Modeling chronic kidney disease (CKD) in type 2 diabetes (T2D) is complex owing to the interplay of numerous risk factors, biomarkers, outcomes, and competing risks. One approach to capturing this interplay is the use of flexible frameworks such patient-level simulation (PLS) based on multivariable regression. The flexibility afforded by regression-based PLS allows a diverse array of risk factors to be captured and modeled. The present study aimed to review the markers, outcomes, risk factors, and sequelae captured in PLS models of CKD in T2D.

Ten regression-based renal PLS models were identified. End-stage kidney disease (ESKD) was the most frequently included kidney outcome, followed by micro- and macroalbuminuria. Doubling of serum creatinine and estimated glomerular filtration rate were the most frequently included markers of CKD. The most frequent clinical risk factors were blood pressure and glycemic control, the most frequent demographic factors were sex and age. Treatment-related factors included anti-diabetic and anti-hypertensive treatment, and treatment with sodium-glucose co-transporter 2 and dipeptidyl peptidase-4 inhibitors specifically. CKD was modeled as a risk factor mainly for mortality as well as cardiovascular and cerebrovascular disease.

Regression-based PLS models of CKD in T2D commonly predicted ESKD, but not, for example, time to dialysis, which is associated with substantial qualify-of-life impairments and costs. Furthermore, models differed in their choice of risk factors (particularly treatment-specific effects) and sequelae. Regression-based PLS modeling of CKD may benefit from greater standardization, including improved justification of endpoint and risk factor selection.