Should first-line empiric treatment strategies cover coagulase-negative staphylococcal infections in severely malnourished or HIV-infected children in Kenya?

Background Bloodstream infection is a common cause of morbidity in children aged <5 years in developing countries. In studies reporting bacteremia in Africa, coagulase-negative Staphylococci (CoNS) are commonly isolated. However, it is currently unclear whether children who are highly susceptible to infection because of severe acute malnutrition (SAM) or HIV should be treated with antimicrobials specifically to cover CoNS. We aimed to determine the clinical significance of CoNS amongst children admitted to a rural hospital in Kenya in relation to nutritional and HIV status. Methods Systematically collected clinical and microbiological surveillance data from children aged 6–59 months admitted to Kilifi County Hospital (2007–2013) were analysed. Multivariable regression was used to test associations between CoNS isolation from blood cultures and SAM (MUAC <11.5cm or nutritional oedema (kwashiorkor)), and HIV serostatus; and among children with SAM or HIV, associations between CoNS isolation and mortality, duration of hospitalization and clinical features. Results CoNS were isolated from blood culture in 906/13,315 (6.8%) children, of whom 135/906 (14.9%) had SAM and 54/906 (6.0%) were HIV antibody positive. CoNS isolation was not associated with SAM (MUAC<11.5cm (aOR 1.11, 95% CI 0.88–1.40) or kwashiorkor (aOR 0.84, 95% CI 0.48–1.49)), or a positive HIV antibody test (aOR 1.25, 95% CI 0.92–1.71). Among children with SAM or a positive HIV antibody test, CoNS isolation was not associated with mortality or prolonged hospitalization. Conclusion In a large, systematic study, there was no evidence that antimicrobial therapy should specifically target CoNS amongst children with SAM or HIV-infection or exposure.


Introduction
Infections are amongst the leading causes of morbidity and mortality in children under the age of five years in developing countries despite recent advances in preventive and therapeutic interventions such as vaccines and antimicrobial drugs. [1] Several studies have been conducted in developing countries aimed at determining the causes of bacteraemia in children aged <5 years. [2][3][4][5][6] Findings from these studies have guided vaccine introduction and management of paediatric bacteremia in these settings. However, there still remains a paucity of comprehensive data of potential pathogens causing invasive disease in children in many areas since most health care facilities lack adequate microbiological diagnostic capacity. [3] Studies of the causes of bacteraemia in hospitalised young children in developing countries have reported that cogulase-negative Staphylococci (CoNS) are frequent blood culture isolates. [2][3][4][5][6][7][8][9][10][11][12][13] CoNS are variably reported as 'culture contaminants' [2, 5-7, 9, 10] or true pathogens. [7,8,[11][12][13] However, CoNS are common skin and mucous membranes commensal flora, [14] with relatively low virulence, but possess factors that aid in adherence, biofilm formation, tissue colonization and immune evasion. [15,16] CoNS are recognized as a cause of clinically significant infections, mostly in preterm infants [17], in critically ill patients with indwelling medical devices [15,[17][18][19] and in immunocompromised states. [7,8] It is unknown if CoNS may cause clinically significant infection in children in sub-Saharan Africa with severe acute malnutrition (SAM) and human immunodeficiency virus (HIV) infection who are at an increased risk of serious invasive infection and mortality. [7,11,12,[20][21][22][23][24] One published report from Jamaica has suggested that CoNS are important pathogens amongst children with SAM, and are associated with respiratory signs; [8] and this is referenced in national guidelines for the management of SAM, designed for health facilities that do not have access to microbiological diagnosis. [25] WHO recommends penicillin or ampicillin and gentamicin as first-line antibiotics for complicated severe malnutrition and HIV infected children suspected to have invasive disease. WHO also recommend that teatment regimens need to be adapted to local resistance patterns. [26] A separate study done in Jamaica reported that 6/15 (40%) of blood cuture isolates from 150 severely malnourished children were CoNS, with greater sensitivity to amoxicillin-clavulanic acid than to crystalline penicillin, suggesting that current empiric antimicrobial therapy could be improved. [27] If CoNS are common pathogens in SAM or HIV, then specific treatment with agents such as vacomycin, daptomycin or teicoplanin may be warranted since resistance to beta lactam antibiotics has been demonstrated. [28] Currently, limited data and inconsistent reporting mean that the clinical significance of CoNS is unclear.
To resolve the question of whether CoNS is clinically significant amongst hospitalized children with SAM and/or HIV infection in Africa, we undertook a retrospective analysis of a large systematically collected clinical and microbiological surveillance dataset from hospitalized children in rural Kenya. We aimed to describe the association of CoNS isolation with SAM and HIV status, and association in these groups with clinical signs, mortality, and duration of hospitalization.

