Objective: To estimate the effect of delaying antiretroviral treatment (ART) for 15, 30, or 60 days after tuberculosis (TB) treatment initiation on mortality and virological suppression.
Design: Cohort of 573 ART-naive HIV-infected children initiated on TB treatment at an outpatient clinic in South Africa between April 2004 and March 2008.
Methods: Hazard ratios for mortality and viral suppression were estimated using marginal structural models and multivariate Cox models, respectively.
Results: During follow-up (median 9.64 months), 37 HIV-infected children died after a median of 62 days of TB treatment. ART was initiated in 461 children at a median of 17 days after TB treatment initiation, 415 (90%) achieved viral suppression. The hazard ratios of death for initiating ART more than 15, more than 30, or more than 60 days of TB treatment compared with initiating within 15, 30 and 60 days, respectively, were 0.82 (95% CI: 0.48, 1.41), 0.86 (95% CI: 0.46, 1.60), and 1.32 (95% CI: 0.55, 3.16). Hazard ratios for analysis restricted to severely immunosuppressed children were: 0.92 (95% CI: 0.51, 1.63), 1.08 (95% CI: 0.56, 2.08), and 2.23 (95% CI: 0.85, 5.80), respectively. Hazard ratios for viral suppression were 0.98 (95% CI: 0.76, 1.26), 0.95, (95% CI: 0.73, 1.23), 0.84 (95% CI: 0.61, 1.15), respectively and did not change with restriction to children severely immunosuppressed.
Conclusion: In this observational study, we found that delaying ART for 2 months or more in children diagnosed with TB may be associated with poorer virological response and increased mortality, particularly in children with severe immunosuppression. These findings should be confirmed in a randomized controlled trial.
aDepartment of Epidemiology, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA
bWits ECHO, Harriet Shezi Children's Clinic, Chris Hani Baragwanath Hospital, University of the Witwatersrand, Johannesburg, South Africa
cDepartment of Medicine, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA.
Received 21 November, 2009
Revised 10 February, 2010
Accepted 17 February, 2010
Correspondence to Marcel Yotebieng, MD, MPH, PhD, Research Assistant Professor, Department of Epidemiology, Gillings School of Global Public Health, University of North Carolina, McGavran-Greenberg Hall, CB #7435, Chapel Hill, NC, USA. E-mail: email@example.com
Worldwide, about 1200 new pediatric HIV infections occur daily and more than 90% of new infections happen in sub-Saharan Africa . Children who are infected early in life progress more rapidly to AIDS and death. Tuberculosis (TB) is a major cause of morbidity and mortality among HIV-infected children in Africa . Following TB infection, children (particularly young infants), have a higher risk of progression to disease (extrapulmonary and pulmonary) and death irrespective of their HIV status . However, HIV is also a major risk factor for childhood TB . The burden of TB in HIV-infected children is not well known because of the difficulties in diagnosing TB in this population . Though it might be greatly influenced by selection bias, reported prevalence of HIV among TB-infected children range from less than 5% in industrialized settings to over 50% in some African settings [6–8]. In HIV/TB coinfected children, diminished cell-mediated immunity increases the risk for disseminated TB disease, especially in advanced stages of HIV infection, resulting in high mortality [9,10]. Restoration of cellular immunity with antiretroviral therapy (ART) reduces the susceptibility to TB .
The WHO recommends initiation of TB treatment as soon as the diagnosis of TB is suspected in an HIV-infected child. The management of HIV-infected children with TB is complicated as rifampicin, the backbone of any TB treatment regimen, has been shown in adults to reduce the serum concentrations of most protease inhibitors by about 80% or more, and that of nonnuclease reverse transcriptase inhibitors by 20–60% . In addition, TB drugs and ART have overlapping toxicities. To avoid drug interactions and to differentiate side effects for TB drugs and ART, WHO advises that initiation of ART be delayed for at least 2–8 weeks after TB treatment initiation [12,13]. In South Africa, national guidelines recommend completion of TB therapy before starting ART in children not severely immunosupressed [CD4 percentage (CD4%) ≥25% in children, ≤11 months, ≥20% for children aged 12−35 months, or ≥15% for older children], or delay of ART for 4–8 weeks in children with severe immunosuppression . The potential effects of delaying ART on survival or virological response to ART are unknown.
