Аңдатпа
Objective: To evaluate whether serial biomarker changes during corticosteroid therapy were associated with in-hospital mortality among patients with severe or critical COVID-19. Materials and Methods: This single-center retrospective cohort included 286 adults treated with corticosteroids; 186 survived and 100 died in hospital. C-reactive protein (CRP), procalcitonin, lactate dehydrogenase (LDH), ferritin, D-dimer, neutrophils, and lymphocytes were assessed on treatment days 1, 3, 5, 7, and 9; the neutrophil-to-lymphocyte ratio and systemic immune-inflammation index were derived from these measurements. Percentage changes were examined using receiver operating characteristic (ROC) curve analysis and separate logistic regression models adjusted for age, sex, baseline intensive care unit location, any comorbidity, and the corresponding day-1 biomarker value. Results: Among the seven biomarker-interval combinations carried forward into the primary regression table, CRP change between days 5 and 7 showed the greatest discrimination (AUC = 0.75). An exploratory same-cohort derivation cutoff > −31.4%, representing a decrease of less than 31.4% or any increase, yielded 75.5% sensitivity and 69.9% specificity. Within the available adjustment set, changes in CRP, LDH, and ferritin were associated with in-hospital mortality, whereas changes in procalcitonin, neutrophils, and lymphocytes were not statistically significant. Restricted cubic spline analysis suggested a nonlinear association between the percentage change in CRP from days 5 to 7 and in-hospital mortality. Exploratory analyses suggested that the associations of day 5–7 CRP and LDH changes with mortality differed across calendar periods. Conclusions: Serial biomarker trajectories were associated with outcome but were not adequate as standalone decision tools. Because the temporal ordering of biomarker changes and clinical deterioration was unavailable, the CRP cutoff cannot be interpreted as a prospective treatment or escalation threshold. The same-cohort derivation cutoffs require prospective external validation, and the findings should not be interpreted as evidence of corticosteroid nonresponse or causal treatment effects.