Meeting Brochure and registration form      SMDM Homepage

Wednesday, October 24, 2007
P4-41

DIABETIC RETINOPATHY PROGRESSION IN TYPE 2 DIABETICS: A MULTI-PARAMETER EVIDENCE SYNTHESIS

Ba Pham, MSc1, Stephanie A. Duench, PhD, (Candidate)2, John Gonder, MD3, Irene Hramiak, MD4, Maggie Hong Chen, MMath5, Brenda Lee, MSc6, Minh Do, PhD, (Candidate)5, Andrea Anonychuk, MSc7, and George Tomlinson, PhD5. (1) THETA - Toronto Health Economics & Technology Assessment Centre, Toronto, ON, Canada, (2) University of Waterloo, Waterloo, ON, Canada, (3) University of Western Ontario, London, ON, Canada, (4) Lawson Health Research Institute, University of Western Ontario, London, ON, Canada, (5) University of Toronto, Toronto, ON, Canada, (6) I3 Innovus, Burlington, ON, Canada, (7) University Health Network, Toronto, ON, Canada

Purpose: Diabetic retinopathy (DR) is a common complication in type 2 diabetics (T2Ds). Current guideline recommendations regarding early diagnosis and prevention of DR are based upon limited data. Continuing reliance on DR progression data from large but out-dated studies does not reflect current management of T2Ds. We conducted a multi-parameter evidence synthesis (MPES) to quantify and predict the progression of DR.

Methods: Systematic review: A validated literature search was used: MEDLINE (1996 – Dec. 2006) and EMBASE (1980 – Dec. 2006) were searched to identify studies of 1) T2Ds, 2) follow-up duration > 1 year, and 3) reporting DR incidence data. Broad-screening and full-text reviewing were conducted independently by two reviewers. Baseline characteristics were extracted for each reported cohort. Incidence data was first abstracted according to the DR classification system used in the study reports (e.g., ETDRS, AAO) and subsequently reconciled to a 4-state classification using their clinical features.

MPES: A 4-state Markov-chain tunnel model (S0: No DR, S1: Background, S2:Pre-Proliferative and S3:Proliferative) was used. Observed state-specific incidence data were re-expressed as functions of transition rates and follow-up durations. Bayesian progression and prediction estimates were derived according to Welton et al. SMDM 2005 and using Bayesian simulations in WinBugs.

Results: A total of 39 studies published in 1980-2006 (n=22,387 participants; males: 51%; mean age: 54 years; mean T2D-duration: 8 years; mean HbA1C: 9.1%, mean DBP/SBP 80/136; and mean follow-up duration: 5.4 years) reporting data on 59 cohorts (North American n=9, NA Indians n=6, Europe n=28, and Asia n=16) were included. Data was available for high baseline A1C >= 9% [low A1C]: 18 cohorts (n=6,515 T2Ds) [18 cohorts (n=5,612)].

Annual transition probability estimates (%) were S0:S1: 7.5 (95% credible interval: 6.4-8.7); S1:S2: 9.6 (7.9-11.7); and S2:S3: 12.4 (9.4-16.2). The corresponding estimates for high baseline A1C [low A1C] were 7.4 (5.6-9.7) [7.1 (5.3-9.4)]; 12.9 (7.3-21.6) [6.0 (4.2-8.5)]; and 12.5 (7.4-20.6) [6.4 (4.3-9.5)]. Progression estimate from No to Proliferative DR after 10 years [20] was 8.2 (6.2-10.6) [32.3 (26.3-38.7)] and modified by baseline A1C.

Conclusion: Lack of glycemic control is a disease modifying risk for DR in T2Ds. The consolidated data supports early diagnosis of background disease and timely intervention to prevent further disease progression. Both uncertainty propagation and data synthesis was simultaneous in this MPES framework.


See more of Poster Session IV
See more of The 29th Annual Meeting of the Society for Medical Decision Making (October 20-24, 2007)