To compare the diagnostic accuracy of white blood cell-surface biomarkers (CD64, CD11b and HLA-DR), C-reactive protein (CRP) and hematological parameters to diagnose definite sepsis among pre-term neonates presenting with suspected late-onset neonatal sepsis (LONS). This was a prospective, single-gate, diagnostic study in a Level III neonatal unit. Fifty-three neonates (gestation, <34 weeks) with LONS (onset, >72 age), were enrolled. Cell-surface biomarkers, CRP and haematological parameters were assayed at 0 and 48 h after onset. The reference standard was definite sepsis, defined as a positive blood culture with a non-contaminant organism. The index tests (cell-surface biomarkers, CRP and haematological parameters) were compared between subjects with or without 'definite sepsis'. The area under the receiver operator characteristics curves (AUC) generated for each index test at 0 and 48 h was compared. Level III neonatal unit in a tertiary care institute. Of 53 enrolled pre-term infants, 24 had definite sepsis. Among all the index tests evaluated, CRP at 48 h had the highest AUC [0.82 (95% confidence interval, 0.69, 0.92)]. The expression of CD11b and HLA-DR was significantly reduced among the septic neonates. Among the cell-surface biomarkers, the maximum AUC was recorded for HLA-DR at 48 [0.68 (95% CI, 0.54, 0.81)]. Comparisons between index tests were not statistically significant. C-reactive protein is superior to other sepsis screen biomarkers and white blood cell-surface biomarkers in diagnosing culture-positive LONS among pre-term infants. CD64, CD11b and HLA DR as diagnostic tests in this group have limited discriminatory value. The diagnosis of neonatal blood stream infections is a challenge. In response to bacterial blood stream infections, white blood cells are known to produce an excess of certain types of specialized proteins on their surface, including CD64, CD11b and HLA-DR. In this study we evaluated the concentration of these cell-surface proteins for diagnosing blood stream infections in pre-mature newborn babies, whose onset of infection was beyond 72 h of life. We compared these tests against standard tests that are currently in clinical use, such as C-reactive protein and blood white cell counts. All tests were performed at the time of initially suspecting the infection and 48 h later. The gold standard against which all these tests were evaluated was blood culture, in which the offending bacteria are grown in specialized laboratory media. Of 53 pre-mature babies with suspected infection, 24 had blood culture-proven infection. Among all tests, C-reactive protein at 48 h had the best ability to distinguish definite infection from no infection. The expression of CD11b and HLA-DR was significantly reduced among infected neonates. We conclude that C-reactive protein is superior to white blood cell-surface proteins and white cell count in diagnosing definite late-onset infections among pre-term infants.