Prospective multicentre study of host response signatures in neonatal sepsis in Sub Saharan Africa

Few biomarkers for sepsis diagnosis are commonly used in neonatal sepsis. While the role of host response is increasingly recognized in sepsis pathogenesis and prognosis, there is a need for evaluating new biomarkers targeting host response in regions where sepsis burden is high and medico-economic resources are scarce. The objective of the study is to evaluate diagnostic and prognostic accuracy of biomarkers of neonatal sepsis in Sub Saharan Africa. This prospective multicentre study included newborn infants delivered in the Abomey-Calavi region in South Benin and their follow-up from birth to 3 months of age. Accuracy of transcriptional (CD74, CX3CR1), proteic (PCT, IL-6, IL-10, IP-10) biomarkers and clinical characteristics to diagnose and prognose neonatal sepsis were measured. At delivery, cord blood from all consecutive newborns were sampled and analysed, and infants were followed for a 12 weeks’ period. Five hundred and eighty-one newborns were enrolled. One hundred and seventy-two newborns developed neonatal sepsis (29.6%) and death occurred in forty-nine infants (8.4%). Although PCT, IL-6 and IP-10 levels were independently associated with sepsis diagnosis, diagnostic accuracy of clinical variables combinations was similar to combinations with biomarkers and superior to biomarkers alone. Nonetheless, CD74, being the only biomarkers independently associated with mortality, showed elevated prognosis accuracy (AUC > 0.9) either alone or in combination with other biomarkers (eg. CD74/IP-10) or clinical criterion (eg. Apgar 1, birth weight). These results suggest that cord blood PCT had a low accuracy for diagnosing early onset neonatal sepsis in Sub Saharan African neonates, while association of clinical criterion showed to be more accurate than any biomarkers taken independently. At birth, CD74, either associated with IP-10 or clinical criterion, had the best accuracy in prognosing sepsis mortality. Trial registration Clinicaltrial.gov registration number: NCT03780712. Registered 19 December 2018. Retrospectively registered. Participants were delivered in two sub-urban health centres (sub-urban arm) and three urban University hospitals (hospital arm) in the Abomey-Calavi, Sô-Ava and Cotonou districts in the South Benin region where malaria is hyper-endemic34. In both arms, only infants born from mother living in the Abomey-Calavi district were recruited to facilitate the follow-up and minimize effect of geographical origins. In the sub-urban arm, that includes only normal geatation with low-risk delivery (no maternal risk factors for infection), all consecutive births were included, whereas in the hospital arm only newborns born from mothers with maternal-foetal risk factors for infection (prematurity, prolonged rupture of the membrane, maternal fever) were included. In both arms, the exclusion criteria included maternal HIV positive status, major congenital malformation and refusal of consent. All children from both arms were followed clinically on a bi-monthly base during the first 3 months of life. The follow-up consisted of scheduled home visits and unscheduled emergency visits if the infant was ill. The study protocol was approved by the Comité d’Ethique de la Recherche – Institut des Sciences Biomédicales Appliquées (CER-ISBA 85-5). Written informed consent was obtained from parents. All methods were performed in accordance with the relevant guidelines and regulations. The exposure were occurrence of gestational malaria (GM). GM was defined as a malaria infection during pregnancy or at delivery. For women in the suburban arm, malaria screening was performed at each scheduled prenatal visit using a thick blood smear. Mothers from the Hospital arm were screened only at the time of delivery. In this group, antenatal malaria was established on the basis of mother’s anamnesis. For both study arms, placental blood smear and mother’s peripheral blood smear were performed. The Lambaréné technique was used to quantify parasitaemia with a detection threshold of five parasites per microliter34. Neonatal sepsis was suspected in neonates with more than two of the following criteria being present: neutrophil count < 7500/mm3 or > 14 500/mm3, band form > 1500/mm3, immature/total neutrophils ratio > 0.16, platelets count < 150 000/mm3 and CRP > 10 mg/L. Suspected neonatal sepsis was considered as clinical sepsis when the following clinical signs were associated: temperature irregularity; respiratory distress or apnoea; seizures, altered tonus, irritability or lethargy; vomiting, altered feeding pattern, ileus; skin perfusion alteration, haemodynamic signs (tachycardia, hypotension); hypoglycaemic/hyperglycaemic, hyperlactatemia or identification of focal infection such as soft tissue infection or conjunctivitis19,57,58. All newborns with a clinical sepsis were subsequently adjudicated by one independent pediatrician (PT) and sorted into ‘presumed sepsis’ and ‘definite clinical sepsis’ grouped as “adjudicated sepsis”. In discordant cases, a second independent pediatrician (ULT) performed the final adjudication with access, in addition to the full medical file review, to microbiological cultures results. Parallel to microbiological culture (BACT/ALERT® system), specific BioFire® FilmArray® panels (bioMerieux, Marcy-l’Etoile, France) were run for all positive blood cultures (Blood Culture Identification (BCID) panel), cerebrospinal fluids (meningitis/encephalitis panel), respiratory and gastrointestinal samples (Pneumonia and Gastro-Intestinal panels). All studied biomarkers were kept blinded for the adjudication. At birth and at follow-up visits, the clinical examination data of the children were collected. Blood samples were obtained at birth, then at week (W)1, W4, W8 and W12. The study protocol has been described in detail elsewhere34. In brief, 18 mL cord blood at birth and 2 mL peripheral blood were collected in PAXgene™ Blood RNA tubes (PreAnalytix, Hilden, Germany) for transcriptomic biomarkers study. For protein biomarkers study 9 mL cord blood at birth and 500 µL to 1 mL peripheral blood were collected in heparin tubes. PAXgene™ Blood samples were stabilized at least 2 h at room temperature after collection and frozen at − 80 °C following the manufacturer’s guidelines. Heparin blood samples were centrifuged at 1500 to 2000 rpm for 5 min. The plasma obtained was stored at − 80 °C. Total RNA was extracted from cord blood by QIAsymphony SP/AS (QIAGEN Hilden, Germany) using PAXgeneTM Blood RNA Kit (PreAnalytix, Hilden, Germany) according manufacturer guidelines. The RNA integrity was measured prior to RNA amplification using a Bioanalyser 2100 (Agilent Technologies, Palo Alto, CA) in accordance with the manufacturer’s instructions (RNA integrity number ≤ 6 were excluded [min 6.5–max 9.4]). The expression level of CX3CR1 and CD74 was measured by RT-qPCR using ABI7500 thermocycler (Applied BioSystems®, California,USA) from 10 ng of RNA samples using prototype Argene® kit (bioMerieux, France) following the manufacturer’s instructions. The determination of the copy concentration of each gene is performed using a calibration curve. The results are expressed as the ratio of the concentrations of CX3CR1 or CD74 to those of HPRT1 (Hypoxanthine Phosphoribosyltransferase 1) used to normalize the gene expression results. PCT, IL-6, IL-10 and IP-10 proteins in plasma were analyzed in a batch by Simplex AssaysTM according to the manufacturer’s instructions as previously described59. The Ella microfluidic analyzer (Protein Simple, San Jose, CA, USA) was used to assess cytokine concentrations60. The primary outcome was the diagnosis of clinical neonatal sepsis, and secondary outcome was mortality within the first 3 months of life. Neonatal sepsis diagnosis was established by the local paediatrician based on the clinical examination of the child and initial laboratory workup including haemogram, C-reactive protein (CRP) and microbiological cultures (blood, cerebral fluids and urine). Neonatal sepsis that occurred within the first 72 h following birth was considered as an early onset neonatal sepsis (EONS), and late onset (LONS) thereafter (for detailed algorithm for sepsis diagnosis, see published study protocol34). An independent statistician (FB) (Soladis Inc. Lyon, France; https://www.soladis.com) and IRD biostatistician (GA) supported the statistical methodology and performed all analysis. Statistical analyses were performed using R software version 3.6.1. The variables were assessed for normality using Kolmogorov Smirnov test. Numbers and frequency were used for qualitative data and medians and IQR (inter-quartile range: [Q1–Q3]) for quantitative data. Qualitative variables were compared using the Chi-squared test (or Fisher’s exact test for small expected numbers). The distribution of quantitative data was compared using Student’s t-test (or the Mann–Whitney t-test when distribution was not normal or Welch test when homoscedasticity was rejected) if 2 groups were compared. If more than 2 groups, the distribution of quantitative data was compared using Anova test (or the Kruskal–Wallis test when distribution was not normal or when homoscedasticity was rejected). To evaluate their diagnostic accuracy, data-driven analysis was performed. Three models were used: Model 1 encompassed only Hospital patients either with or without clinical sepsis; Model 2 enriched the non-sepsis group with unseptic patient from the suburban arm. Model 3 refined the sepsis group with only the culture positive sepsis considered. Selection of cut-off values or discrimination values defining the positive and negative test results were performed. Several methods for selecting optimal cut-off values in diagnostic tests are proposed in the literature depending on the underlying reason for this choice. Here, we selected a cut-off to have a sensitivity of 0.95 and maximize specificity. This choice of cut-off was the same in the rest of the publication61. CD74/IP-10 was the score corresponding to the division of CD74 gene expression level by IP-10 serum concentration. We used three datasets to test more complex models, either genes (noted CX3CR1 & CD74), proteins (noted Protein biomarkers) or sets of biomarkers (noted All biomarkers). To avoid overfitting and to compare our models, we used random sampling which takes place within each class and must preserve the overall distribution of data by class. To do this, we created a distribution, repeated 200 times, of 75/25% of the data. This distribution was used to optimize hyperparameters with package caret. Then we compared the average AUC and select the best average AUC for