Multidrug resistance bacteremia in neonates and its association with late- onset sepsis and Coagulase-negative Staphylococci

Introduction: This study aimed to assess the association between multidrug resistance (MDR) and late-onset sepsis (LOS) among newborns with bloodstream infection (BSI). Methodology: In this cross-sectional study, we routinely tested every newborn with a presumptive diagnosis of sepsis admitted to the largest reference maternity hospital in Lima, Peru for BSI over an 18-month period. We tested every isolate for MDR by using the disk-diffusion method and assessed its associated factors by using a robust Poisson regression analysis with a particular focus on its association with LOS (vs. early-onset sepsis, EOS). Results: We analyzed a total of 489 subjects, including 340 (69%) newborns with LOS, and estimated an MDR rate of 80% (95% confidence interval, CI: 76%-83%), which was significantly higher (p-value < 0.001) among LOS (85%; 95% CI: 81%-89%) than EOS cases (67%; 95% CI: 59%-75%). The primary isolate was coagulase-negative Staphylococci (CoNS) (60%), which exhibited a limited subset of antibiotic MDR patterns, most of which were characterized by their resistance to cefoxitin, gentamicin, and clindamycin and levofloxacin. Overall, the prevalence of MDR was higher among LOS compared to EOS cases (adjusted prevalence ratio [aPR] = 1.28; 95% CI: 1.14-1.45), and among BSI due to CoNS compared to other bacteria (Apr = 1.10; 95% CI: 1.01-1.20). Conclusions: MDR among newborns with sepsis is exceptionally high, being even higher among those with LOS than newborns with EOS, and among those infected with CoNS compared to other bacteria. Furthermore, CoNS exhibited a limited subset of MDR patterns, which could be used to guide therapeutic decisions.


Introduction
Neonatal sepsis is a significant cause of neonatal morbidity and mortality in Latin America and worldwide [1].Clinically, neonatal sepsis can be classified as early-onset sepsis (EOS) or late-onset sepsis (LOS) depending on whether the sepsis clinically manifests within or after the first 72 hours of life, respectively [2]. Regardless of the type, this disease is difficult to diagnose since it often presents with nonspecific signs and symptoms, making it is essential to procure proper medical care and effective antibiotic treatment as soon as possible [3].
Currently, the increasing levels of antibiotic resistance are a public health concern worldwide. Neonatal sepsis is increasingly more frequently associated with bacteria showing multidrug-resistance (MDR) , such as extended-spectrum beta-lactamase (ESBL), and carbapenemases producers [4]. This fact makes it difficult to implement adequate empirical antimicrobial therapy in newborns, thereby extending the length of hospital stay [5] and increasing the risk of mortality [6]. To prevent such outcome, it is crucial to determine the epidemiology of EOS and LOS, which seems to related to bacterial colonization by vertical transmission [7] and direct contact with contaminated environments by horizontal transmission [8], respectively.
As in other low-middle income countries, in Peru, antibiotic resistance levels continue to rise, becoming an every emerging threat to health [9]. Different reports have shown that bacterial antibiotic resistance levels are changing, varying by age, sample origin, geographical distribution, and several other not necessarily modifiable factors [10][11][12]. The present study aimed to assess the epidemiology of neonatal sepsis and its potential association between multidrug resistance and its classification (EOS/LOS) in the Peruvian context, in order to identify modifiable risk factors to help reduce both.

Study design and Population
Following a cross-sectional study design, we analyzed every case of neonatal sepsis with positive bacteremia diagnosed at the National Maternal Perinatal Institute (INMP) of Lima during the period January 2017-June 2018. At the INMP, which is the largest maternity reference hospital in Peru (> 20,000 births annually), every newborn has full insurance coverage, including critical care if needed. We aimed to enroll every eligible subject, and thus, the study inclusion criteria were newborns (< 28 days of life) with a diagnosis of neonatal sepsis confirmed with positive bacteremia tested by blood culture. Furthermore, to prevent information bias, the exclusion criteria included: contaminated blood cultures and samples collected from patients receiving antibiotic therapy during sample collection. Appropriate ethical approvals was obtained by National Maternal Perinatal Institute, in Peru.

