University of California, San Francisco Logo

University of California, San Francisco | About UCSF | Search UCSF | UCSF Medical Center

Home > Global Health Literature Digest > Predictors of Failure
Predictors of failure of first-line antiretroviral therapy in HIV-infected adults: Indian experience
Global Health Sciences Literature Digest
Published October 15, 2007
Journal Article

Rajasekaran S, Jeyaseelan L, Vijila S, Gomathi C, Raja K. Predictors of failure of first-line antiretroviral therapy in HIV-infected adults: Indian experience. AIDS 2007 Jul;21 Suppl 4:S47-53.


To determine incidence and risk factors for treatment failure with first-line antiretroviral therapy among HIV-infected adults attending the largest public HIV care center in India, and to estimate the number who will need a second-line treatment regimen in the next 3.5 years.

Study Design

This was a retrospective cohort study abstracting medical records of HIV-infected adult patients enrolled in a government-sponsored antiretroviral therapy program.


A country-wide program providing free HIV care and ART to eligible patients accessing care in Tambaram, Chennai at the Government Hospital of Thoracic Medicine (GHTM), the largest public HIV care center in India. Tambaram is a suburb situated 27 km south of Chennai in the state of Tamil Nadu. It is considered a hub to all major destinations in south Tamil Nadu.


Among the 4,355 HIV-infected adult patients enrolled in GHTM from 1 April 2004 to 31 December 2006, 1,370 fit the eligibility criteria of the study: adult treatment-na�ve with a minimum of three CD4 T cell measures obtained within a minimum 12 (+/-1) month follow-up. Twenty-two percent of eligible patients were 15-29 years old and 69.4% were 30-44 years old. Fifty-four point two percent were male and 60.9% of the patients were from rural areas. CD4 counts were less than 50 cells/µl in 27.7% of the patients, between 50-99 cells/µl in 24.4%, between 100-199 cells/µl in 32.8%, and less than 200 cells/µl in 15.1%. Absolute lymphocyte counts were less than 1000/mm3 and greater than 1600/mm3 in 45.4% and 13% of patients, respectively. Body weight was less than 40kg in 26.6% of the patients and the hemoglobin concentration was less than 9g/dl in 23.4%. Previously, 60% of the patients had received anti-tuberculosis treatment.


There was no intervention. Patient data during a period of 33 months were obtained retrospectively from several sources, including the GHTM electronic medical record system, patient-level information on ART, socio-demographic data, and laboratory results, and linked through an anonymous patient identification number. Those enrolled in the ART program receive free treatment and clinical check-up during monthly visits to GHTM. The approved ART regimens were zidovudine (AZT) or stavudine (D4T), lamivudine (3TC), and nevirapine (NVP) or efavirenz (EFV). Drug adherence was assessed using the pill-count method. Body weight was measured during every monthly visit, whereas hemoglobin concentration and absolute lymphocyte count were measured every month during the first year and once every three months subsequently.

To estimate the number of patients who will need second-line treatment regimen in the next three years, EPP2005 and Spectrum was utilized. As the incidence of treatment failure was expected to increase over time, the progression for the months 30, 36, 42 and 48 was estimated using exponential growth models, based on the cumulative probability of treatment failure at 12, 18, 24 months.(2)

Primary Outcomes

The primary outcome was treatment failure, assessed using the WHO definition of either a decrease in CD4 count to or below the baseline before treatment, or a 50% decrease from the on-treatment peak value (if known), or CD4 concentrations persistently less than 100 cells/µl.(1) Secondary measurements were negative changes in absolute lymphocyte count, hemoglobin concentrations, and body weight.


A total of 54 patients experienced treatment failure between 1 April 2004 and 31 December 2006. The cumulative probabilities of treatment failure at 12, 18, 24, and 30 months were 1.9%, 4.9%, 6.8%, and 12.4%, respectively. Male patients had a 3.5 (95% CI 1.6 -7.4) times greater multivariate adjusted hazard ratio for treatment failure compared with female patients (p<0.001).

Negative changes in absolute lymphocyte count, hemoglobin concentration, and body weight occurred in 11.7%, 15.4%, and 16% of the patients, respectively. The patients who exhibited a negative change in absolute lymphocyte count during follow-up had 3.2 (1.6-6.2) times greater hazard ratios than those in whom the change was positive over time (p<0.001). Patients exhibiting a negative change in hemoglobin concentration had 3.1 (1.6-6.2) times greater hazard ratio values for treatment failure compared with those in whom there was a positive change (p<0.001). Similarly, the patients who had a negative change in body weight over time had 3.5 (1.9-6.4) times higher hazard ratio for treatment failure compared with those with a positive change in body weight (p<0.001). The patients who had previous history of anti-tuberculosis treatment had 1.6 (0.9-3.0) times greater hazard ratios for treatment failure compared with others (p=0.09). Patients from urban areas had 1.9 (1.1-3.2) greater hazard ratios for treatment failure compared with those from rural areas (p=0.021).

