This thesis assesses factors associated with a number of short and long-term outcomes in HIV-infected patients receiving antiretroviral treatment in Asia. Analyses in this thesis were based on two cohorts of HIV-infected patients; The Treat Asia HIV Observational Database (TAHOD), a multi-centre prospective observational cohort from countries in Asia-Pacific region, and the HIV Netherlands Australia Thailand (HIV-NAT) collaboration cohort, a cohort of patients treated with antiretroviral treatments at HIV-NAT in Bangkok, Thailand. We examined factors associated with time to immunological failure endpoints, such as CD4≤ 200 cells/??L, CD4≤ 100 cells/ ??L, and CD4 return to baseline, and with the virological failure endpoint, detectable viral load defined as a value greater than 500 copies/mL. Multivariate Cox proportional hazard models were used. Results showed that CD4 count at baseline and changes in CD4 strongly predicted immunological failure. For virological failure, detectable viral load at baseline was the strongest predictor. As a step to developing simplified monitoring strategies, in which patients with a low risk of failure could have their monitoring CD4 count and viral load tests deferred, we developed predictive models for each immunological and virological failure endpoint. Models were developed on the HIV-NAT cohort, and validated on the independent TAHOD cohort. For predictive models, the complementary log-log transformation for each endpoint was fitted appropriate to the interval censored nature of the data. To assess goodness-of-fit, cut-offs were defined for the predicted risks that separated patients from low risk to high risk. Overall, the observed versus expected failures from HIV-NAT data agreed quite well across all endpoints, probably reflecting that the HIV-NAT database was the data we built the models upon. Not only did these models fit the HIV-NAT database well, they also discriminated patients from low to high risk groups. When we validated models with TAHOD data, the observed and expected failures agreed well only in the model for CD4 count return to baseline. For most of the endpoints, the predictive models overestimated the number of failures, with predicted values larger than observed. However, the proportions of failures were lowest in the low risk group and highest in the high risk group, indicating that our models did discriminate between patients at high and low risk, and that the predictive models might still be of use for the purpose of simplified monitoring strategies. With CD4 count and viral load monitoring tests now comprising a large component of the cost of HIV treatment in resource limited settings, we developed and assessed a simplified monitoring strategy that aimed to reduce the numbers of monitoring tests performed. The predictive models developed earlier were used to calculate the probabilities of failure in TAHOD patients. We assumed that patients would have their CD4 and viral load assessments annually, at baseline and at one year, predicted risk of failure at ensuing clinical visits, week 12, 24 and 36. For patients at low predicted risk of failure at ensuing clinical visits, we assessed the effect of deferring monitoring tests, both in terms of blood tests avoided, and in terms of delaying detection of failure in some patients. A number of levels for the predicted risk of failure that lead to deferral of testing were evaluated. The results suggested that predicted probabilities of failure of 10% - 20% gave the best results across all failure endpoints. These cut-offs could save a median of 598 (51.6%) (range 37 (2.6%)_-1,218 (81.9%) ) blood tests over the first year of treatment, but would fail to detect 29 (18%) (range 10 (7.4%) - 128 (39.3%) ) failures. The median time from failure to detection in those patients who did fail and had deferred monitoring tests was 28 weeks. Rates of antiretroviral treatment change in TAHOD were examined. We identified patterns and factors associated with the rate of treatment change. Median time to the first treatment change was 3.2 years. Factors predicting rate of treatment change in TAHOD were treatment combination, being on second or third combination, number of drugs available in each site and being an injecting drug user. The overall rate of treatment change in TAHOD was 29 per 1OO-person-year. Around 30% of patients stopped their treatment due to adverse events. These rates of treatment change are lower than have been seen in patients in western countries. This may be due to patients in developing countries having access to fewer antiretroviral drugs than patients in developed countries.
|Srasuebkul, Preeyaporn, Public Health & Community Medicine, Faculty of Medicine, UNSW
|Publisher:University of New South Wales. Public Health & Community Medicine
|Australiasian Digital Theses Program
Page generated in 0.0023 seconds