Effect of disease duration in a randomized Phase III trial of rintatolimod, an immune modulator for [ME/CFS], 2020, Mitchell et al

Sorry for my skepticism, but this looks like a lesson in 'how to get around the problems created by a blinded placebo controlled trial'. Answer 1: 'very detailed post hoc stratification'

The measure, exercise treadmill tolerance, as measured once at baseline and once at trial end seems less than ideal. Activity monitoring would be a much better measure.




To be fair, there does seem to be an improvement in the seconds Rintatolimod participants spent on the treadmill compared to the placebo.

But what's going on with the standard deviations and the 95% confidence intervals? The report says the results were distributed normally, and so a 95% confidence interval should be about two times the standard deviation either side of the mean*. Look at the top left result (ITT population, with the Rintatolimod treatment). The mean change was about 96, with a huge standard deviation of 251. So wouldn't the 95% confidence interval be from (96 -502 = 406) seconds to (96 +502= 598) seconds? Instead, it's reported as 64 to 229 seconds. It look weird to me; I'm happy to be corrected through.

I'm wrong (see Trish's note below). But the standard deviations are still very big relative to the means.

 

It's a post hoc analysis, which means it wasn't part of the original plan to analyse the data in this way. It seems to me that they realised that overall the results weren't significant, so they looked at the data and found most of those who showed some improvement were in the 2 to 8 years illness duration. So they looked just at those and found half of them improved a bit. Seems pretty unimpressive to me, but maybe I'm missing something.

 

I think you need to divide the SD by the square root of the sample size, to get the standard error of the mean.
https://www.statisticshowto.com/pro...l for the mean is a way of,and 7 is the upper.

 

Answer #2 in how to market results from a blinded placebo controlled trial: focus on something entirely different from the planned outcomes.

Here are the main results, comparing improvements in cohorts with the treatment and a placebo:

So, it looks as though more patients treated with rintatolimod were able to increase their time on the treadmill by more than 25% than those patients treated with the placebo. Even if you don't buy the post hoc stratification and look at the left-hand bars for the whole trial populations, rintatolimod seems a bit better.

But then there is a chart (using exactly the same purple and grey colours) and lots of discussion comparing those patients who were able to increase their treadmill time by 25%, not with controls, but with those who were not able to increase their time by 25%.

For example, see this:


So, they are telling us that those patients treated with rintatolimod who increased their time on the treadmill by more than 25% also had improvements in two secondary endpoints, not against the controls, but against the patients treated with rintatolimod who didn't increase their time by more than 25%.

They don't report the secondary endpoints for controls at all. And one of those secondary endpoints is just a part of the SF-36, so who knows how many secondary endpoints were measured and not reported.

To me, it looks like a very modest average improvement (in an outcome that isn't that relevant to ME/CFS) (that may possibly include a very small group who really did respond, but might just be chance or measurement bias), wrapped up in a lot of marketing.

 

Here are the scatter plots so you can make up your own mind.

The first one is for the so-called 'target population', the second one is for everyone else. Illness duration is on the x axis, with different scales for each chart. % change in the time on the treadmill (from baseline to end of treatment) is on the y axis.

The blue dots are the patients treated with rintatolimod, the red dots are the placebo.

 

One of the criteria to be included in the 'Target Population' was 'PEM lasting more than 24 hours'. The participants in the non-target population presumably either didn't have PEM or had PEM that lasted less than 24 hours. It's not clear how the PEM criteria was determined, but it looks as though it was patient self-report during trial eligibility screening.