By the year 2010, the Joint United Nations Program on HIV/AIDS (UNAIDS) estimate that there are 1,138,020 people living with HIV/AIDS in Thailand [1]. The mathematic model describing the epidemic trends using the Asia Epidemic Model software projected that there will be 10,835 new HIV cases each year [1]. In Thailand, it was only after the establishment of the National Access to Antiretroviral Program for People living with HIV/AIDS (NAPHA) in 2002 that combination antiretroviral therapy (cART) became widely available free of charge throughout the country [2]. In a previous study from Thailand, treatment with GPO-VIR^® (a locally-produced generic fixed-dose combination of stavudine, lamivudine, and nevirapine) resulted in 62.7% and 93.3% of 90 HIV treatment-naïve patients achieving undetectable HIV viral load at 24 and 48 weeks, respectively [3]. Despite the significant reduction in morbidity and mortality among HIV-infected patients receiving cART, a considerable number of patients fail to achieve a sustained virological and immunological response to therapy [4–7]. The incidence and predictors of treatment failure have generally been analyzed worldwide, [8–10] however, these results may not be applicable to more heterogeneous cohorts such as in Thailand, where non-nucleoside reverse transcriptase inhibitor (NNRTI)-based regimen is the first line of treatment, the majority of HIV-1 virus is subtype E (CRF01_AE), and cART is usually initiated at a very low CD4 count. The data regarding treatment failure from Thailand are still limited. Furthermore, the number of long-term cART treatment cohorts is scarce. Hence, our primary objective was to study the incidence and risk factors for treatment failure in a large cohort in northern Thailand. Such data would assist physicians in Thailand as well as in similar resource limited settings predicting and detecting treatment failure earlier in the course of cART.

This is a large cohort of HIV-infected patients with middle or low socioeconomic status from a tertiary care hospital in northern part of Thailand. During the study period of 6 years and a median follow-up duration of 3.63 years (IQR1.66, 5.19), the cumulative incidence of treatment failure was 12.3% and 9.6% by intention-to-treat and on treatment analyses, respectively. This result is similar to failure rates found in a previous study by Anekthanonon et al from Bangkok, Thailand (2.4% by on treatment and 19.8% by intention- to -treat analysis)[11] and in the same range of 3-27% from many cohorts [10, 12, 13]. The wide range of virological failure from many cohorts could be attributable to differences in definition of treatment and virological failure between studies as well as the length of follow-up. However, in this study in which majority of patients were infected with HIV-1 subtype CRF01_AE, the immunological and virological outcomes are comparable to those from many US and European cohorts in which majority of patients were infected with HIV-1 subtype B, and those from African cohorts in which majority of patients were infected with HIV-1 subtype C [10, 14–16]. The incidence rate of virological failure in our study was 2.79 cases per 100 person-years which is relatively low compared to the range of 3.36 cases per 100 person-year to 13.6 cases per 100 person-years in other cohorts [9, 17]. This high response rate may be from the very high adherence rate in our cohort.

We found that adherence of < 95% and age < 40 years were the independent predicting factors for virological failure. Although adherence has been demonstrated to be important in many clinical cohorts, [18–21]there is discordance of results depending on types of adherence measurement. In our study, the data on adherence were collected according to the patient's self report of the amount of medication taken between visits without standardized set of questions. Therefore, the result may not be completely reliable. However, it is still demonstrated to be an independent factor for treatment failure. The finding that younger age at cART initiation was the risk factor for virological failure is similar to other studies [14, 22]. Increased age may be a marker for greater maturity, lifestyle stability, and disease-specific education; these factors are likely to affect long-term adherence to therapy [14, 23]. However, younger age and poor adherence were both independent risk factors for virological failure.

The SWISS HIV Cohort Study showed that only 39% of individuals treated with cART reached CD4 T-lymphocyte count of 500 cells/mm³[15]. Many HIV cohorts demonstrated factors potentially affect the recovery of CD4 T- lymphocyte including age, HCV serology, duration of HIV-1 infection, CDC stage, baseline HIV-1 RNA levels, baseline CD4 count, and adherence of cART [8–10, 13, 24]. Our study found that there was significant lower proportion of patients achieving CD4 cell count of ≥ 350 cell/mm³ after 2 years of cART between those with baseline CD4 cell count ≤ 100 cell/mm³ and those with higher baseline CD4 count (59.0% vs. 79.8%). It has been hypothesized that patients who initiated cART late in the course of disease may have thymic damage and limitation of T cell production and proliferation [25, 26]. We could also demonstrate that the increment of the CD4 cell count of < 50 cells/mm³ after the first 6 month of cART may predict immunological failure. Failure to achieve early CD4 recovery could be another signal of thymic damage. Our study did not find the difference in CD4 recovery between sexes, which was similar to other reports [15, 16, 25, 27]. However, some studies reported better immunological response among female patients [22, 28]. Whether HCV co-infection adversely affects HIV treatment outcomes remains controversial [16, 29, 30]. In this study, HCV seropositivity was not associated with higher risk of immunological failure. Due to the high adherence rate among patients evaluable for immunological failure (98.7% had adherence ≥ 95%), we could not demonstrate the association between poor adherence and immunological failure.

Our study has several limitations. First, HIV-1 RNA was not monitored at baseline in all patients so we couldn't evaluate the higher baseline HIV-1 RNA viral load as a predictor for virological failure. Second, the HIV-1 RNA monitoring was not regularly performed in all patients, therefore, we could not calculate the accurate time of treatment failure. It is also possible that some patients may already have virological failure at the time we evaluated the immunological criteria and thus misinterpret the true CD4 cell count response. Third, assessment and documentation of patient's adherence to cART was not standardized. However, this reflected the usual clinical practice and adherence was still the strongest predictor for virological failure. Fourth, this is a retrospective cohort study that the accuracy of analysis depends on the completeness of data entry into the database. Finally, this study is limited to a single cohort, the findings should be confirmed in other clinical settings.

In summary, this analysis of a cohort of HIV-infected patients under usual clinical care showed that predictors of immunological and virological failure can be identified using data routinely recorded. The risk factors of virological failure were poor adherence of cART and younger age, whereas those of immunological failure were low baseline CD4 count and slow CD4 cell recovery. This study demonstrated that even in resource-limited settings, the high rate of success could be expected in the cohort with good and sustainable drug adherence.