Strain identification
We cultured every blood sample in bottles which were incubated in a BD BACTEC automated blood culture system for seven days before reporting no growth. Then, we identified the etiological agent by performing Gram stains and sub-cultures in selective media, both according to conventional microbiology protocols [13]. During this process, we considered coagulase-negative Staphylococci (CoNS) as a plausible etiological agent only when patients tested CoNS positive in two separate blood cultures.

Antimicrobial susceptibility testing
We assessed the antimicrobial susceptibility of each isolate using the standard disk-diffusion method in Mueller-Hinton agar plates as well as the Kirby Bauer method with Escherichia coli ATCC 25922 and Staphylococcus aureus ATCC 25923 as quality control. Then we interpreted the results as resistant, intermediate, or sensitive according to the recommendations of the Clinical and Laboratory Standards Institute (CLSI) guidelines [14]. Bacteria resistant to at least one antibiotic in three or more drug classes were considered to be MDR [15]. When requested, we assessed ESBL expression using a conventional ESBL disk synergy test, containing cefotaxime, amoxicillin plus clavulanic acid, and ceftazidime in Mueller-Hinton agar [16].

Data Registry and Quality Control
Data registry was conducted using the WHONET software, which is a free Windows-based database software developed by the World Health Organization Collaborating Centre for Surveillance of Antimicrobial Resistance to facilitate the registry and analysis of antimicrobial susceptibility data [17]. With this software, we incorporated several quality control procedures, including specific categories, range values, double-check, and two-pass verification with an independent reviewer (GS).

Statistical Analysis
First, we performed a descriptive data analysis summarizing each categorical variable by its absolute and relative frequencies and every numerical variable by its mean and standard deviation. Second, we performed a bivariate comparison between newborns with EOS vs. LOS using the Chi-square test and the Student's t-test for proportion and mean comparisons, respectively. Third, we explored the antibiotic susceptibility of the top five most frequent etiological agents responsible for EOS and LOS cases separately, considering as resistant any isolate tested as either resistant or intermediate in the antimicrobial susceptibility test. Fourth, we assessed the rates of isolates with a positive ESBL phenotype, carbapenemases, methicillin-resistant Staphylococcus aureus (MRSA), and MDR. Fifth, we estimated the MDR rate for each etiological agent and compared MDR rates between EOS and LOS cases. Sixth, we performed a sub-analysis among newborns infected with CoNS to assess the main antibiotic-resistant patterns using only those antibiotics with levels of resistance over 50% per antibiotic tested. Finally, we used a by generalized linear model with a Poisson distribution, link log, and robust error variance to model the prevalence of MDR and identify its main associated factors, especially focused on its association with LOS. We conducted these analyses using STATA™ MP version 14.0 (Stata Corp., College Station, TX) and a 95% confidence interval (95% CI).

Study population
We analyzed a total of 489 newborns diagnosed with sepsis: 340 (69%) classified as LOS (13.5 ± 7.3 days of age) and 149 as EOS (1.5 ± 1.0 days of age) ( Table 1). Overall, we tested most newborns (96%) hospitalized at the INMP, including 38% from the intensive care unit, and only a small fraction of newborns (4%) were tested upon transfer from other hospitals. When comparing the demographic data and the clinical characteristics of the EOS vs. LOS groups, we only found significant differences in terms of preterm birth (22% vs. 34%, p-value = 0.011).

Etiological strains among newborns with EOS or LOS
Among newborns with EOS, the leading etiological agent identified in the blood cultures was by far CoNS (64%) ( maltophilia (2%), and three other bacteria (Enterococcus spp., and P. aeruginosa and Enterobacter), each with a frequency of 1% or less (Table 1).     antibiotic resistance rates (range: 0% to 100%). The most common etiological agent (Top 1) among EOS was CoNS, which showed high resistance rates (> 50%) to penicillin, ciprofloxacin, gentamicin, clindamycin, levofloxacin, and trimethoprim/sulfamethoxazole, with values to cefoxitin close to 60% of resistance. The second most frequent etiological agents (top 2) were other Staphylococcus, which was found to be highly resistant (> 50%) to cefoxitin, gentamicin, ciprofloxacin, levofloxacin, and trimethoprim/sulfamethoxazole. The third and fourth most frequent etiological agents were Enterococcus (top 3) and Streptococcus (top 4), which were highly resistant (> 50%) to a few commonly used antibiotics. Finally, E. coli (top 5) also showed high resistance rates (67% to 100%) to standard antibiotics such as ampicillin, amoxicillin/clavulanic acid, ciprofloxacin, levofloxacin, and trimethoprim/sulfamethoxazole, and cephalosporins (cefepime and cefotaxime), but with high sensitivity (100%) to carbapenems (imipenem and meropenem). We also observed that CoNS (top 1) and Staphylococcus spp. (top 5) isolates showed resistance rates greater than 50% to most of the antibiotics tested, except for amikacin, rifampin, and tetracycline (all < 36%) ( Table 2,3).

