David Tuller: Trial By Error: Professor Sharpe’s Pre-Hearing Briefing for Monaghan

I'll try again...

 

I didn't understand this at all. Why make things up?

 

There is definitely a correlation - but what happens is that the intervention (+ regression to the mean) just shifts everything [well, not everything, but you know what I mean] to the right. There's also a big hole where I suspect the missing data goes.

 

I prefer the gif because it loops back to the beginning. With YouTube video, you lose that.

 

The trouble is, you will probably get a significant simple correlation, because overall, the more severely affected people will tend to rate themselves more severely (on the physical function and Chalder fatigue measures) and will also tend to have have more trouble with the 6MWT.

Could we get around this problem this way? I'm thinking out loud really.

1. Using the SMC arm, calculate the regression equation that expresses the relation between each outcome measure at the start of the trial and the corresponding outcome measure at 52 weeks.

2. Using this equation, express each person's 52-week score on your chosen outcome measure as a residual based on the equation that you just calculated. These new scores will express the degree of change over the course of the trial, relative to what you'd expect for that person's level of severity if there were no therapy. A positive residual would indicate greater change than expected, and a negative residual would indicate less change than expected. The residual scores have overall severity factored out, which will solve a lot of problems.

3. See whether residual scores for your subjective outcome measure(s) correlate with those for your objective measures. Are the correlations tighter for some trial arms than others?

4. There seems to be a fourth step needed to test statistically whether trial arm significantly modulates the correlations you get from step 3. I'm not sure about the maths of this one.

 

I stand corrected by my honorable friend.

 

So I don't clutter up this thread with my gifs any more, I'm going to start a new thread for PACE graphs and gifs...

 

I was too, but thought I might seem like a stickler.

 

Nonetheless, I was poaching your logic TiredSam.

 

I am happy to unashamedly stickle. Especially when Michael Sharpe constructs a sentence that I could use to teach my beginners with:

There is / there are, much and many, has and have, singular / plural, present perfect - all the classic mistakes Germans make when they start learning English. If I didn't know better I'd suspect ... oh no, better steer well clear of Godwin's Law.

 

But asking patients to rate their ability to do things isn't as reliable as actually measuring how well they can do things. Because patients are susceptible to wishful thinking or may not like to admit how ill they are.

I not surprised that in picture posted earlier, physical function seems to increase more over time than walking distance. Walking distance better reflects real physical function. Unfortunately it's surely also more affected by missing data. And to account for PEM it would have to be several walking distance tests over a few days (or until the patient refuses).

 

OK - they *thought* it was a good proxy. And yeah, is easily shifted by persuasion. I thought that's what I was trying to say. Brain not working today.

[I've deleted the post]

 

500m is about a third of a mile? So in 6 minutes that would equate to around a normal walking pace for a healthy person who isn't in any particular hurry. I agree it doesn't tally with an SF-36 of 15 where I would expect walking to be quite restricted, if possible, and significantly slower than a "normal" walking pace, again, if possible at all.

There'd have to be a pot of gold, on a hover-sled, at the end, and a few hungry tigers behind me to get me to go that far in 6 minutes, and 98 times out of 50 I'd be lunch.

 

I'm defo having an "off" day. I'm talking rubbish. I get extremely frustrated when i can't understand stats any more.

 

No, you are not - I find that implies lack of attention to detail and disrespect in communication towards the audience - as if not bothering to re-read what the person has written before sending it off. Not how you would expect a professor at a world class university to communicate.

Personally, when I receive e.g. commercial communication from a company with bad grammar/spelling, I just disregard the company. (Unless it is obvious they are foreign and did their best in English.)

 

The argument that GET is only harmful if carried out by the wrong person or in the wrong way goes back to the collaboration between the PACE doctors and AFME.

Here's a link to a letter in the BMJ co-authored by Prof Sharpe and Sir Peter Spencer (who was then the new CEO of AFME) in 2007.

https://www.bmj.com/content/335/7617/411

They tried to explain away patient surveys in that way

I'm sorry that I don't have a link to the whole paper

 

Quite. Scientific writing is a thing - I have a friend who teaches courses on it. You would expect precision and accuracy to be second nature to Michael Sharpe if he was half the scientist he claims to be. It's also interesting that the briefing was not signed by the other PACE authors, just purportedly sent in their name. If any one of them had read it before it was sent, surely they'd have spotted such a howler and suggested he correct it.