In this study, we aimed to estimate the effect of delaying ART initiation for at least 15, 30, or 60 days in children initiating TB treatment on virological suppression and survival. We hypothesized that children in whom ART is delayed will be at higher hazard of death but will be more likely to achieve viral suppression once they are initiated on ART.
We used routinely collected data from an observational cohort of HIV-infected children, who sought care at the Harriet Shezi Children's Clinic, an outpatient pediatric HIV clinic at Chris Hanni Baragwanath Hospital in Soweto, South Africa, during the first 4 years (April 2004 to March 2008) of the government ART program . TB was routinely diagnosed on clinical grounds including severe failure to thrive, prolonged (more than 2 weeks) cough, suspicious chest radiograph with or without a positive contact history. Bacteriological confirmation was attempted in older children who can produce a sputum sample. Children were treated according to national guidelines with a combination of rifampicin, isoniazid, and pyrazinamide for the initial 2 months followed by rifampicin + isoniazid for the remaining 4 months .
According to the South Africa National Guidelines , all HIV-infected children diagnosed with TB are eligible for ART. The first line ART regimen for children receiving TB treatment consists of stavudine, lamivudine, and ritonavir-boosted lopinavir (LPV/r) for children 3 years or younger; or stavudine, lamivudine and efavirenz for those over 3 years and over 10 kg of weight. Double doses of ritonavir dose were given during antitubercular treatment [18,19].
Children initiated on ART were clinically assessed at 1 month, at 3 months, and every 3 months thereafter or as needed clinically. Laboratory investigations (i.e., viral load and CD4 cell count) were done at baseline and every 6 months or whenever indicated. Clinical and laboratory information were entered in an electronic database.
Children were included in this analysis if their first visit in the clinic occurred between 1 April, 2004 and 31 March, 2008; they were initiated on TB treatment prior to ART initiation; and were 15 years or younger at the time of TB treatment initiation. Children who were already on TB treatment at first visit in the clinic were excluded. For analysis of time to viral suppression, eligible children had to also have at least one viral load measurement post-ART initiation.
Two main outcomes were considered. Survival was defined as time from TB treatment initiation to death. For six children whose exact date of death was unknown, the last clinic visit date was used. Time to viral suppression was defined as time from ART initiation to the date of first viral load measure below 400 HIV RNA copies/ml.
For presentation of results, time from TB treatment initiation to ART initiation was categorized as more than 15, more than 30, or more than 60 days. The 15, 30, and 60 days cut-offs were chosen to correspond to the minimum 2–8 weeks delay between TB treatment and ART initiation recommended by the WHO .
Additional covariables considered were level of immunosuppression, viral load (dichotomized at about the median of 5 log10 copies/ml), age at TB initiation (<18 months, 18–35 months, 36–59 months, 60 months or older), and weight-for-age Z score (WAZ) (greater than −2SD (no undernutrition), −2SD to −3SD (undernutrition), and less than – 3SD (severe undernutrition) [20,21].
Baseline characteristics (at time of TB treatment initiation) were compared using Wilcoxon Rank-Sum test for continuous variables and Pearson chi squared for categorical variables.
