Abstract / Summary
Background: Chronic kidney disease (CKD) is linked to cognitive, psychological and functional decline, yet few studies track these domains together around the first detection of incident disease. We examined three longitudinal outcomes while preserving observed CKD onset and ordinary study time for participants who remained free of CKD. Methods: We analysed harmonised longitudinal survey and biomarker data from CHARLS (China, 2011–2018), HRS (United States, 2006–2016) and SHARE (Europe and Israel, 2015–2022) among adults aged 50 years or older. The main comparison included 75,044 participants (2480 with incident CKD and 72,564 free of CKD). Global cognition, depressive symptoms and functional limitation were three prespecified co-primary outcome domains, not a composite. Missing baseline covariates were imputed separately by cohort with multiple imputation by chained equations (MICE; m = 20, 20 iterations). Random-intercept restricted maximum likelihood (REML) mixed models (Model 1 and Model 2) were fitted in each completed data set and combined with Rubin’s rules. Model 2 adjusted for seven baseline chronic disease indicators, body mass index, smoking and alcohol use. For each outcome, inference focused on the change in the annual outcome trajectory associated with incident CKD occurrence (the trajectory change estimate), with Benjamini–Hochberg correction across the three Model 2 tests. Results: Rubin-pooled Model 2 trajectory change estimates were global cognition, −0.019 (−0.031 to −0.007) (BH q = 0.003); depressive symptoms, 0.018 (0.005 to 0.032) (BH q = 0.008); functional limitation, 0.075 (0.063 to 0.087) (BH q < 0.001), in standardised outcome units per year. An exploratory joint study-time module identified three trajectory classes and compared class membership by CKD status. Lagged Cox models used CKD status at the interval start for four non-frailty, visit-detected proxy endpoints. Conclusions: Incident CKD was associated with directionally and quantitatively distinct longitudinal changes in cognition, depressive symptoms and functional limitation. These results are adjusted associations, not instantaneous causal effects. Joint trajectory classes and lagged event models were exploratory; frailty index and stacked three-domain GEE analyses were supplementary.