In sub-Saharan Africa, millions of HIV-infected people will start antiretroviral therapy (ART) within the next few years [1–3]. Mortality within the first months following ART initiation was shown to be higher in sub-Saharan Africa than in industrialized countries .
Scaling-up the access to ART in African settings where tools for diagnosing and treating opportunistic infections are less available than in high-resource settings entails the necessity to prevent patients who have started ART dying from preventable or curable diseases. It is thus important to document the causes of morbidity that continue to occur in patients receiving ART, and to identify patients at higher risk of dying or developing severe morbidity once they have started ART. Previous studies have already identified tuberculosis, acute sepsis, cryptococcosis and toxoplasmosis as key causes of mortality in patients starting ART with low CD4 cell counts [5–8].
We estimated the incidence and risk factors of all causes of severe morbidity and mortality within the first months following early ART initiation in 792 HIV-infected adults in Abidjan, the economic capital city of Côte d'Ivoire.
In December 2002, a multicentre randomized trial (Trivacan ANRS 1269 trial) was launched in Abidjan . The protocol was approved by the ethics committee of the Ivorian Ministry of Health and the Institutional Review Board of the Agence de Recherche sur le SIDA et les Hépatites Virales (ANRS). The main objective of this trial was to assess various structured ART interruptions strategies. The trial was designed in two phases. Patients were included in the first phase (‘ART initiation phase’) if they met the following criteria: age ≥ 18 years, no past history of ART, CD4+ T-cell (CD4) count between 150 and 350 cells/μl or CD4 percentage between 12.5 and 20.0%, living in Abidjan and written informed consent. At inclusion, all patients started continuous ART and were followed under cohort procedures.
After at least 6 months in the ART initiation phase, patients with undetectable viral load and CD4 cell count > 350 cells/μl were randomized into the ‘ART interruptions strategies’ phase. The criteria for being randomized had to be reached before 18 months of treatment. This explains why the CD4 cell count criterion for being included in the ART initiation phase was atypically high, in comparison with the CD4 cell count starting measure in the international guidelines at the time when the trial was started.
We present here data on severe morbidity that occurred during the ART initiation phase of the Trivacan trial.
The procedures of the ART initiation phase have been previously described [10,11]. In summary: at enrolment, patients started either zidovudine (ZDV)–lamivudine (3TC)–efavirenz or ZDV–3TC–indinavir/ritonavir (800/100 mg twice daily). CD4 cell count (True Count technique on FACScan; Becton Dickinson, Aalst-Erembodegem, Belgium) and plasma HIV-1 RNA (real-time PCR on Taq Man technology Abi Prism 7000; Applied Biosystems, Courtaboeur, France; threshold of detectability at 300 copies/ml)  were measured every 3 months. Cotrimoxazole was systematically given to all patients. All clinical events were reviewed by an event documentation committee. The diagnostic criteria were the same as those used in the ANRS 1203 cohort study [13,14]. Tuberculosis was defined as definitive when a mycobacterium of the tuberculous group was isolated, and as presumptive when all the following criteria were met: (1) consistent clinical picture; (2) presence of acid-fast bacilli on sputum sample, bronchoalveolar lavage or normally sterile body fluid or tissue from a site other than lungs or no other clinically significant pathogen isolated; (3) unsuccessful response to standard antibiotherapy; and (4) successful response to standard antituberculous therapy. Bacterial diseases with positive blood culture for nontyphi Salmonella were classified into World Health Organization (WHO) clinical stage 4. Bacterial diseases with positive blood culture for other clinically significant pathogens were classified into WHO clinical stage 3. When blood cultures were negative, the following bacterial diseases were classified into WHO clinical stage 3: pneumonia, pleurisy, enteritis with stool cultures positive for nontyphi Salmonella or Shigella sp., salpyngitis, pyelonephritis, prostatitis, orchi-epididymitis, meningitis, endocarditis, pyomyositis, pericarditis, and deep-abscess. Diseases were taken into account only if the date of the first symptoms was after the date of ART initiation.