Carbapenemase screening
Overall, 13 isolates were tested for the presence of carbapenemases, all of which were from newborns with LOS. Among these isolates, 62% (8/13) were carbapenemase positive, all of which were Acinetobacter spp. strains (8/8). The five isolates that tested negative to carbapenemases included three K. pneumoniae, one E. coli, and one P. aeruginosa.

MDR patterns
We performed a sub-analysis among newborns infected with CoNS to compare the antibiotic MDR patterns between EOS and LOS ( Table 3) and observed that a small subset of antibiotic MDR patterns characterized most of the EOS and LOS cases among newborns (40% and 65%, respectively). Specifically, we identified that the most frequent antibiotic MDR pattern was resistance to cefoxitin, cefepime, clindamycin, gentamicin, levofloxacin, and trimethoprim/sulfamethoxazole. This pattern alone characterized 15% of EOS cases and 43% of LOS cases. Furthermore, overall, the CoNS isolated exhibited similar patterns of antibiotic MDR patterns, which were characteristically resistant to cefoxitin, gentamicin, and clindamycin, or levofloxacin.

MDR-associated factors
In our study, we identified that MDR was significantly associated with LOS and CoNS infection. In our multivariable regression analysis, we observed that the prevalence of MDR was higher among cases with LOS compared to those with EOS (adjusted prevalence ratio [aPR] = 1.28; 95% CI: 1.14 -o 1.45), and among BSI due to CoNS compared to those by other bacteria (aPR = 1.10; 95% CI: 1.01 -1.20). Bivariate analysis showed that prematurity might represent another risk factor for MDR among newborns. However, the multivariate regression analysis showed that both LOS and CoNS represented the most critical MDR risk factors in our study population (Table 4).

