The global burden of tuberculosis is closely linked to the HIV epidemic. In 2000, 7% to 12% of the estimated 8.3 million new tuberculosis cases worldwide and 26% of new cases in the United States were attributed to HIV infection.1 HIV dramatically increases the risk of progression from latent Mycobacterium tuberculosis infection to active disease,2,3 and HIV-infected tuberculosis patients have a high annual incidence of death and opportunistic infections,4-8 in part, because tuberculosis accelerates HIV disease progression.9-11 As a result, tuberculosis caused 11% of all adult AIDS deaths in the year 2000.1 The risk of death is substantially greater among persons with <200 CD4+ lymphocytes/mm3 compared with persons with higher CD4+ lymphocyte counts.5,7,8,11
Highly active antiretroviral therapy (HAART) dramatically decreases the incidence of opportunistic infection and death in patients with advanced HIV infection, including those with tuberculosis.12-17 Tuberculosis Trials Consortium (TBTC) Study 23 demonstrated a 12-month mortality rate of 4.9% in HIV-infected tuberculosis patients during the HAART era, which had decreased from 20% in the pre-HAART era; the risk of mortality plus opportunistic infections decreased from 38.9% to 15.4%.18,19
HAART in HIV-infected tuberculosis patients is complicated by an increased risk of drug toxicity and immune reconstitution inflammatory syndrome (IRIS),20,21 particularly in persons in whom HAART is started within 2 months of antituberculosis therapy.22,23 The risk of these toxicities is greatest in persons with <200 CD4+ lymphocytes/mm3, but so is the decrease in mortality risk. Given these competing risks and benefits, the optimal timing of HAART initiation in relation to the start of antituberculosis therapy is unknown, particularly in persons with <200 CD4+ lymphocytes/mm3. The World Health Organization (WHO) states that in persons with <200 CD4+ lymphocytes/mm3, it may be necessary to start antiretroviral therapy concomitantly with antituberculosis therapy. The WHO provides 3 options: (1) defer antiretroviral therapy until completion of antituberculosis therapy; (2) defer antiretroviral therapy until completion of the initial phase of antituberculosis therapy, and then use ethambutol plus isoniazid in the continuation phase; and (3) treat tuberculosis with a rifampicin-containing regimen plus efavirenz-based antiretroviral therapy.24 There have been no randomized clinical trials to assess this issue. We therefore conducted a decision analysis to gain insight into the optimal timing of HAART initiation in HIV-infected persons receiving antituberculosis therapy.
The software program DATA 3.5 by Treeage (Williamstown, MA) was used to perform the decision analysis; the decision trees are illustrated in Figure 1. Standard decision analysis techniques were used.25-28 Hypothetical cohorts of 1000 HIV-infected adults with tuberculosis and <200 CD4+ lymphocytes/mm3 were studied. Data were obtained from studies in which patients received standard rifamycin-based directly observed antituberculosis therapy for 6 months. This included 8 to 10 weeks of standard 4-drug induction therapy with isoniazid, rifampin or rifabutin, pyrazinamide, and ethambutol, followed by isoniazid and rifampin or rifabutin for 4 months. HAART consisted of a regimen that included 2 nucleoside reverse transcriptase inhibitors plus an HIV-1 protease inhibitor, nonnucleoside reverse transcriptase inhibitor, or third nucleoside reverse transcriptase inhibitor.
The analysis was run under 3 conditions: early HAART (HAART started during the first 2 months of antituberculosis therapy); deferred HAART (HAART started during months 2 through 6 of antituberculosis therapy); and no HAART (no HAART during antituberculosis therapy). The extent to which patients in the no-HAART group received HAART after completion of antituberculosis therapy varied in the studies cited. Two outcomes were assessed. The first was all-cause mortality within 12 months of starting antituberculosis therapy. The second outcome, referred to as the “combined outcome,” included all-cause mortality, new AIDS-defining illness, severe IRIS, and severe (defined as grade 3) or life-threatening (defined as grade 4) drug toxicity within 12 months of starting antituberculosis therapy. Drug toxicities included elevated hepatic transaminases, leukopenia, anemia, and rash.