All care was free-of-charge, and patients were reimbursed for the cost of transportation without making any distinction between scheduled and unscheduled visits. In the study centres, participants were attended by nurses, physicians, and social workers who were dedicated full-time to the study. Patients with any severe symptom were attended to first, irrespective of whether the visit was unscheduled or scheduled.
All patients included in the ‘ART initiation phase’ of the Trivacan trial were eligible for the present study. They were excluded from the analyses if they were HIV-2 infected, if they had undetectable viral load on ART initiation, or if they had an ongoing episode of severe morbidity at the time when they started ART. Baseline was the date of ART initiation. The date of the end of study was the date of the 18-month visit following ART initiation, or the date when the patient was randomized into the ART interruption strategies if less than 18 months after the date of ART initiation.
First, we estimated, overall and by baseline characteristics: the incidence of mortality, the incidence of severe morbidity, the Kaplan–Meier probability of reaching virological success of ART and the Kaplan–Meier probability of responding immunologically to ART. Severe morbidity was defined as all WHO stage 3 or 4-classifying morbidity events (with the exception of oral candidiasis). Virological success was defined as the first time when a viral load was measured as undetectable. Immunological response was defined as the first time when a gain in CD4 cell count of at least 50 cells/μl since baseline was observed. The trends for incidence of mortality and morbidity over time were tested using Poisson regression.
Second, we studied factors associated with death and with severe morbidity using a multivariate Cox proportional-hazard regression analysis. Baseline explicative variables were age, sex, body mass index, WHO clinical stage combined with past history of successfully treated tuberculosis (stage 1 or 2 versus stage 3 or 4 without tuberculosis past history versus stage 3 or 4 with tuberculosis past history), baseline haemoglobin, baseline CD4 cell count and baseline plasma HIV-1 viral load. Follow-up explicative variables were follow-up CD4 cell counts and follow-up viral loads (detectable versus undetectable). In a first set of analyses, only baseline variables were included in the multivariate model. In a second set of analyses, both baseline and follow-up variables were included in the model. For mortality analysis, severe morbidity was included in the second set of analyses as a time-independent explicative variable.
Finally, we studied whether severe morbidity was associated with virological and immunological outcomes. The association of each biological outcome with follow-up severe morbidity was studied using a multivariate Cox analysis, adjusting on all characteristics associated with the corresponding outcome with a P < 0.20 in univariate analysis.
For the morbidity incidence rate calculation and for the analyses of factors associated with morbidity, we only took into account the first event.
All Cox proportional-hazard models were checked for the proportional hazards condition.
Patients and follow-up
Of the 840 adults enrolled in the Trivacan trial, 48 were excluded from the present study because they were infected with HIV-2 (n = 16), they had an undetectable viral load at baseline (n = 9), or they had a prevalent episode of tuberculosis at the time when they started ART (n = 23). The 792 patients who remained in analyses predominantly had a CD4 cell count > 200 cells/μl (71%) and were at WHO stage 1 or 2 (64%) (Table 1). During follow-up, 370 (47%) patients were randomized in the Trivacan trial at 7 months, 155 (20%) at 10 months, 51 (6%) at 13 months, 28 (3.5%) at 16 months, and 188 (24%) were never randomized. Of the latter, nine were lost-to-follow-up before study termination.
Causes of severe morbidity and mortality
During follow-up, 18 patients died (including eight with at least one documented episode of severe morbidity between inclusion and death) and 59 patients presented 65 episodes of severe morbidity. Of these episodes, 43 (66%) were recorded for the first time at unscheduled visits versus 22 (34%) at scheduled visits.
