University of the Aegean
Statistical methods for the evaluation of diagnostic biomakers in the presence of censoring
Abstract
dc:descriptionBiomarkers play a significant role in decision making and are considered as the key to early diagnosis which may in turn lead to complete cure of a disease or the limitation of its progress. It is desired that the development of new biomarkers will contribute to obtaining more predictive information regarding the disease status of a patient as well as to providing a better understanding concerning the biological mechanism of various diseases. The evaluation of diagnostic markers is a necessary procedure before a medical test is used. In this thesis we will introduce new methodologies that contribute to the evaluation of a marker regarding mainly its diagnostic accuracy in the presence of censoring.Censoring is a phenomenon primarily met in survival analysis in which patients aremonitored over time until they experience an event. This event is usually death whenfatal diseases are involved. In many cases, due to practical, psychological or other rea-sons many patients decide to leave/quit the study (for example due to their relocation toanother country or due to the end of a study when some patients are event-free). Thetime at which these patients will experience the event is unknown to the clinical researchers and the only information available is that they were event-free until a given time. It might be the case that a marker’s measurements change with time and are used to predict the future disease status of a subject. For such biomarkers it is expected that measurements taken closer to the event will be more indicative of the disease or its progression so it is natural to assume that biomarker measurements are related to the time to event variable. Hence, the phenomenon of censoring naturally arises in the concept of evaluating such a biomarker. Censoring can also occur on the biomarker value itself. There are cases where due topractical reasons or technological limitations marker measurements cannot be providedbelow or beyond some known limit, usually called limit of detection (LOD) which mayvary from batch to batch. In such cases the biomarker is itself subject to left or rightcensoring respectively. In this thesis we examine settings where censoring may occur onthe time to event as well as settings where the biomarker measurements themselves aresubject to censoring.
Degree
thesis:*- Grantor dc:publisher
- University of the Aegean
- Year dc:date
- 2013
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Bantis, Leonidas
- Μπαντής, Λεωνίδας
Subjects
dc:subject × 16Rights
dc:rights- Statement dc:rights
-
- BY_NC_ND
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/36698