İzlem Süresi İçerisinde Belirteçlerin Performanslarını Değerlendirmede Zamana Bağlı Roc Eğrisinin Kullanımı
xmlui.mirage2.itemSummaryView.MetaDataShow full item record
The ROC (Receiver Operating Characteristic) method commonly used to evaluate the performance of biological markers which often used to help diagnosis and treatment. The performance of the marker should be evaluated with time dependent ROC curve analysis that predicts the subjects with and without the event during the time interval [0, t]. The first aim of this study is to combine several biomarkers from real data set, which are used to predict the event using time dependent ROC curves, through Cox proportional hazard regression method and compare the performance of the combined biomarker to individual biomarkers within follow-up time. Another aim is to investigate optimal cut-off points of composite markers for changes over time. Lastly, identifying covariates, if exists, that may influence the performance of markers and make adjustments for time dependent ROC curve where necessary is another objective. In application, the performances of Troponin T, Myoglobin, BNP and CK-MB biomarkers for estimating death by any cardiological reason were evaluated. The follow-up time of 410 patients was taken as 240 hours after admission to the hospital. Myoglobin and BNP markers were found to be statistically significant with Cox proportional hazard regression. A composite marker was formed by combining these biomarkers by the same method. It was shown that, composite marker showed a higher performance than individual biomarkers within approximately first 180 hours of follow-up time. Our results indicated that, optimal cut-off points of composite marker which were used to discriminate between the subjects with and without the event changed during follow up time therefore the risk status in the follow-up time of individuals should be determined according to these cut-off points. On the other hand, there were no individual characteristics that might affect the performance of the biomarkers. As a result, it is shown that ROC curves, which are usually performed for diseases observed from long-term follow-up studies in literature, can also be used for diseases which might be observed in short term follow-up studies. This method may play a role in the early diagnosis of high-risk patients and the making a decision for the time of discharge from hospital for patients with less risk patients.