The episodes of severe morbidity were invasive bacterial diseases (28 patients, 32 episodes), tuberculosis (23 patients, 25 episodes), and other WHO stage 3 or 4 classifying diseases (eight patients, eight episodes). The latter included isosporiasis (n = 1), cerebral toxoplasmosis (n = 1), Kaposi's sarcoma (n = 2), unexplained chronic diarrhea (n = 2), chronic genital herpes simplex virus infection (n = 1) and cryptosporidiosis (n = 1). The 32 episodes of bacterial diseases were 16 pneumonia, five isolated bacteraemia, four enteritis, three pyelonephritis, one pyomyositis, one deep abscess, one prostatitis and one salpyngitis. At the time when the first bacterial episode occurred, the median last available CD4 cell count was 281 cells/μl [interquartile range (IQR), 197–406] and the median time since ART initiation was 3.0 months (IQR, 1.3–4.5). A pathogen was isolated in 18 of these 32 episodes, including in blood culture in 12 episodes. The 18 bacterial strains were Streptococcus pneumoniae (n = 6), nontyphi Salmonella (n = 6), Escherichia coli (n = 4), Staphylococcus aureus (n = 1) and Enterobacter cloacae (n = 1). The 25 episodes of tuberculosis (definitive 68%), were pulmonary (64%), extra-pulmonary (24%), and both (12%). At the time when the first tuberculosis episode occurred, the median last available CD4 cell count was 235 cells/μl (IQR, 167–315) and the median time since ART initiation was 3.7 months (IQR, 0.5–5.4).
In the 18 patients who died, a cause of death could be identified in nine cases: tuberculosis (n = 3), cerebral toxoplasmosis (n = 1), Kaposi's sarcoma (n = 1), isolated bacteraemia (n = 1), malaria (n = 1), cardiac insufficiency (n = 1) and trauma (n = 1). Of the remaining nine patients, three died with symptoms of undiagnosed origin (one pleuritis, one unexplained chronic diarrhea, one acute unexplained fever), and the other six died at home or elsewhere without having attended the care centre before death.
Overall incidence and risk factors of severe morbidity and mortality
During the entire follow-up, the incidence of severe morbidity was estimated at 10.6/100 patient-years (P-Y) [95% confidence interval (CI), 8.10–13.10] overall and 13.3/100 P-Y, 9.5/100 P-Y and 7.9/100 P-Y in patients with pre-ART CD4 cell count < 200, at 200–350 and > 350 cells/μl, respectively.
The incidence of death was estimated at 2.6/100 P-Y (95% CI 1.5-4.1) overall, and 5.0/100 P-Y, 1.7/100 P-Y, and 0.0/100 P-Y in patients with pre-ART CD4 cell count < 200, at 200–350 and > 350 cells/μl, respectively.
Figure 1 shows the incidence of severe morbidity (Fig. 1a) and mortality (Fig. 1b) by baseline characteristics. In univariate analysis, both mortality and severe morbidity were significantly associated with body mass index, haemoglobin, and WHO clinical stage. In addition, mortality was significantly associated with baseline CD4 cell count although the association between severe morbidity and CD4 cell count did not reach significance, and severe morbidity was significantly associated with baseline viral load level whereas the association between mortality and baseline viral load did not reach significance.
Tables 2 and 3 show the results of the multivariate analysis of the association between severe morbidity or mortality with baseline characteristics on the one hand and with follow-up characteristics (adjusting on baseline characteristics) on the other hand. As shown in these tables, severe morbidity was independently associated with an advanced baseline WHO stage, a low baseline haemoglobin, a high pre-ART viral load, a low follow-up CD4 cell count and a persistently detectable viral load during follow-up. When considering separately the main groups of severe morbidity, the two variables most strongly associated with the risk of tuberculosis were a past history of tuberculosis and a low follow-up CD4 cell count, whereas the variable most strongly associated with severe bacterial diseases was a high pre-ART viral load (Table 2). Mortality was independently associated with a low baseline body mass index, a low baseline or follow-up CD4 cell count, a persistently detectable viral load during follow-up, and the occurrence of at least one episode of severe morbidity during follow-up (Table 3). There was no significant association between age or sex and mortality or severe morbidity.
