Survival status and its predictors among multi-drug resistance tuberculosis treated patients in Ethiopia: Multicenter observational study

Background Although substantial progress has been made in combating the crisis of multi-drug resistance tuberculosis (MDR-TB), it remained the major public health threat globally. Objective To assess patients’ survival and its predictors among patients receiving multi-drug resistance tuberculosis treatment at MDR-TB treatment centers of southern and southwestern Ethiopia. Methods A multicenter retrospective observational study was conducted from April 14 to May 14, 2019, among patients receiving MDR-TB treatment at three MDR-TB treatment centers, Butajira, Arbaminch and Shenengibe Hospitals, located in south and southwestern Ethiopia. A total of 200 records were reviewed using a check list adopted from the national MDR-TB treatment charts and other relevant documents. Data were entered into Epi-Data version 4.2.0 for cleaning and exported to STATA-13 for analysis. Descriptive analysis was carried out and results were presented by text, tables, and charts. Kaplan-Meier (log-rank test) and Cox regression were used to compare baseline survival experience and to determine predictors of patients’ survival (death), respectively. The adjusted hazard ratio (AHR) was used to measure the strength of association and a p-value of <0.05 was considered to declare statistical significance. Results Of 200 patients, 108 (54%) of them were males. The mean (+ standard deviation) age of the study population was 32.9±9.5years. During follow-up, 22 (11%) deaths were reported. The overall incidence density of death was 11.99, 95% CI [7.89–18.21] per 100,000person-years. The median (interquartile range (IQR)) survival time was 375(249–457) days. Comorbidity (AHR = 23.68, 95% CI [4.85–115.46]), alcohol consumption (AHR = 4.53, 95% CI [1.21–16.97]), and history of poor adherence (AHR = 12.27, 95% CI [2.83–53.21]) were independently associated with patients’ survival (death). Conclusion In this study, the incidence density of mortality was very high. Alcohol consumption, poor adherence, and the presence of comorbidity were independently associated with death. Hence, alcohol users, patients with comorbidity and poor adherence should be given due attention during therapy.


Introduction
Multi-drug resistant tuberculosis (MDR-TB) is a major concern at global, regional and country levels. According to the 2019 global TB report, there were 3.4% new cases and 18% previously treated cases of MDR-TB in 2018. In Bangladesh, among reported MDR-TB cases, 1.5% of them were new cases and 4.9% of them were previously treated TB cases. The incident rate of MDR-TB cases in this region was 3.7%. In the Democratic Republic of Congo (DRC), 1.7 % of new cases and 9.5% of previously treated TB cases were reported. The overall incidence of MDR-TB in DRC was found to be 7.2%. Ethiopia ranked 8 th among the 30 high MDR-TB burden countries with 2700 MDR-TB cases each year. The estimated prevalence of MDR-TB in the country is 0.71% among newly diagnosed patients and 16% in patients under retreatment(1).
Moreover, MDR-TB was responsible for a sizeable number of TB-related deaths globally. A study from the United Kingdom reported the death rate of 6.4% (2). According to Peter et al MDR-TB was claimed for 3.9% mortality (3). Findings from India and South Africa indicated the mortality rate of 17% (4) and 20 % respectively (5). In Tanzania, 6.5% of mortality was reported among MDR-TB patients (6). Two studies from Ethiopia revealed a mortality rate of 24.4% (16) and 18.3% among MDR-TB patients (7).
The mortality rate due to MDR-TB was amplified by comorbidities (8). The human immune virus (HIV) co-infection was the major risk factor (6,9). In one study, 31.3% of patients have died at 12 months of follow-up. Moderate to severe anemia and being smear positive were significantly associated with death (10). In Ethiopia, the mortality rate of patients was higher in the earlier stages of treatment. Complications, drug-resistance, and smoking had contributed to an increased risk of mortality (11). Though fewer studies had explored the rate of mortality in some parts of Ethiopia, they were single centered and hence, difficult to conclude the incidence 2 of national mortality. Therefore, this multi-center study was aimed to assess the incidence of mortality and its predictors among patients receiving MDR-TB treatment at selected treatment centers in the south and southwestern regions of the country.

Study design and setting
A multicenter retrospective observational study was conducted from April 14 to May 14, 2019, among patients receiving MDR-TB treatment at Butajira, Arbaminch and Shenengibe General Hospitals, all located in south and southwestern Ethiopia. They are about 113km, 505km, and 329km respectively away from Addis Ababa, the political center of Ethiopia.
Butajira General Hospital is located at the Gurage zone (southern Ethiopia) and currently serving around 5 million population. It started the MDR-TB treatment service in 2015.

Study population and patient enrollment
All adult MDR-TB patient charts who fulfilled the eligibility criteria at selected health care facilities were enrolled in the study and there is no special sampling technique employed. 3 We include all adult patients with the confirmed diagnosis of MDR-TB based on Xpert MTB/RIF® assay (12) and enrolled in the MDR-TB treatment program since January 2013.
Charts that reported complete baseline and follow-up data and the intended treatment outcomes were included. We excluded charts of patients with incomplete data and charts of patients transferred to other facilities. Accordingly, 213 charts were assessed for eligibility and 200cahrts were included in the final analysis ( Fig. 1)

Data collection procedures and study variables
The data was collected by using a structured checklist prepared from different kinds of literature, WHO guidelines (13,14,15) and national MDR-TB treatment follow up chart. The checklist contains several variables. Patient-related variables include Age, sex, residence, pregnancy, marital status, smoking status, educational level, height, weight, and body mass index (BMI). Disease-related variables such as the category of MDR-TB, drug resistance status, and co-morbidities were also included in the checklist. Furthermore, the checklist also contains drug-related data including type of medication and drug regimen. To capture the impact of time on the treatment outcome, the time at which treatment was initiated and mortality occurred was recorded. All the above data were extracted from patient charts. The data regarding mortality was obtained from the Physicians mortality summary notes as confirmed by the caring physician's name and signature.

