Trends in Streptococcus pneumoniae Antimicrobial Resistance in US Children: A Multicenter Evaluation

Abstract Background Antimicrobial resistance (AMR) poses a significant challenge for treating pneumococcal disease. This study assessed AMR trends in Streptococcus pneumoniae from US children. Methods We evaluated antibiotic resistance, defined as facility antimicrobial susceptibility reports of intermediate/resistant, in 30-day nonduplicate S pneumoniae isolates from children (<18 years of age) with invasive (blood or cerebrospinal fluid/neurological) or noninvasive (respiratory or ear/nose/throat) isolates at 219 US hospital inpatient/outpatient settings in the BD Insights Research Database (January 2011–February 2020). We used descriptive statistics to characterize the percentage of antimicrobial-resistant isolates and generalized estimating equations to assess variations in resistance over time. Results Of 7605 S pneumoniae isolates analyzed, 6641 (87.3%) were from noninvasive sources. Resistance rates were higher in noninvasive versus invasive isolates. Isolates showed high observed rates of resistance to ≥1 drug class (56.8%), ≥2 drug classes (30.7%), macrolides (39.9%), and penicillin (39.6%) and significant annual increases in resistance to ≥1 drug class (+0.9%), ≥2 drug classes (+1.8%), and macrolides (+5.0%). Conclusions Among US children over the last decade, S pneumoniae isolates showed persistently high rates of resistance to antibiotics and significant increases in ≥1 drug class, ≥2 drug classes, and macrolide resistance rates. Efforts to address AMR in S pneumoniae may require vaccines targeting resistant serotypes and antimicrobial stewardship efforts.

The introduction of pneumococcal conjugate vaccines (PCVs) against Streptococcus pneumoniae infections transformed the epidemiology of pneumococcal disease (PD) and associated health outcomes in both children and adults [1,2]. Initial decreases in PD rates following introduction of PCVs were accompanied by reductions in cases of antimicrobial-resistant invasive pneumococcal disease (IPD) in children in the United States (US) and globally [3][4][5][6][7]. PCVs can reduce antimicrobial resistance (AMR) directly by reducing the incidence of resistant pneumococcal infections as well as indirectly by reducing antibiotic use, thereby relieving selective pressure on antibiotic-resistant strains [8,9]. However, increased AMR over time has been observed in nonvaccine serotypes in pediatric IPD [10] and in both vaccine and nonvaccine serotypes in noninvasive PD pediatric samples [11].
Currently, the Centers for Disease Control and Prevention (CDC) estimates that approximately 30% of IPD cases are caused by S pneumoniae isolates resistant to 1 or more clinically relevant antibiotics. Drug-resistant S pneumoniae isolates caused an estimated 900 000 infections and 3600 deaths per year in all age groups in 2014 and has been designated a serious threat by the CDC [12]. Worldwide, S pneumoniae is the fourth leading pathogen in terms of deaths associated with or attributable to resistance [13]. Current data on AMR in S pneumoniae are essential not only for clinical management, but also to assess evolutionary AMR trends in this important pathogen and prioritize efforts to reduce resistance. However, there is a lack of information on S pneumoniae resistance in the pediatric population, which does not have as many treatment options as adult patients [14]. In particular, there is a dearth of information on resistance in pediatric noninvasive PD infections, which constitute the majority of PD cases [2]. We therefore evaluated AMR in S pneumoniae isolates obtained from children with invasive or noninvasive PD.

