Study design and Setting
A retrospective cohort study was conducted. We reviewed patient chart, clinical record and ART data base for five years from 2009 to 2014. The study was conducted in selected ART clinics found in southwest Ethiopia. South West Ethiopia encompasses five zones namely, Jimma, Illu Ababora, Kafa, Sheka and Bench-Maji.
- In our study we defined Tuberculosis case as: (1) Smear positive result of acid fast bacilli or (2) TB confirmed by culture or (3) Clinical sign and symptoms compatible with TB infection criteria by WHO or (4) Chest radiography confirmation.
- Children on HAART was defined as those who took combination therapy of 3 antiretroviral drugs that included two non-nucleoside reverse transcriptase and one protease inhibitor.
- Children, who were lost, died, or transferred out or did not develop the events until the last visit were considered as censored.
Source and study population
All children younger than 15 years having a follow up care in ART clinic in south west Ethiopia were source population. While all randomly selected HIV positive under 15 years old children registered from September, 2009 to August, 2014 in south west Ethiopia were study population.
- Exposure- treatment with highly active antiretroviral treatment (HAART) for at least 2 month
- Outcome- TB illness
Inclusion and exclusion criteria
- Inclusion- All under 15 years old children on HAART or Pre-HAART follow up who were registered from September 2009 to August 2014.
- Exclusion – All under 15 years old children who started anti-TB treatment at the beginning of follow up and/or diagnosed as TB patient
Sample size and Sampling procedure
844 sample size was calculated with double population proportion formula by considering the following assumption
α =Type one error (0.05)
Z α/2= Critical value at 95 % level of significance
Z1-β= standard normal distribution value corresponding to power (90%)
Fresh list of hospitals in each zone namely Bench Maji, kefa ,Sheka, Jimma and Ilubabor were prepared. After having the list of ART clinics that provide ART service for less than 15 years of age, and with proportional allocation methods, samples were selected from each ART clinic with systematic random sampling technique for both ART and Pre ART cohort.
Data collection tool and procedure
A standardized tool which has been adapted from existing literature was used. Adapted tools was translated to local language by language expertise. Then relevant data was collected from patients’ pre-ART and ART follow up log books, Data base and other clinical records. Data was collected by six data clerk who work in ART clinic after receiving 2 days training on how to extract data from these records. To assure quality of data extraction, a pretest was conducted in 5% of the sampled population.
Data entry and Analysis
The collected Data were coded and double entry was made in EpiData version 3.1 statistical package. Then the data were exported to the Statistical Package for the Social Sciences (SPSS) version 23 software for analysis. Before analysis, the data were processed and cleaned by running frequency, sorting and cross tabulation to check completeness, outliers and data entry errors. Descriptive statistics analysis were done to describe the characteristics of the study subjects.
To compare proportions of TB between HAART naïve and HAART cohort, we used the chi-square test, whereas we compared medians by the Mann-Whitney test. The Incidence rate was calculated with an Open Epi software and were expressed per 100 child-years of follow-up on HAART. The TB free survival probability were constructed by the Kaplan-Meier method and compared between HAART naïve and HAART groups by the log-rank test. The comparisons was two- tailed with p values < 0.05.
We estimated the effect of HAART on TB incidence by adjusting for confounders measured at baseline and time-varying intermediates. We did this by fitting weighted pooled logistic regression model to construct stabilized inverse probability of treatment and censoring (IPTC) weights. Then, these IPTC weights were used in a weighted pooled logistic regression model to approximate the parameters of a marginal structural Cox model(26). Each patient in the above logistic models received a time-varying weight inversely proportional to the estimated probability of having his/her own observed history of HAART initiation, as described elsewhere. For each child and visit, the denominator of the weight can be viewed as the probability that they received their actual treatment history and remained uncensored up to that time, conditional on their past treatment and covariate history. Confounders were selected based on previous studies. The covariates included in these models were age, sex, baseline clinical stage and baseline CD4 as baseline covariates. The time varying covariates were CD4, Clinical stage, follow-up status, adherence and CPT.
Bivariable and multivariable Cox proportional hazard regression models were used to see independent predictors of TB incidence. Variables with p-value <0.2 in bivariable analysis were transferred to Multivariable cox proportional hazard model. In Multivariable analysis, Variables with P-value <0.05 at 95% confidence level was considered as statistically significant predictors of TB incidence. The results were expressed as hazard ratios (HRs) with 95% confidence intervals (CI).