Skip to main content

Archived Comments for: Treatment outcome of tuberculosis patients at Gondar University Teaching Hospital, Northwest Ethiopia. A five - year retrospective study

Back to article

  1. Outcome misclassification and incorrect odds ratios

    Aaron Kipp, Vanderbilt University Medical Center

    5 November 2009

    In the ongoing struggle against TB transmission and drug resistant TB, information on treatment outcomes and what factors affect these outcomes are vitally important. Tessema et al take the first step toward that goal by reporting for the first time on treatment outcomes in northwest Ethiopia since the introduction of directly observed therapy in 2000 at the Gondar University Teaching Hospital.

    However, I believe there are some serious limitations and errors that may mislead readers and adversely affect conclusions. First, more than 40% of patients diagnosed at the study hospital are transferred out to other DOT clinics. This is not a problem in and of itself, and may facilitate treatment success by allowing patients to more easily access DOT. However, in the analyses and conclusions, the authors combine these patients with other patients who remained at the study hospital but did not have successful treatment. As a result, the authors report an “unsatisfactory” success rate of 29.5%. However, it is likely that many of the patients who transferred out successfully completed treatment at other clinics and therefore should not be classified with those who did not complete treatment. By my calculations, the treatment success rate among patients receiving DOT at the study hospital was 51% (1181/2320). This figure is nearly identical to the 49% found by Shargie et al in southern Ethiopia to which the authors refer. This is further exemplified in the odds ratio (95% confidence interval) of 0.266 (0.230, 0.309) observed for the comparison of treatment success among rural patients versus urban patients (Table 4). It is not surprising that rural patients are more likely to “not be treated successfully” because these patients are likely to be transferred out to a rural DOT clinics closer to their residence. A very important analysis would be to compare treatment outcomes between DOT received at the study hospital and the outpatient DOT clinics, but it does not appear that the authors have this information available.

    Second, there are numerous incorrect crude odds ratios reported in Table 4 for age and TB type (I could not verify the adjusted ORs). As an example, the authors report an OR (95% CI) of 0.985 (0.595, 1.630) for the association of ages 15-24 with successful treatment compared with ages 0-14. While the reported result indicates no difference in treatment success, the tabular data indicates that treatment was successful in 41% of the 15-24 year olds and in only 23% of the 0-14 year olds. By my calculations, the correct OR (95% CI) comparing the two age categories is 2.435 (1.960, 3.025). I found the remaining age and TB type ORs to be similarly incorrect, and it is my assumption that the corresponding adjusted ORs are also incorrect.

    Finally, the authors fail to make any conclusions based on their study data with the exception that older age is associated with an outcome of death (no statistical analysis is reported, however). Rather than expand on the results from their statistical analyses of treatment success, they repeat conclusions from other authors, whom they reference. It would have been nice to see the authors discuss their own findings in the context of the current literature.

    I hope these comments are usefule to the authors as they continue their work on teatment outcomes. Their work to identify and improve upon barriers to successful TB treatment is important and will be helpful in the work to reduce TB transmission and drug resistance.

    Aaron M. Kipp, PhD

    Research Instructor
    Vanderbilt University Medical Center
    Division of Epidemiology

    Competing interests

    no competing interests

  2. Response to Dr. Aaron Kipp’s comments

    Belay Tessema, Institute of Clinical Immunology, Medical Microbiology and Epidemiology of infectious Diseases

    14 January 2010

    Dear Editor,
    First of all, we would like to thank Dr. Aaron Kipp for reading and forwarding his personal view/opinion regarding to our article. In response to the points he raised, we put forward the following answers.
    First point (Outcome misclassification). As it is clearly stated on the methods part of the article, treatment outcome classification including treatment success was made based on the standard definitions of the National Tuberculosis and Leprosy Control Program guideline (NLCP) adopted from WHO( Ref. 8). In addition, the possible effect of the relatively higher patients transferred out rate on the calculated value of low treatment success rate has already well explained on the discussion part of this article (paragraph 1). Thus, Dr Aaron Kipp’s comments are rather personal opinion; different researchers can make analysis in different approach by including or excluding different groups, there is no hard and fast rule. For us the most important thing is that to show clearly about what is included and what is excluded in the analysis.
    Second point (incorrect crud odds ratios). We have checked the results of crud odds ratios and adjusted odds ratios to age and TB type (table 4) and we found that the adjusted odds ratios, more important than crud odds ratios are correct, however, there was a mistake of labeling the reference categories after transferring the output results of crud odds ratios of age and TB type in to the table. Nevertheless, as the adjusted odds ratios are correct and since we didn’t use the results of crud odds ratios for age and TB type for our discussion as well as to draw our conclusions, we believe that this minor error doesn’t affect the important findings, discussions as well as conclusions of this article.
    Last point (Conclusion). As it is indicated on the discussion part of the article, unlike his personal view, we have made conclusions effectively based on our pertinent findings such as low treatment success rate (29.5%), high death rate (10.1%) and high default rate (18.3%) among tuberculosis patients in the study area.

    Belay Tessema

    Competing interests

    None declared