95 4 Diabetes-tuberculosis comorbidity in a low-income setting blood glucose between visits [25]. Blood pressure was measured using an electronic sphygmomanometer provided by MOH. Statistical analysis We described patient characteristics using mean with standard deviation or median with interquartile range for continuous descriptors and proportions for categorical descriptors. This was disaggregated by the presence or absence of elevated blood glucose and presented in a table (Table 1). The baseline prevalence of elevated blood glucose was defined as the proportion of all participants with elevated blood glucose. This prevalence was presented overall and by region, sex, age category and HIV status. A sub-analysis for the prevalence of elevated baseline blood glucose was conducted for 33 patients with a fasting blood glucose measurement using a cut-off of >5.5 to 6.9 mmol/l for pre-DM and ≥7.0 mmol/l for blood glucose measurement in the diabetes range. Unpaired t-test and Kruskal Wallis test compared blood glucose measurements and diastolic and systolic blood pressure changes between baseline, 2nd, and 5th month visits. We used a logistic regression model to assess the predictors of elevated baseline blood glucose and adjusted for age, sex, HIV status and weight/BMI (selected apriori) in the final multivariate model. Variable selection was based on the forward and backward elimination process at p=0.2, verified with the adaptive Least Absolute Shrinkage and Selection Operator (LASSO) variable selection approach. A repeat analysis using blood glucose measurements at months 2 and 5 was impossible as very few participants had elevated measurements. Nonetheless, we conducted a sensitivity analysis employing a nested multi-level logistic regression model incorporating random effects at two levels to assess the influence of regional factors on the risk of elevated baseline blood glucose. This analysis accounted for the clustering of individuals within specific regions. TB treatment outcome was the secondary outcome of this study. This was described and classified into two - favourable TB treatment outcome (defined as patients with either cured or completed outcome assigned at the end of treatment) or unfavourable TB treatment outcome (defined as those who died, lost to follow-up, stopped treatment, transferred out, re-initiated treatment, or treatment failure outcome at the end of treatment). Participants receiving treatment for DRTB were excluded from the reclassification since they were still on treatment at the end of the study. The differences in TB treatment outcome and any possible association between blood glucose and TB treatment outcome were assessed. The baseline predictors of unfavourable TB treatment outcome were evaluated using a logistic regression with the new binary TB treatment outcome variable – favourable or unfavourable TB treatment outcome. We used the forward and backward elimination method at p=0.2 to identify variables for inclusion in
RkJQdWJsaXNoZXIy MTk4NDMw