Antimicrobial resistance of Streptococcus pneumoniae from invasive pneumococcal diseases in Latin American countries: a systematic review and meta-analysis

Background Invasive pneumococcal disease has declined since pneumococcal conjugate vaccine introduction in Latin America and the Caribbean (LAC). However, serotype distribution and antimicrobial resistance patterns have changed. Methods We conducted a systematic review to evaluate the frequency of antimicrobial resistance of Streptococcus pneumoniae from invasive disease in LAC. Articles published between 1 January 2000, and 27 December 2022, with no language restriction, were searched in major databases and gray literature. Pairs of reviewers independently selected extracted data and assessed the risk of bias in the studies. The quality of antimicrobial resistance (AMR) studies was evaluated according to WHO recommendations (PROSPERO CRD42023392097). Results From 8,600 records identified, 103 studies were included, with 49,660 positive samples of S. pneumoniae for AMR analysis processed. Most studies were from Brazil (29.1%) and Argentina (18.4%), were cross-sectional (57.3%), reported data on AMR from IPD cases (52.4%), and were classified as moderate risk of bias (50.5%). Resistance to penicillin was 21.7% (95%IC 18.7–25.0, I2: 95.9), and for ceftriaxone/cefotaxime it was 4.7% (95%IC 3.2–6.9, I2: 96.1). The highest resistance for both penicillin and ceftriaxone/cefotaxime was in the age group of 0 to 5 years (32.1% [95%IC 28.2–36.4, I2: 87.7], and 9.7% [95%IC 5.9–15.6, I2: 96.9] respectively). The most frequent serotypes associated with resistance were 14 for penicillin and 19A for ceftriaxone/cefotaxime. Conclusion Approximately one-quarter of invasive pneumococcal disease isolates in Latin America and the Caribbean displayed penicillin resistance, with higher rates in young children. Ongoing surveillance is essential to monitor serotype evolution and antimicrobial resistance patterns following pneumococcal conjugate vaccine introduction.


Introduction
Invasive disease caused by Streptococcus pneumoniae (IPD) is one of the leading causes of morbidity and mortality in children and older adults worldwide (1).IPD included mainly meningitis, bacteremia, and bacteremic pneumonia (2).
The invasiveness and pathogenesis of S. pneumoniae are defined by capsular composition; currently, one hundred serotypes have been identified (3).
In Latin America and the Caribbean, it is estimated that every year, IPD is responsible for up to 28,000 deaths, 182,000 hospitalizations, and 1.4 million outpatient consults (4).
Serotype distribution in IPD changes over time by age group, clinical manifestation, and regional location (7).SIREVA is an official regional laboratory surveillance program (SIREVA) that reports information about serotype distribution and antimicrobial resistance (AMR) in IPD in the pediatric and adult populations.
S. pneumoniae infections are frequently associated with inappropriate antimicrobial prescriptions both in the community and in the hospital.IPD rates have decreased since the implementation of national immunization programs.However, serotype distribution and resistance patterns have been modified (8).
Antimicrobial resistance has emerged in S. pneumoniae during the last years with a high impact on global health.Worldwide the highest rates of resistance to penicillin and erythromycin were found in serotypes 6B, 6A, 9 V, 14, 15A, 19F, 19A, and 23F (9).However, information about serotype distribution and antimicrobials in IPD is scarce.
This systematic review aimed to describe reported data about antimicrobial resistance and associated serotypes in IPD from Latin American and Caribbean countries.

Methods
The analysis presented here was part of a broader systematic review that included epidemiological data on the burden of pneumococcal disease in LAC.The findings on AMR are presented in this article.We conducted a systematic literature review and metaanalysis of AMR in IPD in LAC during the last 20 years following Cochrane methods (10), the MOOSE guidelines for observational studies (11), and the PRISMA statement for reporting systematic reviews and meta-analyses (12).The protocol is registered in PROSPERO CRD UK (registration number: CRD42023392097).

Inclusion criteria
Studies from any LAC countries, regardless of age or sex, risk groups, with at least 20 culture-confirmed cases from sterile sites (e.g., blood, cerebrospinal fluid, pleural fluid) were eligible for inclusion.IPD clinical presentations included sepsis/bacteremia, meningitis, bacteremic pneumonia, empyema, peritonitis, osteoarticular infection/septic arthritis, and endocarditis.
Cohort studies, case-control studies, cross-sectional studies, case series, epidemiological surveillance reports, hospital-based surveillance studies, interrupted time series (ITS), and controlled ITS (CITS) studies were included.Systematic reviews and meta-analyses were only considered as sources for primary studies.When data or data subsets reported in more than one publication were found, the one with the larger sample size or the most recent were selected.

