A word of caution: one should take into account the coverage of drug susceptibility testing when interpreting results
Susan van den Hof, KNCV Tuberculosis Foundation
16 January 2013
We very much welcome the publication of results on tuberculosis epidemiology from Kazakhstan in BMC Infectious Diseases by researchers from within the country. It is important that insights from Kazakhstan and other Central Asian countries are shared with the international, non Russian speaking, community.
Kazakhstan is a country with a declining incidence of tuberculosis, but with a high prevalence of multi-drug resistant tuberculosis (MDR-TB). It is therefore important to study and know the trends in the MDR-TB burden, and understand risk factors. As we find it an important field of research with important implications, we would like to comment on this manuscript.
The authors state that the incidence of MDR-TB in Kazakhstan has increased in recent years, from 2007 to 2010. This statement is based on the number of diagnosed MDR-TB cases. We would like to point out that the increase as calculated by the authors is not statistically significant (P=0.12). More importantly however, one should realize that the coverage of drug susceptibility testing (DST) and its reporting in the TB register has increased during these years. Although all oblasts have been able to perform culture and DST for years and they have made tremendous progress in recent years, the actual coverage and reporting of results has not been and still is not complete. Increased DST coverage from 2007 onward may also partly explain the trends within oblasts presented in this article: in 2007 DST results were not included in the electronic TB register for 34% of all notified TB patients compared to 47% in 2010. In a study with a different focus but based on the same routine TB surveillance data from Kazakhstan over the years 2007-2011, and accepted for publication in the International Journal of Tuberculosis and Lung Disease (IJTLD), we show that overall MDR-TB prevalence among those with DST results available remains similar over these years. With a decreasing overall TB incidence, this means that the MDR-TB incidence rate in fact probably has gone down.
While taking into account DST coverage, it would have been advisable to stratify the data by treatment history, as this is a main determinant for MDR-TB and could show trends in MDR-TB rates for new and retreatment patients separately. This is important as drug resistance among new TB patients is a marker for transmission of MDR-TB, while MDR-TB prevalence among retreatment patients at least in part is due to acquired resistance.
Most of the patients (83-87%) did not have one of the predefined risk factors registered. The authors seem to interpret this as if these patients have an existent but unknown risk factor, which needs further research. However, especially in a TB endemic country, it is very well conceivable that many TB patients do not have one of the major risk factors for TB.
From the methods and results presented, we are not confident about the statistical methods used. For instance, we are not sure whether and if yes, at which geographical level the notification rate of new TB was correlated to the proportion of patients with a certain risk factor. Also, the authors use variables in the analysis on risk factors that are not appropriate for this analysis. For instance, non-adherence occurs during treatment, after diagnosis, and thus could be used as a risk factor for unsuccessful treatment outcomes but is not a risk factor for the occurrence of new TB.
KNCV Tuberculosis Foundation is an NGO founded in the Netherlands with a representative office in Almaty, Kazakhstan. Over the last years, we have provided technical assistance to national tuberculosis programs in Central Asia and as such also have gained knowledge and experience with the national electronic tuberculosis register in Kazakhstan. While we highly respect the efforts of the authors to share their insights with the international scientific community, we feel compelled to caution readers when interpreting the conclusions presented here.
Kind regards,
Susan van den Hof (1)
Aigul Tursynbayeva (2)
Svetlana Pak (2)
1. KNCV Tuberculosis Foundation, The Hague, The Netherlands
2. KNCV Tuberculosis Foundation, Representative office for Central Asia, Almaty, Kazakhstan
Competing interests
We declare not to have any competing interests.
Response to Susan van den Hof
Assel Terlikbayeva, GLOBAL HEALTH RESEARCH CENTER OF CA
24 January 2013
1) We worked with data supplied to us from the Kazakh National TB Control Program and the National Institute of Geography. We did not generate the data ourselves, and we cannot attest to the accuracy of how it was collected and how accurate it is. As was pointed out in your comment the "actual coverage and reporting of results has not been and still is not complete", and thus, to a certain extent the data are estimates to a greater degree than one might prefer in an ideal circumstance. Thus, any statements about trends in the overall incidence of TB and particularly drug resistant TB have to be taken with a degree of uncertainty. it is possible that the increase in MDR cases reported in the data represents an increase in detection without a true increase in the number of new cases--time will tell. However, in a circumstance in which by all accounts a significant number of MDR cases remain undetected, it is reasonable to assume that such untreated cases are being transmitted in the community and causing secondary cases. Even for detected cases within the NTP, there are still long delays in obtaining drug susceptibility results (months, generally) and this could also contribute to ongoing transmission. The introduction and wider use of the Hain test and GenExpert may alleviate this situation significantly, and great attention should be given to their integration into the NTP.
