This evaluation provides the first evidence from India regarding the operational challenges of delivering WHO-recommended mortality-reducing TB-HIV interventions under general field conditions. It also provides information on the daunting problem of the high mortality of HIV-infected TB patients. In this resource-limited setting, nearly 1 in 5 HIV-infected individuals diagnosed with TB died during TB treatment. While previous studies have reported high mortality among HIV-infected persons with tuberculosis 
, the high death rate we observed in Andhra Pradesh was particularly concerning given the provision of free CPT and reasonable levels of ART uptake. These findings imply that very high levels of ART uptake may be required to substantively reduce tuberculosis case fatality rates.
With no additional resources or efforts beyond provider trainings and monitoring by regular HIV and TB programme staff, more than 95% of all detected HIV-infected TB patients were initiated on CPT. Mixed adherence may have reduced the effectiveness of CPT in this population, as just half of patients collected >60% of their cotrimoxazole. With increased awareness of CPT, adherence may improve; additional measures the programmes should consider include use of peer-counselor networks and non-governmental organizations to support and encourage patients and providers.
We found that patients provided ART during TB treatment had less than half the risk of death as those not provided ART. This finding was independent of sex or previous TB treatment history, and is consistent with findings from several other settings 
. This evidence from India strongly reinforces the interpretation that ART is the single most potent intervention available for public health programmes to reduce death during TB treatment among HIV-infected individuals. Regardless of CPT availability, in this setting the linkages to ART were arguably inefficient. Of the 600 HIV-infected TB patients not already on ART, despite evidence of referral, only 229 (38%) started ART during TB treatment. Even when we included patients already on ART before TB treatment, in total only 50% of TB patients received any ART during TB treatment. While all of the remaining patients may not have been immediately eligible for ART, given the strong protection against death during TB treatment, the TB and HIV programmes must urgently seek improvements in ART referral efficiency and treatment uptake.
HIV-infected males experienced higher death rates than females; the reasons for this finding are unclear. Overall case-fatality rates among males with tuberculosis in India are modestly higher than in females, and postulated reasons for higher mortality among males include differences in treatment adherence, age distribution, smoking behavior, and comorbidities 
. Regardless, ART was associated with similar protection from mortality in males and females alike. The 104 patients registered as retreatment cases suffered from poor treatment success (50%) and high case-fatality (30%). No information was available regarding patient co-morbidities or drug resistance, but the very high death rates among previously-treated patients urgently warrants further exploration.
Our evaluation was subject to several important limitations. This was a retrospective evaluation of a demonstration conducted under routine programme conditions; available information was limited to that recorded on standard patient records which had been modified to include additional HIV-related information. With ascertainment of HIV-infection among TB patients based on selective referral of those persons with HIV risk factors, the patients identified might not be representative of the larger population of HIV-infected TB patients. All 3 districts practiced the same selective referral policy, but in Vizianagaram district a simultaneous population-based survey of HIV infection among TB patients provides insight into the effectiveness of the ascertainment of HIV infection 
. In Vizianagaram, the survey estimated a 6.5% (95% CI 4.5–9.4%) prevalence of HIV infection in TB patients, which would imply that there were approximately 116 (81–168) HIV-infected TB patients among the cohort of 1788 registered from Vizianagaram in this study period. This evaluation documented 95 HIV-infected TB patients from Vizianagaram, or 82% of the expected number, suggesting that the ascertainment as sufficiently efficient to expect that these results could be generalized to all HIV-infected TB patients in these districts.
Treatment outcomes were recorded by many providers, and were not independently validated. Information about other biomedical risk factors for death that may have confounded this evaluation, such as severity of TB disease, drug resistance, severity of immune suppression, or comorbidities, were not available. No ‘control’ areas without CPT were assessed because we did not seek to evaluate the efficacy of CPT as an intervention, and observational analysis of the association of CPT and mortality were not possible since almost all patients were exposed to CPT during TB treatment. The efficacy of CPT, however, has been amply demonstrated in multiple settings. The recording of ART information was dependent on feedback from the ART centre. Some patients may not have had ART consumption recorded on their TB treatment card, so we may have misclassified these patients as not provided ART, and thus underestimated the protective effect of ART in this population. We were also unable to collect adequate information about sub-type of extrapulmonary TB, patient CD4 count, or the specific timing of ART initiation. Future studies should seek to collect this information to better understand the finding of lower overall mortality in extrapulmonary TB, and the protective effect of ART against mortality during TB treatment.
Translating findings into policy and practice
On the basis of this experience, decentralized CPT delivery for HIV-infected TB patients has been included in national TB-HIV policies, and is being presently implemented in high HIV-prevalence settings throughout India 
. Joint TB-HIV trainings now strongly emphasize the referral of HIV-infected TB patients to ART centres. TB patient records and RNTCP registers have been modified to include information on HIV status, CPT, and ART. These recording enhancements will allow for routine programme monitoring and supervision of the local programme efficiency in detecting HIV-infection among TB patients and linking these patients to CPT and ART. Operational research is ongoing in other districts in India to understand weaknesses in referral to ART centres and ART uptake.
Cotrimoxazole prophylaxis can be delivered to HIV-infected TB patients under programmatic conditions in India. Despite the availability of free cotrimoxazole locally and ART from referral centres, death was common and was strongly associated with the absence of ART during TB treatment. These findings have been translated into national policy and programme implementation through the expansion of cotrimoxazole availability and training on the importance of ART. To minimize death, the TB and HIV programmes should target high levels of ART uptake and closely monitor progress in implementation.