Discussion
In the present study, we observed that the MDR rate among newborns with sepsis was exceptionally high. Furthermore, it was of note that this rate was even higher among newborns with LOS compared to those with EOS, and among newborns infected with CoNS compared to those infected with other bacteria. Additionally, we observed that CoNS was associated with a limited subset of MDR patterns, which could be useful to guide therapeutic decisions.
In Peru, neonatal sepsis represents a significant cause of neonatal morbidity and mortality, with an estimated incidence of 4.1 per 1000 live births (95% CI: 2.7 -5.5) [18]. The etiological agents of sepsis vary significantly from country to country and even from institution to institution. However, the role of Enterobacteriaceae and Acinetobacter as the leading etiological agents worldwide is commonly accepted. In our study, we observed that the most common cause of neonatal sepsis was CoNS, followed by Enterobacteriaceae and Gram-negative nonfermenters, especially Acinetobacter spp.
We also observed isolates of Klebsiella with high levels of antibiotic resistance. This is consistent with a previous Peruvian study that reported high levels of resistance to trimethoprim/sulfamethoxazole and ciprofloxacin (73% and 65%, respectively) as well as high positivity to ESBL (75%) [9]. Recently, Klebsiella is in the spotlight due to the increase of the extremely antibiotic resistance associated with significant clinical outbreaks worldwide, including in Peru [19,20].
In the present study E. coli isolates with MDR rates (~87%) were found, being higher than rates reported previously by Palma N et al. [21] in Peru (~80%). These researchers reported E. coli isolates with high levels of resistance to ampicillin (93%), nalidixic acid (66%), and trimethoprim/sulfamethoxazole (66%) [21]. However, in our study, we reported even higher levels of resistance to E.coli showing 100% resistance to ampicillin, amoxicillin plus clavulanic acid, levofloxacin, 92% resistance to trimethoprim/sulfamethoxazole and 65% to gentamicin. This is of particular concern, since ampicillin plus gentamicin is the most common empiric antibiotic combination used in neonates. Only last resort antibiotics such as colistin are available to treat extensively drug-resistant Acinetobacter (> 97% of resistance to all antimicrobials tested); it is important to highlight the elevated rates of antimicrobial resistance in these Acinetobacter found in these neonates with bacteremia that could be a a clone installed in the NICU with similar antibiotic resistance pattern. Thus, Acinetobacter spp is usually related to intestinal carriage and is linked to neonatal outbreaks in Neonatal Intensive Care Units (NICU) [22]. These bacteria are associated with increased mortality in NICU, and active surveillance is recommended as a strategy to control neonatal colonization and avoid posterior infection or dissemination for the hospital setting [22]. In general, the high antibiotic resistance rates of Gram-negative bacilli in bacteremia in neonates are in accordance with similar studies performed in other low-and middleincome countries such as southeast Asian [23].
In our study, three S.aureus isolates were identified in newborns with neonatal sepsis who were positive to specific test for MRSA (100%); we only have the data of 10 S. aureus tested for cefoxitin, being 60% resistant to this antibiotic indicating near 60% MRSA. Attending to CoNS the resistance to cefoxitin, was 68% in EOS and 82% in LOS. These rateswere higher than recent Peruvian reports, at least regarding MRSA, which described 54% of MRSA positivity [24]. Moreover, a high proportions of nasal carriage of MRSA have been found among health care workers in previous studies made in Lima, which reported MRSA from nasal carriage to belong to similar lineages that were recovered from their patients with bacteremia [25].
It is essential to highlight that CoNS is not only the most frequent bacteria isolated from human skin and mucous, but it also represents an increasing cause of health care-associated bacteremia [26]. Moreover, what is more worrisome is that CoNS is strongly associated with MDR. In our study, CoNS was the most common cause of neonatal sepsis. It was associated with high resistance , historically topenicillin, but also to cefoxitin, gentamicin, clindamycin, ciprofloxacin, levofloxacin, and trimethoprim/sulfamethoxazole, all being higher among LOS than among EOS cases. Antibiotic resistance of skin isolates with CoNS has been reported to be increased during neonatal hospitalization (mainly during the first week) [27]. Although the levels of MDR in EOS with CoNS are considerable, they may be related to neonatal colonization with resistant CoNS, especially in the gut, immediately after birth [27].
We observed that the prevalence of MDR was higher among LOS compared to EOS cases (85% vs. 67%), in relation to the length of hospitalization, with the levels of resistance increasing with longer hospital stays. Usually in clinical settings, the most common sources of CoNS are related to the patient's microbiota and the hands of healthcare workers, thus, this could also mean that the bacteria related to this infection possibly come from these hospital environments compared to EOS, the origin of which is more linked to bacteria transmitted through the mother [28]. This finding is important because although MDR is multifactorial, the excessive use of broad-spectrum antibiotics in NICUs is a key MDR risk factor [29]. Apart from the use of antimicrobials themselves [30] and the important role played by CoNS in MDR risk, medical care personnel could act as a reservoir for antibiotic-resistant bacteria [31]. Furthermore, the presence of MDR bacteria, including ESBL-producing microorganisms, has commonly been isolated in health worker devices, such as mobile phones and workwear [32].
Peru, as other low-middle income coutries, presents a weak and fragmented health system, which some health institution do not have semi-authomatic systems to evaluate Minimum Inhibitory Concentration, for this reason we did not have information about vancomycin susceptibility levels, as an important limitation. Nonetheless, this study shown important data of antibiotic resistance levels in vunerable population such as neonates, contributing to improveantibiotic resistance survillance and control. Moreover it is recommended a antimicrobial use optimization programs (PROA), for a better antibiotic resources management, which reduces and controls the antimicrobials use in the NICU. The high relevance of CoNS in the present study, also, suggests a reinforcement in the handwashing and other hygiene measures of health workers, because they could become the reservoirs and / or dispersers of antimicrobial resistant bacteria.

Conclusions
In conclusion, MDR is high among Peruvian newborns with neonatal sepsis and is associated with the presence of LOS and CoNS. These findings are important, mainly because of the lack of a robust health surveillance system in Peru. Furthermore, considering the strong association between MDR and LOS, we strongly recommended introducing antibiotic resistance surveillance systems (infection and colonization), at least in every NICU with a high MDR rate, as well as prioritizing increased hygiene measures and good practices for medical care personnel.