For the analysis of mortality, inverse probability-of-treatment-and censoring (IPTC) weighting of marginal structural models was used to adjust for time-dependent confounding [22,23]. Time-dependent confounding could occur because, even though per guidelines all children with TB were eligible for ART, the decision to initiate or delay ART for a given child was likely to be influenced by clinical and immunological characteristics, with children with advanced disease being more likely to be initiated on ART earlier. Stabilized weights for treatment and censoring were estimated as described in reference . Time from TB treatment to ART initiation or to the end of follow-up due to administrative censoring (at 36 months of follow-up or at last visit before 31 March, 2008 whichever came first) or informative censoring (loss of follow-up defined as no clinic visit for at least 6 months and unsuccessful contact tracing by phone and/or home visit) were treated as failure times. Using pooled logistic models, we estimated the probabilities of initiating ART (treatment weights) or remaining in care (censoring weights) at each visit with baseline level of immunosuppression, log viral load, WAZ, age at TB treatment initiation, and the interaction term between WAZ and level of immunosuppression as fixed covariates for the numerators of weights. For the denominators, time-dependent level of immunosuppression, log viral load, WAZ, and ART initiation were added as covariates in the models. Restricted cubic spline terms for time from TB treatment (with 5 knots at 5th, 25th, 50th, 75th, and 95th percentile of its distribution) were also included in these models. The stabilized weights for ART initiation, administrative censoring, and informative censoring were multiplied together to obtain the final IPTC weights. Assuming no misspecification of the models for IPTC weight estimation, and provided that data are available on all important time-dependent and baseline predictors of ART initiation, drop out and mortality, then IPTC weighted estimation of a marginal structural model provides a valid method to account for confounding and selection bias [22–24]. A necessary condition for correct model specification is that the stabilized weights have a mean of one . The mean of the IPTC weights was 1.01 and its values ranged from 0.24 to 5.60. Both unweighted and IPTC-weighted (trimmed at the first and 99th percentiles) Cox proportional models for mortality were fit for each of the three categories of time between initiation of TB treatment and ART.
To assess hazards of viral suppression children had to be on ART. Hence, there were no time-dependent confounders and the only potential source of confounding that had to be adjusted for in the estimation of the effect of timing of ART initiation on viral suppression were the fixed baseline characteristics. Cox proportional models were used to obtain the crude and adjusted effect of the timing of ART initiation on virological suppression. All covariates included in the models were formally assessed for the proportional hazard assumption using the Kolmogorov-type supremum test .
To assess if the effects of delaying ART on mortality or viral suppression vary by the level of immunosupression at baseline, each analysis was repeated for each stratum of level of immunosuppression when there were sufficient data. All analyses were conducted using SAS 9.1 (SAS Institute, Cary, North Carolina, USA) and all tests were conducted using a two-sided 0.05 significance level, without correction for multiple comparisons. All confidence intervals (CI) were estimated using robust variances .
Between April 2004 and March 2008, 3187 HIV-infected children initiated care in the clinic. In 909 children, TB treatment was initiated after their initial visit in the clinic. Of those 909, 336 were excluded from this analysis eight were older than 15 years and 328 were on ART at the time of TB treatment. The 573 children included in the analysis were followed for a median time of 9.6 [interquartile range (IQR): 1.9 to 23.1] months, 75 (13%) were lost to follow up: 38 prior to ART initiation and 37 while on ART. Children lost to follow-up did not differ from those still in care by any of the baseline characteristics or timing of ART initiation.
The median age of the 573 children included in this analysis, was 3.5 (IQR: 1.4 to 6.8) years (Table 1). At baseline, the median CD4% was 11.9% (IQR: 6.6–18.3%), and the majority (75.2%) were severely immunosuppressed (CD4% below 25% for children younger than 1, or below 20% for children 1 to 2 years, or below 15% for children 3–15 years ); the median viral load was 5.2 (IQR 4.5–5.9) log copies/ml; and the median WAZ was −2.3 (IQR: −3.6–−1.3) with 34.7% of children severely underweight-for-age.
Most children (461/573 80.5%) were initiated on ART. ART was started at a median time of 17 (IQR 0–49) days after TB treatment initiation. Time between the start of TB treatment and initiation of ART was more than 15 days in 55% (275/501) of children, more than 30 days in 42% (206/498), and more than 60 days 23% (111/494) (denominator decreases with increased duration of follow-up time due to censoring). Thirty one percentage (141/461) were started on ART the same day TB treatment was initiated and 6% (27/461) were initiated on ART 6 months after TB treatment initiation and 13 of the 112 children (12%) who were never initiated on ART have longer that 6-months of follow-up. Children who were initiated on ART earlier (within 1 month of starting TB treatment) had lower CD4 cell count, and lower WAZ, than those in whom ART was delayed for more than 1 month (Table 1).