Search strategy for identification of studies and data sources
Records published between 1 January 2000, and 27 December 2022, in the following databases: PubMed, Embase, CINAHL (Cumulative Index of Nursing and Allied Health Literature), LILACS (Latin American and Caribbean Health Science Literature)/ ScieLO, EconLIT, Global Health, and Web of Science, with no language restriction, were searched.Strategies search, and terms for each database are presented in Supplementary Appendix A. The reference lists of the articles were hand-searched for additional information.We contacted the original authors to obtain any missing information or clarification, but for this analysis it was not necessary.
SIREVA, other regional or national surveillance databases, including antimicrobial resistance databases, and relevant sources like regional MoH, PAHO, and reports from referral hospitals were searched.Databases containing regional proceedings, congresses' annals, doctoral theses, websites from regional scientific meetings, experts, and related associations were also conducted.

Outcomes of interest
We explored outcomes by type of IPD, serotype distribution, and antimicrobial resistance reported during the same period.

Selection of articles and data extraction
Publications were screened by two reviewers using title and abstract according to the eligibility criteria.Discrepancies were solved by consensus of the entire work team.Potentially eligible articles were  13), a web-based platform designed to process systematic reviews.One reviewer performed data extraction and verified by a second one using a pre-specified extraction online form previously piloted in 10 studies.From eligible articles, the research team extracted the following study information: publication and study characteristics (type of publication, year published, authors, geographic location, study design including domains for risk of bias assessment), study population characteristics (age, sex, sample size, population risk, inclusion and exclusion criteria), and outcomes (frequency of AMR and frequency of serotypes associated with AMR).

Risk of bias assessment
Included studies were assessed for risk of bias by two independent reviewers, with discrepancies resolved by consensus with the whole team.The risk of bias in observational studies and the control arm of trials was assessed using the checklists developed by the U. S. National Heart, Lung, and Blood Institute, which classify studies as high risk of bias (POOR), moderate risk of bias (FAIR), and low risk of bias (GOOD).For the assessment of cohort studies and cross-sectional studies, the tool comprises 14 items, while nine items apply to the case series studies (14).

Quality assessment of AMR studies
The quality of AMR studies was evaluated according to WHO recommendations (15).

Statistical analysis 2.7.1 Primary analysis
To analyze the data, descriptive statistics and performed a proportion meta-analysis, using metaprop {meta} package with R software version 4.2.2. were used (16,17).We applied an arcsine transformation to stabilize the variance of proportions (Freeman-Tukey variant of the arcsine square root of transformed proportions method) (18).We applied DerSimonian-Laird weights for the random effects model where heterogeneity between studies was found.We calculated the I 2 statistics as a measure of the proportion of the overall variation in the proportion that was attributable to between-study heterogeneity.An I 2 > 60-70% was considered as substantial heterogeneity, and below 30% as low level of heterogeneity (19).Selective reporting within studies was assessed by comparing available protocols with the reports.

Subgroup analysis, sensitivity analysis, and investigation of heterogeneity
We conducted subgroup analyses classifying the studies by fiveyear calendar period, country, age group (0-5 years, 6-64 years, 65 or more years old), and by antibiotic resistance.Both types of analyses could contribute to the investigation of heterogeneity causes.

Literature search and study selection
We identified 8,600 records in seven different databases.After eliminating duplicates, we screened the remaining 4,533 by title and abstract.We made a full-text assessment of 414 considered relevant to determine eligibility.Finally, 103 studies met the inclusion criteria (Figure 1).

Characteristics of included studies
The characteristics of the included studies are summarized in Supplementary Table 3 in Appendix A. There were 98 full texts, four abstracts, and one thesis; 59 (57.3%) were cross-sectional studies, 42 (40.9%)case series, one (0.9%) prospective cohort, and one (0.9%) non-comparative cohort.
These studies were published between 2001 and 2022, with 38.8% (n = 40) published between 2000 and 2009 and 50.5% (n = 52) between 2010 and 2019.Only 10.7% (n = 11) were published in the last 3 years (2020 to 2022).The inclusion period of participants was from 1987 to 2019, with 41.7% (n = 43) of the studies that included participants before 2000, 47.6% (n = 49), which included participants between 2000 and 2009, and 10.5% (n = 11) included participants from 2010 onwards.Only two (1%) studies did not report the inclusion period.The reported duration of the studies ranged from 12 to 299 months.
Of the 103 studies, 54 (52.4%) reported data on AMR from IPD, 23 (22.3%) from meningitis (including two studies from IPD that only reported AMR from meningitis cases), 22 (21.4%) from pneumonia, and four (3.9%) from bacteremia.A total of 49,660 positive samples of S. pneumoniae for AMR analysis were processed, with a range of samples evaluated between 6 to 11,377, including 40,889 samples from IPD (range 17-11,377), 4,743 samples from pneumonia (range 11-2,629), 3,633 samples from meningitis (range 6-854), and 395 samples from bacteremia (range 56-167).In one study, the total number of samples processed was not reported.
AMR was tested for three or more antibiotics in 54.4% (n = 56) of the studies, two antibiotics in 23 studies (22.3%), and in the Only in a single study (0.9%) molecular techniques (WGS-based assessment) were used to predict antibiotic resistance from genomic data by detecting resistance genes.Also, in 25.3% (n = 26), combined methods were used.The method used was not reported in 15.6% (n = 16) of the studies.The references of the included studies are in Supplementary Appendix A.

