1536-P: Parametric Survival Analysis of Complications from Type 1 Diabetes
Aim: To develop novel models of the risk of fatal and non-fatal complications for individuals with type 1 diabetes (T1D). Models used two 30 years’ follow-up studies: Diabetes Control and Complications Trial (DCCT)/Epidemiology of Diabetes Interventions and Complications (EDIC) and Epidemiology of D...
Gespeichert in:
Veröffentlicht in: | Diabetes (New York, N.Y.) N.Y.), 2019-06, Vol.68 (Supplement_1) |
---|---|
Hauptverfasser: | , , , , , , , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | Aim: To develop novel models of the risk of fatal and non-fatal complications for individuals with type 1 diabetes (T1D). Models used two 30 years’ follow-up studies: Diabetes Control and Complications Trial (DCCT)/Epidemiology of Diabetes Interventions and Complications (EDIC) and Epidemiology of Diabetes Complications (EDC).
Method: Models were developed for: Microalbuminuria, Macroalbuminuria, End-Stage Renal Disease, Impaired Glomerular Filtration Rate, Non-proliferative Diabetic Retinopathy, Proliferative Diabetic Retinopathy, Clinically Significant Macular Edema, Ketoacidosis, Hypoglycemia, Ulcers, Diabetic Peripheral Neuropathy, Amputations, Cardiovascular (CVD) Major Adverse Cardiac Event (MACE), CVD non-MACE, Death after CVD MACE, Death after CVD non-MACE, and Death with no history of CVD. Each model considered 44 risk factors, including demographics, time-varying clinical measures, and other T1D complications. Bivariate associations, followed by stepwise regression, were used for variable selection. Interval-censored Weibull and Gompertz survival models were chosen for non-fatal and fatal complications, respectively. Final models combined optimal statistical prediction and expert knowledge for model interpretability.
Results: All models included age, gender, diabetes duration, and HbA1C, with HbA1c significantly increasing the risk of most complications. Across all models, the highest predictor of a complication was history of another complication(s). Blood pressure (e.g., systolic, diastolic), pulse rate, and lipid variables (e.g., triglycerides, LDL, HDL, total cholesterol) were significantly predictive of most complications. Cumulative hazard plots comparing predicted and actual hazards indicated good accuracy of prediction for all models.
Conclusion: Our risk equations highlight the importance of accounting for the time-varying nature of many risk factors and the impact that history of diabetes complications has on a different complication. |
---|---|
ISSN: | 0012-1797 1939-327X |
DOI: | 10.2337/db19-1536-P |