Abstract #93  -  Factors associated with non-adherence to long-term HAART: a 10-year follow-up analysis with correction for the bias induced by missing data
Authors:   Presenting Author:   Mrs Marie Préau - University of Nantes
Additional Authors:  Mrs Camelia Protopopescu, Mrs Marie Préau, Pr François Raffi, Mrs Perrine Roux, Mrs Fidéline Collin, Pr Jacques Reynes, Pr Pierre Dellamonica, Pr Catherine Leport, Mrs Maria-Patrizia Carrieri, Pr Jean-Paul Moatti, Dr Bruno Spire,
Aim: The aim of this study was to estimate the effect of treatment-related and psychosocial factors influencing non-adherence over a 10-year follow-up of the APROCO-COPILOTE ANRS CO8 cohort during the maintenance phase of highly active antiretroviral therapy (HAART), while taking into account the bias induced by the presence of missing adherence data.
Method / Issue: APROCO-COPILOTE ANRS CO8 is a cohort of patients started on a PI-containing regimen between 1997 and 1999. Overall, 1010 patients participated in this analysis, each having had at least 12 months of follow-up after HAART initiation and at least one self-reported adherence measure available during the follow-up period (months M12-M120). Data collection was based on clinical records and self-administered questionnaires which gathered details of patients psychosocial characteristics and experience with HIV disease and treatment. Patients were classified as non adherent if they reported taking less than 100% of prescribed medications during the previous four weeks. First, a Generalized Estimating Equations probit model was used to identify predictors of non-adherence. Second, a selection model, based on a Heckman two-stage approach, was used in order to account for missing data bias and to measure the extent to which this could affect the results of the first model.
Results / Comments: During the study period 263 patients (26%) reported being always highly adherent. Non-adherent behavior was reported by 747 patients, corresponding to 2070 (35%) of the 5956 visits with valid adherence data. After correcting for the bias due to missing data, non-adherence was independently associated with the number of self-reported side-effects, a t.i.d. or higher dosing regimen, HIV clinical stage B or C and HIV diagnosis before the HAART era. Non-adherence was more likely among patients who were younger, had children, were born in the EU, had depressive symptoms, consumed alcohol daily and declared a lack of support from their main partner. In our study, adjusting for missing outcome data using the Heckman approach clearly showed that ignoring missing data can significantly change the pattern of predictors of non adherence and introduce a significant bias in the estimates, the effect of which would have been to underestimate the probability of non adherence. Certain factors, such as being born in the EU, having children, clinical stage B or C and HIV-diagnosis before the HAART era, were significant only when the bias due to missing data was taken into account.
Discussion: Our findings in this study underline the importance of using adapted specific statistical methods to the type of data analyzed. We also found that non-adherence is associated with both psycho-social conditions and treatment-related characteristics. To improve long term patient outcomes for those at risk of adherence failure, tailor made sychosocial interventions and regimen-based strategies with better toleration need to be improved.
Go Back