Risk of bias assessment
For cross-sectional and cohort studies, 40 (65.6%) were assessed as being at moderate (fair) risk, 15 (24.6%) were assessed as low (good) risk, and only six (9.8%) as high (poor) risk.The most frequent domains that did not meet the evaluation objectives were related to sample size justification, the evaluation of exposures more than once or different levels of exposure, and the blinding of the evaluators.For case series studies, 29 (69%) were rated as low risk, 12 (28.6%)as moderate risk, and only one (2.4%) as high risk.The domains that did not meet the objectives more frequently were whether the cases were consecutive and the description of the statistical methods.The complete evaluation of risk bias assessment by study design is in Supplementary Tables 4, 5 in Appendix A.

Quality assessment of AMR studies
The results of the quality assessment of AMR studies are summarized in Supplementary Table 6 in Appendix A. The quality was scored as high in 87.4% (n = 90) of the studies and moderate in the remaining 12.6% (n = 13).The most frequently missed items were not specifying whether a reference/control strain was included when assessing antimicrobial susceptibility (64.7%, n = 66), including less than 100 isolates, or not reporting the number of isolates evaluated (43.7%, n = 45).
Only four studies reported antimicrobial susceptibility from bacteremia cases, with 395 isolates evaluated for penicillin, 163 for ceftriaxone/cefotaxime, and 56 for both trimethoprimsulfamethoxazole and erythromycin.The percentage of resistance to penicillin ranged from 0% (Argentina; s17) to 20.6% (Argentina; s2).Two studies reported the resistance to ceftriaxone/cefotaxime, it was 0% in both (Argentina and Chile; s2, s51).One study reported the percentage of resistance to trimethoprim-sulfamethoxazole and erythromycin, with 17.9% of resistance each (s51; Supplementary Table 11 in Appendix A).
The full assessment of susceptibility of S. pneumoniae is in the Supplementary Tables 12-15 in Appendix B.
For ceftriaxone/cefotaxime and erythromycin, the most frequent serotypes associated with resistance was 19A (58.48%, n = 193 and 46.27%, n = 397 respectively).In the case of ceftriaxone/cefotaxime, another five serotypes were reported associated with resistance but with a lower frequency, all included in the PVC10 vaccine (Figure 2B).Regarding erythromycin, except for serotypes 1, 4, 18C, and 22F, all the other serotypes included in the PVC10, PVC13, 529 and PVC20 vaccines were associated with resistance, with a frequency less than 5% (Figure 2C).
The complete assessment of serotypes associated with resistance is in Supplementary  are in Appendix B.

Proportion meta-analysis for resistance to penicillin and ceftriaxone/cefotaxime
A proportion meta-analysis for resistance to penicillin and ceftriaxone/cefotaxime was performed.Subgroup analyses were conducted by five-year calendar period, country, and age range (0-5 years, 6-64 years, and 65 or more years; Table 1).

Data analysis from regional laboratory surveillance database reports (SIREVA)
Among the countries of the region of the Americas most represented in SIREVA reports were Argentina, Brazil, Chile, and Colombia, which between 2013 and 2018 reported more than 600 results from children under 5 years old.Before 2013, non-PCV13 serotypes were not listed.Resistance to penicillin and cefotaxime did not present significant differences between 2013 and 2018 in any of the countries analyzed.Regarding erythromycin resistance, in Argentina, a significant increase (p < 0.05) was observed between 2013 and 2018.Analyzing the main serotypes related to antimicrobial resistance, when comparing 2013 with 2018, a significant increase in 19A was observed in Chile, as well as 24A in Argentina and Chile (Figure 5).