IRIS was defined as a transient worsening of tuberculosis-related disease after initiation of an appropriate antituberculosis regimen or HAART. Severe IRIS was defined as any of the following: worsening of neurologic symptoms because of meningitis or expansion of an intracranial tuberculoma; tracheal narrowing attributable to compressive adenopathy; expanding lymphadenitis requiring surgical drainage; need for therapeutic paracentesis, thoracentesis, or pericardiocentesis; or permanent discontinuation of HAART or antituberculosis therapy. The proportion of IRIS cases that progressed to severe IRIS was termed the IRIS-related severe event rate. The proportion of IRIS cases that progressed to death was termed the IRIS-related mortality rate. The proportion of severe or life-threatening (grade 3 or 4) drug toxicity cases that progressed to death was termed the drug toxicity-related mortality rate.
The probability estimates used in the analysis are provided in Table 1. All estimates were based on data from primary source studies. When data from more than 1 study were available, the baseline probability estimate was determined via random effects inverse of variance analysis of the combined data, using the 95% confidence intervals as the range of probability estimates. When just 1 study was available, 95% confidence intervals of the event rate were reported as the range of probability estimates.
For the primary outcome, mortality was assigned a utility of 0 and survival was assigned a utility of 1. Based on the probability at each branch in the decision tree, the program determined the expected utility of each of the 3 treatment strategies. The treatment strategy with the highest expected utility (ie, numeric value) represented the therapeutic option with the most favorable outcome. One-way and 2-way sensitivity analyses were performed using the range of probability estimates to identify the impact, if any, on the favored treatment strategy.
For the outcome of all-cause mortality, there were 33, 48, and 147 deaths in the early HAART, deferred HAART, and no-HAART groups, respectively, among hypothetical cohorts of 1000 persons (Table 2). For the combined outcome of mortality, new AIDS-defining illness, severe IRIS, and severe or life-threatening drug toxicity, there were 497 events in the early HAART group and 501 events in the deferred HAART and no-HAART groups. Most of the events in the early and deferred HAART groups were attributable to drug toxicity, whereas most events in the no-HAART group were deaths and AIDS-defining illnesses (see Table 2). Although the early HAART group had the lowest number of deaths and AIDS-defining illnesses, it also had the highest number of severe IRIS events.
The causes of death are listed in Table 3. Most of the deaths were HIV-related, but the proportion of HIV-related deaths varied according to the timing of HAART initiation. HIV-related death accounted for 58% of the deaths in the early HAART group but 94% of the deaths in the no-HAART group.
One-way sensitivity analyses were performed for all variables included in the mortality analysis (Table 4A). The number of deaths that could occur based on the range of probability estimates for each variable is provided. In the early HAART group, the variables with the greatest influence on the results (based on the size of the range of deaths and listed from greatest to least) were HIV mortality, IRIS-related mortality, tuberculosis relapse or treatment failure, and drug toxicity-related mortality. For the first 3 variables, there was potential overlap in the number of deaths compared with ranges for the same variables in the deferred HAART group. Assuming that all other probability estimates were at baseline, the event rate thresholds in the early HAART group (above which deferred HAART would be favored) were 0.035, 0.046, and 0.107 for HIV-related mortality, IRIS-related mortality, and tuberculosis relapse or treatment failure, respectively. No threshold was identified for drug toxicity-related mortality; early HAART was always favored. In the deferred HAART group, the variables with the greatest influence on the number of deaths were HIV mortality, tuberculosis relapse or treatment failure, and drug toxicity-related mortality. In the no-HAART group, HIV mortality had the greatest influence on overall mortality. The range of probability estimates for HIV mortality resulted in overlap in the possible number of deaths in the early HAART, deferred HAART, and no-HAART groups. There were no other clinical events that had probability estimate ranges affecting the results between deferred HAART and no HAART.
