CD4 trajectory among HIV positive patients receiving HAART in a large East African HIV care centre Agnes N. Kiragga 1, Beverly Musick 2 Ronald Bosch, Ann Mwangi 3, Constantin T. Yiannoutsos 2 Background CD4  cell-count testing is a marker for disease progression in HIV infected patients [1]. For patients on antiretroviral therapy (ART) in resource-limited settings who do not have access to routine viral load testing, changes in the CD4+ T cells are an important method for monitoring treatment response [2]. In resource-limited settings, with the new advent of easy access to Antiretroviral Therapy (ART), it is important to derive accurate estimates of the immunological impact of ART among HIV-positive patients. Previous studies have reported immunological response among patients who are active and still in care. However, as up to 50% of the patients discontinue from care within the first two years after ART initiation, it is important to factor this in the analysis, because the CD4 counts among those still in care may not be representative of the CD4 counts among all patients who started therapy. We sought to study the CD4 trajectory accounting for the fact that some patients drop out of care and may be dead. Correspondence: Phone: Fax: References 1.Hughes MD. Evaluating surrogate endpoints. Controlled Clin Trials 2002; 23: 703– WHO. Antiretroviral therapy for HIV infection in adults and adolescents revision. Rapid advice [Accessed 15 March 2011]. 3. Lok, J. J., Bosch, R. J., Benson, C. A., Collier, A. C., Robbins, G. K., Shafer, R. W., et al. Long-term increase in CD4+ T-cell counts during combination antiretroviral therapy for HIV-1 infection. AIDS, 24(12), Conclusions  The results show a much lower median CD4 count over the first five years from ART initiation, when accounting for loss to follow-up and death.  Estimating CD4 cell count response to ART using patients in active follow-up may lead to significant overestimates and response to therapy.  This overestimation may be even higher given that patient mortality is incompletely observed. Methods which account for incomplete observation of patient mortality must be developed. weights. Subjects who dropped out were replaced by those with the same covariate profile who remained in care. Two models were used, one to predict early lost to follow-up in the first interval using baseline covariates among subjects who were on treatment and another to predict using subjects in follow up who were off treatment. A sensitivity analysis was performed considering varying fractions of dropouts who would be expected to remain on treatment after dropout from their current clinic. Results A total of 25,261 patients were included in the analysis; 16,071 (63.6%) were female, median age at ART initiation was 37 years [inter-quartile range (IQR)31-44], median CD4 cell was 112 (cells/µL) [42-181]. At follow-up, 3944 (15.6%) patients died within the first interval. Higher age, calendar year of ART initiation, higher current CD4 cell count and being on second line therapy were significantly associated with a lower probability of being lost to follow-up. In general, there were significant differences in the medium unweighted CD4 count cells with the IPCW- weighted CD4 count, as shown in the tables above. Figure 1: Non-weighted and IPCW-weighted CD4 counts by baseline CD4 count Methods Patients We included patients aged ≥15 years, receiving routine HIV care and initiating ART after 02/2004 at the Academic Model Providing Access to Healthcare (AMPATH), Kenya. Patients who initiated ART within 9 months of database closure and those who had no follow-up visits after initiating ART were excluded. We estimated the CD4 cell count trajectory represented by the median CD4 cell count after ART initiation at six-month intervals. Statistical analysis We estimated the CD4 cell count trajectory represented by the median CD4 cell count after ART initiation at six-monthly intervals. We used Inverse probability of censoring weighting (IPCW) to address the problems of selective follow-up similar to methods developed by Lok et al., (AIDS, 2010). Briefly, CD4 counts of patients who died were replaced with -1 for the subsequent intervals after death and up to the potential duration of follow-up (defined as start of ART to database closure date). We used a pooled logistic regression model to estimate the probability of remaining in follow-up. Factors included in the model were age, sex, year of ART initiation, World Health Organization (WHO) stage, CD4 count at ART initiation, PMTCT uptake prior to ART initiation, time-updated CD4 cell count, initiation of second-line therapy and being on ART during their last interval. We used distinct models to predict the weights for the first interval from baseline and for subsequent intervals and stabilized the weights by dividing numerator and denominator Figure 2: Non-weighted and IPCW-weighted CD4 counts Table 1: The non-weighted and IPCW-adjusted median CD4 cell counts over time. Interval (months after ART initiation) Median CD4 count (cells/μl) IPCW-adjusted Median CD4 count (cells/μl) ( 3- 9) ( 9-15) (15-21) (21-27) (27-33) (33-39)348214