Reader Comments

Post a new comment on this article

Concerns about the conduct and analysis of this study

Posted by rwaldridge on 27 Nov 2013 at 13:55 GMT

At a time when there is considerable interest in the potential for the fluoroquinolones to shorten treatment for drug-sensitive tuberculosis, the trial conducted by the National Institute for Research in Tuberculosis (NIRT) in Chennai[1] could represent an important contribution to our understanding of whether this is achievable. Unfortunately, we have serious concerns about the conduct and analysis of this study which makes the interpretation of the results challenging.

Because moxifloxacin was not available at the start of the study, patients were only enrolled into the gatifloxacin and control regimens during the first 12 months. Thereafter, an attempt was made to equalise the numbers in each treatment arm by changing the allocation ratio to 2:1:1 in favour of the moxifloxacin arm. However, this meant that probably only around half of the 170 patients on the control arm were enrolled concurrently with patients allocated to the moxifloxacin arm. In addition the gatifloxacin arm was terminated 10 months before the moxifloxacin and control arms. No attempt appears to have been made to compare regimens during concurrent enrolment periods of the study in what would be a properly randomised comparison, in order to examine whether the conclusions would have been altered. Known or unknown differences in patient management or between patients enrolled in the different periods of enrolment could have an impact on the differences between regimens in the analyses presented.

The Data Safety and Monitoring Board (DSMB), whose members and independence are not noted in the report or the study protocol, made a recommendation in February 2006 to terminate the gatifloxacin arm of the study and in October 2006 to terminate the moxifloxacin arm. No information is given either on the guidelines under which the DSMB made this decision, or the results that were available to the DSMB in each case; only that that the decision was due to ‘high TB recurrence rates in these two arms compared to the control arm’[1]. However, most of the data on recurrence (from 24 months follow-up) would have accrued after the study had terminated. The final reported results comparing moxifloxacin with the control arm would suggest there was minimal difference between the outcomes of these two regimens, and that there was no evidence for a difference in the recurrence rate between these two arms (P=0.38). These results suggest the termination of the study is very likely to have been premature. Early termination introduces bias and, as noted in Pocock and White in relation to another trial: ‘the premature pulling out of the trial on insufficiently convincing evidence inhibits the ability of clinical science to resolve an important therapeutic issue’[2].

The culture results at two months are of particular interest since they suggest that 10% more patients receiving moxifloxacin converted at that time compared to the control arm. The authors fail to comment on what this means for the predictive ability of two month culture conversion for successful treatment shortening, a subject recently addressed by other publications in this journal.[3,4]

The results of this study might appear to support those of the OFLOTUB randomised controlled trial, presented at the recent conference of the IUATLD in Paris, which failed to show a shortened daily gatifloxacin based arm was non-inferior to the standard six month daily regimen. However, as we have pointed out, there are several serious limitations in the NIRT trial report, in particular the premature closure of enrolment, which need to be recognised when assessing the conclusions of the study and any future meta-analyses which includes these data.

Rob Aldridge on behalf of the North London TB Journal Club (http://northlondontb.org/).
North London TB Journal Club meets monthly; it is organised by the London School of Hygiene and Tropical Medicine, University College London and MRC Clinical Trials Unit. The points above represent concerns raised during a discussion of this paper at a meeting of the Journal Club on 12th November 2013.

1. Jawahar MS, Banurekha VV, Paramasivan CN, Rahman F, Ramachandran R, et al. (2013) Randomized clinical trial of thrice-weekly 4-month moxifloxacin or gatifloxacin containing regimens in the treatment of new sputum positive pulmonary tuberculosis patients. PLoS One 8: e67030.
2. Pocock S, White I (1999) Trials stopped early: too good to be true? Lancet 353: 943-944.
3. Phillips PP, Fielding K, Nunn AJ (2013) An Evaluation of Culture Results during Treatment for Tuberculosis as Surrogate Endpoints for Treatment Failure and Relapse. PLoS One 8: e63840.
4. Wallis RS, Wang C, Meyer D, Thomas N (2013) Month 2 Culture Status and Treatment Duration as Predictors of Tuberculosis Relapse Risk in a Meta-Regression Model. PLoS One 8: e71116.

No competing interests declared.