During follow-up, 92.7% of the patients had drug adherence greater than 95%. The percentage of treatment failure was 3.2% in patients receiving stavudine, lamivudine and nevirapine, 7.7% in those receiving stavudine, lamivudine and efavirenz, 4.2% in those given zidovudine, lamivudine and nevirapine, and 15.8% in those given zidovudine, lamivudine and efavirenz. The percentage of treatment failure was significantly greater (p<0.05) in the regimens that included efavirenz, which were mostly given to patients receiving rifampicin as part of their treatment for concomitant tuberculosis. Mortality increased significantly among patients who had treatment failure (26.1%) compared with patients among whom treatment did not fail (3.6%).

According to NACO, a projected 300,000 adult patients with HIV will be receiving ART by 2011. It was planned that 100,000 adult patients should be receiving ART by 2007, with 50,000 adult patients enrolled for ART in each subsequent year. The number of expected treatment failures was computed using the incidence of treatment failure at 1, 2, 3, and 3.5 years; therefore, the number of patients needing second-line ART in 2008, 2009, and 2010 is expected to be 5956 (95% CI 3,860-8,052), 15,782 (10,372-21,152), and 34,527 (22,034-47,075), respectively. At the end of 3.5 years, there would be an estimated 50,977 (32,088-70,129) patients needing second-line drugs.


The authors conclude that a significant number of patients in India (approximately 51,000) will need second-line ART therapy in the next 3.5 years, and that, therefore, developing an appropriate policy for second-line treatment is urgent. They suggest monitoring decreases in hemoglobin concentration, body weight, and absolute lymphocyte counts could be inexpensive laboratory indicators to help physicians predict treatment failure.

Quality Rating

Based on the Newcastle-Ottawa rating system for observational cohorts, this study is of good quality. The study was limited by not being able to measure drug resistance and viral load, the gold standard for determining treatment failure. A possible weakness through selection bias of the inclusion criteria was not discussed by the authors. Requiring three CD4 counts within one year resulted in exclusion of 68% of the treatment-na�ve patients visiting the hospital. Such a high degree of selectivity may have resulted in a study sample more adherent, or different in other ways affecting treatment failure, than the overall clinic population. The study did look at biological factors related to treatment failure and found associations with hemoglobin, weight, and lymphocyte count that might be useful surrogates, but no formal attempt to predict treatment failure with these variables was presented, and it is unclear how well they would perform in prediction models.

In Context

There have been limited data on the incidence and risk factors for treatment failure associated with the generic fixed ART regimen used in India. In Uganda, 36% phenotypic drug resistance was reported among patients in an ART program.(3) Data from EuroSIDA, a prospective, international, observational cohort of HIV-infected patients in Europe, reported incidence of treatment failure at 12 months was 11.6 per 100 person-years of follow-up, reducing over time.(4) This study reported a lower rate in the first 12 months, but differences in the definition of treatment failure make it difficult to compare the two. The current study also differed from two previous studies that found baseline CD4 count was significantly associated with treatment failure.(4,5) The extrapolations for future numbers of cases expected on second-line therapy also appears to differ from estimates from a WHO survey that projected a 2% annual increase in the cumulative number on second-line therapy in low- and middle-income countries outside Latin America (6).

Programmatic Implications

High cost, lack of reliable laboratory facilities, and inadequately trained personnel inhibit the assessment of treatment failure by viral load in resource-limited settings. For this reason, WHO guidelines were utilized along with other biological indicators. Patients with negative changes in nutritional factors during follow-up, such as absolute lymphocyte count, hemoglobin concentration, and body weight, experienced higher instances of treatment failure than those with positive changes. It is possible these simple standard tests could help practitioners predict both treatment failure and mortality, but additional study evaluating how well they can predict those outcomes is needed.


  1. World Health Organization. Scaling up antiretroviral therapy in resource limited settings: treatment guidelines for a public health
  2. approach. 2003 revision. Geneva: World Health Organization; 2003.
  3. Ramakumar R. Technical Demography. New Delhi: Wiley Eastern United; 1986. pp. 32-34. (No abstract available.)
  4. Weidle PJ, Malamba S, Mwebaze R, Sozi C, Rukundo G, Downing R, et al. Assessment of a pilot antiretroviral drug therapy programme in Uganda: patients' response, survival, and drug resistance. Lancet 2002 Jul 6;360(9326):34-40.
  5. Dragsted UB, Mocroft A, Vella S, Viard JP, Hansen AB, Panos G, et al. Predictors of immunological failure after initial response to highly active antiretroviral therapy in HIV-1-infected adults: a EuroSIDA study. J Infect Dis 2004 Jul 1;190(1):148-55.
  6. Deeks SG, Barbour JD, Grant RM, Martin JN. Duration and predictors of CD4 T-cell gains in patients who continue combination therapy despite detectable plasma viremia. AIDS 2002 Jan 25;16(2):201-7.
  7. Renaud-Th&#233;ry F, Dongmo NB, Vitoria M, Lee E, Graaff P, Samb B, Perriens J. Use of antiretroviral therapy in resource-limited countries in 2006: distribution and uptake of first- and second-line regimens. AIDS 2007 Jul;21 Suppl 4:S89-95.