Mixed Effects Regression Models for Estimated Glomerular Filtration Rate in Glomerular Diseases

Background: Disease progression in non-diabetic glomerular diseases is traditionally modeled assuming linear decline in estimated glomerular filtration rate (eGFR), despite known non-linearity in a subset of patients. Methods: Adults and children with MCD, FSGS, MN, and IgAN/V from the Cure Glomerulonephropathy Network (CureGN) with at least 2 eGFR measurements were included. Mixed-effects linear regression models with random intercepts and slopes were fitted, stratified by age, sex, APOL1 risk alleles, hypertensive status, and weight status. Potential improvement of model fit with the inclusion of a quadratic term was assessed using likelihood ratio tests. Results: Among 2790 participants, 34% were pediatric, 43% were female, 27% and 35% were overweight and obese, respectively, and 44% were hypertensive. Overall eGFR decline was -1.3 ml/min/1.73m2 per year, and was slowest in MCD (-0.2 per year) and fastest in FSGS (-2.0 per year). eGFR decline was faster in overweight and obese participants (-1.6 and -1.4 versus –0.8 per year, respectively, p=0.02) and those with Stage 1 and 2 hypertension (-1.5 and -2.2 versus -0.7 per year, p<0.001). Participants with MCD, those under 12 years old at enrollment, and those with normal weight and normal or elevated blood pressure had statistically significant quadratic terms for time (range -0.5 to –0.12), indicating periods of initial stability and subsequent decline, on average (Figure). Among the subset with genetic testing, those with 2 APOL-1 risk alleles (121/1,959) had faster progression (-3.2 ml/min/1.73m2 per year) and a significant quadratic mean trajectory (beta = 0.42), indicating steep initial decline followed by stabilization. Conclusion: These findings underscore that eGFR decline is not uniform across patients with glomerular diseases but varies by demographic, genetic, and health factors. Non-linear modeling of eGFR in certain subgroups may improve risk prediction. Funding: NIDDK Support