Patient's disposition is shown in [Fig ]. Of the 434 patients selected to receive ART between June and September 2006, 405 (93%) attended the baseline visit and 312 (72%) were included in the study. Reasons for non participation of eligible patients were imminent transfer to another treatment centre (n = 20), first antiretroviral prescription by a physician not participating in the study (n = 15), a physical health status rendering long interviews difficult for the patient (n = 11), refusal to take part (n = 8) and other reasons (n = 39).
Among the 312 enrolled patients at the start of ART, the mean age (± SD) was 37 ± 9 years, with a majority of women (63%) [Table ]. Although 97 participants (31%) reported no income-generating occupation, 76 (78%) of these received occasional or regular financial help from relatives. Thirty-two patients (10%) benefited from treatment support from the Global Fund, which was associated with neither socio-economic status nor with income levels (data not shown). Forty-six per cent of the cohort had the lowest financial income (monthly income of less than USD50). About two-thirds of the patients (70%) lived close to the clinic, with only 15 (5%) living more than 4 hours from the Day Hospital by public transport.
Socio-demographic and clinical data at start of ART of the 312 patients studied
Most patients (n = 195 or 63%) had been followed at the clinic for less than one month before initiating ART. However, only 36 patients (12%) reported the use of alternative therapies for HIV in the past. At the time of starting ART the majority of the cohort were classified as symptomatic (64% at CDC stage B and 14% at stage C), the median baseline CD4 count was 104 cells/mm3 (interquartile range 50 - 177 cells/mm3). All patients received two nucleoside reverse transcriptase inhibitors (NRTIs) and one non-nucleoside reverse transcriptase inhibitor (NNRTI) with the exception of one patient who received two NRTIs and one protease inhibitor.
When asked about social support and stigmatisation, 275 participants (88%) reported having disclosed their HIV status to at least one person and 260 (83%) declared satisfied by family or friends support. Most patients (n = 258 or 80%) had optimistic expectations of the effects of ART on their health status, a sizable minority believed that treatment could cure HIV (44%) and 118 patients (38%) were unaware of possible ART side effects.
Treatment outcomes and adherence measures
Of 312 patients in the initial cohort, 219 (70%) were still coming to the pharmacy to refill their medication after 6 months of follow-up, while 51 (17%) were lost, 28 (9%) had died and 14 (4%) were had been transferred to other care centres. The incidence rate of loss to follow-up was 40.1 per 100 persons years (95% CI: 30.5 - 52.8) and mortality rate was 21.2 per 100 persons years (95% CI: 15 - 31). Drop-out occurred at a median of 44 days (IQR: 28 - 80 days) and death occurred at a median of 55 days (IQR: 28 - 81 days) after treatment initiation.
The first routinely scheduled follow-up CD4 cell count was performed by only 71 patients (23% of patients still followed-up). The median rise in CD4 cell count from baseline to this point was 117 cells/mm3 and 16 patients (22%) with available follow-up CD4 cell count met criteria for immunological treatment failure. Using phone call tracing alongside tracking of patients during regular follow-up visits, we obtained viral load values for 206 patients. Virological treatment failure (viral load value of > 400 copies/ml) at 6 months was experienced by 26 patients, representing 13% of patients with available viral load value and 8% of the entire cohort.
Self-reported adherence data after one month of therapy were obtained for 238 patients (76%). The majority of respondents (78%) claimed not to have missed a single dose during this period. The proportion of participants that reported full adherence to treatment during the past month decreased from 83% at one month to 57% at 6 months.
Adherence based on pharmacy refill charts after 6 months of follow-up could be analysed for 278 participants. Twenty-eight subjects were excluded because of their short follow-up duration (<2 months) and 6 patients had incomplete data on the refill chart [Fig ]. Two-thirds (64%) of patients assessed came regularly every month within two weeks of the pharmacy-appointed dates and qualified as adherent in the analysis. Non-adherence (interruptions or discontinuations) was defined for 23% of the participants with available data: 49 subjects interrupted their treatment for at least 3 weeks during the follow-up period; 14 patients lost to follow-up could be traced by phone calls and confirmed they had discontinued ART. Thirty-seven patients (14%) could not be traced by phone call after loss from pharmacy follow-up.
