Statistical Modeling of Survival of Tuberculosis Infected HIV Patients Treated with Antiretroviral Treatment: A Case of Felege Hiwot Referral Hospital, Bahir Dar, Ethiopia

Senait Cherie Adegeh^* and Essey Kebede
^*Correspondence: Senait Cherie Adegeh, Department of statistics, College of Science, Bahir Dar University, Bahir Dar, Ethiopia, Email:

Keywords

HIV/TB Co-infection • Survival Analysis • Antiretroviral Treatment • Cox proportional hazards model

Introduction

Background

The dual infections of HIV and Tuberculosis remain serious health issues in the world. Globally, the incidence of opportunistic infections among AIDS patients has declined after ART introduction. However, tuberculosis, among other infectious diseases remains a major cause of morbidity in developing countries. Without treatment, as with other opportunistic infections, HIV and TB can work together to shorten the life of the person infected [1]. The world health organization recommends that all HIV infected persons should be screened for TB and HIV infected persons without active TB disease be evaluated for treatment of latent TB infection. The impact of HIV/AIDS on TB is significant. With 24% of all TB deaths being associated with HIV, 13% of new TB cases are related to HIV/AIDS at the global [2].

The world health organization (WHO) estimates that 11.1 million people are coinfected with TB and HIV, over 90% of these dually infected individuals reside in developing nations. In 2011, 430,000 people are estimated to have died of TB and HIV co-infection [3,4]. In the same year, Seventy-nine percent of the HIV positive TB cases were in the African region. The burden of disease through HIV/TB co-infection is particularly high in sub-Saharan African [4]. ART alone somewhat reduces the risk of TB disease; however, even after ART is started, the risk of TB still remains many times higher than the general population, especially during the first few years of ART. Even among those who have already had TB in the past and been cured, they are still extremely susceptible to recurrent TB disease. Ethiopia was among the first few African countries to introduce ART in 2003, in selected health facilities following the issuance of the National Antiretroviral Drugs (ARVs) Supply and Use Policy in 2002. The Ethiopian free ART scheme was launched in 2005 [5].

TB is the most common opportunistic infection among HIV infected individuals including those who are taking ART and co-infected individuals are at high risk of death. TB may occur at any stage of HIV disease and is frequently the first recognized presentation of underlying HIV infection. As compared to people without HIV, people living with HIV have a 20-fold higher risk of developing TB and the risk continues to increase as CD4 cell counts progressively decline [6].

In Ethiopia, TB is the leading cause of morbidity, one of the three major causes for hospital admission, and the second killer next to malaria [7]. TB and HIV co-infection are associated with special diagnostic and therapeutic challenges and constitute an immense burden on healthcare systems of heavily infected countries like Ethiopia [7,8]. The TB-HIV co-infection rate in Ethiopia is 41% [3]. This rate is still high; consequently, it needs further study to identify factors related to high rates of HIV/TB co-infection and low survival times of these patients.

Most of the studies in Ethiopia focused on the prevention and factors that increase the chance of contracting the disease. It can be said that, less attention was given on the survival of HIV positive patients infected with tuberculosis taking ART. The extent of relevance of socio-demographic, clinical and risk behavior factors for survival of HIV/TB in Ethiopia setting is not yet well described by previous studies. This study motivates to identify the major factors affecting survival time of HIV/TB co-infected patients under ART and to predict the survival probability of these individuals. Thus the main objective of this study is to model the survival time of people living with HIV under antiretroviral therapy who had screened positive for tuberculosis.

Methods and Materials

The study area

This study was conducted at Felege Hiwot referral hospital, which is located in Bahir Dar city. The city is located approximately 565 km northwest from the capital city of Ethiopia, Addis Ababa. Based on the 2007 census conducted by the Central Statistics Agency of Ethiopia, Bahir Dar city has a total population of 221,991, of whom 108,456(48.8%) are men and 113,535(51.14%) women. Felege Hiwot Referral Hospital provides general outpatient and inpatient services, including surgical and obstetric emergency care. Infectious diseases account for most of the inpatient and outpatient visits. It has been providing voluntary counseling and testing (VCT) services. In 2003, the ART clinic of the Referral Hospital started its activity; after the Ethiopia government launches free ART in 2005, the Referral Hospital started to provide free service to patients.

