Multi-drug resistant tuberculosis (MDR-TB), defined as resistance to at least isoniazid and rifampicin is a major cause of mortality in tuberculosis-HIV co-infected patients. The number of reported MDR-TB cases in South Africa rose from 6795 in 2008⁽¹⁾ to 10 085 in 2011,⁽²⁾ among the highest in the world. Of the 6795 reported MDR-TB cases in 2008, 28% (N=1866) occurred in the province of KwaZulu-Natal.⁽³⁾ 76% were co-infected with HIV. ^{(4, 5)}

The SAPiT trial⁽⁶⁾ showed that the initiation of antiretroviral therapy (ART) during tuberculosis treatment in patients with mostly drug susceptible tuberculosis and HIV co-infection reduced mortality by 56%. Based on these and other findings,^{(7, 8)} the 2010 South African National ART guidelines were changed; stating that patients co-infected with MDR-TB and HIV should be commenced on ART irrespective of CD4+ cell count.⁽⁹⁾

Empiric evidence for survival benefit of early initiation of ART in patients with MDR-TB is lacking. The SAPiT trial offers an opportunity to provide such information.⁽⁶⁾ The purpose of this subgroup analysis of the SAPiT trial was to assess the impact, through a randomized controlled trial design, of early ART initiation on survival, in HIV-infected patients with MDR-TB.

The SAPiT trial enrolled 642 HIV-tuberculosis co-infected patients and 489 (76%) had sputum culture and drug susceptibility testing for M tuberculosis performed. Prior to 2007, 56 patients had susceptibility testing performed when drug resistance was suspected. After 2007, when drug susceptibility testing was routinely available, 389 were tested for resistance at enrolment and 44 were tested at a follow-up visit as part of the routine testing. 23 of 489 (5%) patients were diagnosed with MDR-TB (Figure 1), of which 11 of 23 had no past history of tuberculosis. At baseline, patients with MDR-TB had lower Karnofsky scores⁽¹¹⁾, were more often WHO stage 4 and presented more frequently with extra-pulmonary tuberculosis (Table 1).

At baseline, MDR-TB patients in the combined integrated treatment arm (14/322 = 4.3%) and the sequential treatment arm (9/167 = 5.4%; Table 1) were similar. Of the 14 patients in the combined integrated treatment arm, nine were diagnosed with MDR-TB on their enrolment sputum specimen, one during the study (but had no enrolment susceptibility testing) and four had susceptible M tuberculosis at enrolment and subsequently developed MDR-TB. Of the nine patients in the sequential treatment arm, two were diagnosed with MDR-TB on their enrolment sputum specimen, 3 during the course of treatment (but had no enrolment susceptibility testing) and four had susceptible M tuberculosis at enrolment and subsequently developed MDR-TB. One of the 14 patients in the combined integrated treatment arm and none in the sequential treatment arm had extensively drug resistant tuberculosis.

Seven of the 23 (30%) MDR-TB patients died; two in the combined integrated treatment arm and five in the sequential treatment arm. Death rates were similar in the two treatment arms until 6 months (Figure 2). After 6 months, more patients died in the sequential treatment arm compared to the integrated treatment arm (Figure 2). The mortality rate in MDR-TB patients was 11.9 per 100 person-years (95%CI: 1.4–42.8) in the combined integrated treatment arm compared to 56.0 per 100 person-years (95%CI: 18.2–130.8) in the sequential treatment arm (HR: 0.14; 95% CI: 0.02–0.94; P=0.04, adjusting for baseline CD4+ cell count and whether MDR-TB treatment was initiated). The mortality rate in MDR-TB patients was approximately 4-fold higher (P<0.001) than mortality in patients who did not have MDR-TB (Table 2a). Irrespective of whether patients received specific treatment for MDR-TB, mortality rates were lower in the combined integrated treatment arm (Table 2b).