Seven (1.2%) children died before ART initiation (3 within 30 days of TB treatment and 4 after 60 days) and 30 died after ART initiation (22 in the first 2 weeks of TB treatment, 25 in the first 30 days, and 26 in the first 60 days of TB treatment. Children who died were more likely at baseline to be younger (median age 1.4 vs. 3.7 years, P < 0.01), to be severely underweight for age (median WAZ −3.8 vs. −2.2, P < 0.01), to have higher viral load (median 5.9 vs. 5.2, P < 0.01), and to have severe immunosuppression (8.7% vs. 3.2, P = 0.04)) (Table 2). Kaplan–Meier survival cures show that while the survival probability curve of children in whom ART was delayed was initially slightly higher, it declined continuously and eventually past below that of those initiated on ART immediately, particularly in severely immunosuppressed children (Fig. 1).
Adjusted hazard ratios (aHR) for the effect on mortality of delaying ART for more than 15, more than 30, or more than 60 were 0.82 (95% CI 0.48, 1.63), 0.86 (95% CI 0.46, 1.60), and 1.32 (95% CI 0.55, 3.16), respectively. All aHRs account for the time-varying level of immunosuppression, log viral load, WAZ, and censoring prior to ART, and time fixed age at TB treatment initiation (Table 3). Analysis with IPCT weights trimmed at the first and 99th percentiles produced results with similar aHR for the first two estimates: [0.76 (95% CI 0.43, 1.34), 0.72 (95% CI 0.37, 1.38)], but a different estimate for the aHR for delay of at least 60 days [0.86 (95% CI 0.36, 2.09)].
Full stratification of the analysis of mortality by level of immunosuppression was not possible because only four deaths occurred among the 125 children not severely immunosuppressed at baseline (one immediately after TB treatment initiation and before ART and three after ART initiation). Restriction of the analysis to children with severe immunosuppression resulted in stronger effects for delaying ART initiation for more than 60 days, with an aHRs of 2.23 (95% CI 0.85, 5.80) (Table 3). Analysis with the IPCT weights trimmed at the first and 99th percentiles did not change these results.
Of the 461 children initiated on ART, 324 (70%) had at least one viral load measurement post-ART initiation (after a median time of 5.51 months of ART initiation) of whom 291 (90%) ever achieved viral load below 400 copies/ml. Children who achieved virologic suppression were more likely at the time of TB treatment initiation to be older (median age 4.0 vs. 1.6 years, P < 0.01), to have higher weight-for-age (median WAZ −2.2 vs. −3.1, P < 0.02), and to have lower viral load (median log 5.2 vs. 5.8, P = 0.01).
Delaying ART for more than 15 or more than 30 days had no effect on viral suppression [aHRs = 0.98 (95% CI 0.76, 1.26) and 0.95 (95% CI 0.73, 1.23), respectively], but delaying more than 60 days after TB treatment initiation tended towards lower hazards of achieving virological suppression [aHR 0.84 (95% CI 0.54, 1.10)]. All hazard ratios were adjusted for level of immunosuppression, log viral load, WAZ, and age at TB treatment initiation (Table 3).
Full stratification of the analysis of viral suppression by level of immunosuppression was again not possible because only 63 of the 324 children with a viral load measurement on ART were not severely immunosuppressed at baseline and only six of them had not yet achieved virological suppression at the time of analysis. Restricting the analysis to only severely immunosuppressed children did not change the results substantially the [aHRs 0.96 (95% CI: 0.73, 1.26), 0.91 (95% CI: 0.67, 1.24), and 0.77 (95% CI: 0.50, 1.18) for delay of at least 15, at least 30 or at least 60 days, respectively].
