The Analysis and Reporting of Cardiovascular and Neoplastic Event Data in Type 2 Diabetic Populations with Diabetic Nephropathy: A Systematic Review

Background: Type 2 (T2) diabetes and diabetic nephropathy are risk factors for a range of vascular and neoplastic outcomes. The competing nature of these outcomes means that the use of standard survival analysis techniques (like Kaplan-Meier (KM) curves and Cox proportional hazards (PH) regression) could introduce bias in risk estimates. The literature was reviewed to determine whether authors accounted for competing risks (CRs) when reporting time-to-event analyses in populations with T2 diabetes and nephropathy. Methods: Medline, EMBASE and CINAHL databases were searched for studies with ≥5 years follow up reporting vascular or neoplastic outcomes in patients with T2 diabetes and nephropathy. Results: The search returned 3,886 abstracts. Of these, 179 studies were eligible for full-text review, with 29 papers eligible for data extraction. No study accounted for CRs. Analyses liable to incorporate bias through ignoring competing outcomes were present in 21 studies, of which: all studies used Cox-PH regression, 6 studies used KM curves and 2 studies produced cumulative incidence estimates from Cox-PH analysis. The remaining 8 studies did not use time-to-event analyses. Conclusions: Studies investigating cardiovascular or neoplastic outcomes in populations with DN fail to account for CRs in their analyses. Using standard analyses when populations are at increased risk of multiple outcomes is liable to introduce an unknown level of bias in risk estimates. Absolute risk estimates from KM and Cox-PH regression are well known to be susceptible to this, while many have identified relative risk measured from Cox-PH regression as also being susceptible.