Prognostic significance of serum resolvin D1 levels in patients with acute supratentorial intracerebral hemorrhage: A prospective longitudinal cohort study

Abstract

ObjectiveResolvin D1 (RvD1) has anti-inflammatory properties and may be neuroprotective. This study was designed to assess usability of serum RvD1 as a prognostic biomarker after intracerebral hemorrhage (ICH). MethodsIn this prospective, observational study of 135 patients and 135 controls, serum RvD1 levels were measured. Its relations to severity, early neurologic deterioration (END) and poststroke 6-month worse outcome (modified Rankin Scale scores of 3–6) were determined via multivariate analysis. Predictive effectiveness was evaluated based on area under receiver operating characteristic curve (AUC). ResultsPatients had markedly lower serum RvD1 levels than controls (median, 0.69 ng/ml versus 2.15 ng/ml). Serum RvD1 levels were independently correlated with the National Institutes of Health Stroke Scale (NIHSS) [β, −0.036; 95 % confidence interval (CI), −0.060--0.013; VIF, 2.633; t = -3.025; P = 0.003] and hematoma volume (β, −0.019; 95 % CI, −0.056--0.009; VIF, 1.688; t = -2.703; P = 0.008). Serum RvD1 levels substantially discriminated risks of END and worse outcome with AUCs at 0.762 (95 % CI, 0.681–0.831) and 0.783 (95 % CI, 0.704–0.850) respectively. A RvD1 cut-off value of 0.85 ng/ml was effective in predicting END with a sensitivity of 95.0 % and specificity of 48.4 % and its levels <0.77 ng/ml distinguished patients at risk of worse outcome with a sensitivity of 84.5 % and specificity of 63.6 %. Under restricted cubic spline, serum RvD1 levels were linearly related to risk of END and worse outcome (both P > 0.05). Serum RvD1 levels and NIHSS scores independently predicted END with odds ratio (OR) values of 0.082 (95 % CI, 0.010–0.687) and 1.280 (95 % CI, 1.084–1.513) respectively. Serum RvD1 levels (OR, 0.075; 95 % CI, 0.011–0.521), hematoma volume (OR, 1.084; 95 % CI, 1.035–1.135) and NIHSS scores (OR, 1.240; 95 % CI, 1.060–1.452) were independently associated with worse outcome. END prediction model containing serum RvD1 levels and NIHSS scores, and prognostic prediction model containing serum RvD1 levels, hematoma volumes and NIHSS scores displayed efficient predictive ability with AUCs at 0.828 (95 % CI, 0.754–0.888) and 0.873 (95 % CI, 0.805–0.924) respectively. Such two models were visually shown via building two nomograms. Using Hosmer-Lemeshow test, calibration curve and decision curve, the models were comparatively stable and had clinical benefit. ConclusionThere is a dramatical declination of serum RvD1 levels after ICH, which is tightly related to stroke severity and is independently predictive of poor clinical outcome, implying that serum RvD1 may be of clinical significance as a prognostic marker of ICH.