Could MSB’s ARDStrial be adapted? Hi all, withthe expectation...

Could MSB’s ARDStrial be adapted?

Hi all, withthe expectation that MSB’s ARDS trial data will soon be unblinded, perhaps thisweek, this post is to explore (well really to speculate on) a scenario where MSB negotiatesFDA approval to adapt and continue with the current trial, as opposed tostarting a new one, if the unblinded data shows some efficacy.

FDA guidancedocuments relating to trial adaptations

The FDA havepublished guidance on Adaptive Designs for Clinical Trials of Drugs andBiologics. (Nov 19; https://www.fda.gov/media/78495/download).This document provides commentary adaptive trials, including prospectivelyplanned Bayesian interim analysis and, relevant to this scenario, changes to the design ofa trial based on comparative interim results.

Page 23indicates that the FDA may accept well motivated changes to trial design (evenunplanned) as long as there is discussion with the FDA and that there is controlof the Type 1 error rate to maintain the trial integrity. Further to this, page24 discusses how the FDA may assess this in the context of new information thatarises from sources external to the trial (e.g. disease pathphysiologyor patient characterisation) that may have changed the trial design had thatinformation been known at the trial commencement.

Of course, weare in an emergency pandemic with a novel virus and also a rapidly changingtreatment landscape. The FDA have also published a separate document which elaborateson this guidance, within in the specific context of COVID-19: COVID-19eveloping Drugs and Biological Products for Treatment or Prevention Guidancefor Industry (May 2020; https://www.fda.gov/media/136238/download).

This documentgenerally acknowledges that the pandemic has created challenges in conductingtrials and also that emerging developments may result in a need for protocolchanges. The following extract in italics (my underlining) is from Page 8:

If a trialincorporates the possibility of early stopping for evidence of benefit orany adaptations to the sample size, dosing arms, or other design features,sponsors should prospectively plan the design in a manner to ensure control ofthe type I error rate and reliable treatment effect estimation. An independentcommittee, such as a DMC, should be tasked with providing any recommendationsfor early termination or design adaptations based on unblinded interimdata.

FDAanticipates events that occur outside of an ongoing trial may provide importantnew information relevant to the ongoing trial (e.g., changes to the standardof care) and may motivate revisions to the trial design. Well-motivatedchanges based on information external to the trial can be acceptable andsponsors are encouraged to discuss these changes with the FDA.

Could MSB have prospectivelyplanned for an adaptation to the ARDS trial?

In terms of prospectiveplanned adaptations to their trial, MSB’s only specific announcement wasthat a DSMB would review the trial at 30%, 45% and 60% for overwhelmingefficacy or futility.

However, we also know that MSB consulted with the FDA and changed their sample sizewhen initiating it. MSB confirmed on 30 April that they had set a sample size based on specific guidance provided bythe FDA for robust statistical analysis...to 300 ventilator dependentpatients.

It is likelythat these discussions with the FDA in April would have involved discussionsregarding treatment effect estimation and control of the type I errorrate. If at the time of this discussion, the background mortality rate was60%, and the treatment effect estimation was at least a reduction inmortality of 43% (the Mt Sinai pilot data indicated a much greater affect thanthis) then a sample size of less than only 200 subjects would have adequatelypowered the study with a Type I error rate of 5% (this would also have had aType II error rate of 5% - see Appendix 1).

So why wasthere an increase in the sample size to 300? Were the FDA driving a particularlyrigorous control of the Type 1 error rates and/or really require such dramatic reductionsin mortality? In an emergency environment, one would have thought any reductionin mortality would be welcome so why so ambitious?

It's possible that the treatment effect estimation was lower than 43% (hence needing a greater sample size). Another speculation is that the trial was powered allowing for adaptation.

Let’s considera scenario of a 43% treatment effect for n=300. A basic 2x2 Fisher table indicatesthat if a 43% reduction in mortality was achieved with either a 60% or 40%background mortality rate, the p-value of a null hypothesis would be < 0.002(see Appendix 2). A treatment affect approaching this would certainly leave some‘wiggle room’ to adapt and control the Type 1 error rate in the eventthat an adaptation was required at an interim review.

The wordingof the recommendation from the DSMB

MSB’sannouncement on 18 December 2020 regarding the DSMB interim analysis had thiswording (with my underlining):

The DSMBreported that there were no safety concerns and noted that the trial is notlikely to the meet the 30-day mortality reduction endpoint at the planned300 patient enrolment. The DSMB recommended that the trial completewith the currently enrolled 223 patients, and that all be followed-up asplanned.

As others havecommented, the language is certainly not ‘stopped for futility’. The DSMB are infact recommending a form of continuation that will enable maximum use of thedata gathered so far. IMHO, this wording is quite suggestive that some efficacy hasbeen observed, but perhaps that there is a need to change the trial design and/or samplesize.

What externalinformation has arisen since the trial was initially designed?

Since the trialhas commenced, there has been new information external to the trial which,if known and understood at the time of the trial commencement, would quitepossibly have influenced the enrolment criteria. Some issues that have beendiscussed by SI or others are:

1. Changesin medications (especially steroids such as dexamethasone)

2. Changesto the Standard of Care with respect to proning, oxygentation and

3. atendency to delay commencement of intubated ventilation given (1) and (2) above and also thatventilation was observed to be having adverse longer term outcomes in some patients

4. studiesshowing correlations of very high and rising CRP levels with poor clinicaloutcomes

Meanwhile,whilst it has not specifically all been related to Rem-L or acute diseases, other trial results released since May last year may have presented learnings on MSB's stem cells Mechanism of Action. A common theme is that MSB’stherapeutics seems to work better when introduced early in the onset of adisease. i.e. their beneficial effect in modulating damaging inflammation seems to have more of a therapeutic benefit in prophylaxis as opposed to regeneration. Theimportance of timing revealed in chronic might also be relevant in principle to acute forms of inflammatory disease.

A possiblescenario for adapting the trial

My workinghypothesis is that Rem-L will have a noticeably higher efficacy in reduction of30 day mortality when applied early to a ARDS patient populations with elevated/ sharply increasing CRP levels, and perhaps particularly to those patients whohave not responded to steroids. Due to the changes in SoC, and particularly thetendency to delay ventilation and hence enrolment of ventilator dependent patients, these populations may have been treated quickly in the early stages of the pandemic, but treated later in the course of the disease as time went on. This may explain why early interim analysis passed but the last did not.

If unblindeddata is stratified for steroid treatment/response, CRP levels wrt time of Rem-Ltreatment, this might reveal this target population. Could this together with the information external to the trial form a basis to adapt the trial enrolment criteria, increase the sample size andcontinue??

Funding

There has beenmuch discussion on MSB’s cash runway and some implied ‘dependence’ onongoing support from Novartis for another all-cause ARDS trial or a capital raise.

However, if the existingtrial were approved to adapt and continue, would MSB not continue to receive funding supportfrom NIH?

Really lookingforward to hearing about what the data reveals soon!

GLTAH

Appendix 1 –Sample size estimation less than 200 at 95% power based on 43% reduction inmortality from 60% to 34%




Appendix 2 – 2x2 Fisher table analysis of 43% reduction in mortality at n=300