We sought to estimate the effect of delaying ART initiation in children receiving TB treatment on mortality and virologic response to ART. The analysis of a cohort of 573 South African children showed that a delay of 30 days or less between TB treatment and ART initiation had no effect on mortality or viral response to ART. Although not statistically significant, the effect of delaying ART for more than 60 days tended towards higher hazards of death and lower hazards of viral suppression. Severely immunosuppressed children at time of TB treatment initiation in whom ART was delayed for more than 60 days were twice as likely to die (aHR 2.23, 95CI 0.85, 5.80).
This is the first report on the effect of timing of ART initiation in children diagnosed with TB. In a recent analysis of a cohort of HIV/TB coinfected adults in Spain, simultaneous ART and TB treatment (ART initiated within ±2 months of TB treatment) was associated with improved survival . In a randomized trial of early (within 14 days) versus deferred ART initiation in adults with opportunistic infections other than TB, early ART resulted in less AIDS progression or death with no increase in adverse events including the immune reconstitution inflammatory syndrome (IRIS) . In young infants, it has been shown that, independent of TB status; early initiation of ART even in asymptomatic infants is associated with improved survival .
The principal justifications for the current recommendation to delay ART in children initiating TB treatment is to allow differentiation between side effects from ART and TB drugs, to avoid occurrence of IRIS [31,32], and to improve the viral response to ART by avoiding drug interactions . The observation in this cohort that the hazard of viral suppression among children for whom ART was delayed did not differ from those with early initiation suggests that delaying ART in children initiating TB treatment has little or no effect on virological response to ART. Similarly, in a randomized controlled trial of immediate versus delayed ART initiation in adults with TB meningitis, the proportions with viral suppression in the two arms did not differ by 9 months of ART even though the median time to virological suppression was significantly longer in the deferred arm .
Our analysis had several strengths. First, we used data from one of the largest pediatric HIV clinics in sub-Saharan Africa, resulting in a large sample size of 573 HIV-infected children with TB of whom 461 initiated ART. Second, we used appropriate methods to measure the effect of delaying ART on mortality in an observational study, by controlling for a bias due to confounding by indication and selective drop out using a marginal structure model. We acknowledge that even the most appropriate statistical method may not adequately control for such confounding, especially in the routine clinic setting where initiation of ART is most likely closely linked to severity of disease. Third, by considering different cut-offs for the definition of delay, we were able to elicit a ‘dose–response’ relationship, with a tendency of increasing negative effect on both mortality and viral response to ART with increasing length of delay. Fourth all children in our cohort came from a single clinic. This is important as initiation of TB treatment and assessment of ART involves some subjectivity; having the same set of clinicians making those decisions, limits the variability. Fifth, only 8% of children were lost to follow-up after they were initiated on ART, and loss to follow-up was not associated with timing to ART initiation or the baseline covariates.
Our study also suffered from limitations. First, this is a retrospective analysis of data from an observational clinical cohort. As result, we do not have data on side effects, drug interactions, or IRIS, individual reasons for initiation or delay of ART and adherence to TB treatment or ART. Compared with asymptomatic children, children in more advanced stage of HIV disease at ART initiation have been shown to be more likely to be adherent to treatment . In our cohort, children with more advanced HIV disease at TB initiation were more likely to be initiated earlier on ART than those with less advanced disease. Differential adherence between children with less and those with more advanced HIV disease could therefore underlay some of our findings on hazard ratio for virological suppression . Second, children receiving rifampin were treated with efavirenz (if older than 3 years) or ritonavir-boosted lopinavir (if 3 years or younger) to achieve optimal virological outcomes [18,19]. These regimens are often not available in resource-poor countries, limiting the generalizability of our results. Third, due to the small number of children not severely immunosuppressed at baseline and particularly the limited number of events among them, it was not possible to fully assess effect of delaying ART on mortality or viral suppression among those children. It is possible that the effect of delaying ART on mortality differs by the level of immunosuppression.
In conclusion, results suggest that delaying ART initiation in children with TB for up to 30 days is not associated with increased mortality, while delaying for more than 60 days might decrease survival, especially in severely immunosuppressed children. Given the burden of TB in HIV-infected children, it is important that these results be confirmed in a randomized controlled trial.