Location and participants
The study was conducted at Kilifi County Hospital (KCH), located at the Kenyan coast and serving a mainly rural population of low social economic status. Children requiring hospitalization are admitted to either the general pediatric ward or high dependency unit (HDU) which provides close patient monitoring, but does not have capacity for central venous catheters, arterial lines, parenteral nutrition or invasive ventilation.
All pediatric admissions were systematically assessed and investigated with blood cultures prior to initiation of antibiotics (normally benzylpenicillin or ampicillin plus gentamicin) as previously described. [2] Children presenting with minor trauma or those scheduled for elective surgery did not have blood cultures done and were therefore not included in this analysis. HIV testing was offered to all admissions by trained counsellors, as per Kenya national guidelines. [29] HIV serologic testing was performed using a standard algorithm after consent was obtained from parents or guardians. A positive serologic test [KHB Colloidal Gold (KHB Shanghai Keshaun Bio-engineering Co, Shanghai, China)] was confirmed by a second serologic test [First Response™ 1-2.0 (PMC Medical Pty. Ltd, Daman, India)]. Children with HIV infection or exposure and their families were counselled and referred to the hospital's HIV clinic for further investigation, management and follow-up. Confirmatory HIV test results by PCR were not consistently available, and we therefore used HIV serologic results for the analysis. Medical care is provided according to World Health Organization (WHO) guidelines for fluids, antimicrobials, nutritional therapy for SAM and other components of supportive care. [26] Clinical and laboratory data were entered routinely onto an electronic data entry system. All children aged between 6 and 59 months admitted to KCH between January 2007 and December 2013 with blood culture results available were eligible for inclusion into this analysis.

Clinical definitions
All children underwent anthropometry at admission. SAM was classified according to WHO by the presence of kwashiorkor (nutritional edema), or by mid-upper arm circumference (MUAC) <11.5cm. [26] Features of respiratory distress included nasal flaring, chest wall indrawing, difficulty in breathing, deep respirations, hypoxia or tachypnea. Hypoxia was defined as an oxygen saturation 90% in room air using pulse oximetry. Tachypnea was defined as a respiratory rate !50 breaths per minute in children aged <12 months, and !40 breaths per minute in children aged 12-59 months. [26] Impaired consciousness was defined as Blantyre Coma Scale <4 in children up to 9 months and <5 in children aged !9 months. [30] Microbiologic methods Blood samples for culture were collected on admission after children's skin was cleaned with 70 percent ethanol and allowed to dry as previously described. [2] Regular phlebotomy training was done to reduce blood culture contamination rates. Blood samples were inoculated into commercial vials containing enriched soybean-casein digest broth with CO 2 (BACTEC Peds Plus/F medium, Becton Dickinson), and incubated in an automated blood culture system (BACTEC 9050, Becton Dickinson). Manufacturer's recommended blood volume for culture is between 0.5 and 5.0 mL, with optimum results obtained for volumes between 1.0 and 3.0 mL. Blood culture vials were weighed before and after sample inoculation to determine the volume of blood cultured. Culture vials were incubated for 5 days, or until the system indicated the culture was positive. Aliquots of culture sample from positive vials were Gram-stained and sub-cultured, and isolates identified by standard microbiological methods. [2] CoNS, Bacillus species, coryneforms, Micrococcus species, viridans group streptococci and Candida species were presumed to be contaminants and no targeted therapy for these was given; empiric antibiotic therapy was continued based on clinical resolution of illness/recommended WHO guidelines for use antimicrobials. [26] CoNS species were not determined and CoNS isolates were not stored or tested for antibiotic susceptibilities. All laboratory procedures were internally controlled and KEMRI-Wellcome Trust Research Laboratories were externally monitored for quality assurance by the United Kingdom External Quality Assessment Service and Good Clinical Laboratory Practice (GCLP) accredited by Qualogy, UK.