Evolution of morbidity and mortality over time
During the first, second and third quarter following ART initiation, the overall incidence of severe morbidity was 16.6 (95% CI, 10.9–22.4), 10.2 (95% CI, 6.1–15.9) and 6.6/100 P-Y (95% CI, 2.8–12.9), respectively (trend over time: P = 0.03), and the incidence of mortality was 3.1 (95% CI, 1.1–6.6), 1.0 (95% CI, 0.1–3.7) and 1.5/100 P-Y (95% CI, 0.2–5.5), respectively (trend over time: P = 0.33). Figure 2 shows the evolution in time of the incidence of mortality, overall severe morbidity, tuberculosis and bacterial diseases since ART initiation. The incidence rate of morbidity was 25.2 (first quarter), 13.2 (second quarter) and 4.7/100 P-Y (third quarter) in patients with pre-ART CD4 cell count < 200 cells/μl (trend over time: P = 0.02), 15.1 (first quarter), 8.3 (second quarter) and 8.2/100 P-Y (third quarter) in patients with pre-ART CD4 cell count at 200–349 cells/μl (trend over time: P = 0.27) and 8.0 (first quarter), 10.9 (second quarter) and 5.6/100 P-Y (third quarter) in patients with pre-ART CD4 cell count ≥ 350 cells/μl (trend over time: P = 0.80).
Association between severe morbidity and virological or immunological outcomes
The probability of reaching viral load undetectability at least once was estimated at 0.82, 0.95 and 0.97 at 6, 12 and 18 months, respectively. In multivariate analysis, patients who had at least one severe morbidity event during follow-up had an adjusted hazard ratio of reaching viral load delectability of 0.68 (95% CI, 0.51–0.91, P = 0.01). Of the 59 patients with at least one episode of severe morbidity, 41 had their first episode before viral load undetectability was noticed, and 18 had their first episode after having reached viral load undetectability at least once.
The probability of having a gain in CD4 cell count > 50 cells/μl since baseline at least once was estimated at 0.79, 0.94 and 0.96 at 6, 12 and 18 months, respectively. In multivariate analysis, patients who had at least one severe morbidity event during follow-up had an adjusted hazard ratio of reaching a gain in CD4 > 50 cells/μl of 0.64 (95% CI, 0.48–0.85; P = 0.002).
Our patients started ART with a much more moderate level of pre-ART immunosuppression in comparison with the vast majority of adults who have been starting ART in sub-Saharan Africa in recent years [4–6,8,15–22]. They were followed under cohort procedures with a standardized documentation of all morbidity events and a low rate of loss-to-follow-up.
The ART-LINC team previously reported that patients starting ART in low-income settings had a higher hazard of mortality than those starting ART in high-income settings during the first few months following ART initiation . In our study, the rapidly declining rate over time of tuberculosis and invasive bacterial diseases may partly explain these findings.
Our patients were followed under trial conditions, with all care free of charge. The rate of mortality would have probably been higher if these patients had been followed-up in field conditions. Despite optimal conditions of access to care, the rate of mortality was not null in patients with pre-ART CD4 cell count at 200–350 cells/μl. Furthermore, even if most patients who experienced severe morbidity did not die, the occurrence of at least one severe morbidity event after ART initiation was significantly associated with impaired medium-term immunological and virological outcomes. These data suggest that starting ART before the CD4 cell count drops below 350 cells/μl in HIV-infected adults in Côte d'Ivoire may prevent short-term mortality on ART and improve short term immunological and virological treatment outcomes by reducing the rate of severe morbidity.