Two-way sensitivity analyses were then performed for all variables that had a significant impact in the 1-way sensitivity analyses (see Table 4B). In addition, sensitivity analyses of IRIS and drug toxicity rates were performed, because these 2 events are important in clinical decision making regarding HAART initiation. Comparing early versus deferred HAART, the combinations of clinical event probabilities that had the greatest impact on the outcome were HIV mortality plus mortality attributable to IRIS or drug toxicity and HIV mortality plus tuberculosis relapse or treatment failure. The same combinations had the greatest influence on the number of deaths in the deferred HAART and no-HAART groups. For the early and deferred HAART groups, 2 additional combinations were important: IRIS-related mortality plus the IRIS event rate or the drug toxicity-related mortality rate. Of note, however, the analysis favored early HAART initiation over deferred HAART (35 vs. 48 deaths), even with the highest rates of IRIS (70%) and drug toxicity (56%) in the early HAART group.
For HIV-infected tuberculosis patients with <200 CD4+ lymphocytes/mm3, the greatest mortality benefit occurred when HAART was started during the first 2 months of antituberculosis therapy. The mortality benefit was most pronounced when comparing either HAART group with the no-HAART group, but mortality was also lower with early HAART initiation compared with deferred HAART (33 vs. 48 deaths). The analysis included all causes of mortality assessed: HIV, tuberculosis, IRIS, and drug toxicity. Because survival is the most important clinical outcome, the impact of treatment strategies on mortality is of paramount importance.
When determining the optimal time to initiate HAART, however, one must also take into account AIDS-defining illnesses and the adverse effects of nonfatal but severe IRIS and drug toxicity. Of note, for the combined endpoint, which included all the previous causes plus mortality, the benefit was also greatest when HAART was started during the first 2 months of antituberculosis treatment. Although the difference in the absolute number of events between the 3 groups was small, the greatest overall benefit was with early HAART initiation. In addition, the contribution of each of the 4 events to the overall event number varied substantially among the 3 groups. The most severe endpoints of mortality and AIDS-defining illness were lower in early versus deferred HAART and were substantially lower in both HAART groups compared with the no-HAART group. The number of severe drug toxicity events was substantially higher in the early and deferred HAART groups than in the no-HAART group. It should be noted that the mortality rate associated with severe drug toxicity in these patients is low, however.19
These findings demonstrate that the overall benefits of early HAART outweigh the adverse effects. This argues for the initiation of antiretroviral therapy within 2 months of antituberculosis therapy in HIV-infected tuberculosis patients with <200 CD4+ lymphocytes/mm3.
The results of the baseline analysis do not necessarily account for specific scenarios that clinicians might encounter when determining the best time to initiate HAART in an individual patient, however. It is also important to determine the clinical events that are most influential in the timing of treatment so as to identify situations in which it may be more beneficial to defer HAART initiation. The sensitivity analysis is helpful for both of these situations.
In the 1-way sensitivity analysis, the variable with the greatest influence on the results was HIV-related mortality. Based on the data from the clinical study used for this analysis, HIV mortality rates were lowest when HAART was initiated earliest. Given the profound impact of HAART on lowering mortality risk, this is not surprising. Although more data on the mortality rate according to early versus deferred HAART initiation are needed, it is unlikely that there would be differences in the rank order of HIV mortality according to early versus deferred HAART versus no HAART initiation. If the HIV-related mortality rate exceeded 3.5% in the early HAART arm and the baseline rate in the deferred HAART arm remained 3.0%, however, there would be more deaths than in the baseline deferred HAART arm.