Study Design
This was a retrospective study of antibiotic susceptibility of specified nonduplicate S pneumoniae isolates (first noncontaminant S pneumoniae isolate within 30 days) collected from hospitalized and ambulatory pediatric patients (aged <18 years at time of culture collection) between January 2011 and February 2020 as part of routine clinical management. IPD cases were defined as those with S pneumoniae isolates obtained from cerebrospinal fluid (CSF)/neurology samples or blood (including valve and ventricle catheter tip sources). Noninvasive PD cases were defined as those with S pneumoniae isolates obtained from respiratory samples or the ear, nose, or throat (ENT). Skin/wound, urine, and other nonsterile sources not listed above were evaluated but excluded from statistical modeling as they are not commonly associated with S pneumoniae infections. Cultures likely to be associated with contamination or colonization, such as environmental/surveillance specimens from rectal or nasal swabs, were excluded by use of a previously described algorithm [15].
Reporting institutions consisted of US hospitals in the BD Insights Research Database (Becton, Dickinson and Company, Franklin Lakes, New Jersey), which includes small and large hospitals in urban and rural areas and provides geographical representation across the US [16,17]. The study was approved as involving use of a limited retrospective data set for an epidemiology study and deemed exempt from consent by the New England Institutional Review Board/Human Subjects Research Committee (Wellesley, Massachusetts) and conducted in compliance with Health Insurance Portability and Accountability Act requirements.
AMR was evaluated in S pneumoniae isolates based on facility reports; minimal inhibitory concentration (MIC) breakpoints were not standardized across facilities. Antibiotic resistance was assessed using the following definitions: • Penicillin resistant: Intermediate (I) or resistant (R) to penicillin. • Macrolide resistant: I or R to erythromycin, azithromycin, or clarithromycin. • Fluoroquinolone resistant: I or R to levofloxacin or moxifloxacin. • Extended-spectrum cephalosporin (ESC) resistant: I or R to ceftriaxone, cefotaxime, or cefepime. • Tetracycline resistant: I or R to doxycycline or tetracycline. • Any drug class or ≥2 drug class resistant: I or R to any of the tested antibiotics or ≥2 of the drug classes listed above, respectively.

Outcomes
For each category of resistance defined above, we evaluated the percentage of resistance (mean number of resistant isolates per total isolates tested) overall, by invasive versus noninvasive PD, and by year, sex, age group, and hospital characteristics.

Statistical Analysis
Descriptive statistics of the percentage of resistant isolates over time were presented by cross-tabulation. For the multivariate analyses, the generalized estimating equation (GEE) method

Antimicrobial Resistance in S pneumoniae Isolates
The majority (56.8%) of isolates were resistant to at least 1 drug class (≥1 drug class) during the study period, and 30.7% of isolates were resistant to ≥2 drug classes ( Figure 1; Table 2). There were too few isolates with resistance to ≥3 drug classes for meaningful analyses (n = 637). Approximately 40% of isolates were resistant to macrolides (39.9%) or to penicillin (39.6%) ( Table 2). Tetracycline resistance was observed in 11.0% of isolates, ESC resistance in 6.8%, and fluoroquinolone resistance in 0.4% of the isolates. The remainder of the analyses were focused on resistance to macrolides or penicillin, as tetracycline is generally not recommended in pediatric patients aged <8 years [14] and the proportions of isolates with ESC and fluoroquinolone resistance were low.

Association of Demographic and Hospital Characteristics With S pneumoniae Resistance
In multivariate analyses of total S pneumoniae isolates, significant differences among age groups were observed for resistance to ≥1 drug class, ≥2 drug classes, macrolides, and penicillin ( Table 3). The <2-year-old age group had the highest levels of resistance and the 5-to 17-year-old age group had the lowest. As mentioned above, resistance was higher in noninvasive PD (respiratory, ENT) than in IPD (blood, CSF/neurological) isolates. Other characteristics associated with higher rates of S pneumoniae resistance included hospital bed size (≥1 drug class), hospital-onset cultures (all drug categories), and geographic region (all drug categories) (Supplementary Table 2).
Noninvasive PD isolates showed similar age patterns to total isolates. Macrolide resistance rates were significantly higher for respiratory versus ENT isolates, but resistance rates for ≥1 drug class, ≥2 drug classes, and penicillin were similar for these sources. Other characteristics associated with higher resistance rates in noninvasive PD isolates included hospital bed size (≥2 drug classes and macrolides), hospital-onset cultures (all drug categories), urban/rural location (macrolides), teaching versus nonteaching hospitals (macrolides), and geographic region (all drug categories) (Supplementary Table 2).
In contrast to the age pattern observed for total and noninvasive PD isolates, for IPD isolates the 2 to 4-year-old age group had the highest levels of resistance to ≥1 drug class, ≥2 drug classes, macrolides, and penicillin (Table 3). Resistance to all evaluated drug categories was higher for blood versus CSF/neurological isolates. Other characteristics associated with higher resistance rates in IPD isolates included hospital bed size (≥1 drug class and macrolides), hospital-onset cultures (≥2 drug classes, macrolides, and penicillin), teaching versus nonteaching hospitals (≥1 drug class), and geographic region (all drug categories) (Supplementary Table 2).

