On 2015 Dec 03, Siddharudha Shivalli commented:

I read the article titled ‘Smear positive pulmonary tuberculosis among diabetic patients at the Dessie referral hospital, Northeast Ethiopia’ by Amare H et al, with a great interest. Authors’ efforts are praiseworthy. In their single centre hospital based study, authors highlight the prevalence of TB among known diabetics and factors associated with it. However, following are some issues and concerns.

For cross sectional study, adequacy and representativeness of study sample size are essential to ensure the validity of the study findings. Authors have justified the adequacy by calculating the sample size (n=236), however, I am not sure about the representativeness. Do 236 study participants selected consecutively over a period of only 3 months (February 2012 to April 2012) represent the diabetic patients who visit the Dessie referral hospital (average diabetic patient number approximately 1,700)? Systematic random sampling would have been more apt for this i.e. including every 4th or 5th eligible patient depending upon the weekly or monthly patient input.

Another limitation of the study is the inclusion criterion as authors have studied only pulmonary tuberculosis (PTB) suspected diabetic patients. If one wants to estimate the prevalence, all the diabetics should have been studied. Hence, reported prevalence in this study may be an underestimation. In addition reporting of prevalence should have been done with 95% confidence intervals (6.2%, 95% CI: 3.7-10.25). In addition, reported associations between prevalence of PTB among diabetes patients and study variables in this study may not imply cauasality owing to cross sectional study design.

In this study, variables which had a p-value of less than 0.20 were taken to multivariate logistic regression. However, it is recommended to assess and report the adequacy of applied regression model. Failure to do so may lead to misleading or incorrect deductions. Although the study sample was relatively large (n=236), a word about R2 (explaining the variance in prevalence of PTB) of the applied regression model would have been more affirmative.

None the less, I must congratulate the authors for investigating an important public health problem.

Competing interests: The author declares that there is no conflict of interest about this publication.

On 2015 Dec 03, Siddharudha Shivalli commented:

I read the article titled ‘Smear positive pulmonary tuberculosis among diabetic patients at the Dessie referral hospital, Northeast Ethiopia’ by Amare H et al, with a great interest. Authors’ efforts are praiseworthy. In their single centre hospital based study, authors highlight the prevalence of TB among known diabetics and factors associated with it. However, following are some issues and concerns.

For cross sectional study, adequacy and representativeness of study sample size are essential to ensure the validity of the study findings. Authors have justified the adequacy by calculating the sample size (n=236), however, I am not sure about the representativeness. Do 236 study participants selected consecutively over a period of only 3 months (February 2012 to April 2012) represent the diabetic patients who visit the Dessie referral hospital (average diabetic patient number approximately 1,700)? Systematic random sampling would have been more apt for this i.e. including every 4th or 5th eligible patient depending upon the weekly or monthly patient input.

Another limitation of the study is the inclusion criterion as authors have studied only pulmonary tuberculosis (PTB) suspected diabetic patients. If one wants to estimate the prevalence, all the diabetics should have been studied. Hence, reported prevalence in this study may be an underestimation. In addition reporting of prevalence should have been done with 95% confidence intervals (6.2%, 95% CI: 3.7-10.25). In addition, reported associations between prevalence of PTB among diabetes patients and study variables in this study may not imply cauasality owing to cross sectional study design.

In this study, variables which had a p-value of less than 0.20 were taken to multivariate logistic regression. However, it is recommended to assess and report the adequacy of applied regression model. Failure to do so may lead to misleading or incorrect deductions. Although the study sample was relatively large (n=236), a word about R2 (explaining the variance in prevalence of PTB) of the applied regression model would have been more affirmative.

None the less, I must congratulate the authors for investigating an important public health problem.

Competing interests: The author declares that there is no conflict of interest about this publication.