Differential Analysis & Reporting of Longitudinal Event Data in Type 2 Diabetic Nephropathic Study Populations: A Systematic Review
Feakins BG., Stevens RJ., McFadden E., Farmer A.
Background: Type 2 (T2) diabetes and diabetic nephropathy (DN) are risk factors for multiple vascular and cancerous outcomes. The competing nature of these outcomes mean that the use of standard survival analysis techniques can result in bias and overestimation of the risk; particularly in cases where follow-up is long or the competing outcome is common. I reviewed the literature to determine whether authors account for competing risks (CRs) in reporting time-to-event risk measures of outcomes in individuals with T2-DN. Methods: Medline, EMBASE and CINAHL databases were searched for trial and cohort studies reporting occurrences of vascular or neoplasm events in T2 diabetic nephropathic populations. Articles were excluded if median follow-up was <5yrs or <1,000 subjects with T2-DN were present at baseline. Articles published <1980 or not available in English were also excluded from the analysis. Results: From 3,886 abstracts identified, 179 studies were eligible for full-text review. Of 16 papers currently identified for data extraction: only 1 implemented a true CRs model, 14 used single outcome models or composite end-points and 1 study provided only general incidence statistics. More results will be available. Conclusions: Preliminary results suggest most studies fail to account for CRs in their analyses. The use of both composite end points and serial single-outcome models can produce bias through either assuming the same hazard for several outcomes or failing to accurately define the population at risk, respectively. Simulation work is currently underway to explore these biases and their potential effect on the risk measures generated.