• 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.

Williams, White et al: PACE: Heterogeneity in CFS - empirically defined subgroups [...]

user9876

Senior Member
Messages
4,556
They chose a set of variables and then saw clusters. I would assume an amount of playing with data etc to get the clusters they wanted. Given other papers they have written I would be suspect about any stats they use.

Even if it were a good methodology, finding clusters in symptoms or other collected data in no way helps suggest mechanism. At best clusters could provides suggestions to look at not conclusive results.
 

Hajnalka

Senior Member
Messages
910
Location
Germany
The largest, 'core' subgroup (33% of participants), had relatively low scores across all domains and good self-efficacy.
Not sure, if I understand that right. Does that mean, that the largest subgroup is mentally strong and free of false illness beliefs (even from the authors' viewpoint)? That they have a biomedical illness and CBT would do nothing for them?
 
Last edited:

user9876

Senior Member
Messages
4,556
Using PCA to reject variables may well remove outlier clusters which represent a pattern for just a few patients. The way the variables explain the overall variance may be low but it may stick out for a few samples.

What seems strange is they turned continuous variables into binary ones using the median as a threshold. So the PCA is really done on whether some variables are above or below the median which seems strange to me.

But perhaps the variance in the data set just represents the values they have chosen to collect - rather than anything that is an actual effect.

The labelling of clusters is probably the most interesting thing as it represents their interpretation of the features in the cluster and hence expose
 

user9876

Senior Member
Messages
4,556
Good heavens! How could you possibly suspect them of something so underhand as that! :p;)

Given their comment

we acknowledge that the subgroups identified will be limited by the variables selected for inclusion, and that the majority of the decisions relating to variable reduction were undertaken post-hoc.

Similarly a post-hoc decision was made when choosing the number of classes to include in the final latent class model. By using a combination of statistical indices and clinical experience, as in previous literature, however, we have been able to derive clinically recognizable and subsequently validated subgroups of CFS.

I think my comment is fair.

I take the last quote as saying that they interpreted their data with respect to their existing biases.
 

Snow Leopard

Hibernating
Messages
5,902
Location
South Australia
What seems strange is they turned continuous variables into binary ones using the median as a threshold. So the PCA is really done on whether some variables are above or below the median which seems strange to me.

It is indeed strange to me, but I don't know how common this may be in psychology.

non-binary variables for the remaining 541 patients were then dichotomized via median split to produce a binary categorical dataset

Have the all the medians for the continuous variables they've used been published in other papers?

Is this really a valid idea if the data for that variable is normally distributed?

Should the split be defined a priori so that it is not inherently dependent on the specific dataset (because it means any other data set they analyse could have quite different split values)?
 

user9876

Senior Member
Messages
4,556
Have the all the medians for the continuous variables they've used been published in other papers?

Is this really a valid idea if the data for that variable is normally distributed?

Should the split be defined a priori so that it is not inherently dependent on the specific dataset (because it means any other data set they analyse could have quite different split values)?

I don't think medians have been published. The reason for using the median and not the mean is it is more robust and hence good for differing distributions hence used in a lot on non-parametric stats.

But PCA works on continuous variables (assuming they mean principle component analysis)
 

Snow Leopard

Hibernating
Messages
5,902
Location
South Australia
Also,

Further validation was undertaken comparing the groups using age, sex, illness duration and whether they met the CDC or London criteria for CFS (The National Task Force on Chronic Fatigue Syndrome, 1994; Reeves et al. 2003) (Table 4).
Of the patients, 62% met the CDC criteria for CFS, but differences across subgroups were not statistically significant (p = 0.12). In contrast, 53% of all patients met the London criteria for ME, and differences across subgroups could not be explained by chance alone (p < 0.001). This relationship, however, appeared to be driven by the low proportion of patients in the mood and polysymptomatic groups.This was to be expected, given that the high levels of co-morbid anxiety and depression that characterized these subgroups would, by definition, result in exclusion from the London criteria (White et al. 2007).
There was a significant difference between subgroups for sex (p = 0.014), with the FSS group, and to
a lesser extent avoidant–inactive, having a relatively greater proportion of women. Age and illness duration did not differ across subgroups (data not shown).

I know certain people will certainly be frowning after reading this...

It's kind of ironic that this study is proving what some people were saying: many of the patients included were not the Typical ME patients we know and love, but rather mildly affected "chronic fatigue" patients with mood/anxiety disorders.
 

soti

Senior Member
Messages
109
Yes, for LCA if I recall correctly you need to know how many groups you want. Essentially it involves squinting at the data and seeing if, e.g., the solution looks better with 3 or 4 clusters. So it works best in a case where the real solution is much better than the non-solutions.
 

Sidereal

Senior Member
Messages
4,856
I think the PACE dataset is quite interesting in the sense that it demonstrates that CBT is an ineffective treatment even for depression/anxiety-related fatigue given that it looks like only about 11% had actual ME/CFS. I have for a long time criticised clinical psychology for being a pseudoscience pushing bogus psychotherapies. In order to raise their status/income they strenuously oppose the more effective (albeit still largely unsatisfactory) biological treatments for psychiatric conditions, constantly trying to undermine the biological model. In recent years the quackery has really reached fever pitch levels / reductio ad absurdum with their advocacy of useless/harmful nonsense like CBT for psychosis which is a neurodevelopmental brain disease. It's all the more strange seeing fringe medical doctors like White/Sharpe pushing psychosocial quackery. They really should know better.
 

Ben H

OMF Volunteer Correspondent
Messages
1,131
Location
U.K.
Great study. So great. Truly amazing. Done by one of of the best researchers that God ever created. This study is good. It is actually so good that the patients will call the researchers and say: "Please, researchers, stop doing all these good studies. All the good research is just too much. You're making us too healthy with you're research. You've spent your research grants too good the last years. We can't stand it anymore. Patients in U.K are sick of all the greatness seen in ME research in the U.K." They make so much difference for the patients. If they continue like this they won't have ME patients left in about 200000000 years. That would be a pity.

Best regards from Donald JT

I had a good laugh at this, thanks! 10/10 sarcasm :D


B
 

K22

Messages
92
They clearly believe those of us who believe are symptoms are physical and exacerbation could mean harm still have faulty illness beliefs. What is any non ME sufferer, say a skeptic dr going to surmise from this paper other than at least some of symptoms aren't physical and that if we say we are too sick to do GET we just have "unhelpful avoidance behaviour". I for one used to have a GP who, I'm sure on the basis of this type narrative, concluded my main issue wasn't over riding disability but was a treatable lack of motivation and unhelpful beliefs about my illness and rehabilitation prospects - que much stress as she tried to get help in to motivate me.