Introduction
The problem with subsets is their ability to mess up CFS research efforts. If, for instance, CFS patients with immune or neurological problems are present in a study then elucidating either will be difficult if not impossible. The recognition that subsets may pose a problem in CFS is not new.
The authors of the CDC ‘Fukuda’ definition of CFS in 1994 noted that the ‘looseness’ of definition would probably result in the inclusion of different kinds of CFS patients. Indeed it is the vague definition of CFS that makes the subset problem possible.
Researchers have never been clear on what constitutes CFS. The 1994 definition was promulgated to produce a more or less consistent baseline for research studies. A consensus definition created by a panel of experts based on anecdotal reports; i.e. their conception of what constituted CFS, it seemed more like a stopgap measure that a permanent entity.
Now the standard criteria in CFS research studies, this rather vague definition (severe fatigue of at least six months duration and 4/8 symptoms; post-exertional fatigue, sore throat, tender lymph nodes, headaches of a new type, unrefreshing sleep, muscle pain, multi-joint pain, impaired concentration) has certainly brought consistency to the field but at a cost of reduced precision.
It seems unrealistic to expect that this definition would not evolve over time, yet it has remained unchanged for over 10 years now, and the CDC, the agency that created it – and the only organization with the standing to alter it – has, until recently, been committed to it.
The types of subset in CFS and the negative effects they may have on research finding are hot topics right now. Jason presented a major paper last year that argued that differentiating subsets is absolutely critical if we are to make further progress in CFS.
Likewise, Vance Spence, a researcher associated with CFS research group MERGE, stated in his recent presentation “Making the Difference in CFS”, that there was no more critical problem in CFS than identifying subsets. Spence talked of how strange patterns of data seen in CFS studies often leave CFS researchers, as he put it, ‘scratching their heads.
Given their problems subset can cause for CFS research and the potential for their occurence it remarkable how little work has been done to ferret them out. Most attempts to do so have used symptom and clinical data rather than laboratory data. The two studies before us are amongst the first to use both clinical and physiological data
Conna, U., Aslakson, E. and P. White. 2006. An empirical delineation of the heterogeneity of chronic unexplained fatigue in women. Pharmacogenomics 7, 355-364.
Aslakson, E., Wollmer-Connar, U. and P. White. 2006. The validity of heterogeneity in chronic unexplained fatigue. Pharmacogenomics 7, 365-373
Note that two of the three researchers here are psychiatrists. One, Dr. White, is rather notorious for his views that CFS is a biopsychosocial phenomena. He has, however, also done research on the role infection plays in triggering CFS. Be prepared for a psychological orientation to some of their conclusions.
Several studies have attempted to identify the subsets in CFS but until 2006 all had attempted to differentiate CFS patients on the basis of symptoms not laboratory data. In the most extensive attempt yet to uncover the subsets present in CFS this group used the data gathered during the 2-day hospital study done by the CDC WIchita, Kansas in 2003.
Note that title of this paper says chronic fatigue not chronic fatigue syndrome. Like many of the other Pharmacogenomics studies this study contained three study groups only one of which contained chronic fatigue patients; a CFS group (n=55), an idiopathic fatigue group (n=53), and an un-fatigued group that had been age, sex, race and BMI matched to the CFS group. These groups only contained women. Unusual for a study of this sort CFS patients with major depressive disorder were included.
Although about 15% of the original study group were males they were excluded because some of the variables used to create these subsets were impacted by gender. Including then would have reduced the sensitivity of those criteria and undermined the ability of the researchers to create their subsets.
Method
This group took all the data points gathered by the CDC (@500) and winnowed them down to 38 measures that best explained the variability found in the data set. They included clinical (3), symptom (15) and lab data (21). The lab data consisted of the following categories; sleep (6), endocrine (6), immune (3), red blood cell (3) and tilt data (1).
They then did a latent class analysis (LCA) to create classes or groups. These ‘patterns analysis’ statistical programs attempt to find correlations in large sets of data that are used to produce natural groups. They are typically used to uncover new patterns that researchers can then examine more closely to determine if they are valid. They do not determine cause and effect.
Findings
The basic questions one asks about such studies are a) did they create groups, and b) did the groups make sense? The answer in this case was a qualified yes.
The PCA analysis identified six groups that explained about 40% of the variance in the data set. I have been told this is a satisfactory result. The researchers then did a latent class analysis (LCA) to find ‘loci in the multidimensional space of measurements (clinical and biological) where subjects cluster together.
The subjects that are close to each other belong to the same class” (?). This also produced a group of six subsets of groups with distinct symptom and laboratory findings.
The presence of a satisfactory 6-class suite was encouraging. It was vitally important, however, for these researchers to show that these groups also differentiated CFS from idiopathic fatigue from healthy controls – that CFS patients dominated some groups, IF patients others, and the healthy controls others. Otherwise it would be delineating something other than CFS.
