1. audio file - high quality (mp3 format, 47.1 MB)
2. audio file - low quality (mp3 format, 23.6 MB)

Dr Hosseinipour presented an important study from Malawi which examined the ability of clinical and immunological definitions of ART failure that are in operation in Malawi to detect virologic failure. 152 cases of failure (75% immunological, 21% clinical, 4% both) were detected. Of these 59% were confirmed virologically (VL>400 copies). Multivariate predictors of failure were being on ART more than 3y, having a CD4 count below 200 (both of which were positive predictors of failure) and having KS (negative predictor). The conclusion from this study is that really need VL to confirm failure, especially when failure is based on clinical events because there is a substantial risk of misclassification.

Apollo Basenero presented a study with a similar aim to the first presentation. This study was done in a clinic in Infectious Diseases Institute at Mekerere University in Uganda. In this setting CD4 counts are done 6 monthly for assessment. The objective was to validate the utility of a clinician based consensus meeting to predict failure. Patients were classified into 3 categories: clinical and immunological failure with good adherence (cat 1); the same with poor adherence (cat 2); and an inconclusive group who had immunologic failure with clinical stability (cat 3). 100 patients were evaluated. All those classified in category 1 had a VL>400. Of those in category 2 who did not have a CD4 rise after adherence counselling (38%), all had VL >400. Of those in inconclusive category, 56% had VL >400. In conclusion, viral load is needed to accurately identify people with failure.

In a study with a similar aim, but with slightly different design, Dr Anh evaluated the accuracy of the WHO 2003 criteria for failure in a clinic population in Cambodia.  They did a cross sectional study of 399 patients attending a single center in Cambodia and evaluated CD4 failure, clinical failure and also measured a number of other potential predictors including change in total lymphocyte count. In contrast to the previous 2 studies, VL was measured in all patients so sensitive and specificity could be assessed. VL failure was defined as VL >50 copies. The sensitivity of both the clinical and immunologic criteria was low (18% in both cases) and only 30% when combined. Specificity for predicting failure was much higher.    

In a departure from the previous presentations, Dr Couto-Fernandez, described a national laboratory network for HIV-1 resistance measurement in Brazil (RENAGENO), and presented some data on 2309 samples in Rio. Subtype B was the most prevalent subtype, followed by F and several recombinant forms and miscellaneous others. The prevalence of resistance mutations reflected, as expected, the pattern of drugs in use in Brazil over the period of the study. 

Ronald Cantrell presented data from Lusaka, Zambia assessing the utility of two measures of adherence in predicting virologic failure, the Medication Possession Ratio (MPR) based on delays in pharmacy refill visits, and self reported adherence in the previous 3 days. The study population was 753 patients who had either WHO defined clinical or immunologic failure but not both, and who therefore went on to have a VL as part of the local management algorithm. Virologic failure was defined as >400 copies. There was a strong relationship between virologic failure and the MRP but not with self-reported adherence, suggesting that the former measure may be a better indicator of true levels of adherence in this population. Although interesting, this was a selected group of patients and the extent to which this can be generalized to other settings is uncertain.