Trends in AMR Over Time
In adjusted analyses, S pneumoniae resistance to ≥1 drug class and ≥2 drug classes increased significantly from January 2011 to February 2020 ( Table 4). The annual rate of increase was +0.9% per year for ≥1 drug class and +1.8% per year for ≥2 drug classes over this time period (P < .001 for both analyses). Noninvasive PD isolates showed similar increases (+0.6% per year for ≥1 drug class and +2.2% per year for ≥2 drug classes; P < .001 for both analyses). The rate of IPD isolates with resistance to ≥1 or ≥2 drug classes fluctuated during this time period, but trends were not significant. For total S pneumoniae isolates, resistance to macrolides increased (+5.0% per year, P < .001) and resistance to penicillin decreased (−1.1% per year, P < .001) ( Table 4). The same pattern was observed for noninvasive PD (+5.7% per year for macrolides, P < .001; −1.1% for penicillin, P < .001) and IPD isolates (+1.4% per year for macrolides, P = .005; −1.0% for penicillin, P = .006).

DISCUSSION
Streptococcus pneumoniae remains an important cause of pediatric infections despite marked decreases in disease incidence following introduction of PCVs. Effective antibiotics against pneumococci are available, but management of this pathogen can be complicated by AMR. As documented in this study, more than half (56.1%) of 7605 S pneumoniae isolates from US children were resistant to at least 1 drug class in the adjusted analyses, and there were persistently high levels of AMR, particularly to macrolides and penicillin, over the last decade. Multivariate analyses demonstrated significant increasing trends in the proportion of total and noninvasive PD isolates resistant to ≥1 drug class and ≥2 drug classes, and a significant increasing trend in macrolide-resistant isolates for both noninvasive PD and IPD sources. The magnitudes of increasing resistance to ≥2 drug classes (+1.8%/year) and macrolides (+5.0%/ year) were particularly alarming. High and increasing S pneumoniae resistance to macrolides has also been documented in adults with PD [16,17]. More encouraging results were observed with penicillin resistance, which showed a significant decreasing trend, although the approximately 35% resistance rate for total isolates in the most recent data (2019-2020) is still quite substantial. Our findings are consistent with a metaanalysis that reported high levels of penicillin and macrolide resistance in S pneumoniae in children globally [7].
A notable result from this study was higher resistance rates in noninvasive PD isolates compared with IPD isolates, regardless of the drug category evaluated. A similar observation has also been made in adults [17]. Although noninvasive PD cases are generally less severe, they contribute to widespread antibiotic use [18][19][20] and exert a heavy burden on the healthcare system, including an estimated 382 182 pneumonia hospitalizations and 93 million office visits for otitis media in US children between 1997 and 2019 [2]. It is likely that the observed resistance in pediatric noninvasive PD isolates is related to high levels of antibiotic use for outpatient respiratory infections, as has been suggested by some studies [21,22]. We also found higher resistance rates in children under the age of 2 compared with older age groups for total and noninvasive PD isolates. This finding is of significant concern, as children <2 years of age are more vulnerable to S pneumoniae infections and deaths than older pediatric age groups [23,24].
Drug-resistant S pneumoniae infections represent one of the few antibiotic-resistant threats that can be mitigated by vaccines [12]. Although US childhood PCV vaccination rates are generally high (82%-92%), lagging vaccination rates are observed in some populations, including uninsured children and children living below the poverty level, and coronavirus disease 2019-related disruptions in vaccination schedules have further exacerbated these disparities [25]. Pneumococcal vaccines reduce AMR by decreasing the frequency of S pneumoniae serotypes that are more likely to be antibiotic resistant [3,7] and by reducing respiratory infections that result in antibiotic use [9]. Reductions in rates of PD are also associated with reductions in antimicrobial-resistant S pneumoniae caused by serotypes covered by vaccines [3]. However, there has been a recent increase in antimicrobial-resistant S pneumoniae due to serotypes not covered by 13-valent PCV (PCV13) [3].
Accordingly, another potential defense against antimicrobialresistant pneumococcal infections is expanded-valency vaccines. New PCVs covering additional serotypes (15-valent PCV [PCV15] and 20-valent PCV [PCV20]) have recently been approved for adults in the US [26,27], and PCV15 has also been approved for children in the US [27]. PCV15 and PCV20 cover several additional serotypes that contribute to resistant IPD [3].