M.Y. was an NIH Fogarty Fellow sponsored by grant DHHS/NIH/FIC 5 D43 TW01039–08 AIDS International Training and Research Program at the University of North Carolina at Chapel Hill (UNC) and received additional support from the UNC Center of Global Initiatives and the American International Health Alliance. Wits Echo in addition to the South African government funding received support from USAID/PEPFAR. These funding sources had no role in the design and conduct of this study. None of the authors had any conflict of interest. M.Y., T.M., H.M. have full access to the data. We are extremely grateful to the dedicated staff of Harriet Shezi Children's clinic for their amazing work.
1. UNAIDS, AIDS epidemic update
. 2009, Joint United Nations Programme on HIV/AIDS (UNAIDS) Geneva, Switzerland. p. 1–50.
2. Chakraborty R. HIV-1 infection in children: a clinical and immunologic overview. Curr HIV Res 2005; 3:31–41.
3. Newton SM, Brent AJ, Anderson S, Whittaker E, Kampmann B. Paediatric tuberculosis. Lancet Infect Dis 2008; 8:498–510.
4. Mukadi YD, Wiktor SZ, Coulibaly IM, Coulibaly D, Mbengue A, Folquet AM, et al
. Impact of HIV infection on the development, clinical presentation, and outcome of tuberculosis among children in Abidjan, Cote d'Ivoire. AIDS 1997; 11:1151–1158.
5. Marais BJ, Graham SM, Cotton MF, Beyers N. Diagnostic and management challenges for childhood tuberculosis in the era of HIV. J Infect Dis 2007; 196(Suppl 1):S76–S85.
6. Coovadia HM, Jeena P, Wilkinson D. Childhood human immunodeficiency virus and tuberculosis co-infections: reconciling conflicting data. Int J Tuberc Lung Dis 1998; 2:844–851.
7. Nelson LJ, Schneider E, Wells CD, Moore M. Epidemiology of childhood tuberculosis in the United States, 1993-2001: the need for continued vigilance. Pediatrics 2004; 114:333–341.
8. Nelson LJ, Wells CD. Global epidemiology of childhood tuberculosis. Int J Tuberc Lung Dis 2004; 8:636–647.
9. Elenga N, Kouakoussui KA, Bonard D, Fassinou P, Anaky MF, Wemin ML, et al
. Diagnosed tuberculosis during the follow-up of a cohort of human immunodeficiency virus-infected children in Abidjan, Cote d'Ivoire: ANRS 1278 study. Pediatr Infect Dis J 2005; 24:1077–1082.
10. Palme IB, Gudetta B, Bruchfeld J, Muhe L, Giesecke J. Impact of human immunodeficiency virus 1 infection on clinical presentation, treatment outcome and survival in a cohort of Ethiopian children with tuberculosis. Pediatr Infect Dis J 2002; 21:1053–1061.
11. Kampmann B, Tena-Coki GN, Nicol MP, Levin M, Eley B. Reconstitution of antimycobacterial immune responses in HIV-infected children receiving HAART. AIDS 2006; 20:1011–1018.
12. World Health Organization, Antiretroviral Therapy of HIV Infection in Infants and Children in Resource-Limited Settings: Towards Universal Access
, W.H. Organization, Editor. 2006, World Health Organization: Geneva, Switzerland.
13. Mofenson LM, Oleske J, Serchuck L, Van Dyke R, Wilfert C. Treating opportunistic infections among HIV-exposed and infected children: recommendations from CDC, the National Institutes of Health, and the Infectious Diseases Society of America. MMWR Recomm Rep 2004; 53:1–92.
14. South African National Tuberculosis Guidelines
15. Department of Health. Operational Plan for Comprehensive HIV and AIDS Care, Management and Treatment for South Africa
. 2003 19 November 2003 [cited 2009 April 2].
16. National Tuberculosis Control Programme Practical Guidelines
, Health, Editor. 2000.
17. Meyers T, Eley B, Leoning W. Guidelines for the management of HIV-infected children – 2005,
Health, Editor. 2005, Jacana Media.
