Analysis for patient data data in a European HIV database for indications of treatment success and adherence

Since the discovery of human immunodeficiency virus (HIV) and acquired immune deficiency syndrome (AIDS) in the early eighties, much research has been carried out into all aspects of the virus and its life cycle. A number of HIV databases have been collected and collated for research purposes, which tend to contain data collected in small numbers from a large number of sources. A large proportion of these are research strains. There are a number of clinical databases, dotted around the world, containing relatively small numbers of patients collected from a restricted geographical area. In comparison to this the EuResist project database contains patient data collected from three geographically distinct regions, with different migration patterns, over an extensive period. The EuResist project has formed one of the largest clinical patient databases in the world, with a view to developing a tool for predicting treatment success or failure based on the mutational status of the patient's HIV population. This work interrogates the data to validate the EuResist integrated database and to examine a number of aspects of the data contained within. A hypothesis has been developed and a preliminary analysis carried out to identify a novel methodology for identifying periods of patient non-adherence to treatment. A logistic analysis has been carried out to examine the relative effect of each drug within a treatment regime on the treatment outcome. The effect of a number of other factors on both adherence and treatment outcome are also examined. These analyses showed that there are differences between the three original datasets, which need to be taken account of in the work of the EuResist project. We showed that both adherence and treatment outcome are affected by Patient risk group, but that this relates to specific risk groups, with mother-to-child transmission patients being less likely to have a successful treatment outcome and having lower levels of adherence. We showed that, with the exception of Sequinavir, the use of Ritanovir to 'boost' the effect of protease inhibitors (PIS) did not make any significant difference to treatment outcome within this dataset. We also showed that, within our cohort of patients taking the World Health Organization recommended PIS, Nelfinavir had a significantly lower likelihood of treatment success than patients on other drugs. The phylogenetic analysis offered support for the principal that patients who are geographically-isolated would develop similar mutations in response to the same treatment regimens. The distribution of patient risk groups and viral subtypes amongst the phylograms agrees with the generally known theory that the HIV infections across European is formed by two epidemics, one of subtype B virus infecting homo/bisexual (Hom/Bi) patients and a second epidemic of non-B subtypes spreading amongst heterosexual (Het) patients.