Data from CD4-guided, trials of structured treatment-interruption strategies performed in industrialized and low-resource countries showed that patients ‘off’ ART had higher severe morbidity and mortality rates than those ‘on’ ART, even at CD4 cell counts higher than those currently recommended in international guidelines for ART initiation [9,23]. This led some experts to suggest that the issue of ‘when to start ART’ should be rapidly addressed through trials comparing the current starting criteria with earlier criteria . The spectrum of HIV-morbidity in sub-Saharan Africa gives specific arguments to suggest that ART should be started earlier than currently recommended. In a cohort study of HIV-infected, untreated adults in South Africa, the risk of AIDS in patients with a CD4 cell count at 250–350 cells/μl was 1.9 times greater than the risk previously reported from European cohorts . In sub-Saharan Africa before the ART era, the rates of HIV-associated tuberculosis and invasive bacterial diseases and, consequently, the rate of severe morbidity at early stages of immunosupression, were shown to be higher than in high-income countries. The high rate of bacterial diseases, among others, led WHO experts to recommend that cotrimoxazole prophylaxis should be started earlier in Africa than in industrialized countries [26–28]. In this context of early severe morbidity, starting ART at an earlier stage of immunosupression would be logical.
To refine the criteria for starting ART in Africa, the pre-ART factors associated with the outcomes need to be better documented, in order to treat first and foremost patients who will be at higher risk of failing treatment. The CD4 cell count is not the only pre-ART factor to predict the events occurring after ART initiation. In our study, patients with low haemoglobin level, low body mass index and advanced WHO stage were at higher risk of dying and/or developing severe morbidity, as previously reported by others [4,5,16–18]. The CD4 cell count and plasma viral load measurements performed after ART initiation were stronger risk factors for the on-ART severe morbidity and mortality than the pre-ART CD4 cell count and viral load values, even during the first year of treatment [8,29,30]. A high plasma viral load was more strongly associated with the occurrence of severe bacterial diseases than with the occurrence of tuberculosis. Finally, a past episode of successfully treated tuberculosis before ART initiation was independently associated with the risk that a new episode of tuberculosis would occur once ART has been started. This is the second time that an association between past history of tuberculosis and incident tuberculosis on ART has been reported in Côte d'Ivoire . Interestingly, though, such an association was not found in other studies performed in South Africa [8,31]. There could be two explanations for this divergence. First, the history of tuberculosis is not an easy variable to record exhaustively in routine care conditions . The longer the time since a past episode of tuberculosis occurred, the higher the risk that a patient may omit to declare it but also the higher the risk that a relapse may occur . Second, in the local context of tuberculosis care, for example, the percentage of patients who self-administer their antituberculosis treatment in comparison with the percentage of those who receive directly observed therapy, may lead tuberculosis to be more likely to relapse on ART in some settings compared with others [33,34].
The main limit of our study was that patients who reached criteria for ART success after 7 months were randomized into the Trivacan trial and therefore left the study. Thus, after 7 months, the follow-up of the healthier patients was censored whereas the sickest patients remained in follow-up. This could have led to an overestimation of the mortality and morbidity rates after 7 months. If no follow-up censoring had been made, the decreasing rate of mortality and morbidity over time would have probably been even more marked.
In conclusion, this large cohort study of the mortality and morbidity in adults who started ART at intermediate stages of immunosuppression in West Africa gives new arguments to reinforce the hypothesis that, in this region, ART should be started before the CD4 cell count drops below 350 cells/μl. Further studies should assess whether patients with low BMI, low haemoglobin, high viral load, clinical stage3 or past history of active tuberculosis should start ART earlier.
We would like to thank Joanna Orne-Gliemann (INSERM U593) for her grammatical and editing contributions.
Contributors: R. Moh, C. Danel, X. Anglaret, E. Bissagnene and R. Salamon sought funding for the study and were responsible for overall study coordination. E. Messou, C. Seyler, A. Anzian, Y. Abo and S. Eholié were responsible for clinical care of patients at the study centre. T. Ouassa led all biological aspects. D. Gabillard was the study statistician. X. Anglaret, R. Moh, D. Gabillard and C. Danel drafted the manuscript, which all authors subsequently reviewed, edited and approved.