RE: Concerns about the conduct and analysis of this study

MSJawahar replied to rwaldridge on 11 Dec 2013 at 07:09 GMT

Thank you for these comments. We agree that in hindsight the recommendation of the DSMB to suspend recruitment to the gatifloxacin regimen initially, and in particular to the moxifloxacin regimen later appear premature. However, these decisions were made on data available at those particular points in time and we were bound to accept the recommendation of the DSMB. The DSMB was an independent body constituted according to established guidelines and consisted of external subject experts and a biostatistician and the only institutional member was the Member-Secretary. In February 2006, there were 13 TB recurrences out of 96 eligible patients in the gatifloxacin arm (18.8/100 PYs) and 3 out of 103 in the control arm (4.9/100 PYs). In October 2006, the available data showed 6 TB recurrences out of 66 in the moxifloxacin arm (14.6/100 PYs) and 3 out of 133 in the control arm (1.9/100 PYs). Since these differences were considered statistically significant the DSMB recommended suspension of recruitment, first to the gatifloxacin arm and then to moxifloxacin and control arms. However patients who had already completed treatment were followed up for 24 months and this is the data we have presented in our paper. And as we have discussed in our paper the TB recurrence in the test regimens (gatifloxacin and moxifloxacin regimens) occurred earlier in the post-treatment follow-up while some of the TB recurrence in the control regimen occurred later during follow-up. Twenty percent (2 out of 10) recurrences in the control regimen occurred beyond 12 months. Thus at the time of the DSMB recommendation the difference in the TB recurrence rates between the gatifloxacin and control regimen and the moxifloxacin and control regimen were larger than they ultimately turned out to be.

The suggestion to look at concurrent comparisons within the study is well merited. We have had a fresh look at the data to address this question and have made concurrent comparison of the 2 quinolone regimens with the control regimen, i.e. from the start for the study to the termination of the gatifloxacin arm, May 2004 - February 2006 (gatifloxacin vs. control), and from the start of enrollment to the moxifloxacin arm to the end of enrollment, May 2005 – October 2006 (moxifloxacin vs. control, 2:1) For the former set there were 19 TB recurrences out of 121 (15.7%) in the gatifloxacin arm vs. 6 out of 128 (4.6%) in the control arm (p = 0.008). For the second set there were 11 TB recurrences out of 107 in the moxifloxacin arm (10.3%) and 6 of 51 in the control arm (11.8%). These data broadly conform to what we have reported overall, while the difference in the TB recurrence rates were significantly higher in the gatifloxacin arm compared to the control arm, no difference was observed in TB recurrence rates between the moxifloxacin arm and the control arm.

With reference to 2nd month sputum culture conversion even though there is a body of opinion that suggest it might serve as a predictor of post-treatment TB recurrence it has been our experience that it does not correlate with post-treatment TB recurrence rates. In two earlier studies at NIRT we have observed that even when the 2nd month culture negative rates were as high as 91% and 92 to 98%, TB recurred in 20% and 8 to13% respectively.1,2

Notwithstanding the limitations in the interpretations of the data from this truncated study, that you have rightly pointed out, and that which we have acknowledged in our paper, the information from the OFLOTUB study presented in the recent IUATLD Conference would seem to support the conclusion we have drawn with reference to the gatifloxacin regimen. While the regimen used in the OFLOTUB study was a daily 4-month regimen, ours was a thrice weekly regimen. It is not surprising that the regimen failed. Due to early termination of our study we could not arrive at a firm conclusion on the efficacy of the thrice-weekly moxifloxacin regimen. And as we have discussed in our paper our current study with 4-month moxifloxacin regimens at the NIRT and the ReMox study may provide the answer for the feasibility of a 4-month regimen with currently available TB drugs, while the world waits for the advent of new TB drugs.

1. Tuberculosis Research Centre. Shortening short course chemotherapy: a randomised clinical trial for treatment of smear-positive pulmonary tuberculosis with regimens using ofloxacin in the intensive phase. Ind J Tuberc, 2002; 49:27–38.
2. Tuberculosis Research Centre, Madras, and National Tuberculosis Institute, Bangalore. A controlled clinical trial of 3- and 5-month regimens in the treatment of sputum-positive pulmonary tuberculosis in South India. Am Rev Respir Dis, 1986; 134: 27–33.

M S Jawahar

No competing interests declared.