The second most important variable in the analysis was the IRIS-related mortality rate. In a setting with a high IRIS mortality rate with early HAART (eg, >4.6%, the threshold value), such as tuberculosis meningitis or pericarditis,29 the number of deaths in the early HAART group would exceed that seen in the deferred HAART group, and deferring HAART until after 2 months of antituberculosis therapy (but before the completion of antituberculosis therapy) would be preferred. The third most important variable for assessing early versus deferred HAART was the tuberculosis failure and relapse rate. More data (ie, clinical trials) on this rate according to the timing of HAART initiation are needed. If they confirm the data available to date,30 however, this would provide further support for early initiation. Of note, tuberculosis relapse risk in HIV-infected persons seems to be attributable primarily to a low CD4+ lymphocyte count.31 Therefore, with earlier HAART initiation and concomitant CD4+ lymphocyte count increases, one would expect a lower relapse risk.
In the 2-way sensitivity analyses, HIV mortality continued to have the greatest impact on outcome (see Table 4). In addition, however, with high rates of IRIS and IRIS-related mortality among persons who initiate HAART early, there could potentially be more deaths than among those who defer HAART initiation. High rates of grade 3 or 4 drug toxicity plus high rates of toxicity-related mortality could also result in a large number of deaths, and therefore favor deferred HAART initiation. Even with the highest rates of IRIS (70%) and drug toxicity (56%) in the early HAART group, however, early HAART initiation would be preferred over deferred HAART (35 vs. 48 deaths). High rates of IRIS mortality plus drug toxicity mortality with early HAART could favor the strategy of deferred HAART initiation.
The no-HAART group pertained to patients who did not receive HAART during antituberculosis therapy, but they could have received HAART after completion of antituberculosis therapy. Thus, this was not a group that never received HAART.
The clinical and public health implications of this decision analysis for the developing and developed world are that HIV treatment should be integrated into tuberculosis care so that HIV-infected tuberculosis patients with <200 CD4+ lymphocytes/mm3 can receive concomitant antituberculosis and antiretroviral therapy and be monitored appropriately for adverse events.
There are several limitations of this decision analysis. The most important probability estimate for influencing outcomes in our analysis was the rate of HIV-related mortality. TBTC Study 23 was the first study to compare prospectively HIV-related mortality and AIDS-defining illness in persons with tuberculosis and <200 CD4+ lymphocytes/mm3 who were treated with HAART versus those who were not treated with HAART. This study was limited by its small sample size and lack of randomization of HAART use, however.19,30 We accounted for this by performing a sensitivity analysis with a broad range of possible event rates. In addition, there are limited data on drug toxicity rates and associated mortality rates when HAART is started during the first 2 months of antituberculosis therapy. We were able to address this problem via the 1-way and 2-way sensitivity analyses. It would be helpful to have specific information on drug discontinuation rates for antiretroviral therapy and antituberculosis therapy according to the timing of HAART initiation. Discontinuing such complex treatment regimens might increase HIV-related mortality or tuberculosis relapse and treatment failure. Finally, our probability estimates took into account compliance with HAART and antituberculosis therapy, because the estimates were based on data obtained in the clinical setting. If adherence to either regimen were to differ substantially from that in the studies cited, however, the risks and benefits of such therapy might differ, and that could affect the outcome of the analysis.
In our analysis, HAART started during the first 2 months of antituberculosis treatment decreased total mortality and the combined endpoint of mortality, new AIDS-defining illnesses, severe IRIS, and severe drug toxicity in HIV-infected tuberculosis patients with <200 CD4+ lymphocytes/mm3, compared with initiating HAART after 2 months of antituberculosis therapy. The benefit of HAART persisted even when extremely high rates of drug toxicity or IRIS were assumed. High rates of IRIS-related mortality with early HAART initiation would favor deferring HAART until at least 2 months after initiation of antituberculosis therapy, however. These results support early initiation of HAART during tuberculosis treatment in HIV-infected tuberculosis patients with <200 CD4+ lymphocytes/mm3, except in persons at high risk of IRIS-related mortality.