Statistical analysis
Statistical analysis was performed using STATA version 13 (Stata Corp., College Station, TX, USA). Data were extracted from the database, checked for any biological implausible measurements that may have resulted from transcription errors, and resolved prior to analysis. Blood culture results were indicated as no growth, CoNS, other isolates presumed non-pathogenic as listed above, or bacterial species that are usually regarded as pathogens (referred to from here as 'known pathogens'). CoNS were analysed separately from the group of presumed contaminants described above. Children with mixed CoNS and presumed contaminants on blood culture were excluded from this analysis.
Data obtained from initial admission to hospital was included in this analysis; hospital readmission data was not included in this analysis.
Children were grouped by sex, nutritional status (MUAC (!13.5 cm, 12.5 to 13.4 cm, 11.5 to 12.4 cm, <11.5 cm) or the presence of kwashiorkor) and HIV serologic test result (negative, positive or not performed).
Continuous variables were described using median and interquartile range (IQR) and compared using a Wilcoxon rank sum test. Categorical variables were described using proportions and compared using Chi-squared or Fisher's exact test as appropriate.
Multivariable logistic regression analysis was used to identify putative risk factors of CoNS, and the association of CoNS with clinical signs and symptoms, and mortality. Multivariable linear regression was used to assess the association of CoNS with duration of hospitalization (days). Transformation of age (natural log) and blood volume (square root) was done for the linear regression models.
Regression models were set up using backward selection; predictors with p<0.05, and predictors known a priori to act as potential confounders regardless of the p value were included in the final model. For both logistic and linear regression models, adjustment for a priori confounders: age, sex and volume of blood culture was done. Inadequate blood samples for culture may increase the likelihood of isolation of CoNS and presumed contaminants, especially in younger children and severely malnourished children with difficult venipuncture. SAM, HIV and malaria are associated with mortaliy in hospitalized children hence these variables were adjusted for in the regression models. Clinical features found to have a p value of 0.1 on univariable regression were entered into multivariable regression models to determine the association of CoNS with clinical presentation at admission.
The primary analysis compared children with CoNS isolated to all other children, since the presence of CoNS may have masked growth of known pathogens. As secondary analyses, we also compared children with CoNS isolated to those with no growth on blood cultures, and blood culture isolation of CoNS versus known pathogens. Comparison of children with CoNS to children with presumed contaminants was also done. Sub-group analyses of the relationship between CoNS isolation and HIV status above and below 18 months of age were done as a positive antibody test above 18 months confirms HIV infection, whilst under 18 months, infection and maternal exposure may not be distinguished.

Ethical considerations
This study was approved by the Kenya Medical Research Institute's Ethical Review Committee, Nairobi (SSC/ERC 2906). This analysis was a retrospective review of admissions to hospital based on routine clinical data and laboratory findings.

Participants and blood culture isolates
Overall, 13,315 children were included in the study (S1 Fig (Table 1). CoNS were isolated from blood culture in 906 (6.8%) children, known pathogens were isolated from 564 (4.2%) and 11,044 (82.9%) children had no growth on blood cultures. Presumed contaminants were isolated from blood culture in 801(6.0%) children. 1,714/2,301 (74.5%) of all blood culture isolates were either CoNS or presumed contaminants. 23 children had more than one organism isolated from blood culture ( Table 2). None of culture results had CoNS/presumed contaminants isolated together with known pathogens. CoNS isolation declined during the period studied (Table 3).

Association between CoNS and SAM
The prevalence of SAM was similar between children with CoNS and all other children (14.9% vs. 12.7%, p = 0.16). There was no evidence of an association between either low MUAC or kwashiorkor and CoNS isolation (S2 Table and Table 3) in multivariable analyses, adjusting for age, sex, blood volume sampled, HIV status, malaria, and year of admission.
However, HIV antibody positivity was more common amongst children with CoNS than in those with negative cultures, (aOR 1.49, 95% CI 1.09-2.04) but much less common in children with CoNS than in those with known pathogens (aOR 0.31, 95% CI 0.21-0.45) ( Table 3).

