AI Summary & Key Takeaways
AI InsightThe overall association was insensitive to the substitution value used for the 27.3% of measurements below the assay limit of quantification (LOQ, 3 ng/ml) and to multiple imputation; however, in the complete-case analysis restricted to quantifiable concentrations no dose response gradient remained (OR, 0.950; p=0.797), indicating that the signal is driven by concentrations below the LOQ rather than by a continuous gradient across the measurable range.
📌 3 Key Takeaways
Abstract
Aims: Hypovitaminosis D is near-universal in critical illness, but whether it reflects a modifiable micronutrient deficit or the acute-phase response is unclear, and no accepted 25-hydroxyvitamin D 25(OH)D threshold exists. We examined admission 25(OH)D and 28-day mortality in sepsis, testing dependence on cut-off choice and on nutritional status. Methods: We retrospectively studied 267 adults admitted to a general intensive care unit (ICU) with sepsis as the admission diagnosis, who stayed at least 48 hours and had 25(OH)D measured on admission as part of the routine panel. The primary exposure was log-transformed continuous 25(OH)D; pre-specified Institute of Medicine categories and tertiles were secondary. Logistic regression was adjusted for age, gender, severity scores, comorbidity, albumin, and lactate. Results: Median 25(OH)D was 10.0 ng/ml; 75.7% were below 20 ng/ml. Mortality at 28 days was 52.1%. The overall association was insensitive to the substitution value used for the 27.3% of measurements below the assay limit of quantification (LOQ, 3 ng/ml) and to multiple imputation; however, in the complete-case analysis restricted to quantifiable concentrations no dose response gradient remained (OR, 0.950; p=0.797), indicating that the signal is driven by concentrations below the LOQ rather than by a continuous gradient across the measurable range. Each doubling of 25(OH)D carried 22% lower adjusted odds of death [odds ratio (OR), 0.775; 95% confidence interval (CI), 0.645-0.931; p=0.006]. Risk was confined to the lowest tertile (≤5.0 ng/ml; adjusted OR, 2.460; 95% CI, 1.168-5.184; p=0.018), with middle and highest tertiles indistinguishable. The pre-specified ≤12 ng/ml category was non-significant after adjustment (p=0.097), whereas moving the boundary to <12 or ≤10 ng/ml made it significant. Effect sizes were near-identical in hypoalbuminemic and normoalbuminemic strata. Discrimination was poor (area under the curve, 0.570). Conclusion: Low admission 25(OH)D is independently associated with 28-day mortality in sepsis, but the signal is driven by values below the assay limit of quantification rather than by a gradient across the measurable range, and conventional cut-offs provided no prognostic separation. The limited available nutritional proxies do not explain it, and 25(OH)D is not usable as a stand-alone prognostic test.