• Welcome to Phoenix Rising!

    Created in 2008, Phoenix Rising is the largest and oldest forum dedicated to furthering the understanding of and finding treatments for complex chronic illnesses such as chronic fatigue syndrome (ME/CFS), fibromyalgia (FM), long COVID, postural orthostatic tachycardia syndrome (POTS), mast cell activation syndrome (MCAS), and allied diseases.

    To become a member, simply click the Register button at the top right.

PACE Trial follow-up: Here's the table looking at the effects of having CBT or GET after 52 weeks

Sean

Senior Member
Messages
7,378
However, another option is to use lines, as below, to indicate CUD. I could then drop both the difference and the CUD bar (if I did that, I could probably do both fatigue and function on the same graph too).
For CUD I presume the technically correct thing is to show the full range, not just the upper limit, even though is what the effect would have to reach to be certain of being a CUD. In this case CBT/GET doesn't even reach the bottom of that range, so you lose nothing by using the full range, and also avoid the risk of being accused of cherry-picking.

Still think you should ditch the Difference bar, if it is not significant. Then shift the CUD bar over next to the None bar, and also take out the green line and the text immediately above it. I'd take out the dashed line too, I think it is very clear from the 3 remaining bars alone that CBT/GET fell short of CUD, and doesn't need extra highlighting.

You are only trying to make two points. That CBT/GET offered no advantage over APT/SMC, and that CBT/GET did not reach CUD. That is the whole guts of this paper. Everything else is irrelevant distracting clutter.

By higher res I meant maybe 50% increase in linear dimensions, and might also help if you use a crisper font, that one is a little fuzzy, to my eyes at least.
 
Last edited:

Simon

Senior Member
Messages
3,789
Location
Monmouth, UK
Thanks, Sean
For CUD I presume the technically correct thing is to show the full range, not just the upper limit, even though is what the effect would have to reach to be certain of being a CUD. In this case CBT/GET doesn't even reach the bottom of that range, so you lose nothing by using the full range, and also avoid the risk of being accused of cherry-picking.
Um, didn't really understand that point about range - how should I represent CUD?

Here's my latest effort. The text turned out to be in dark grey, so have made black. I made the fonts bigger, but as the graph is bigger too, perhaps not by enough.


upload_2015-11-1_15-46-51.png
 

Sean

Senior Member
Messages
7,378
That is exactly what I was thinking. :)

Definitely scale up the fonts, the numerals in particular have to be clear and proportional. Bold text also sometimes looks better de-bolded when scaled up, so try that too.

Possibly also narrow the bars a bit, around 25-30%.

Remove the black border on the CUD box.

Maybe darken the None bar a little to increase contrast from the background, or a slightly different colour.

Otherwise, bewdiful. The whole paper exposed in one simple clean image. :thumbsup:

Coyne it too. :)
 

SOC

Senior Member
Messages
7,849
@Simon, just my 2 cents --

The CUD box to the side probably won't read to a lay audience. I'd either stack the box with an internal descriptor (CUD written inside the box) or use boundary lines only.

A question -- if the scale is 0-33 as listed in the title, is using only 0 to -6 on the dependent scale making the difference (although small) look more significant than it is? I'm trying to look at this as a lay reader rather than a technical one. Is that your audience -- the general public? If so, Mean Improvement with positive numbers might read more clearly to them than Mean Change with negative numbers. The general public doesn't understand negative numbers very well, especially when it's a double negative -- a negative change is good.

Again, it's about whether this is for a technical audience or the general public. For a technical audience you can assume more ability to understand the subtleties of graphical presentation and stick more strictly to the data, like using negative change as positive. For a lay audience, you need to simplify the graphic to get the point across and not worry about changing negative change scores to positive improvement scores as long as you maintain accuracy of information. It's about the message and the audience.
 

Simon

Senior Member
Messages
3,789
Location
Monmouth, UK
Thanks, Sean
How did you get the CUD range, or was that given in the paper?
Clinically useful difference from original PACE paper (White, 2011).

Think I've done most of your changes.

Thanks @SOC
A question -- if the scale is 0-33 as listed in the title, is using only 0 to -6 on the dependent scale making the difference (although small) look more significant than it is? I'm trying to look at this as a lay reader rather than a technical one. Is that your audience -- the general public?
Re 0-33, think that would make the graph too hard to read; have rescaled to max of 12, which realisitcally about the highest change possible given the scores at the end of one year, and makes the changes look more to scale. Could push this up a bit if needed. Will be a similar position for SF36 Physical function.

Not quite the general public, those interested in research who would have some kind of feel for a graph, but have gone for positive numbers as you suggest.

The CUD box to the side probably won't read to a lay audience. I'd either stack the box with an internal descriptor (CUD written inside the box) or use boundary lines only.
Didn't quite understand this. See below for some changes.
Note I've moved CUD to the middle, and added back the dashed line, which I hope is clearer. IF not that can move back, but is this the kind of thing you had in mind?

upload_2015-11-1_18-52-40.png


Caption:
"Improvement in Chalder Fatigue scores (note a change of -2.0, ie a drop in fatigue score, becomes an improvement of +2.0). Source: my analysis of data from long-term follow-up for PACE trial, (Sharpe, 2015. Appendix Table C). *Clinically useful difference from original PACE paper (White, 2011)."

And thanks to @wdb for checking my numbers.
 

SOC

Senior Member
Messages
7,849
Thanks, Sean

Clinically useful difference from original PACE paper (White, 2011).

Think I've done most of your changes.