Association between CoNS and outcomes in children with SAM or HIV
Amongst children with SAM and children with HIV, there was no association between isolation of CoNS and mortality compared to all other children or children with no growth on

Clinical features of CoNS in children with SAM or HIV
Amongst children with SAM, no clinical signs or symptoms at admission were more frequently observed amongst children with CoNS than all other children after adjusting for age, sex, blood volume sampled, HIV status, malaria, and year of admission (Table 5). Delayed capillary refill was less frequent amongst children with CoNS than amongst all other children.
Compared to children with known pathogens, children with CoNS less frequently presented with a prolonged capillary refill time (Table 5). Amongst children with positive HIV antibody test, no clinical features were more frequent amongst children with CoNS isolation compared with all other children and those with no

Discussion
In rural Kenya, CoNS were commonly isolated from blood culture amongst children admitted to hospital, as has been reported in several studies in developing countries. [9,13] However, CoNS was no more common among children with SAM and/or a positive HIV antibody test than among other children. CoNS isolation was not associated with inpatient mortality, or All admissions e 3.0 HIV, human immunodeficiency virus. a n/N (%), number with variable of interest/total group size (percentage). b Adjusted for age, sex, blood volume, HIV Ab status, nutrition status, malaria, and year.
prolonged duration of hospitalization amongst those discharged alive. In children with SAM or HIV there was no clinical phenotype associated with CoNS blood culture isolation. Thus, take the approach of a large epidemiological study, our findings suggests that CoNS are not commonly behaving as pathogens in children with HIV or SAM in this setting, and overall do not warrant targeted empiric first-line therapy. However, we cannot exclude that in a very small number of individuals CoNS may have been pathogenic.
Our conclusions differ to a study conducted among 336 severely malnourished children in Jamaica where CoNS [24/69 (35%)] were the most frequent blood culture isolates. CoNS were reported to be more frequent amongst children with the poorest nutrition status (kwashiorkor and marasmic kwashiorkor) compared to children with marasmus only. Children with CoNS were reported to be more likely to have radiographic pneumonic consolidation than children with bacteremia due to other organisms (8 (22%) p<0.02). [8] CoNS isolation was therefore considered to be clinically significant requiring appropriate antimicrobial therapy. However, surprisingly given the other findings, isolation of CoNS was not associated with mortality. The differences from our study may be due to chance given the small sample size in Jamaica, confounding by differential skin contamination or because children were recruited from a different population.
CoNS have also been reported as common blood culture isolates and of clinical significance amongst hospitalized severely malnourished children in Kenya, [11] Ethiopia, [13] and amongst HIV-infected children in South Africa [12] and Zimbabwe. [7] However, not all of these studies These findings from resource-poor settings similar to ours reflect the challenges of whether or not to institute definitive management based on interpretation of culture results. Determining whether CoNS are clinically significant in an individual child remains challenging, even when culture facilties exist. For this reason, we chose to examine the risks for CoNS and its outcomes in a large population of sick children in order to inform guidelines used in the vast majority of settings where sick children are treated that do not have access to microbiological diagnosis.
Our study primarily compared children with CoNS to all other children because CoNS may have masked the growth of known pathogens. In addition, we compared children with CoNS to those with either known pathogens or no growth on blood culture. Comparison of CoNS to presumed contaminants did not yield more information than the results we have presented as mortality amongst children with SAM (aOR 0.55, 95%CI 0.25-1.17) and HIV-positive children (aOR 0.83, 95%CI 0. 15-4.47), and clinical features did not differ between the two groups. There were limitations to our study. We did not routinely perform repeat blood cultures when CoNS was isolated or undertake speciation. Our study principally examined blood cultures taken at admission to hospital and therefore may not reflect hospital acquired infections or settings where indwelling devices or invasive ventilation are used. Antibiotic treatment initiated at admission may have influenced outcomes amongst hospitalized children with CoNS if they were sensitive to beta lactams. However, this treatment data was not available and we were therefore not able to adjust for this in the multivariable analysis. We lacked confirmatory PCR test results for HIV viral DNA for children aged <18 months. To mitigate this we conducted a separate analysis adjusting for age (below and above 18 months old) (S3 Table and S4  Table), and results were similar to those obtained including all children with a positive HIV antibody test.
In rural Kenya, although CoNS were common blood culture isolates amongst children with SAM or a positive HIV antibody test, but there was no evidence that CoNS were commonly clinically significant in either of these groups, nor to suggest that specific treatment is required as standard practice.