Sponsorship: This study was supported by the French Agence Nationale de Recherches sur le SIDA et les hépatites virales (ANRS, Paris, France, grant ANRS 12136), and the Ivoirian Ministry of Public Health within the collaborative programme PAC-CI.
4. Braitstein P, Brinkhof MW, Dabis F, Schechter M, Boulle A, Miotti P, et al
. Mortality of HIV-1-infected patients in the first year of antiretroviral therapy: comparison between low-income and high-income countries. Lancet 2006; 367:817–824.
5. Etard JF, Ndiaye I, Thierry-Mieg M, Gueye NF, Gueye PM, Laniece I, et al
. Mortality and causes of death in adults receiving highly active antiretroviral therapy in Senegal: a 7-year cohort study. AIDS 2006; 20:1181–1189.
6. Seyler C, Anglaret X, Dakoury-Dogbo N, Messou E, Touré S, Danel C, et al
. Medium-term survival, morbidity and immunovirological evolution in HIV-infected adults receiving antiretroviral therapy, Abidjan, Côte d'Ivoire. Antivir Ther 2003; 8:385–393.
7. Lawn SD, Bekker LG, Myer L, Orrell C, Wood R. Cryptococcocal immune reconstitution disease: a major cause of early mortality in a South African antiretroviral programme. AIDS 2005; 19:2050–2052.
8. Lawn SD, Myer L, Bekker LG, Wood R. Burden of tuberculosis in an antiretroviral treatment programme in sub-Saharan Africa: impact on treatment outcomes and implications for tuberculosis control. AIDS 2006; 20:1605–1612.
9. Danel C, Moh R, Minga A, Anzian A, Ba-Gomis O, Kanga C, et al
. CD4-guided structured antiretroviral treatment interruption strategy in HIV-infected adults in west Africa (Trivacan ANRS 1269 trial): a randomised trial. Lancet 2006; 367:1981–1989.
10. Danel C, Moh R, Anzian A, Abo Y, Chenal H, Guehi C, et al
. Tolerance and acceptability of an efavirenz-based regimen in 740 adults (Predominantly women) in West Africa. J Acquir Immune Defic Syndr 2006; 42:29–35.
11. Moh R, Danel C, Sorho S, Sauvageot D, Anzian A, Minga A, et al
. Haematological changes in adults receiving a zidovudine-containing HAART regimen in combination with cotrimoxazole in Cote d'Ivoire. Antivir Ther 2005; 10:615–624.
12. Rouet F, Ekouevi DK, Chaix ML, Burgard M, Inwoley A, Tony TD, et al
. Transfer and evaluation of an automated, low-cost real-time reverse transcription – PCR test for diagnosis and monitoring of human immunodeficiency virus type 1 infection in a West African resource-limited setting. J Clin Microbiol 2005; 43:2709–2717.
13. Anglaret X, Messou E, Ouassa T, Toure S, Dakoury-Dogbo N, Combe P, et al
. Pattern of bacterial diseases in a cohort of HIV-1 infected adults receiving cotrimoxazole prophylaxis in Abidjan, Cote d'Ivoire. AIDS 2003; 17:575–584.
14. Seyler C, Toure S, Messou E, Bonard D, Gabillard D, Anglaret X. Risk factors for active tuberculosis following antiretroviral treatment initiation in Abidjan. Am J Respir Crit Care Med 2005; 172:123–127.
15. Ferradini L, Jeannin A, Pinoges L, Izopet J, Odhiambo D, Mankhambo L, et al
. Scaling up of highly active antiretroviral therapy in a rural district of Malawi: an effectiveness assessment. Lancet 2006; 367:1335–1342.
16. Stringer JS, Zulu I, Levy J, Stringer EM, Mwango A, Chi BH, et al
. Rapid scale-up of antiretroviral therapy at primary care sites in Zambia: feasibility and early outcomes. JAMA 2006; 296:782–793.
