Analysis plan and code: do composite inflammatory-nutritional indices add prognostic information beyond their components?

Purpose. Composite inflammatory–nutritional indices such as the neutrophil-to-lymphocyte ratio are widely reported as predictors of postoperative death, but a ratio is informative only if its components act on risk in equal and opposite proportion, which is rarely tested. This study asks whether six routine indices (NLR, PLR, SII, FAR, CAR and PNI) carry prognostic information beyond the laboratory values they are built from. Design. Secondary analysis of INSPIRE v1.4.2 (PhysioNet, doi:10.13026/1eay-yc85) , a de-identified perioperative dataset from Seoul National University Hospital, 2011–2020. The cohort comprises 12,890 adults at their first orthopaedic operation, excluding pre-existing kidney failure; 356 died within 1 year and 548 within 2 years. For each index, logistic models of 1-year death compare the index with its own components (likelihood ratio test of the ratio constraint, Holm-adjusted across indices), with its dominant single component, and with a 20-df clinical model (optimism-corrected change in AUC, calibration and decision curves). Missing values are handled by multiple imputation. Death within 2 years is the secondary outcome; fifteen sensitivity analyses are prespecified. Expected outcomes. We hypothesised that for most indices the components would carry information the ratio discards, and that no index would exceed its dominant component by 0.01 in AUC. The increment over the clinical model is estimated without a directional hypothesis and judged against prespecified thresholds. Registration status. This is a retrospective registration. The analysis plan and all code were frozen by an MD5 manifest on 11 October 2026 at 17:52 CST, and the primary analysis was completed the same evening, before submission; Q18 gives the full timeline. The plan, code, and pre-analysis outputs are archived on Zenodo, doi:10.5281/zenodo.23298642 .