Adherence measurements and virological treatment failure
There were significant differences between the ability of different measures of adherence to predict virological treatment failure after 6 months of therapy [Table ]. Pharmacy-refill irregularity was the most powerful predictor (OR, 12.40; 95% CI, 4.75-32.40; P < 0.001). In the sub-sample of 194 patients whose 6-month viral load and pharmacy refill charts were available, 38% of the pharmacy non-adherent patients demonstrated virological treatment failure while 95% of pharmacy adherent patients presented with virological suppression. Immunological treatment failure was also significantly associated with virological treatment failure (OR, 7.78; 95% CI, 1.68-36.0; P = 0.002). However, pharmacy adherence estimated by pharmacy refill charts had greater accuracy for detecting virological treatment failure than CD4 count changes at 6-months: the sensitivity was higher (72% versus 53%) with approximately the same specificity (82% versus 88%).
Potential predictors associated with virological treatment failure (>400 HIV RNA cop/ml) after 6 months of initiating ART
Notably, self-reported adherence at 1 month was not significantly associated with virological treatment outcomes.
Associations between patient characteristics, retention and mortality
Univariate analysis of individual patient factors [Table ] showed that the most significant factor associated with survival was the improvement of early ART side-effects, (OR: 0.11; 95% CI: 0.02-0.52). Patients starting treatment with CD4 cell count below 100 cells/mm3 were at significantly greater risk of death during the follow-up period (OR: 2.69; 95% CI: 1.12-6.44). We found only a trend for association of HIV CDC stage with mortality. However, HIV CDC clinical stage at the beginning of treatment significantly predicted (P < 0.001) loss to follow-up: compared with asymptomatic patients CDC stage A, CDC stage B patients (OR: 5.72; 95% CI: 1.33-24.70) and specially CDC stage C patients (OR: 16.90; 95% CI: 3.58-80.30) had greater rates of loss to follow-up.
Logistic regression of patient characteristics associated with treatment outcomes in the first 6 months after initiating ART
There was a non-significant trend to lower rates of interruption of follow-up for women (OR: 0.58; 95% CI: 0.31-1.08) and for those reporting improving ART side-effects after one month of treatment (OR: 0.49; 95% CI: 0.19-1.26). None of the socio-economic determinants studied appeared to influence patient retention and mortality: neither age, nor level of education, nor economic situation and social support were significantly associated with loss to follow-up or with death.
Associations between patient characteristics and pharmacy adherence
As patients lost to follow-up may have introduced bias into the analysis, we analysed the association of individual patient characteristics with pharmacy adherence in a sensitivity analysis, using two scenarios, a best-case and a worst-case scenario [Table ]. In the best-case scenario, where all participants lost to follow-up were analysed as adherent, greater age (OR: 0.38; 95% CI: 0.11-1.33), female sex (OR: 0.66; 95% CI: 0.37-1.18) and patients reporting improvement of early ART side-effects (OR: 0.51; 95% CI: 0.23-1.10) tended keep pharmacy appointments less irregularly (P < 0.2). In this scenario, a monthly middle income was significantly associated (P = 0.01) with greater pharmacy adherence. Low (OR: 2.49; 95% CI: 1.03-6.01) or high (OR: 3.76; 95% CI: 9.58) incomes groups showed a higher risk for pharmacy non-adherence. In the worst-case scenario, female sex (OR: 0.60; 95% CI: 0.35-1.01) and improving early ART side-effects (OR: 0.47; 95% CI: 0.24-0.95) were characteristics associated (P < 0.05) with lower pharmacy non-adherence. CDC stage B patients (OR: 2.68; 95% CI: 1.26-5.68) and specially CDC stage C patients (OR: 5.13; 95% CI: 2.02-13.00) had higher risk of pharmacy non-adherence than asymptomatic patients.
Patient characteristics associated with pharmacy non-adherence during the first 6 months after initiating ART
In the multivariate analysis [Table ], women presented lower risk for non-adherence both in the best case (adjusted OR: 0.56; 95% CI: 0.29-1.07; P = 0.08) and in the worst case scenarios (adjusted OR: 0.56; 95% CI: 0.31-1.02; P = 0.05). Economic status, in particular patients with the highest monthly income when compared with monthly middle income (OR: 3.24; 95% CI: 1.24-8.46; P = 0.04), was retained as a predictor of poor adherence only in the best case scenario. When compared with asymptomatic patients, the multivariate analysis confirmed a marked risk of non-adherence for CDC stage B patients (OR: 2.75; 95% CI: 1.23-6.18) and CDC stage C patients (OR: 5.07; 95% CI: 1.87-13.80) in the worst-case scenario (P = 0.003).
Multivariate logistic regression of baseline characteristics associated (P < 0.2) with pharmacy non-adherence in the univariate analysis