17. Laurent C, Ngom Gueye NF, Ndour CT, Gueye PM, Diouf M, Diakhate N, et al
. Long-term benefits of highly active antiretroviral therapy in Senegalese HIV-1-infected adults. J Acquir Immune Defic Syndr 2005; 38:14–17.
18. Zachariah R, Fitzgerald M, Massaquoi M, Pasulani O, Arnould L, Makombe S, et al
. Risk factors for high early mortality in patients on antiretroviral treatment in a rural district of Malawi. AIDS 2006; 20:2355–2360.
19. Calmy A, Pinoges L, Szumilin E, Zachariah R, Ford N, Ferradini L. Generic fixed-dose combination antiretroviral treatment in resource-poor settings: multicentric observational cohort. AIDS 2006; 20:1163–1169.
20. Weidle PJ, Wamai N, Solberg P, Liechty C, Sendagala S, Were W, et al
. Adherence to antiretroviral therapy in a home-based AIDS care programme in rural Uganda. Lancet 2006; 368:1587–1594.
21. Wester CW, Kim S, Bussmann H, Avalos A, Ndwapi N, Peter TF, et al
. Initial response to highly active antiretroviral therapy in HIV-1C-infected adults in a public sector treatment program in Botswana. J Acquir Immune Defic Syndr 2005; 40:336–343.
22. Coetzee D, Hildebrand K, Boulle A, Maartens G, Louis F, Labatala V, et al
. Outcomes after two years of providing antiretroviral treatment in Khayelitsha, South Africa. AIDS 2004; 18:887–895.
23. Strategies for Management of Antiretroviral Therapy (SMART) Study Group. CD4+ count-guided interruption of antiretroviral treatment. N Engl J Med
25. Badri M, Lawn S, Wood R. Short-term risk of AIDS or death in people infected with HIV-1 before antiretroviral therapy in South Africa: a longitudinal study. Lancet 2006; 368:1254–1259.
27. Mermin J, Lule J, Ekwaru JP, Malamba S, Downing R, Ransom R, et al
. Effect of co-trimoxazole prophylaxis on morbidity, mortality, CD4-cell count, and viral load in HIV infection in rural Uganda. Lancet 2004; 364:1428–1434.
28. Anglaret X, Chêne G, Attia A, Toure S, Lafont S, Combe P, et al
. Early chemoprophylaxis with trimethoprim-sulphamethoxazole for HIV-1-infected adults in Abidjan, Côte d'Ivoire: a randomised trial. Lancet 1999; 353:1463–1468.
29. Lawn SD, Myer L, Harling G, Orrell C, Bekker LG, Wood R. Determinants of mortality and nondeath losses from an antiretroviral treatment service in South Africa: implications for program evaluation. Clin Infect Dis 2006; 43:770–776.
30. Anastos K, Barron Y, Cohen MH, Greenblatt RM, Minkoff H, Levine A, et al
. The prognostic importance of changes in CD4+ cell count and HIV-1 RNA level in women after initiating highly active antiretroviral therapy. Ann Intern Med 2004; 140:256–264.
31. Lawn SD, Badri M, Wood R. Risk factors for tuberculosis among HIV-infected patients receiving antiretroviral treatment [letter]. Am J Respir Crit Care Med 2005; 172:1348, author reply 1348–1349.
32. Harries AD, Hargreaves NJ, Kwanjana JH, Salaniponi FM. Relapse and recurrent tuberculosis in the context of a national tuberculosis control programme. Trans R Soc Trop Med Hyg 2000; 94:247–249.
33. Balasubramanian VN, Oommen K, Samuel R. DOT or not? Direct observation of antituberculosis treatment and patient outcomes, Kerala State, India. Int J Tuberc Lung Dis 2000; 4:409–413.
34. Frieden TR. Lack of directly observed treatment affects tuberculosis relapse rates [letter]. Am J Respir Crit Care Med 2006; 173:359, author reply 359.