We have not seen the paper that apparently is forthcoming in the IJTLD, but a recent paper published by the KNCV team in the same journal (Klinkenberg E, et al. Integration of HIV testing in tuberculosis drug resistance surveillance in Kazakhstan and Kenya. INT J TUBERC LUNG DIS 16(5):615-617) indicates that from 2007-2009 43.4% of all TB cases in KZ for which there were DST results were MDR. This is an astoundingly high rate, and we think it underscores our essential conclusion that the situation of MDRTB in KZ is most serious and deserves great attention. Without significant improvements in availability of DST, maintenance of an adequate drug supply, and strong efforts to assure treatment completion among all patients, MDR rates will remain unacceptably high. We look forward to reading your forthcoming paper. If overall TB rates have come down and the percentage that are MDR stayed the same, then perhaps it is true that MDR rates have "probably" come down, as you state. However, "probably" is not the same as "definitely", and many cases of TB in Kazakhstan are still diagnosed clinically (43% of cases in the IJTLD paper were culture negative). It is possible then, for example, that the apparent decline in cases is really just better classification or diagnosis, i.e. that the cases that have gone away never in fact were TB cases to begin with. This more accurate diagnosis could be because of adoption and integration of better diagnostics, continuing education of general and TB physicians, and overall improvements that have occurred in the NTP. In that circumstance, there would be no actual decrease in MDR cases.
Finally, as to the comment that because p=.12, the change in rates is "not significant": "p" is simply an indication of the likelihood that the observed results are due to chance. Thus, a p value of 0.12 means that there is an 88% chance the results are not due to chance, and that there is in fact a high likelihood that the trend is real. A p value of .05 means that there is a 95% chance that the results are not due to chance. That is all that it means. Eighty-eight percent is still a high likelihood that the finding is not random.
2) From the manuscript, data utilized for this analysis does not differentiate between MDR-TB cases among new TB cases or retreatment cases. The authors raise an important point and one that came up during our preliminary review of the surveillance data. Unfortunately, the data provided for the analysis did not include previous treatment history and thus it was not possible in this analysis to stratify by this indicator.
3) Yes, it is "conceivable that many TB patients do not have one of the major risk factors for TB." But perhaps we do not understand or appreciate all of the risk factors of TB. It is precisely why we are doing the case-control study. It might be of great benefit to the KZ NTP to identify previously unidentified demographic, medical or social risks for TB. This might identify new public health and TB control strategies that could have a significant impact on TB rates in the country.
4) As stating in the manuscript: surveillance data from the National Tuberculosis Program (NTP) and the National Institute of Geography (NIG) were obtained for the years 2006 through 2010. NTP surveillance data included all new tuberculosis cases diagnosed and reported to the NTP from January 1, 2006 through December 31, 2010. Further information is provided that, individual risk factors recorded by the NTP and included in this analysis are alcohol use, child or youth from a vulnerable group, known diabetes diagnosis, history of drug use, incarceration history within the past two years, migrant status, non-regular uptake of anti-tuberculosis medication (less relevant for new cases), penitentiary system staff member, recent mother (having given birth within one year of diagnosis), registered contact of a TB or MDR-TB case, TB health care staff member, and unknown individual risk factors. To clarify, the level of notification rate of new TB cases is individual, from the National Tuberculosis Program data. When running the Pearson correlation the data are reviewed at a national level across the years of reporting, 2007 - 2010.
The non-adherence variable is one collected at the intake of a new TB case, as an attempt to identify known risk factors for developing TB, thus it is not an indication of non-adherence to the current case of TB, but rather to previous cases of TB. Unfortunately, this wasn't documented sufficiently to serve as a proxy indicator for previous TB treatment (point 2 above).
