Comparative performance of NAR, SII and PIV in predicting infliximab primary non-response in biologic-naïve ulcerative colitis patients

Abstract

Abstract Background and aims About one third of patients with ulcerative colitis fail infliximab induction. Composite haematological indices — the neutrophil-to-albumin ratio (NAR), systemic immune-inflammation index (SII) and pan-immune-inflammation value (PIV) — are promoted as accessible baseline predictors, but have never been compared formally with one another, or with their own components, in a prospective biologic-naïve cohort. Methods We prospectively enrolled 98 biologic-naïve adults commencing infliximab as elective outpatients for moderate-to-severe ulcerative colitis. Primary non-response was assessed at week 24 by a pre-defined clinical–endoscopic algorithm. ROC curves were compared by DeLong’s test; independent predictors were identified by Firth penalized logistic regression with bootstrap internal validation, calibration and decision-curve analysis. Results Thirty-one patients (31.6%) met criteria for primary non-response. NAR and the absolute neutrophil count discriminated equivalently (AUC 0.818, 95% CI 0.727–0.909, versus 0.815; ΔAUC 0.003, 95% CI − 0.044 to + 0.051), and NAR was not superior to SII (0.785), body mass index (0.781) or PIV (0.742; all P > 0.2). Albumin added nothing to a model containing the neutrophil count ( P = 0.376). NAR remained independently associated with non-response across five adjusted specifications (odds ratio 1.28–1.38 per 0.1 unit), but substituting the neutrophil count gave equivalent models (C-statistic 0.886 versus 0.878; P = 0.515). The optimism-corrected C-statistic was 0.856, with positive net benefit across 20–50% thresholds. At ≥ 1.04, NAR gave sensitivity 87.1%, specificity 71.6% and negative predictive value 92.3%. The components were informative at different times: albumin did not discriminate at baseline (0.586) but its week-24/baseline ratio did (0.775), whereas the neutrophil count showed the reverse (0.815 and 0.524). Conclusions Baseline neutrophil-based markers identify biologic-naïve patients at risk of infliximab primary non-response, with a negative predictive value of 92.3% supporting a rule-out role from routine laboratory data alone. Composite indices offered no advantage over the absolute neutrophil count, and the equivalence was bounded rather than merely undemonstrated. The two components carried information at different points in the treatment course. The threshold is exploratory and requires external validation.