Of the 23 patients with MDR-TB, 14 (61%) started MDR-TB treatment, 64% (9/14) in the combined integrated treatment arm, and 56% (5/9) in the sequential treatment arm (Figure 3). The overall mortality rate in patients who initiated MDR-TB treatment was 10.8 per 100 person-years (95%CI: 1.3–39.0) and overall mortality rate in patients who did not initiate MDR-TB treatment was 68 per 100 person-years (95%CI: 22.3–160.4, P=0.03). After adjustment for baseline CD4+ cell count, patients with MDR-TB who had received MDR-TB treatment had a 93% lower mortality rate at 18 months post-randomization, compared to patients who had not initiated MDR-TB treatment (P=0.02). Initiating MDR-TB treatment (in those patients who survived long enough to receive their MDR-TB diagnosis) or being in the combined integrated treatment arm reduced the risk of dying by 97%, adjusting for baseline CD4+ cell count (P=0.003).

Seven patients initiated MDR-TB treatment within 6 months of starting standard tuberculosis treatment and only 1/7 (14%) died (Figure 3). In contrast, 16 patients either did not start MDR-TB treatment or started MDR-TB treatment more than 6 months after standard tuberculosis treatment, and 6/16 (38%) died (RR 0.38; 95% CI: 0.06–2.60; P=0.37). In this latter group, 1/8 (13%) patients in the combined integrated treatment arm died and 5/8 (63%) patients in the sequential treatment arm died (RR: 0.20; 95%CI: 0.03–1.35; P=0.12).

In the combined integrated treatment arm, the median time from ART to MDR-TB treatment was 113 (range: 40–239) days(n=6) and in the Sequential treatment arm it was 258 days - only one patient started ART before MDR-TB treatment (Figure 3)

The reasons for not commencing MDR-TB treatment in the combined integrated arm were: lost to follow up (n=2), death(n=1) and response to 1st line TB drugs before commencing treatment for MDR-TB(n=2); in the sequential arm, 4 patients died before commencing treatment for MDR-TB.

In the combined integrated treatment arm, one patient was diagnosed with MDR-TB posthumously; he had not received MDR-TB treatment but had received ART for 2.5 months before he died from severe diarrhoea. The other death in this arm was due to suicide and occurred after 4 months of MDR-TB treatment and 5 months of ART. In the sequential treatment arm, four patients were diagnosed with MDR-TB posthumously; none had received MDR-TB treatment while one had received ART for 3 months before he died; the fifth patient had not initiated ART and had been treated for MDR-TB for 19 days prior to death. Clinical outcomes of tuberculosis treatment (some on standard tuberculosis treatment and others on MDR-TB treatment) were available in 19 of the 23 MDR-TB patients; four were lost to follow-up in the combined integrated treatment arm. At 18 months, seven patients had died, six were still receiving MDR-TB treatment (four in the combined integrated arm and two in the sequential treatment arm) and a further six were cured [four in the combined integrated treatment arm and two in the sequential treatment arm (P=1.00)].

At 12 months of follow-up, 78% (N=9) in the combined integrated treatment arm and all four in the sequential treatment arm had undetectable viral loads. The mean increase in CD4+ cell count was 108 and 89.8 cells/mm³ in the combined integrated and sequential treatment arms respectively (Table 3).

Initiation of ART during tuberculosis treatment in patients with MDR-TB was associated with an 86% reduction in mortality. This survival benefit associated with ART initiation was evident even in patients who had not initiated appropriate MDR-TB treatment. Similar survival benefits have also been observed in retrospective studies among XDR-TB patients who were co-infected with HIV and received ART.^{(12, 13)}

The long laboratory delays in the diagnosis of MDR-TB delays initiation of appropriate treatment and contributes to high mortality among MDR-TB patients. Patients with MDR-TB in this study experienced a 4-fold higher rate of mortality compared to those with non MDR-TB. Similar high mortality rates have been observed in patients co-infected with MDR-TB and HIV in rural KwaZulu-Natal, where 40% of MDR-TB cases died within 30 days of sputum collection, most dying before the diagnosis of MDR-TB was confirmed.⁽⁵⁾ A retrospective review conducted in Thailand showed that HIV and MDR-TB co-infected patients were 12-times more likely to die than patients without MDR-TB.⁽¹⁴⁾