Thanks @SOC

Re 0-33, think that would make the graph too hard to read; have rescaled to max of 12, which realisitcally about the highest change possible given the scores at the end of one year, and makes the changes look more to scale. Could push this up a bit if needed. Will be a similar position for SF36 Physical function.

Not quite the general public, those interested in research who would have some kind of feel for a graph, but have gone for positive numbers as you suggest.


Didn't quite understand this. See below for some changes.
Note I've moved CUD to the middle, and added back the dashed line, which I hope is clearer. IF not that can move back, but is this the kind of thing you had in mind?

View attachment 13355

Caption:
"Improvement in Chalder Fatigue scores (note a change of -2.0, ie a drop in fatigue score, becomes an improvement of +2.0). Source: my analysis of data from long-term follow-up for PACE trial, (Sharpe, 2015. Appendix Table C). *Clinically useful difference from original PACE paper (White, 2011)."

And thanks to @wdb for checking my numbers.
Looks good to me. :) You've put a lot of work into this and it now sends a very clear message. :thumbsup:

Minor point to consider -- you could put "Clinically Useful Difference as Defined in the Original PACE paper [reference]" to clarify that you are using their own definition of clinically useful, not yours. Not a big issue.
 

SOC

Senior Member
Messages
7,849
Didn't quite understand this.
Just to clarify :), I got used to making graphs of technical info for politicos whose took their last math class at 15yo, 50 years ago. They don't want to have to think when they see a graphic. They want their first impression to be the final message. So I was thinking that stacking the CUD box on the Any box would say "Here's the baseline. Anything has to be this much bigger to be considered significantly different." What you have is clear. I'm not suggesting you change, just trying to explain.

I may not be making any sense. I've been looking at this stuff too long and my brain is giving out. :depressed:
 

SOC

Senior Member
Messages
7,849
First glance at the graph and I think the "clinically useful difference" just looks confusing and detracts from the overall picture.
I admit I was thinking the same thing... for a non-technical audience.
 
Messages
13,774
Maybe it would be better to just have it say '2.0 was considered to be a clinically useful difference.' ? Having the illustrative bar might be more confusing than useful?
 

Sean

Senior Member
Messages
7,378
Again, it's about whether this is for a technical audience or the general public.
It's for both, I presume.

"Clinically Useful Difference as Defined in the Original PACE paper [reference]" to clarify that you are using their own definition of clinically useful, not yours.
I agree. Maybe like this:

Clinically Useful Difference (as Defined in the Original PACE paper)

••••••••••

1.
On the scale issue, extending it is squashing those bars down and opening up a lot of space in the middle. Put a double lightning bolt axis break between 6.0 and 8.0 and turn the 8.0 into 33.0, (or maybe between the 8.0 and 10.0), and stretch the axis and bars back up.

Then readjust the position of the arrow and Better, and the title, to suit.

2.
Do the axis numerals need the decimal point?

3.
Font could even be a bit bigger, and you missed adjusting the 3.3 in the CBT/GET bar.

4.
Maybe change None to either No CBT/GET, or APT/SMC only.

5.
Not convinced the new arrangement of the bars works. I still prefer the previous arrangement of the bars, it is cleaner to my eye. (But keeping the narrower bars, font & colour adjustment, etc, of course.)

The main comparison or takeaway message that you want everybody to understand is that CBT/GET does not do any better than APT/SMC only. You want that to be the centre feature.

A lot of the message and emphasis comes from the CBT/GET and None bars sitting side by side for easy and even unavoidable visual comparison. If you separate those two bars with the CUD bar, you lose the clarity and impact of that comparison.

I do think you should include the CUD. That is useful additional info for the more technical people to further understand just how poor these scores are absolutely, and the relationship of them to the reality of patients' lives. If you put the CUD bar on the side it will avoid cluttering up the main comparison, but still be in plain sight.
 

Forbin

Senior Member
Messages
966
So, if I understand this correctly, I've achieved all of the benefits of one year of GET and CBT by not engaging in them 2.5 years ago.

Of course, correlation does not imply causation. It's quite possible that some other therapy that I did not engage in 2.5 years ago is responsible for these benefits.
 

Anna Wood

wood/sheridan
Messages
487
fatigue_followup.png

Re-posting this here - it was suggested I indicate what the 'normal' values for each are - is that 0 for fatigue? What about PF - 100? Or is a range (I know I've read this somewhere, but can't think where). I'll do a neater version too.
 

Anna Wood

wood/sheridan
Messages
487

I've not read any of the posts about this, but I thought it might be useful to say how I see the graph for the first time - my immediate thought is I don't understand why the green box sits above the 3.3 line. Both the other two boxes have numbers greater than 2, so surely they are clinically significant.... Should the green box actually got all the way down to the x axis?
 

Sean

Senior Member
Messages
7,378
I've not read any of the posts about this, but I thought it might be useful to say how I see the graph for the first time - my immediate thought is I don't understand why the green box sits above the 3.3 line. Both the other two boxes have numbers greater than 2, so surely they are clinically significant.... Should the green box actually got all the way down to the x axis?
My understanding is that CUD is like statistical significance, in that it is a level that has to be reached, not a measure from 0. It is a range (box) for reasons of statistical uncertainty. That is, we can only say the the CUD value lies somewhere inside that range, and to be 100% sure of reaching a CUD you have to hit the upper end of that range.

It is also called Minimum Clinically Useful Difference, if that helps.

The point of adding it is that CBT/GET definitely did not reach even the minimum possible value for CUD and so cannot said to be a CUD. (And None – no CBT/GET – just reaching into the bottom of the CUD range means only that it probably hit actual CUD, but not definitely.)