Blood culture is currently the gold standard for diagnosing neonatal sepsis, but various biomarkers in serum are commonly used for rapid supportive diagnosis. Among them, serum C-reactive protein (CRP) is the most extensively studied biomarker in neonates and is found in various body fluids like serum and saliva [1,2]. Considering the difficulty with phlebotomy in neonates and the potential problems caused by blood sampling, alternative fluids like saliva could serve as an attractive non-invasive option for detection of CRP. Salivary diagnostic assays have now been well described in various systemic and oral diseases in adults [3,4]. However, translation into neonatal clinical practice is limited and is mostly restricted to cortisol assessment as a stress marker in term and preterm infants [5,6]. The present study was thus planned to investigate if CRP is detectable in saliva, and whether levels correlate with serum CRP, among neonates with sepsis.

Neonates (aged £28 days) of any gestational age, with clinical suspicion of neonatal sepsis or with perinatal risk factors of sepsis, were included in the study, after informed written consent. Neonates with major congenital malfor-mations, oral infections and oral ulcers were excluded from the study. Demographic findings, including intrapartum maternal fever, presence of preterm and or prolonged rupture of membranes, chorio-amnionitis, gestational age, sex, antibiotic treatment, clinical features, and day of presentation were recorded. All subjects underwent a sepsis workup including serum CRP, complete blood counts, and blood culture, in addition to other investi-gations as deemed necessary by the treating physician. Serum CRP was measured using immuno-turbidimetric method using Roche Cobas C series analyzer (Roche Diagnostics). Neonates were then further classified into three groups: Group I – blood culture and sepsis-screen positive sepsis, Group II – blood culture negative but sepsis screen positive (serum CRP ³10 mg/L and one additional parameter - neutropenia based on standardized age based charts or immature: total neutrophil ratios > 0.2), and Group III - only clinical features or perinatal risk factors.

Saliva was collected within 4 hours of collection of the serum sample. Antibiotics were started only after collection of the salivary sample. SalivaBio Infant’s Swab (SIS) (Salimetrics), was used for this purpose [7]. Salivary sampling was done 60 minutes before or after a feed or any oral procedure. The 90 mm SIS was placed between the cheeks and the lower gums after turning the infants head to one side. The swab was left in place for 3 minutes and then removed. The swab was then visually inspected and if found dry, re-introduced for an additional 3 minutes, up to a maximum of two attempts. Only clear saturated swabs were used. The swab was then introduced, saturated end first, into a plunger removed 5 mL syringe and then the re-introduced plunger was used to squeeze out saliva into cryovials. The process was repeated till 0.5 mL volume of saliva was obtained. If sample collected was inadequate, then the saliva collection was repeated with a longer oral stay period for the SIS. In case of any infant discomfort (gag, brow bulge, nasolabial furrow etc.) noticed during procedure, it was immediately discontinued and resumed after signs abated. After collection, labelled cryovials were frozen at –20oC within 4 hours of collection. The saliva collection was done by the treating team members, after structured training. Each member was supervised during initial few sample collections, to ensure standardization.

Assuming that salivary CRP has a minimum expected sensitivity and specificity of 75% to predict serum CRP ³10 mg/L, and an expected prevalence of neonatal sepsis of 40% (as per departmental statistics for the previous year), with an absolute precision of 10% and 90% confidence level, the required sample size was 128 subjects (STATA IC, ver. 13). To account for 5% non-response rate, a total of 136 subjects were included in the final study.

Statistical analysis: Considering the skewed distri-bution of serum and salivary CRP levels, correlation was assessed using Spearman rank order (Spearman’s rho) coefficient. The utility of salivary CRP in predicting elevated seum CRP (³10 mg/L) was assessed by receiver operating characteristic (ROC) curve analysis and an appropriate salivary CRP cut off value was calculated. The sensitivity, specificity, predictive values and likelihood ratios of salivary CRP at the derived cut-off were calculated for predicting serum CRP ³10 mg/L and positive blood culture. P value <0.05 was considered statistically significant. IBM SPSS version 22 (SPSS Inc.) was used for statistical analyses.

A total of 182 neonates satisfied the inclusion criteria. 32 neonates were excluded because of refusal of consent or already receiving antibiotics. Adequate salivary sample could not be obtained in 9 neonates and in 5, the samples were not analyzed because of contamination. Thus,136 neonates were included in the final analysis. The median (IQR) birthweight was 1.98 (1.34,2.57) kg and gestational age was 34.5 (32,37) weeks. Early onset sepsis (£3 days) was seen in 88 (64.7%) of the population. Salivary CRP was detectable in 135 (99%) neonates and the levels increased significantly from Group III to Group I neonates (Table I).

There was a moderate and statistically significant positive correlation between salivary and serum CRP values in the entire study population (r=0.63; P =0.01) and in Group I (r=0.63; P =0.01) and Group II neonates (r =0.5; P =0.01).

The area under the ROC curve for salivary CRP to predict serum CRP ³10 mg/L was 0.861 (95% CI, 0.78 - 0.94, P<0.01), indicating good predictive validity. Based on the co-ordinates of the ROC curve, the cut-off of 0.6 ng/mL was chosen as the optimal salivary CRP cut-off value for predicting serum CRP ³10 mg/L. Salivary CRP ³0.6 ng/mL had a 77% sensitivity, 94% specificity, 99% positive predictive value and 35% negative predictive value, for predicting a serum CRP level of ³10 mg/L; and 75% sensitivity, 58% specificity, 44% positive predictive value and 85% negative predictive value, for predicting a positive blood culture.