The timing of death in patients with tuberculosis who die during anti-tuberculosis treatment in Andhra Pradesh, South India

National TB Control Programmes (NTPs) routinely report treatment outcomes for patients with tuberculosis (TB). This is reflected in national reports and also annual reports from the World Health Organization (WHO) which provide data on treatment outcomes of TB patients from all countries in the world. Treatment outcomes include death, which is defined as death from any cause occurring during the course of anti-TB treatment.

There have been a few publications on the timing of death in patients while on treatment. In sub-Saharan Africa, studies have shown that the majority of deaths occur early during the first 1-2 months of anti-tuberculosis treatment [1–3]. In sub-Saharan Africa, where co-infection with HIV is high, this information has been important for NTPs to plan realistic strategies to reduce death rates during treatment. For example, giving antiretroviral therapy (ART) during the initial phase rather than the continuation phase of anti-tuberculosis treatment is more beneficial in reducing death rates in co-infected TB patients, and is based on a sound knowledge of when HIV-infected TB patients die [4, 5]. Similar reports of early deaths have come from more industrialised countries such as Singapore [6], Russia [7] and Taiwan [8].

India, with a total population of 1.2 billion, has a well established national TB control programme, based firmly on the "DOTS" strategy, and treatment outcomes are reported regularly for patients with all types of TB [9]. India has an estimated 2.0 million incident TB cases each year, and an estimated TB-related mortality of 280,000 deaths per annum [9]. Understanding when in the course of TB treatment patients die is important for determining the type of intervention to be offered and crucially when this intervention should be given. Such information might be useful in further reducing case fatality rates among TB patients. There has been previous work carried out in India between 1999 and 2000 and in 2004 looking at timing of death as part of studies that assessed risk factors for death, failure and default, and between 50-65% of deaths were reported to occur within the initial phase of treatment [10, 11]. However, each of these studies assessed less than 750 patients. As a result, there were small numbers of patients who died, and the findings may not be representative of the wider or more recent picture of timing of deaths during anti-tuberculosis treatment in India within the Revised National TB Control Programme (RNTCP).

The aim of this study was to document the timing of reported death in a large cohort of patients with tuberculosis who die during treatment. The specific objectives were to determine in a defined cohort of TB patients:- i) the treatment outcomes and the number who died while on treatment, ii) the month of death and iii) the characteristics of patients who were recorded as having died early in the initial 8 weeks of anti-tuberculosis treatment.

This study in a large cohort of over 8,000 registered TB patients found a low case fatality at almost 6%. Patients with previously treated TB had higher death rates and those with EPTB had lower deaths rates than those with new pulmonary tuberculosis. The higher death rates in previously treated patients might be explained by more severe and drug resistant disease consequent upon failed first line therapy or initial and undiagnosed multi-drug resistant TB [13, 14]. Although we did not document the types of EPTB during this study, the majority of patients in India with EPTB have lymph node disease [13], which tends to be associated with morbidity but not mortality.

There was a fairly even distribution of deaths during the course of anti-tuberculosis treatment. This is in marked contrast to the situation in sub-Saharan Africa where there is an excess of deaths in the first 1-2 months of treatment, thought to be due to late presentation and therefore severe tuberculosis disease as well as the effects of advanced HIV disease in those who are co-infected [3, 4]. The two main characteristics associated with more frequent "early death" in this Indian study were age above 50 years and new disease compared with previously treated disease. Increasing age has been noted as a risk factor for death in other studies both within India [10, 15] and outside of India [2], and new disease in contrast to recurrent disease was also associated with high early mortality during the first 4 weeks of treatment in Malawi [2]. Why this occurs is not known. Older people may be at higher risk of co-morbid disease which may result in a more serious illness at the time of presentation, diagnosis and treatment, and they may also develop chronic respiratory illnesses resembling and mistaken for smear-negative PTB due to chronic bronchitis and lung cancer [16]. Patients with recurrent disease may also be familiar with the symptoms and signs of TB, and therefore present earlier than those with new disease and as a result have less risk of early death. HIV-serostatus was not associated with "early deaths" in the small sample of patients who were HIV-tested, although many patients in the study were not HIV tested.

Why is there a difference in distribution of deaths in India compared with sub-Saharan Africa? First and most importantly, it will be necessary to repeat this study in other parts of the country and in large numbers of patients to ensure that the results in Andhra Pradesh are nationally representative. If indeed the results are confirmed, then there may be various explanations. India has a problem with initial defaulters [17] i.e., patients who are diagnosed with active TB but fail to get registered and placed on treatment. A high initial default rate will falsely lower early death rates in registered TB patients, obscuring the true picture of patients dying early during the registration and first few weeks of treatment of their disease. Second, HIV co-infection is lower in India than in Africa, with national rates of HIV-infection in TB patients currently at 5-10% [9]. Untreated, advanced HIV disease is therefore not an important factor in India, while in sub-Saharan Africa this has played a major role in case fatality and early deaths, and to some extent this has been mitigated by the introduction of cotrimoxazole preventive therapy and antiretroviral therapy.

If death rates are truly dispersed in an even manner during the course of anti-TB treatment in India then more work needs to be done in this area. Reasons may relate to the treatment of the disease itself, raising concerns about drug adherence, quality of anti-tuberculosis drugs or the presence of undetected drug resistance. They may also relate to co-morbidities which influence mortality. For example, diabetes mellitus has been calculated to account for 15%-20% of pulmonary TB in India [18], and there is growing evidence that diabetes is associated with an increased case fatality in TB patients [19–21]. The timing of death in diabetes patients who have TB is not known and requires active research, but one could speculate that diabetes exerts its negative effects throughout the course of anti-TB treatment as a result of drug-drug interactions, an increased association with anti-tuberculosis drug toxicity and immune suppressive effects of diabetes.

The strengths of this study are that there were a large number of patients enrolled in the cohort and treatment outcomes were tracked and recorded using standardised systems. However, there are a number of limitations. First, this was a record review and it is possible that mistakes were made in the recording of timing and date of death. Second, patients who are recorded as default or transfer out may also have died [15], and such misclassification may affect the results. Third, the records that formed the source of data did not contain valuable information such as results of any culture and drug sensitivity testing, which might have been important in explaining reasons for some of the deaths. Fourth, the 16 selected sites were all supported by LEPRA partnering with RNTCP, and as such may not be representative of other sites where this support was absent.

Whatever the shortcomings of this study, the results should serve to encourage others to repeat similar studies in other parts of the country and should also encourage programme staff to carefully record timing of death and timing of other adverse events such as default and transfer out. A number of potentially important measures might help to reduce death rates by ensuring that i) all patients diagnosed with smear-positive sputum are registered and start anti-TB treatment as soon as possible (thereby cutting down initial default rates) [22], ii) patients who are at risk of drug resistance, such as those previously treated, have culture and drug sensitivity testing so that treatment is appropriately tailored to levels of drug resistance, and iii) elderly patients are investigated for co-morbidities including diabetes mellitus which may increase the risk of death during treatment [23]. A better understanding of when and why the estimated 280,000 annual TB-related deaths occur is essential as India strives to improve programme performance and exceed the new 2015 global targets of treatment success rates of 90% in the years to come [24].

In this large cohort of TB patients registered in Andhra Pradesh, South India, deaths occurred with an even frequency throughout anti-TB treatment. Reasons may relate to i) the treatment of the disease itself, raising concerns about drug adherence, quality of anti-tuberculosis drugs or the presence of undetected drug resistance and ii) co-morbidities, such as HIV/AIDS and diabetes mellitus, which are known to influence mortality. More research in this area from prospective and retrospective studies is needed.