The data

This study is a retrospective cohort study. The study reviews patient’s pre- ART charts, intake forms and follow up charts of HIV/TB co-infected patients in Bahir Dar Felege Hiwot Referral hospital. Each patient has one medical file containing all TB and HIV notes, which includes the patient intake forms and HIV care and follow-up card, prepared by Federal ministry of health (FMOH). Thus, in this study secondary data, which collected from patients follow up records, are used. From this record of the patients, the variables, which are important for the study, are selected by using the patient’s identification number or the laboratory code without any direct contact with the patients. Instead, it is done by communicating with the nurses and counselors to get the medical record and other information important for the study. By the time we collect the data, 15,150 patients have visited the ART clinic. Of these 10441 are on ART since the start of ART and currently active ART patients are 6074. The total number of patients in the target population is 806, with a sample size of 325.

Sampling design and sampling methods

The target populations for this study are patients under the follow up of ART at Felege Hiwot referral hospital from October 2007 to September 2012 coinfected with tuberculosis. To sample those TB tested HIV patients from the population of HIV patients obtained from Felege Hiwot Referral hospital ART clinic data by using their charts and then select HIV/TB co-infected patients by simple random sampling technique. Our analysis was restricted to HIV patients under ART who are positively screened for tuberculosis, whose age is ≥15 years. The study include an HIV infected patients >=15 years of age under ART and diagnosed for TB and receive TB treatment. However the study exclude those patients on ART with age less than 15 years, those patients who are not initiated for antiretroviral treatment, those who are tuberculosis negative, those with unknown TB status, whose diagnosis time was missing and those patients whose death month was missing by looking the patient’s record. Then, for each patient satisfying eligibility criteria the medical file is traced to have explanatory information.

Variables included in the study

Dependent variable

The dependent variable used in this analysis is survival time of HIV-TB coinfected patients. It was measured in months starting from date of ART initiation to death or censored time. Patient Status were coded as ‘1’ if the death time is observed and ‘0’ if censored.

Independent variables

Variables included in the study were selected based on some past studies and those that are expected to be determinant factors of the survival of HIV patients infected with tuberculosis. This study considers several predictors to examine the major factors of survival of HIV/TB co-infected patients. The variables considered in this study were: Age, Gender, Residence of patient, Marital status, Literacy status, Religion, Base line body weight, CD4 count at the start of ART, Regimen type at start of ART, Knowledge about ART, Employment status, Number of living rooms at start, Number of people in the household (family size) at start, WHO clinical stage at start of ART, Alcohol intake, Tuberculosis category, Tobacco use, Drug use, Condom use during sexual intercourse, and Regimen change during TB diagnosis.

Data Analysis

Descriptive methods for survival data

An initial step in the analysis of a set of survival data is to present numerical or graphical summaries of the survival times for individuals in a particular group. This includes survival distribution and Kaplan Meier survival function estimation that are used for the estimation of the distribution of survival time from all of the observations available. In addition, we can make comparisons of the life experience of two or more groups of subjects differing for a given characteristic. Now a day is the Kaplan Meier method for estimating survival curves and the Log Rank test for comparing two-estimated survival curves are most frequently used statistical tools in medical reports on survival data [9].

Semi-parametric models for survival data

The statistical approach to be used in this study is the analysis of time to event data, which is survival data analysis. Survival analysis is an important statistical technique used to describe and model time–to–event data. The most popular of the semi-parametric models is the Proportional hazards model, which has the property that ratio of the hazards of two individuals at time t can depend on the values of their explanatory variables, say β'x = β₁x₁ +β₂x₂+…+β_p x_p, but does not depend on time t [10].

