Questions for the PACE trial PIs

This clinic did not cover questions in regard to funding, research or complaints against the NHS. Such questions need to be addressed to the local GPC (General Practitioner Council). However as there was so much interest on these subjects we have left many of them under this heading for viewing but they are intentionally not answered by the panel.

Moderator: talkhealth

Locked
2 posts
adrian01
Posts: 3
Joined: Mon Aug 19, 2013 6:30 pm
Quote

by adrian01 on Mon Aug 19, 2013 9:23 pm

Questions for the PACE trial PIs

I see that the PACE trial authors have issued a statement that you have put in a locked thread suggesting that they answer questions. Are they also answering questions?

I have some specific questions for them could you put the questions to them or are they only making statements.

1) What mechanism did they use to control for different expectations within the different trial arms. I've read papers saying that
Although active control groups are superior to “no-contact” controls, only when the active control group has the same expectation of improvement as the experimental group can we attribute differential improvements to the potency of the treatment.
From Walter R. Boot, Daniel J. Simons, Cary Stothart, and Cassie Stutts The Pervasive Problem With Placebos in Psychology: Why Active Control Groups Are Not Sufficient to Rule Out Placebo Effects
Perspectives on Psychological Science July 2013 8: 445-454, doi:10.1177/1745691613491271


Yet the APT and SMC group to me seem to be setting up different expectations to the CBT and GET group. This is potentially a serious problem due to the reliance on subjective outcome measures. I believe Wechsler's asthma study showed the dangers of relying on subjective measures where they showed a placebo effect with the subjective measures but not objective ones,

Hence the question how do you know that its not just different levels of placebo effect based on different expectations being measured.

2) In their recovery paper they justify changing the threshold on the short form 36 physical function scale from 85 to 60 with the claim that half the working age population scores under 85. They quote a paper by bowling which is based on the 1992 OPCS Omnibus Survey, which is available in the national data archive. Checking the data shows that the median value (i.e. half) of all working age (less than 65) population is 100. That is at least half of the working age population score 100 according to the study they quote. In fact only 8% of the working age population without long term health issues fall below a score of 85.
Hence there is a question of the validity of the changed thresholds with their mistake being made by failing to take account of the underlying data distribution.
Given this do Profs White, Sharpe and Chadler stick by their revised threshold.
3) Which other trials can they site that have a similar wide scale protocol changes. In developing a statistical plan it may be normal to look at the data distributions to choose the appropriate statistical techniques but I believe that changing outcome measures is less usual and with drug trials would require the approval of the ethics committee that approved the trial.



I would appreciate it if you could get them to answer these questions.

biophile.pr
Posts: 6
Joined: Fri Aug 16, 2013 9:20 am
Quote

by biophile.pr on Tue Aug 20, 2013 2:13 am

Re: Questions for the PACE trial PIs

Very interesting, adrian01. Thankyou for bringing this to attention. I would also like it answered.

As the actual dataset shows that only 8% of the most relevant population sample are excluded when using ≥ 85/100 points as the threshold for normal, not "approximately half" as implied by White et al, this reveals a serious flaw in the PACE threshold for "normal" physical function, which also underpinned their post-hoc weakened definition of recovery. It helps explain how 13% of trial participants had "normal" physical function scores at baseline despite these scores counting then as "significant disability".

Based on their stated reasoning in the Psychological Medicine paper in February 2013 and their subsequent citing of the mean(SD) scores from Bowling et al (1999), their justification for this change from 85 to 60 apparently assumed that the mean (average score) and median (middle score) were about the same.

However it is quite clear that the mean and the median are not in fact similar and therefore the stated reason for changing the threshold was erroneous. The paper by Bowling et al even warned readers that the dataset being reported on had a heavily skewed non-normal distribution, which is also obvious from the histogram given for physical function. Even without accessing the original dataset, when looking at the Bowling et al paper, said histogram suggests the median is either 95 or 100 (top score).

White et al stated in their recent response here that changes to the trial protocol were approved independently by two trial oversight committees. This would include the change to the physical function threshold. It would be rather concerning if such a basic error managed to pass three groups of professionals involved with the PACE Trial, not to mention being unspotted by multiple peer-reviewers in at least two journals, including the Lancet after what its editor in chief described as "endless rounds of peer review".

Locked
2 posts