Modeling studies indicate that these extended-valency vaccines have the potential to substantially reduce the burden of PD in US children [28,29]. The profound effect of vaccines on disease burden is further expanded when economic evaluations of AMR reduction are included [30]. Both the World Health Organization and the US National Vaccine Advisory Committee have endorsed the use of vaccines to decrease levels of antibiotic-resistant bacteria [31,32].
Although vaccines are a critical component of initiatives to combat AMR in S pneumoniae, effective antimicrobial stewardship efforts are also needed to combat resistance in S pneumoniae and other pathogens. Antibiotic overuse continues to be a problem that drives community AMR for S pneumoniae [21,33]. The issue of inappropriate antibiotic use in US children has been well documented [19,[34][35][36]. Although antibiotics   are not routinely recommended for pediatric outpatients with community-acquired pneumonia (CAP), which is often due to viral infections, 73.9% of US children aged 1-6 years with CAP between 2008 and 2015 were nonetheless prescribed antibiotics based on nationwide data [19]. Counter to current guidelines, macrolides are the most common antibiotic class prescribed for ambulatory CAP in children, with approximately 40% of pediatric patients receiving macrolide treatment [19,34,35]. This widespread usage is likely a contributor to the high (adjusted rate of 45.1%) and increasing macrolide resistance rate observed in our study and suggests that targeted antimicrobial stewardship efforts may be required to reduce AMR in S pneumoniae. Both inpatient and outpatient stewardship programs are key to achieving this goal [37,38]. Our data may also be useful in the critical examination of guidelines for treating common respiratory diseases [39]. Our study has limitations, including the use of facility reports to assess AMR, which may have influenced reported resistance rates and contributed to a lack of consistency in resistance thresholds across different facilities. This consideration is particularly relevant for penicillin resistance rates. In 2013, the Clinical and Laboratory Standards Institute (CLSI) increased the nonmeningitis nonsusceptibility breakpoint for parenteral penicillin from an MIC of >0.06 mg/L to >2 mg/L [40]. Use of the lower breakpoint could cause inflated rates of resistance and may have influenced trends in resistance over time. In a recent analysis of US S pneumoniae isolates (n = 7901), primarily (88.5%) from respiratory tract infections, penicillinnonsusceptible rates were 38.8% using the meningitis/oral penicillin breakpoint (>0.06 mg/L) and 6.2% using the parenteral penicillin breakpoint (>2 mg/L) [41]. CLSI changes can take many years to be fully adopted at the clinical laboratory level [42]; a recent survey of 1490 US clinical laboratories reported that between 30% and 62% of laboratories used obsolete breakpoints for at least 1 of the pathogens evaluated, although S pneumoniae was not included in this study [43]. It is therefore possible that delayed adoption of this CLSI change increased the percentage of penicillin-resistant S pneumoniae isolates reported by facilities, which could explain the discrepancy between our data and lower numbers (3.6%) reported by the CDC, which utilizes a central laboratory for resistance rate assessment [24]. Changing resistance breakpoints may also have influenced the decreased resistance to penicillin observed in our study. It is important to note, however, that the data presented here reflect the susceptibility data on isolates available to clinicians during daily management of their patients. Other potential limitations of our study include lack of information on associated infections; our analyses were based solely on culture-positive isolates, and the presence of symptomatic PD was not confirmed. Selection bias, which can affect all microbiologic surveillance studies, may increase estimates of resistance due to a higher likelihood of testing more severely ill patients. We did not evaluate S pneumoniae serotypes or the pneumococcal vaccination status of patients. These data would be valuable additions to a future analysis.
In conclusion, our data document high levels of resistance to macrolides and penicillin in S pneumoniae isolates among US children with IPD and noninvasive PD infections. These data may help inform initial treatment decisions for pediatric outpatients and inpatients with suspected pneumococcal infections. The rates of resistance to ≥1 drug class,≥2 drug classes, and macrolides have increased in pediatric S pneumoniae isolates over the past 10 years, despite overall disease reductions attributable to vaccines. Addressing the high and increasing rates of S pneumoniae resistance may require PCVs with expanded serotype coverage and targeted antimicrobial stewardship efforts [3,37,38].

Supplementary Data
Supplementary materials are available at Open Forum Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.