There are two separate issues here.
The first is the high rate of diarrhoea that was seen in trial participants who were on trofinetide – 80% versus 20% in the placebo group. As has already been discussed here before (a) the diarrhoea was mostly mild to moderate (b) it was successfully managed during the lengthy extension trial (c) it may be viewed as a lesser evil than Rett-associated constipation (also 80%) which routinely necessitates additional medication and can lead to hospitalization and (d) it appears to have been acceptable to most parents as a trade-off to the benefit derived from trofinetide. This leads me to think that this will not be viewed by the FDA as a barrier to approval.
That said, Acadia will almost certainly need to address the issue of diarrhoea and use of concomitant medications in its labelling.
The second issue is the potential that the much higher rate of diarrhoea in the trofinetide trial population caused functional unblinding of the study. If there is a major imbalance in adverse event profiles between active and placebo groups in a trial, it is possible that the patient or the investigator can guess or predict, based on the adverse event profile, which treatment group the patient is on. This can lead to bias.
The potential for functional unblinding is something that needs to be considered in all clinical trials.
The FDA is aware that ensuring perfect blinding is difficult for many treatments, especially where there is an imbalance of side effects between active and placebo groups. This can be seen in two FDA 2019 Draft Guidances:
Rare Diseases: Common Issues in Drug Development Guidance for Industry as well as
Placebos and Blinding in Randomized Controlled Cancer Clinical Trials for Drug and Biological Products Guidance for Industry. The FDA makes it clear that maintaining perfect blinding may be very difficult for some treatments
Treatment-assignment blinding is important to lessen the potential for bias in trial results, but ensuring perfect blinding is difficult for many treatments. An example of potential unblinding is when all patients receiving an experimental drug develop a certain side effect or requires a procedure/surgery, yet no patient in the placebo arm has the same side effect or procedure. When the primary endpoint is clinically meaningful but susceptible to individual interpretation, the trial may benefit from having additional supportive secondary endpoints (e.g., laboratory measurements). Additionally, use of performance outcome assessments (e.g., cognitive tests, ambulation tests), administered by trained health care professionals (blinded to treatment) and standardized across patients and sites, may complement reports from caregivers and patients regarding the relevant aspects of day-to-day functioning….. It is possible for sponsors to assess the adequacy/success of blinding at the end of a trial.Acadia is also aware that the issue of functional unblinding needs to be addressed.
When questioned by analysts about the potential for functional unblinding of the Lavender trial because of the high rate of diarrhoea in the trofinetide population, CEO Steve Davis said that the regulatory body will “automatically” look at this issue.
In response, Davis said that Acadia has done a great deal of statistical analysis and is highly confident that the data proves that there was no functional unblinding.
He stated that data comparisons have been made between those on trofinetide, with and without diarrhoea, and those on placebo, with and without diarrhoea. No meaningful difference in response was found between those participants who had diarrhoea and those who didn’t.
While one might entertain the possibility that Acadia is engaging in BS or bluff here, the company’s massive ramp up to launch, including the hiring of dozens of staff, indicates that it is indeed confident of approval.
Acadia will need to demonstrate to the FDA that it has attempted to lessen the potential for bias in the Lavender results through trial design. It has done that. The trial was placebo-controlled, double-blinded and quadruple-masked.
The potential impact of subjectivity in caregivers in assessing the RSBQ has been addressed by the addition of a second primary endpoint, the CGI-I which was assessed by (blinded) medical professionals. Both primary endpoints were met.
Acadia will also need to demonstrate that it has assessed blinding at the end of the trial and found it to be adequate. Acadia says it has done this thoroughly and is confident in its analysis.
Acadia will likely also use data from the year-long open label extension study to provide further evidence related to the diarrhoea issue. In the Lilac study, it was demonstrated that the diarrhoea could be successfully managed (stated by the CSO) and tolerability issues did not lead to any dropouts (also stated by the CSO). In addition, data from continuing assessment of both RSBQ and CGI-I could be analysed to see if there were any major differences in reported outcomes from the main, blinded trial.
I feel that there’s little more that Acadia could have done to address the issue of potential bias in the Lavender trial. The irony of the diarrhoea issue is that, while it could have been simply managed during the main trial through intervention, such intervention wasn’t possible because of the risk of unblinding the trial!
http://onbiostatistics.blogspot.com/2018/10/potential-unblinding-due-to-imbalance.htmlhttps://www.fda.gov/regulatory-info...mon-issues-drug-development-guidance-industryPlacebos and Blinding in Randomized Controlled Cancer Clinical Trials for Drug and Biological Products Guidance for Industry | FDA