The authors thank Keertan Dheda, Awal Khan, William Burman, and Andrew Vernon for providing data for these analyses. They are grateful to the Antiretroviral Therapy Cohort Collaboration for providing estimates of disease progression according to CD4+ lymphocyte count.
1. Corbett EL, Watt CJ, Walker N, et al. The growing burden of tuberculosis
: global trends and interactions with the HIV epidemic. Arch Intern Med
2. Selwyn PA, Hartel D, Lewis VA, et al. A prospective study of the risk of tuberculosis
among intravenous drug users with human immunodeficiency virus infection. N Engl J Med
3. Bucher HC, Griffith LE, Guyatt GH, et al. Isoniazid prophylaxis for tuberculosis
in HIV infection: a meta-analysis of randomized controlled trials. AIDS
4. Small PM, Schecter GF, Goodman PC, et al. Treatment of tuberculosis
in patients with advanced human immunodeficiency virus infection. N Engl J Med
5. Perriens JH, St. Louis ME, Mukadi YB, et al. Pulmonary tuberculosis
in HIV-infected patients in Zaire: a controlled trial of treatment for either 6 or 12 months. N Engl J Med
6. Connolly C, Reid A, Davies G, et al. Relapse and mortality among HIV-infected and uninfected patients with tuberculosis
successfully treated with twice weekly directly observed therapy in rural South Africa. AIDS
7. Ackah AN, Coulibaly D, Digbeu H, et al. Response to treatment, mortality, and CD4 lymphocyte counts in HIV-infected persons with tuberculosis
in Abidjan, Cote d'Ivoire. Lancet
8. Chaisson RE, Clermont HC, Holt EA, et al. Six-month supervised intermittent tuberculosis
therapy in Haitian patients with and without HIV infection. Am J Respir Crit Care Med
9. Badri M, Ehrlich R, Wood R, et al. Association between tuberculosis
and HIV disease progression in a high tuberculosis
prevalence area. Int J Tuberc Lung Dis
10. Whalen CC, Horsburgh CR, Hom D, et al. Accelerated course of human immunodeficiency virus infection after tuberculosis
. Am J Respir Crit Care Med
11. Chaisson RE, Schechter GF, Theuer CP, et al. Tuberculosis
in patients with the acquired immunodeficiency syndrome: clinical features, response to therapy, and survival. Am Rev Respir Dis
12. Egger M, May M, Chene G, et al. Prognosis of HIV-1-infected patients starting highly active antiretroviral therapy
: a collaborative analysis of prospective studies. Lancet
13. Murphy EL, Collier AC, Kalish LA, et al. Highly active antiretroviral therapy
decreases mortality and morbidity in patients with advanced HIV disease. Ann Intern Med
14. Mocroft A, Vella S, Benfield TL, et al. Changing patterns of mortality across Europe in patients infected with HIV-1. EuroSIDA Study Group. Lancet
15. Leonard MK, Larsen N, Drechsler H, et al. Increased survival of persons with tuberculosis
and human immunodeficiency virus infection, 1991-2000. Clin Infect Dis
16. Palella FJ Jr, Delaney KM, Moorman AC, et al. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. HIV Outpatient Study Investigators. N Engl J Med
17. Garcia DO, Martinez-Gonzalez MA, Cayla JA, et al. Influence of highly active anti-retroviral therapy (HAART) on the natural history of extra-pulmonary tuberculosis
in HIV patients. Int J Tuberc Lung Dis
18. El-Sadr WM, Perlman DC, Matts JP, et al. Evaluation of an intensive intermittent-induction regimen and duration of short-course treatment for human immunodeficiency virus-related pulmonary tuberculosis
. Clin Infect Dis
19. Burman W, Benator D, Vernon A, et al. Acquired rifamycin-resistance with twice-weekly treatment of HIV-related tuberculosis
. Am J Respir Crit Care Med
20. Burman WJ, Jones BE. Treatment of HIV-related tuberculosis
in the era of effective antiretroviral therapy
. Am J Respir Crit Care Med
21. Narita M, Ashkin D, Hollender ES, et al. Paradoxical worsening of tuberculosis
following antiretroviral therapy
in patients with AIDS
. Am J Respir Crit Care Med
22. Navas E, Martin-Davila P, Moreno L, et al. Paradoxical reactions of tuberculosis
in patients with the acquired immunodeficiency syndrome who are treated with highly active antiretroviral therapy
. Arch Intern Med
23. Breen RA, Smith CJ, Bettinson H, et al. Paradoxical reactions during tuberculosis
treatment in patients with and without HIV co-infection. Thorax
24. World Health Organization. TB/HIV. A Clinical Manual
. 2nd ed. WHO/HTM/TB/ 2004 329. Geneva, Switzerland, 2004.