each dataset between all models. AUC accuracies were compared using Bootstrap approach. A p value < 0.05 was considered as significant. For comparison between clinical variables (VC) and biomarkers, we selected the clinical variables of interest following an expert opinion (PT) and corresponding to neonatal risk factors (eg. Gestational age, weight, maternal risk factors, multiple gestation, APGAR score) (Supplementary Table 1). In order to compare clinical and biomarker data, we transformed the data into the same referential to be able to compare them. We transformed the categorical clinical variables into a complete set of dummy variables62. Different transformations were then applied to the data set. A Yeo-Johnson transform is a non-linear transformation that reduces skewness and approximates a normal law. We centered (subtracts the mean of the variable’s data) and scaled data (divides the standard deviation). For the establishment of heatmap, Partial Last Squares (PLS) Regression was used to compare the two datasets. This algorithm comes from the mixOmics package. Biomarkers were deflated with respect to the information extracted/modelled from the local regression on Clinical Variables. Consequently, the latent variables computed to predict Biomarkers from Clinical Variables are different from those computed to predict Clinical Variables from Biomarkers. One matrix Clustered Image Map (CIM) is a 2-dimensional visualization with rows and/or columns reordered according to some hierarchical clustering method to identify interesting patterns. The CIM allows to visualize correlations between variables. Generated dendrograms from clustering were added to the left side and to the top of the image. The used clustering method for rows and columns is the complete linkage method and the used distance measure is the distance Euclidean. We showed only variables with co-variances greater than max(covariance)/262. To study clinical items and biomarkers factors associated with clinical sepsis, and death, we used univariate and multivariate logistic regression models. Three separate models were performed, one for clinical sepsis (Yes/No), the second for confirmed sepsis, corresponding to adjudicated sepsis (Yes/No) (data not shown) and a last one for the death (Yes/No). All variables with a p-value below 0.25 in univariate analysis were selected for the multivariate analysis. In addition, all biomarkers were forced into the multivariate models. Then, a manual backward selection procedure was used to obtain the final adjusted multivariate model, a p-value of < 0.05 was considered statistically significant. Stata version 15 for Windows (Stata Corp., College Station, TX) was used for statistical analyses. We tested several algorithms: Bagged CART (package ipred), CART (package rpart), Naive Bayes (package klaR and naivebayes), Multi-Step Adaptive MCP-Net (package msaenet), Bagged Flexible Discriminant Analysis and Flexible Discriminant Analysis (package earth), Multi-Layer Perceptron (package RSNNS), Regularized Random Forest (package RRF), Quadratic Discriminant Analysis and Linear Discriminant Analysis (package MASS), Robust Linear Discriminant Analysis (package rrcov), Robust Mixture Discriminant Analysis (package robustDA), High Dimensional Discriminant Analysis (package HDclassif), Random Forest (package randomForest and ranger), Stochastic Gradient Boosting (package gbm), Conditional Inference Tree (package party), Neural Network and Neural Networks with Feature Extraction (package nnet), Generalized Additive Model using Splines (package mgcv), Distance Weighted Discrimination with Polynomial Kernel and Distance Weighted Discrimination with Radial Basis Function Kernel (package kerndwd), Support Vector Machines with Radial Basis Function Kernel (package kernlab), Single C5.0 Ruleset (package C50), Boosted Logistic Regression (package caTools), Oblique Random Forest (package obliqueRF), Sparse Partial Least Squares (package mixOmics), k-Nearest Neighbors. For the outcome “sepsis adjudicated”, the model Multi-Layer Perceptron was selected for “All biomarkers”, the model Quadratic Discriminant Analysis was selected for “Protein Biomarkers” and k-Nearest Neighbors for “CX3CR1 & CD74”. For the outcome “clinical sepsis”, the model Generalized Additive Model using Splines was selected for “All biomarkers”, the model Distance Weighted Discrimination with Polynomial Kernel was selected for “Protein Biomarkers” and Support Vector Machines with Radial Basis Function Kernel for “CX3CR1 & CD74”. For the outcome death, the model Sparse Partial Least Squares was selected for “All biomarkers”, the model Sparse Partial Least Squares was selected for “Protein Biomarkers” and Neural Network for “CX3CR1 & CD74”. The random forest model did not select any variables (6Biom + 16 VC). The Stochastic Gradient Boosting model selected “5 Biomarkers and 6 Clinical Variables”. The Flexible Discriminant Analysis model selected “3 Biomarkers and 4 Clinical Variables”. The Multi-Step Adaptive MCP-Net model selected “1 biomarker and 2 Clinical Variables”. See clinical variables legend in Supplementary Table 1. The study protocol was approved by the local institutional review board (Comité d’Ethique de la Recherche de l’Institut des Sciences Biomédicales Appliquées CER-ISBA 85-5). Written informed consent was obtained from parents.