A consequence of the diagnostic and treatment initiation delays resulted in MDR-TB confirmation posthumously in this study. This is not unique. In a post mortem study of 240 individuals (94% HIV-infected), 17% had MDR TB that had not been recognized clinically.⁽¹⁵⁾ Therefore, initiating ART during tuberculosis treatment provides an opportunity to improve survival rates until MDR-TB can be confirmed and appropriate treatment instituted. It is quicker to diagnose HIV, institute ART and improve survival until MDR-TB can be diagnosed and MDR-TB treatment initiated. If MDR-TB can be diagnosed early and both ART and MDR-TB treatment initiated mortality is likely to be reduced even further.^{(12, 13, 16)}

Although MDR-TB patients in this study were more often categorized as WHO stage 4 and had extra-pulmonary tuberculosis, these criteria were insufficient indicators of which patients had MDR-TB. Immune markers such as CD4+ cell count and viral load were similar among MDR-TB patients and those with drug susceptible tuberculosis. Other studies have also shown that the clinical and radiological presentation for patients with MDR-TB and drug susceptible tuberculosis is similar.⁽¹⁷⁾

Another factor contributing to the delay in investigation of possible MDR-TB is that immune reconstitution inflammatory syndrome (IRIS) was considered for MDR-TB patients in the combined integrated treatment arm. In addition to a higher index of suspicion for drug resistance, there is an urgent need for earlier case finding, better diagnostics, as well as integration of tuberculosis and HIV care and treatment.⁽¹⁸⁾

The MDR-TB case load in KwaZulu-Natal has increased 4-fold between 2002 and 2004, with resultant larger numbers of primary MDR-TB cases presenting for tuberculosis treatment. Extremely high rates of primary MDR-TB have been reported from specific parts of KwaZulu-Natal.^{(19, 20)} A high proportion of MDR-TB patients in these areas with high rates of HIV co-infection had no prior history of tuberculosis treatment, in contrast to a recent multi-country study which showed that a history of previous tuberculosis was the strongest risk factor for MDR-TB.^{(21, 22)} It is also important to note that eight patients in our study initially had confirmed drug susceptible tuberculosis and subsequently developed MDR-TB. This sequence suggests either selection of resistant organisms while on tuberculosis treatment or super-infection with MDR-TB,⁽²³⁾ emphasizing the critical importance of strengthening both treatment adherence and infection control in tuberculosis programs.

In the SAPiT study participants in each arm were monitored and evaluated using the same standardized criteria and practices to minimize bias during the study implementation thus preserving external validity. One of the main limitations of this study is the small sample size. Despite this, the results were statistically significant due to the large effect size observed. This was a post hoc analysis and the study was not designed to screen patients for MDR-TB and randomize these patients to early or sequential ART initiation. Newer technology now makes the earlier diagnosis of drug resistance possible.^{(24, 25)} Even though randomization was not stratified on MDR-TB status, the participants who had undetected MDR-TB at baseline constituted a similar proportion of patients in each treatment arm. Thus studying this subgroup of patients preserves the benefits of randomization. When the SAPiT trial was initiated, it was not possible to conduct such a study because of the delays in obtaining drug susceptibility results and many of the patients died (5 of the 7 deaths) even before they were diagnosed with MDR-TB.

The use of all-cause mortality is also a limitation. For example, the death by suicide is treated as an equivalent primary endpoint to tuberculosis meningitis. The SAPiT trial included 18 months of follow up but some patients with MDR-TB were diagnosed late during the study and their clinical outcomes were not known as they were still on the required long course of MDR-TB treatment.

Despite these limitations, our study provides empirical evidence from a randomized controlled trial of a survival benefit when ART is initiated early in patients with MDR-TB. Even though there may be concerns about patient acceptability of integrated therapy, retrospective studies have shown a decreased mortality with integrated TB-HIV treatment in MDR and XDR-TB despite protracted and complex TB treatment regimens.^{(5, 12)}

In this study, although MDR-TB mortality was high compared to non-MDR-TB cases, survival was significantly improved when ART was initiated, even among those who had not yet been initiated on MDR-TB treatment. Early ART initiation is desirable in settings with high MDR-TB prevalence, especially in sick patients and those with low CD4+ cell counts⁽²⁶⁾ because treatment initiation even before MDR-TB is diagnosed has a survival benefit.