Discussion
This study analyzed mainly antimicrobial resistance and associated serotypes in IPD previous and post-introduction of PCVs in pediatric, adult, and mixed populations in healthcare facilities from 15 countries across the LAC region.
Antimicrobial resistance in S. pneumoniae changes over time depending on PCV implementation, serotype distribution, antimicrobial consumption, and other factors showed differences between countries (7,20).
Since the introduction of PCVs in national immunization programs in LAC, a decrease in IPD and changes in serotype distribution and antimicrobial resistance patterns have been observed.We focus on antibiotics useful to treat pneumococcal diseases in children and adults (21,22).In our study, we observed a global penicillin antimicrobial resistance rate was less than 50% in IPD in the LAC region; similar data was reported in other regions (23).
Over the years, penicillin resistance increased until it reached its highest level in 2000-2004, followed by a decline that, despite some fluctuations, coincided with the introduction of PCVs in the different countries of the region between 2008 and 2015 (6,24).The highest rates of resistance were observed in children under 5 years, followed by adults older than 65.Resistance rates between 40 and 50% were reported in the majority before the introduction of PCVs in these countries.In countries from the region, penicillin rates higher than 50% were reported mainly in studies that include only 19A isolates.Despite the introduction of PCVs, serotype 19A remains among the most frequently associated with antimicrobial resistance and multidrug resistance (25).In the pre-PCV period, the studies that reported resistance to penicillin found it mainly associated with serotype 14.
Analysis of serotypes associated with penicillin resistance revealed a prevalence of serotypes 14, 23F, 6B, and other PCV10 serotypes in all studies conducted before the introduction of PCVs, and was In a study conducted in Brazil, the emergence of non-PCV10 serotypes 19A and 6A penicillin-resistant isolates was observed in the post-PCV period (20).As a consequence the study describes an increase of antimicrobial non-susceptibility in a long-term post-PCV10 introduction.
A similar scenario was observed for ceftriaxone/cefotaximeresistant strains, which were associated mainly with serotypes 14, 23F, and 19F in pre-vaccination studies and associated with serotype 19A and other new serotypes in post-vaccination studies.
Interestingly, outcomes from our study showed an increase in erythromycin resistance during the study period mainly related to serotype 19A and other non-vaccine serotypes.No changes in penicillin resistance, increased resistance to erythromycin, tetracycline, and multidrug resistance in the post-PCVs period were observed.In line with our results, the proportion of pneumococci showing resistance to first-line antimicrobials has decreased after vaccination.However, higher rates of resistance to other antimicrobials, mainly macrolides, have been observed in several countries, despite overall reductions in IPD attributable to vaccines, frequently associated with non-vaccine serotypes (27)(28)(29).Serotype 24F appears as an emergent serotype related to multidrug resistance and is characterized by its high invasiveness and probably influenced by antibiotic consumption (30).
One of the primary limitations of the study was the risk of bias of the included studies, which was mainly due to low sample sizes, selection bias, and information bias in the outcome measurement.
Although, the risk assessment for cross-sectional and cohort studies showed that the majority were classified as moderate risk, and case series studies were mostly rated as low risk.Most of the studies included were case series and cross-sectional, which did not provide the best disease estimations.We try to exclude, as far as possible, those articles whose data seemed to be published in more than one article, but it is possible that some AMR data found have been reported by several studies.In Latin American and Caribbean countries (excluding pneumococcal meningitis), it is not mandatory to report to laboratorybased systems like SIREVA, and passive surveillance may not accurately reflect the prevalence of diseases.
According to this study, the treatment of choice for pneumonia remains penicillin or ampicillin and cefotaxime or ceftriaxone for meningitis and bacteremia.
The study provides data on AMR and associated serotypes throughout the LAC counties during pre-and post-vaccine periods, including the late post-vaccination period, which is very important to assess the changes produced by incorporating PCVs in the region.Our results highlight continuous surveillance's importance in assessing the dynamic of serotype distribution and antimicrobial resistance in pediatric and adult IPD from LAC.

Conclusion
The introduction of PCVs in LAC countries has led to changes in pneumococcal serotype distribution and antimicrobial resistance patterns in IPD.There was an overall decline in antibiotic resistance, particularly to penicillin, after PCV implementation.However, concerning trends of increased erythromycin resistance and the emergence of non-vaccine serotypes associated with antibiotic resistance highlight the need for ongoing surveillance.Continuous monitoring of serotype evolution and antimicrobial resistance is essential to evaluate PCV impact and guide treatment recommendations for pneumococcal disease in Latin America and the Caribbean.

FIGURE 5
FIGURE 5Variation over time of resistance to penicillin, ceftriaxone/cefotaxime and erythromycin, and the percentage of serotypes included in the different PCVs.Figure4shows the percentage of resistance to penicillin, ceftriaxone/cefotaxime and erythromycin, and the percentage of serotype 19A and 24F isolates between 2013 and 2018, as well as the percentage of serotypes included in PCV10, PCV13 and PCV20 (dotted lines).PEN: penicillin; CTX: ceftriaxone/cefotaxime; ERY: erythromycin; R: resistance.PCV10 was implemented in Brazil in March 2010 and in Chile in January 2011.PCV13 was implemented in Argentina in January 2012.In Colombia PCV13 was implemented in July 2011 and replaced by PCV10 in January 2012.

TABLE 1
Summary of proportion meta-analyses results for resistance to penicillin and ceftriaxone/cefotaxime, by five-year calendar period, age categories and country.