18. la Porte CJ, Colbers EP, Bertz R, Voncken DS, Wikstrom K, Boeree MJ, et al
. Pharmacokinetics of adjusted-dose lopinavir–ritonavir combined with rifampin in healthy volunteers. Antimicrob Agents Chemother 2004; 48:1553–1560.
19. Ren Y, Nuttall JJ, Egbers C, Eley BS, Meyers TM, Smith PJ, et al
. Effect of rifampicin on lopinavir pharmacokinetics in HIV-infected children with tuberculosis. J Acquir Immune Defic Syndr 2008; 47:566–569.
21. WHO Multicentre Growth Reference Study Group, WHO Child Growth Standards: Length/Height-for-Age, Weight-for-Age, Weight-for-Length, Weight-for-Height, and Body Mass Index-for-Age: Methods and Development
., W. Press, Editor. 2006, WHO: Geneva, Switzerland.
22. Hernan MA, Brumback B, Robins JM. Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men. Epidemiology 2000; 11:561–570.
23. Robins JM, Hernan MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology 2000; 11:550–560.
24. Hernan MA, Lanoy E, Costagliola D, Robins JM. Comparison of dynamic treatment regimes via inverse probability weighting. Basic Clin Pharmacol Toxicol 2006; 98:237–242.
25. Cole SR, Hernan MA. Constructing inverse probability weights for marginal structural models. Am J Epidemiol 2008; 168:656–664.
26. Lin DY, Wei LJ, Ying Z. Checking the Cox model with cumulative sums of martingale-based residuals. Biometrika 1993; 80:557–572.
27. Lin DY, Wei LJ. The robust inference for the proportional hazards model. J Am Stat Assoc 1989; 84:1074–1078.
28. Velasco M, Castilla V, Sanz J, Gaspar G, Condes E, Barros C, et al
. Effect of simultaneous use of highly active antiretroviral therapy on survival of HIV patients with tuberculosis. J Acquir Immune Defic Syndr 2009; 50:148–152.
29. Zolopa A, Andersen J, Powderly W, Sanchez A, Sanne I, Suckow C, et al
. Early antiretroviral therapy reduces AIDS progression/death in individuals with acute opportunistic infections: a multicenter randomized strategy trial. PLoS ONE 2009; 4:e5575.
30. Violari A, Cotton MF, Gibb DM, Babiker AG, Steyn J, Madhi SA, et al
. Early antiretroviral therapy and mortality among HIV-infected infants. N Engl J Med 2008; 359:2233–2244.
31. Lawn SD, Myer L, Bekker LG, Wood R. Tuberculosis-associated immune reconstitution disease: incidence, risk factors and impact in an antiretroviral treatment service in South Africa. AIDS 2007; 21:335–341.
32. Manosuthi W, Kiertiburanakul S, Phoorisri T, Sungkanuparph S. Immune reconstitution inflammatory syndrome of tuberculosis among HIV-infected patients receiving antituberculous and antiretroviral therapy. J Infect 2006; 53:357–363.
33. Torok ME, Yen NTB, Chau TTH, Mai NTH, Phu NH, et al
. Randomised controlled trial of immediate versus deferred antiretroviral therapy in HIV-associated tuberculosis meningitis
. in 49th Interscience Conference on Antimicrobial Agents and Chemotherapy
. 2009. San Francisco, California.
34. Gibb DM, Goodall RL, Giacomet V, McGee L, Compagnucci A, Lyall H. Adherence to prescribed antiretroviral therapy in human immunodeficiency virus-infected children in the PENTA 5 trial. Pediatr Infect Dis J 2003; 22:56–62.
35. Manosuthi W, Kiertiburanakul S, Sungkanuparph S, Ruxrungtham K, Vibhagool A, Rattanasiri S, Thakkinstian A. Efavirenz 600 mg/day versus efavirenz 800 mg/day in HIV-infected patients with tuberculosis receiving rifampicin: 48 weeks results. AIDS 2006; 20:131–132.
Keywords:© 2010 Lippincott Williams & Wilkins, Inc.
ART initiation; children; HIV; South Africa; tuberculosis