Multidimensional nutritional and body-composition phenotypes and their associations with prevalent anaemia and metabolic abnormalities in CKD: protocol for an EHR-based retrospective baseline cross-sectional study with a prospective measurement extension.

Abstract

Chronic kidney disease (CKD) is frequently accompanied by nutritional imbalance, inflammation and metabolic disturbance. Single-marker assessment may not capture the multidimensional nutritional heterogeneity of CKD. This protocol aims to derive nutritional phenotypes in adults with CKD and examine their cross-sectional associations with prevalent anaemia and metabolic abnormalities. This single-centre study comprises an electronic health record (EHR)-based retrospective baseline cross-sectional cohort and a prospective continuous-enrolment baseline extension for standardised measurements. Retrospective records will be deterministically linked at the person level across the hospital information system, laboratory information system and dialysis system using a shared hospital patient identifier. Adults aged 18 years or older with CKD confirmed using Kidney Disease: Improving Global Outcomes (KDIGO)-based criteria will be included. The common-core phenotype will be derived from body mass index, serum albumin, ferritin and 25(OH)D. Haemoglobin, lipid measures, glycated haemoglobin and fasting plasma glucose will be used only for outcome ascertainment and descriptive characterisation, not for latent phenotype derivation. C reactive protein and/or the NLR will be examined as complementary inflammation-related variables in covariate, effect-modification and sensitivity analyses. Waist circumference and standardised body-composition measures will inform a prospective subcohort specific extended phenotype framework. Bayesian latent class/profile models will identify phenotypes, and Bayesian regression models will estimate their associations with prevalent anaemia, dyslipidaemia and prevalent diabetes or diabetes-level hyperglycaemia while propagating phenotype-classification uncertainty. Missing data will be handled according to variable role. Pooled cross-phase analyses will be restricted to the common-core variable framework, with extended phenotyping limited to the prospective subcohort. The study was approved by the Ethics Committee of Cangzhou Central Hospital (approval No. 2025-286-02). The retrospective component will use de-identified EHR data under an approved waiver of individual informed consent, whereas written informed consent will be obtained for the prospective component. Findings will be disseminated through peer-reviewed publications and academic conferences. Open Science Framework: 10.17605/OSF.IO/FZ65E.