Precision of United Kingdom Diabetes Prospective Study Model in Prediction of Risk of Coronary Artery Disease in Type 2 Diabetes Patients from Republic of Macedonia

Smokovski, Ivica (2014) Precision of United Kingdom Diabetes Prospective Study Model in Prediction of Risk of Coronary Artery Disease in Type 2 Diabetes Patients from Republic of Macedonia. PhD thesis, UKIM, Medical Faculty Skopje.

[img]
Preview
Text
Doctor of medical sciences thesis.pdf

Download (389Kb) | Preview

Abstract

Aim: Primary aim of this study was to evaluate the precision of UKPDS model in predicting the CAD risk in type 2 DM patients by quantifying its discrimination and calibration. Material and methods: Observational study with a cohort of 1,404 type 2 DM patients from Republic of Macedonia prescribed insulin treatment in the period from September 2002 till January 2004. UKPDS model uses nine risk factors for CAD risk assessment in type 2 DM patients: age at diabetes diagnosis, duration of diabetes, gender, race, smoking, systolic blood pressure, HbA1c, total cholesterol and HDL cholesterol. Endpoints defined in this study were cases with CAD by 01 January 2012. Discrimination was examined by using the с-statistics (aROC), and calibration by using the Hosmer–Lemeshow χ2 (HLχ2) goodness-of-fit test. Prognostic value was evaluated by the graphic presentation of deciles of predicted and the corresponding observed risk, and by the visual inspection of the calibration plots. Results: Out of the study cohort, 835 patients (59.5% of the cohort) were identified for analysis followed for a mean of 9.4 ± 0.5 years. Observed risk was 36.4% (95% CI 33.2-39.8) compared to the predicted risk for CAD by UKPDS model of 20.8%, i.e. UKPDS model underestimated the observed risk by 43%. Discrimination evaluated by c-statistics was 0.59 (95% CI 0.55-0.63), p<0.001, and the discrimination power of UKPDS model in CAD risk prediction was determined as low (0.5-0.7). Calibration evaluated with the use of HLχ2 test was 5.74 (p=0.68), resulting in generally good calibration of UKPDS model in CAD risk assessment. Graphic presentation of deciles of predicted and corresponding observed risk, and the calibration plots, demonstrated larger deviation between observed and predicted risk in deciles of lower predicted risk, and increased level of concordance in deciles of higher predicted risks. Discussion and conclusion: In our study, UKPDS model demonstrated higher predictive ability in patients with higher predicted risk compared to patients with lower predicted risk. Recalibration of UKPDS model is needed to further improve its predictive ability in diabetic population from Republic of Macedonia, in order to take into consideration the population characteristics, such as very high cardiovascular mortality risk.

Item Type: Thesis (PhD)
Subjects: Medical and Health Sciences > Clinical medicine
Divisions: Faculty of Medical Science
Depositing User: Ivica Smokovski
Date Deposited: 21 Dec 2015 13:09
Last Modified: 21 Dec 2015 13:09
URI: http://eprints.ugd.edu.mk/id/eprint/14635

Actions (login required)

View Item View Item