype isolates have erythromycin MICs of at least 8 µg/mL, an MIC at which treatment failures with clarithromycin and azithromycin have been reported.³⁰ Therefore, careful study and surveillance of the effect of pneumococcal macrolide resistance on clinical outcome are critical to further understanding the clinical relevance of our findings.

Similar to earlier studies for penicillin resistance,^5,7,11 we found that M phenotype was more common among isolates from children younger than 5 years of age, of white race, and those who lived in certain geographic locations. Among children younger than 5 years, factors such as higher incidence of nasopharyngeal carriage of S pneumoniae,^31- 34 day care attendance, and high frequency of antibiotic use may explain higher rates of macrolide-resistant strains. Both high rates of macrolide use and increasing macrolide-resistant pneumococcal infections among young children may contribute to the spread of macrolide-resistant pneumococcus throughout the population. However, children younger than 5 years account for only a small proportion of the overall use of macrolides. Although overall macrolide use did not increase among individuals 5 years and older, macrolide use in this group may also contribute selective pressure for macrolide-resistant pneumococci.

This study has a few limitations. First, we looked at phenotypic expression of macrolide resistance using clindamycin susceptibility as a marker for M or MLS_B phenotype, rather than directly examining the mefE and ermAM genes. Other macrolide resistance mechanisms are rare in pneumococci and the relationship between phenotype and genotype has been confirmed.^11- 12,16 Thus, our approximation should be adequate to describe trends. Second, the MIC panels (highest MIC >16 µg/mL) used did not determine actual MICs for many M phenotype and the majority of MLS_B phenotype isolates. We were, however, able to assess the proportion of M phenotype isolates in the range where treatment failures may occur (MIC ≥8 µg/mL). MLS_B phenotype isolates typically have MICs of 64 µg/mL or greater.¹⁰

Macrolides are important therapeutic agents. Clinical treatment guidelines for empiric therapy for community-acquired pneumonia in adults have recommended using macrolides as first-line agents.^1- 3 The increasing frequency of macrolide resistance among pneumococci and increasing MICs among resistant strains suggest that these treatment recommendations may need reevaluation. Alternative antimicrobial agents for macrolide-resistant infections exist or are in development. For young children who are penicillin-allergic, however, the management of respiratory tract infections may become problematic because doxycycline and fluoroquinolones are not available as alternatives for this age group.

Strategies to control increasing macrolide resistance are needed. A study in Finland demonstrated reduction in mefE-mediated resistance among Streptococcus pyogenes after a practice guideline limiting macrolide prescriptions for pharyngitis was instituted.³⁵ Whether significant reduction in macrolide use is achievable through campaigns promoting appropriate antibiotic use or whether reductions in use will lead to reductions in macrolide-resistant pneumococci is unknown. Total antibiotic prescriptions have decreased since 1993 among all age groups, but our data show that macrolide prescriptions have increased dramatically in children. Given the high proportion of macrolide-resistant pneumococci among the pediatric population, decreasing inappropriate antibiotic use especially with macrolides is important to the success of strategies that will reduce macrolide resistance.

The new heptavalent pneumococcal-conjugate vaccine should provide coverage against most macrolide-resistant pneumococcal strains. Use of this vaccine in children has the potential to significantly change the trends observed but new serotypes not covered by the vaccine may become macrolide-resistant with continued antibiotic pressure. Pneumococci are again showing their remarkable ability to adapt to their environment, and it remains our challenge to utilize macrolides and other antibiotic agents appropriately and promote rapid introduction of effective pneumococcal vaccines.