A word of caution: one should take into account the coverage of drug susceptibility testing when interpreting results
16 January 2013
We very much welcome the publication of results on tuberculosis epidemiology from Kazakhstan in BMC Infectious Diseases by researchers from within the country. It is important that insights from Kazakhstan and other Central Asian countries are shared with the international, non Russian speaking, community.
Kazakhstan is a country with a declining incidence of tuberculosis, but with a high prevalence of multi-drug resistant tuberculosis (MDR-TB). It is therefore important to study and know the trends in the MDR-TB burden, and understand risk factors. As we find it an important field of research with important implications, we would like to comment on this manuscript.
The authors state that the incidence of MDR-TB in Kazakhstan has increased in recent years, from 2007 to 2010. This statement is based on the number of diagnosed MDR-TB cases. We would like to point out that the increase as calculated by the authors is not statistically significant (P=0.12). More importantly however, one should realize that the coverage of drug susceptibility testing (DST) and its reporting in the TB register has increased during these years. Although all oblasts have been able to perform culture and DST for years and they have made tremendous progress in recent years, the actual coverage and reporting of results has not been and still is not complete. Increased DST coverage from 2007 onward may also partly explain the trends within oblasts presented in this article: in 2007 DST results were not included in the electronic TB register for 34% of all notified TB patients compared to 47% in 2010. In a study with a different focus but based on the same routine TB surveillance data from Kazakhstan over the years 2007-2011, and accepted for publication in the International Journal of Tuberculosis and Lung Disease (IJTLD), we show that overall MDR-TB prevalence among those with DST results available remains similar over these years. With a decreasing overall TB incidence, this means that the MDR-TB incidence rate in fact probably has gone down.
While taking into account DST coverage, it would have been advisable to stratify the data by treatment history, as this is a main determinant for MDR-TB and could show trends in MDR-TB rates for new and retreatment patients separately. This is important as drug resistance among new TB patients is a marker for transmission of MDR-TB, while MDR-TB prevalence among retreatment patients at least in part is due to acquired resistance.
Most of the patients (83-87%) did not have one of the predefined risk factors registered. The authors seem to interpret this as if these patients have an existent but unknown risk factor, which needs further research. However, especially in a TB endemic country, it is very well conceivable that many TB patients do not have one of the major risk factors for TB.
From the methods and results presented, we are not confident about the statistical methods used. For instance, we are not sure whether and if yes, at which geographical level the notification rate of new TB was correlated to the proportion of patients with a certain risk factor. Also, the authors use variables in the analysis on risk factors that are not appropriate for this analysis. For instance, non-adherence occurs during treatment, after diagnosis, and thus could be used as a risk factor for unsuccessful treatment outcomes but is not a risk factor for the occurrence of new TB.
KNCV Tuberculosis Foundation is an NGO founded in the Netherlands with a representative office in Almaty, Kazakhstan. Over the last years, we have provided technical assistance to national tuberculosis programs in Central Asia and as such also have gained knowledge and experience with the national electronic tuberculosis register in Kazakhstan. While we highly respect the efforts of the authors to share their insights with the international scientific community, we feel compelled to caution readers when interpreting the conclusions presented here.
Kind regards,
Susan van den Hof (1)
Aigul Tursynbayeva (2)
Svetlana Pak (2)
1. KNCV Tuberculosis Foundation, The Hague, The Netherlands
2. KNCV Tuberculosis Foundation, Representative office for Central Asia, Almaty, Kazakhstan
Competing interests
We declare not to have any competing interests.
Response to Susan van den Hof
24 January 2013
1) We worked with data supplied to us from the Kazakh National TB Control Program and the National Institute of Geography. We did not generate the data ourselves, and we cannot attest to the accuracy of how it was collected and how accurate it is. As was pointed out in your comment the "actual coverage and reporting of results has not been and still is not complete", and thus, to a certain extent the data are estimates to a greater degree than one might prefer in an ideal circumstance. Thus, any statements about trends in the overall incidence of TB and particularly drug resistant TB have to be taken with a degree of uncertainty. it is possible that the increase in MDR cases reported in the data represents an increase in detection without a true increase in the number of new cases--time will tell. However, in a circumstance in which by all accounts a significant number of MDR cases remain undetected, it is reasonable to assume that such untreated cases are being transmitted in the community and causing secondary cases. Even for detected cases within the NTP, there are still long delays in obtaining drug susceptibility results (months, generally) and this could also contribute to ongoing transmission. The introduction and wider use of the Hain test and GenExpert may alleviate this situation significantly, and great attention should be given to their integration into the NTP.