25. Pauker SG, Kassirer JP. Decision analysis. N Engl J Med
26. Inadomi JM. Decision analysis and economic modelling: a primer. Eur J Gastroenterol Hepatol
27. Richardson WS, Detsky AS. Users' guides to the medical literature. VII. How to use a clinical decision analysis. A. Are the results of the study valid? Evidence-Based Medicine Working Group. JAMA
28. Richardson WS, Detsky AS. Users' guides to the medical literature. VII. How to use a clinical decision analysis. B. What are the results and will they help me in caring for my patients? Evidence Based Medicine Working Group. JAMA
29. Centers for Disease Control and Prevention. Tuberculosis
morbidity-United States, 1997. MMWR Morb Mortal Wkly Rep
30. Burman W, Vernon A, Khan A, et al. Timing of antiretroviral therapy
during treatment of HIV-related tuberculosis
[abstract 846]. Presented at: 43rd Annual Meeting of the Infectious Diseases Society of America, San Francisco, CA, 2005.
31. Nettles RE, Mazo D, Alwood K, et al. Risk factors for relapse and acquired rifamycin resistance after directly-observed tuberculosis
treatment: a comparison by HIV serostatus and rifamycin. Clin Infect Dis
32. Dean GL, Edwards SG, Ives NJ, et al. Treatment of tuberculosis
in HIV-infected persons in the era of highly active antiretroviral therapy
33. Dheda K, Lampe FC, Johnson MA, et al. Outcome of HIV-associated tuberculosis
in the era of highly active antiretroviral therapy
. J Infect Dis
34. Wendel KA, Alwood K, Gachuhi R, et al. Paradoxical worsening of tuberculosis
in HIV-infected persons. Chest
35. Olalla J, Pulido F, Rubio R, et al. Paradoxical responses in a cohort of HIV-1-infected patients with mycobacterial disease. Int J Tuberc Lung Dis
36. Orlovic D, Smego RA Jr. Paradoxical tuberculous reactions in HIV-infected patients. Int J Tuberc Lung Dis
37. Ungo JR, Jones D, Ashkin D, et al. Antituberculosis drug-induced hepatotoxicity. The role of hepatitis C virus and the human immunodeficiency virus. Am J Respir Crit Care Med
38. American Thoracic Society, Centers for Disease Control and Prevention, Infectious Diseases Society of America. Treatment of tuberculosis
. Am J Respir Crit Care Med
39. Sterling TR, Alwood K, Gachuhi R, et al. Relapse rates after short-course (6-month) treatment of TB in HIV-infected and -uninfected persons. AIDS
40. Kassim S, Sassan-Morokro M, Ackah A, et al. Two-year follow-up of persons with HIV-1- and HIV-2-associated pulmonary tuberculosis
treated with short-course chemotherapy in West Africa. AIDS
41. Espinal MA, Kim SJ, Suarez PG, et al. Standard short-course chemotherapy for drug-resistant tuberculosis
. Treatment outcomes in 6 countries. JAMA