Study benchmarks inflammation and metabolic markers for 5-year mortality prediction in sleep apnea
Obstructive sleep apnea is linked to higher cardiovascular and all-cause mortality, but identifying which patients face the greatest risk remains a clinical challenge. In a study published July 10 in the Journal of Translational Medicine, researchers led by Jian Liu and corresponding author Xiaoming Li of Shandong Provincial Hospital Affiliated to Shandong First Medical University benchmarked seven inflammation-nutrition and triglyceride-glucose (TyG) related indices as predictors of 5-year all-cause mortality in adults with OSA.
The study found that a combined model incorporating five markers provided the most useful risk signal, though the authors frame their results as “pragmatic risk benchmarking” rather than a ready-for-clinic prediction tool.
What they found
The investigation proceeded in two phases. In the derivation phase, the team analyzed data from 3,503 adults with questionnaire-defined OSA drawn from the National Health and Nutrition Examination Survey (NHANES) cycles 2005-2008 and 2015-2018. Median follow-up was 57.0 months, during which 293 deaths occurred.
The seven candidate biomarkers fell into two categories. Four were metabolic indices derived from the TyG (triglyceride-glucose) index and its variants incorporating body mass index, waist circumference, and waist-to-height ratio, plus the TG/HDL-C ratio. Two were inflammation-nutrition markers: the advanced lung cancer inflammation index (ALI) and the neutrophil percentage-to-albumin ratio (NPAR).
Among individual biomarkers, ALI showed the most robust mortality-related signal across analyses, the authors reported. NPAR showed signal in selected single-marker and nonlinear analyses but was less consistent after full multivariable adjustment. The TyG-related indices were variably associated with mortality on their own and contributed mainly within the combined prediction model.
The best-performing model was a continuous-scale combined model including TyG-BMI, TyG-WC, TyG-WHtR, TG/HDL-C, and ALI, which achieved an AUC of 0.765 (bootstrap-corrected AUC 0.761) in the derivation cohort. The likelihood-ratio test comparing this model to a base model with routine clinical variables yielded p < 0.001, and the net reclassification improvement was also significant at p < 0.001, indicating modest but measurable incremental value.
For external validation, the researchers turned to a multicenter Chinese cohort of 200 patients from six hospitals, all of whom had polysomnography-confirmed OSA and at least 5 years of follow-up. In this cohort, the combined model achieved an AUC of 0.697, compared with 0.672 for the base model. However, calibration was poor, and decision-curve analysis showed net benefit was confined to a narrow higher-threshold range (0.14-0.30).
Why it matters
OSA affects an estimated one billion people worldwide, and the condition carries an elevated mortality risk that existing clinical tools capture only partially. Routinely available blood-based biomarkers that add even modest predictive value could help clinicians stratify risk more effectively without costly or invasive testing.
The study’s systematic head-to-head comparison of seven indices in the same population is valuable for researchers designing risk prediction tools. The finding that ALI outperformed other single markers and that a combination of metabolic and inflammation-nutrition indices performed best provides guidance for future model development.
The authors noted that TyG-related indices and ALI are inexpensive, reproducible, and derived from standard lab panels, making them realistic candidates for integration into clinical workflows if further validated.
Limits
The study has several important limitations. First, OSA was identified by questionnaire rather than polysomnography in the derivation cohort, which may introduce misclassification. Second, the external validation cohort was relatively small (200 patients) with few events, limiting statistical power. Third, calibration was poor in the external cohort, and any net benefit was restricted to a narrow threshold range. The authors describe the external validation results as “preliminary evidence of transportability” rather than definitive support for a clearly superior prediction tool.
Bottom line
A combined model incorporating TyG-related metabolic indices and the advanced lung cancer inflammation index modestly improved 5-year all-cause mortality risk prediction beyond routine clinical variables in adults with OSA. The findings support further research into these biomarkers for risk stratification but do not yet justify clinical implementation.
Source
Liu J, Yang H, Wang Z, Li H, Ahmad S, Zhou S, Zhou L, Lou D, Guo Z, Li X. Benchmarking inflammation-nutrition and TyG-related indices for 5-year mortality risk in adults with questionnaire-defined obstructive sleep apnea: a survey-weighted NHANES derivation cohort with multicenter external validation. J Transl Med. 2026 Jul 10. doi: 10.1186/s12967-026-08540-0. PMID: 42432717.