We have not seen the paper that apparently is forthcoming in the IJTLD, but a recent paper published by the KNCV team in the same journal (Klinkenberg E, et al. Integration of HIV testing in tuberculosis drug resistance surveillance in Kazakhstan and Kenya. INT J TUBERC LUNG DIS 16(5):615-617) indicates that from 2007-2009 43.4% of all TB cases in KZ for which there were DST results were MDR. This is an astoundingly high rate, and we think it underscores our essential conclusion that the situation of MDRTB in KZ is most serious and deserves great attention. Without significant improvements in availability of DST, maintenance of an adequate drug supply, and strong efforts to assure treatment completion among all patients, MDR rates will remain unacceptably high. We look forward to reading your forthcoming paper. If overall TB rates have come down and the percentage that are MDR stayed the same, then perhaps it is true that MDR rates have "probably" come down, as you state. However, "probably" is not the same as "definitely", and many cases of TB in Kazakhstan are still diagnosed clinically (43% of cases in the IJTLD paper were culture negative). It is possible then, for example, that the apparent decline in cases is really just better classification or diagnosis, i.e. that the cases that have gone away never in fact were TB cases to begin with. This more accurate diagnosis could be because of adoption and integration of better diagnostics, continuing education of general and TB physicians, and overall improvements that have occurred in the NTP. In that circumstance, there would be no actual decrease in MDR cases.
Finally, as to the comment that because p=.12, the change in rates is "not significant": "p" is simply an indication of the likelihood that the observed results are due to chance. Thus, a p value of 0.12 means that there is an 88% chance the results are not due to chance, and that there is in fact a high likelihood that the trend is real. A p value of .05 means that there is a 95% chance that the results are not due to chance. That is all that it means. Eighty-eight percent is still a high likelihood that the finding is not random.
2) From the manuscript, data utilized for this analysis does not differentiate between MDR-TB cases among new TB cases or retreatment cases. The authors raise an important point and one that came up during our preliminary review of the surveillance data. Unfortunately, the data provided for the analysis did not include previous treatment history and thus it was not possible in this analysis to stratify by this indicator.
3) Yes, it is "conceivable that many TB patients do not have one of the major risk factors for TB." But perhaps we do not understand or appreciate all of the risk factors of TB. It is precisely why we are doing the case-control study. It might be of great benefit to the KZ NTP to identify previously unidentified demographic, medical or social risks for TB. This might identify new public health and TB control strategies that could have a significant impact on TB rates in the country.
4) As stating in the manuscript: surveillance data from the National Tuberculosis Program (NTP) and the National Institute of Geography (NIG) were obtained for the years 2006 through 2010. NTP surveillance data included all new tuberculosis cases diagnosed and reported to the NTP from January 1, 2006 through December 31, 2010. Further information is provided that, individual risk factors recorded by the NTP and included in this analysis are alcohol use, child or youth from a vulnerable group, known diabetes diagnosis, history of drug use, incarceration history within the past two years, migrant status, non-regular uptake of anti-tuberculosis medication (less relevant for new cases), penitentiary system staff member, recent mother (having given birth within one year of diagnosis), registered contact of a TB or MDR-TB case, TB health care staff member, and unknown individual risk factors. To clarify, the level of notification rate of new TB cases is individual, from the National Tuberculosis Program data. When running the Pearson correlation the data are reviewed at a national level across the years of reporting, 2007 - 2010.
The non-adherence variable is one collected at the intake of a new TB case, as an attempt to identify known risk factors for developing TB, thus it is not an indication of non-adherence to the current case of TB, but rather to previous cases of TB. Unfortunately, this wasn't documented sufficiently to serve as a proxy indicator for previous TB treatment (point 2 above).
Kind regards,
Assel Terlikbayeva (1)
Sabrina Hermosilla (2)
Sandro Galea (2)
Neil Schluger (2)
Nabila El-Basse (2)
1) Global Health Research Center of Central Asia
2) Columbia University in the City of New York
Type your comment here...
Competing interests
None declared