Since no particular form of probability distribution is assumed for the survival times, proportional hazards model is referred to as a semi-parametric or distribution free model. The Cox proportional hazards model is given by:

h(t;X) = exp^(β^' X) h0(t)

where h(t;X) represents the hazard function for the i^th patient; h0(t) is the baseline hazard function at time t; β = (β₁, β₂, …, β_p) is a vector of unknown parameters that are assumed to be the same for all individuals in the study and X= (x_1i,x_2i,…….,x_ki) for i=1,2,…..,n is a vector of explanatory variables for the i^th individual at time t. Consequently, from the proportional hazard function, we can obtain the estimated survivor function, which is given by:

S∧_i (t)= S∧0 (t)exp (β ∧,X)

Where, 0(t) is the baseline survival function. Fitting a proportional hazards model to an observed set of survival data involves estimating the unknown parameters = (β₁, β₂,..., β_p) in the model.

The goal of Cox regression is to explore the association between the survival of patient and many covariates (independent variables or predictors). To accomplish this goal, we need to create a model that includes all variables that are useful in predicting the dependent variable. First, consider the model that will include all the predictors that had a p-value of less than 0.25 in the univariate analyses. That means any variable whose univariate test has a p-value < 0.25 should be considered as a candidate for multivariate analysis.

Conclusion and Recommendations

The objective of the present study was to examine the survival probability of HIV/TB co-infected patients and to identify major factors that affect the survival of patients under ART. To achieve this end, Cox regression analysis was conducted, and the result shows that TB category, CD4 count, employment status, number of living rooms, alcohol and tobacco use were found to be significant factors.

Although, TB is a preventable and curable disease in almost any socioeconomic circumstances especially with the implementation of DOTs, it is associated with high mortality in HIV infection. Effective TB treatment and ART reduces mortality among HIV/TB co-infected patients, but in this study survival of patients co-infected with HIV and TB remains low. So, from this result we recommend the following.

Early admission of the patients is recommended. Physicians are expected to work hard to bring about behavioral changes. Moreover, systematic screening and close follow up of patients for signs and symptoms of TB after starting ART should be given attention. In addition, the limited availability of more sensitive diagnostic tests for TB in HIV/AIDS patients under ART must be addressed. Furthermore, further research is recommended to deal with how to increase the survival probability of HIV/TB co-infected patients treated with ART.

References

1. Weiner Lindsey M, Webb Amy K, Limbago Brandi and Margaret A. Dudeck. “Antimicrobial-Resistant Pathogens Associated With Healthcare-Associated Infections: Summary of Data Reported to the National Healthcare Safety Network at the Centres for Disease Control and Prevention, 2011–2014.” Infection control hosp epidemiol 37 (2016): 1288-1301.
2. Early Detection of Tuberculosis: An Overview of Approaches, Guidelines and Tools. World Health Organization, (2011).
3. Global Tuberculosis Report 2013. World Health Organization, (2013).
4. WHO Policy on Collaborative TB/HIV Activities: Guidelines for National Programmes and Other Stakeholders. World Health Organization, (2012).
5. Guidelines for prevention of mother-to-child transmission of HIV in Ethiopia. Federal Ministry of Health, (2007).
6. Global tuberculosis control: epidemiology, strategy, financing. World Health Organization, (2009).
7. Global Tuberculosis Control: Surveillance, Planning, Financing: WHO report 2008. World Health Organization, (2008).
8. May Susanne and David W. Hosmer. “A simplified method of calculating an overall goodness-of-fit test for the Cox proportional hazards model.” Lifetime data analysis 4 (1998): 109-120.
9. Cox, David R. “Regression models and life-tables.” J Royal Statistical Soc 34 (1972):187-202.
10. Immunologival Consequences of Mycobacterium tuberculosis and Human Immunodeficiency Virus coinfection in Ethiopia. 2007.