Data Quality Assurance
The data collection tool was carefully designed to capture all necessary variables to achieve the study objectives. The charts of each patient were reviewed for inclusion before the data collection. Three clinical pharmacists and three physicians were trained for two days to collect the patient data. The clinical pharmacists collected drug-related information and the patientrelated and clinical variables were collected by the physicians. The pharmacists were also 4 responsible to identify and cross-check adverse drug reactions concerning each anti-TB drug.
At each facility, a senior infectious specialist supervised the data collection process.
Supervisors followed the data collection process and helped by co-relating diagnostic and laboratory findings with the main outcome. Moreover, a pre-test was conducted on 5% of patients' records to test the effectiveness of the data collection tool and the necessary adjustment was made based on the pre-test findings.

Data processing and analysis
Data were checked for completeness and cleaned using EpiData version 4.2 and exported to STAT-13 for analysis. Categorical variables were summarized by counts, graphs and percentages. The baseline characteristics of the patients were compared using Pearson's chisquare (χ2). Normally distributed continuous variables were summarized using mean and standard deviation (SD), whereas median and interquartile range (IQR) was utilized to report non-normally distributed continuous variables. The cumulative survival probability of the patients was estimated using the Kaplan-Meier (KM) curve. A bivariate Cox proportional hazard model was first fitted, and variables with p-value <0.25 in the bivariate analysis were further regressed using multivariable Cox proportional hazard model. Crude and adjusted hazard ratios were calculated to estimate the risk of death. On multivariable Cox proportional hazard regression analysis, variables with a p-value of less than 0.05 were considered to declare statistical significance.

Characteristics of the study groups
Of 213 records screened for eligibility, 13 records were excluded and 200 MDR-TB patients' records were included in the analysis (Fig. 1).

Socio-demographic characteristics
The majority, 108(54%) of the patients were males. The mean ± SD age of the study participants was 32.9±9.5. The highest proportions, 78(39%) of participants were Muslims.
Most, 111 (55.5%) of them were from rural areas and 99 (49.5%) of the participants were married. The highest proportions have a secondary level of education 74(37%). About 62(31%) of the study participants were merchants. Non-smokers and non-alcoholic comprised 190(95%) and 172(86%) respectively. Baseline smoking status, alcohol consumption and body mass index determined patient status (p<0.05) ( Table 1). were found to be positive for HIV. Diabetes mellitus 9(4.5%) and acute kidney injury 7(3.5%) were among the common co-morbidities diagnosed.   days. On the multivariate Cox regression presence of co-morbidity, alcohol consumption and poor adherence were independent predictors of death.
In the current study, the overall incidence of mortality was 11.99 per 100,000 person-years.
The finding was comparable with a similar study conducted in Lithuania that reported 11 per 100,000 person-per years (16). But lower than the study by Girum et al (17) in which the overall reported incidence of death was 7per 100person-years. The differences might be due to the inclusion of a small number of patients and a shorter follow-up period of the former study. The reported death rate (11%) in our study was also lower than the other study done in India 17%, South Africa 20%, and eastern Ethiopia 18.3% (11,18,19), but the higher death rate was reported in Peru 5% (8).
In our study, the median time to death was 375(IQR=249-457) days. This was lower than the study from central Ethiopia which reported the median survival time of 480days (11). But much lower median survival time to death was reported by Kamban et al (10) in which the median survival time to death was 78(IQR=33. 3-154.5) days.
In the current study, patients who had co-morbidity were approximately 24times higher hazard of death (AHR=23.68, 95% CI [4.85-115.46]. As presented earlier in Table 2, HIV/AIDS, diabetes mellitus and acute kidney injury were mostly reported co-morbidities. The presence of such co-morbidities mainly HIV/AIDS (6,9) and diabetes mellitus were strongly related to immune-suppression. Tuberculosis facilitates HIV replication and viral diversification rates through proinflammatory cytokine production. Proinflammatory cytokines increase HIV viral replication and diversity, hence facilitating immune-suppression (20). A study by Chung-Degado et al also showed that patients with co-morbidity were 5.4 times higher hazard of death (8). Our finding is also concurrent with the studies conducted in South Africa (5), Tanzania (6), and Ethiopia (7). Our study was not without limitations. Firstly, The retrospective nature of the data source limited us from tracking the major causes of death. Secondly, the method of patients' adherence assessment was also subjective as it is based on patient reports. Third, most of the patients have no data on sputum smear microscopy results. Lastly, but not least, missing patients' income status, wider confidence intervals and inability to screen out the exact causes of death were some of the major hiccups of this study.

Conclusion
In conclusion, this study found a high rate of mortality among patients receiving MDR-TB treatment in the selected settings. Alcohol use, poor adherence, and the presence of comorbidity were independent predictors of death. This study provided insight into how to provide optimal care of MDR-TB patients with comorbidities, poorly adhered to therapy and habit of alcohol use. However, given all the limitations mentioned above, we urge the readers to interpret the findings of this study cautiously.

Ethical approval
The study approved by the Ethical Review Board of Jimma University and given an IRB number of IHRPG1/565/2019. Because of the retrospective anonymous nature of the study, the need for informed consent was waived. Confidentiality was ensured by removing the name and address of the patients from the data.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.