WHO issues a position statement on the continued use of the shorter MDR-TB regimen following an expedited review of the STREAM Stage 1 preliminary result.
13 April, Geneva - In response to several queries to the World Health Organization (WHO) from member states and other major stakeholders on the continued use of the shorter regimen for multidrug-resistant TB (MDR-TB), following the release of preliminary results of the STREAM Stage 1 phase III randomized controlled trial, the WHO Global Tuberculosis Programme completed an expedited review of data provided by the STREAM trial consortium in January 2018.
The STREAM trial was initiated by The Union in 2012 with its main partner, the Medical Research Council Clinical Trials Unit at UCL. It is mainly funded by the U.S. Agency for International Development (USAID), with additional support from the UK Medical Research Council and the UK Department for International Development (DFID).
Based on the findings outlined in this Position Statement, WHO is advising national TB programmes and other stakeholders to continue using the shorter MDR-TB regimen under the same conditions made in 2016 when WHO first recommended its use.
The STREAM Stage 1 trial recruited 424 study participants from seven sites in four countries (Ethiopia, Mongolia, South Africa and Viet Nam) between 2012 and 2015. Study participants were randomized to receive either the shorter MDR-TB regimen (lasting up to 48 weeks) or a longer regimen of 18-24 months designed according to WHO recommendations. The study was designed to assess “non-inferiority” of the shorter regimen. Non-inferiority trials are intended to show that a novel medicine or regimen is no worse than the usual treatment recommended. Non-inferiority trials do not test whether one regimen is superior to another.
"We appreciate the significant work done by the study investigators to generate randomised controlled clinical trial evidence for the shorter regimen," said Dr Tereza Kasaeva, Director of WHO’s Global TB Programme. “Having evidence-based policies is key to improving care for patients affected by the global public health crisis of MDR-TB.”
The premise of the STREAM Stage 1 trial was that the shorter regimen would have a non-inferiority limit of 10%, i.e. that the difference in efficacy between the shorter and the longer regimen would statistically be no more than 10%.
The preliminary results of the STREAM Stage 1 trial showed that “favourable outcome” in patients treated with longer regimens was marginally higher than in those receiving the shorter regimen (80.6% vs. 78.1% respectively (adjusted value; +2.1%; 95% confidence interval -6.9% to +11.2%). The upper limit (which is conventionally the focus of non-inferiority trials) of 11.2% exceeded the 10% non-inferiority margin specified in the study protocol. Therefore, the preliminary results could not confirm non-inferiority of the shorter regimen when compared with the longer regimen.
The STREAM Stage 1 preliminary data could also not conclusively confirm that the longer regimen was actually better than the shorter regimen, given that the 95% confidence interval exceeded well above and below zero (statistically reflecting “no-difference-in-effect”) and considering the small difference in favourable outcome between the two study arms.
When compared to programmatic data on MDR-TB patient outcomes at the time the STREAM Stage 1 trial started, a higher proportion of patients than expected had a successful outcome on longer regimens, narrowing the originally expected difference in “favourable outcome” between the two study arms. This finding may partly be explained by the careful selection of patients to be treated under trial conditions and the high proportion of patients in both study arms who managed to stay on treatment due to strong support to ensure adherence.
Overall, there was not a statistically significant difference in deaths (from all causes) between the two study arms: 8.5% of the study arm participants had died compared to 6.4% of the control participants. In the subgroup of patients with HIV, the number of observed deaths was higher in the study arm (18/103; 17.5%) compared to the control arm (4/50; 8.0%); the number of observations was small and the difference was not statistically significant.
Most of the deaths in the subgroup of patients with HIV occurred in the South African sites, which also enrolled the majority of the HIV-positive participants in the study. The higher proportion of patients with HIV who died in the study arm may have been confounded by the presence of more advanced HIV-associated disease in that arm, and by site of recruitment. Furthermore, the composition and timing of antiretroviral therapy in the HIV-infected participants were not available for this expedited review, hampering full understanding of the reported mortality. This is a patient group in whom medication adherence support and closer monitoring for clinical response and adverse events is particularly advised. Most importantly, antiretroviral treatment regimens should be initiated early in accordance with WHO recommendations.
“The STREAM Stage 1 trial endpoints do not directly reflect the relative benefits of reducing treatment duration when the shorter regimen is used in eligible patients,” said Dr Karin Weyer, Coordinator of Laboratories, Diagnostics and Drug Resistance at WHO’s Global TB Programme. “As a result, the clear advantage to both patients and national TB programmes may be understated by focusing exclusively on the non-inferiority aspect of the trial. We have therefore called for more data on the use of the shorter regimen to fully assess its feasibility, effectiveness and safety among different patient subgroups".
Further to this expedited review WHO is undertaking an extensive review of its MDR-TB treatment policy guidelines this year, which will result in consolidated drug-resistant TB treatment guidelines and an update of the WHO Companion Handbook for the management of drug-resistant TB.
-  Statistically, findings of clinical trials are usually presented as a single estimate of effect (eg. % difference in efficacy between two regimens). These values are also always presented with 95% confidence intervals, which is the expected range of differences in efficacy around the point estimate.