When a new drug is introduced onto the market, patients treated with it often tend to be sicker than those who are not. Once a drug has been released onto the market, its safety and efficacy can usually only be assessed using non-experimental studies. However, these methods cannot account for the differential assignment of sicker patients to new drugs, which can make a new drug seem detrimental or even dangerous, if these systematic differences are not accounted for. This is a prototypical example of confounding by indication, which is often viewed as an insurmountable problem in pharmacoepidemiologic research. Confounding by indication consists of two components, namely channelling and the risk factor effect. / In this thesis, the index of apparent channelling (IAC) is introduced as a novel tool for the measurement of the degree of channelling. The IAC makes use of propensity scores to quantify the proportion of the variance in treatment assignment that can be accounted for by documented patient characteristics. However, the IAC can only account for channelling due to documented factors. Thus, it is always possible that there be residual channelling due to undocumented factors. Such residual channelling is of concern mostly to the extent that it leads to confounding of the treatment effect. Consequently, the index of residual channelling (IRC) is developed to estimate residual channelling and a novel approach is proposed to assess the strength of the resulting confounding bias. This model-based approach is based on estimating the interaction between treatment effect and the expected strength of residual channeling on treatment assignments of individual patients, measured by the discrepancy between their predicted treatment and the treatment they actually received. / The combination of the index of apparent channelling and the model-based approach to residual channelling provides a practical approach to the problem of assessing the impact of confounding by indication in non-experimental studies in the post-marketing evaluation of the safety and efficacy of new drugs. The application of these methods may enhance the validity of the conclusions drawn in such studies.
| Identifer | oai:union.ndltd.org:LACETR/oai:collectionscanada.gc.ca:QMM.84476 |
| Date | January 2003 |
| Creators | Bjerre, Lise M. |
| Contributors | Abrahamowicz, Michal (advisor) |
| Publisher | McGill University |
| Source Sets | Library and Archives Canada ETDs Repository / Centre d'archives des thèses électroniques de Bibliothèque et Archives Canada |
| Language | English |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Format | application/pdf |
| Coverage | Doctor of Philosophy (Department of Epidemiology and Biostatistics.) |
| Rights | All items in eScholarship@McGill are protected by copyright with all rights reserved unless otherwise indicated. |
| Relation | alephsysno: 002141345, proquestno: AAINQ98209, Theses scanned by UMI/ProQuest. |
Page generated in 0.0019 seconds