This study was mostly successful in this regard. While each class usually contained people from all three groups each class was dominated by one or the other. Of the six classes developed three were dominated by CFS patients (1, 5, 6), two by idiopathic fatigue patients (3, 4), and one was dominated by the healthy controls (2).
The subsets
A class dominated by the well participants
LCA Class II –The ‘Well Group’ group was obese but relative to the other groups had much, much better sleep, zero post-exertional fatigue, and very little sore throat, shortened breath, abdominal pain or fever. They also had low sleep heart rate variability, moderate C-reactive protein (CRP), moderate IL-6 and the lowest cholesterol readings found.
Classes dominated mostly by CFS patients
Class I = ‘Obese hypnoea’ – this group was obese, had poor sleep, high rates of post-exertional fatigue, muscle pain and moderate amounts of joint pain plus lower rates of shortened breath and sore throat and concentration problems than the other CFS groups. They had a high sleepiness score and the highest rates of arousal during sleep, the highest c-reactive protein, the lowest oxygen saturation and normal cortisol.
The correlation between obesity and markers of inflammation (CRP, IL-6) and insulin and the problems with sleep in classes 1 and 3 prompted the researchers to state that obesity in itself plays a ‘prominent role in the production and maintenance of fatigue and other symptoms latent classes 1 (CFS) and 3 (idiopathic fatigue)’.
This was a remarkable statement. The authors are essentially asserting that one class of CFS patients are fatigued simply because they are obese. They note that obesity is associated with increased rates of inflammatory markers, sleep disturbance and depression.
This disregards the fact that the well group were just as obese as the CFS group and had a similar inflammatory profile but didn’t have the sleep problems, the muscle and joint pains, post-exertional fatigue, concentration problems, shortness of breath, etc. found in the obese CFS group.
The inability of obesity to produce the characteristic symptoms of CFS In the well group suggests that far from playing ‘a prominent role’ that it could play at best a secondary role.
Class V – CFS ‘Depressed/Interoception’ – this class had the highest amount of muscle and joint pain (91%), the biggest problems with concentration (86%) and shortened breath (50%), abdominal pain (68%), fever (50%) and almost the highest CRP levels found (66) plus high progesterone and normal cortisol levels.
Problems with concentration were far higher in this group (86%) than in any others (0-45%). This class had a higher percentage of CFS patients in it (73%) than any others. Idiopathic fatigue patients accounted for all the rest of the members. This was the second most debilitated group.
Although their depression scores were not much higher than most of the other groups (54 vs. 48, 50, 51, 54) the authors labelled this group ‘depressed’. They posited the increased muscle and joint pains were due to disrupted interoception, an interpretation Dr. White has championed in the past in CFS.
The levels of IL-6, a pro-inflammatory cytokine, much better differentiated this group that did its depression scores (66- 68, 56, 50, 45, 32). Is this actually the immune group? The high fever, IL-6 and CRP levels could suggest immune activation. Indeed most of the prominent symptoms in this group (sore throat, fever, muscle, joint pains, poor concentration) are emblematic of infection. This scenario, however, was not embraced by the authors.
Class VI – CFS ‘Multi-symptomatic-depressed-stressed-postmenopausal’ – This group had poor sleep, high levels of post-exertional fatigue and muscle and joint pain, photophobia, shortened breath, sore throat and depression and high levels of CRP. They also had low cortisol levels, low sleep HR variability, high c-reactive protein and low testosterone.
The authors labelled this group post-menopausal because they were oldest group (age-55) but the mean age of some other groups was similar (51, 52, 53). Their disability scores indicated this was the most debilitated group.
The authors once again labelled a class depressed whose depression scores were not substantially higher than most of the other classes (55 vs. 54, 51, 50, 48). In contrast to this range look at how much the higher rates of sore throat (73- 59, 28, 17, 13, 8) or shortened breath (45 – 50, 32, 8, 5, 4) than in most other classes.
The low sleep HR variability suggests increased SNS and decreased parasympathetic nervous system activity. Both reduced cortisol levels and reduced PNS activity could result in immune activation and the high CRP levels seen.
Classes mostly dominated by idiopathic fatigue patients
Class III – IF ‘Obese hypnoea, stressed’ – This group was obese, had poor sleep, moderate muscle and joint pains; low sore throat and shortened breath and low to moderate post-exertional fatigue as well as low cortisol , higher insulin and lower sleep HR variability.
As might be expected by the low to moderate amounts of symptoms associated with CFS (post exertional fatigue, sore throat, shortened breath) this group was composed mostly of fatigued but non-CFS patients (60%). It also had about equal numbers of well and CFS patients.
The fatigue and muscle and joints problems in this group could have been due to immune activation secondary to hypocortisolism. Given the few CFS patients in it one wonders if this group is what hypocortisolism looks like without the immune component found in CFS.
A study of a family with genetically derived cortisol impairment found a group much like this one; the family was highly fatigued, overweight, had muscle and joint pain, etc. but evidenced little indication of the immune disruption (sore throat, flu-like symptoms, shortened breath, fever) often found in CFS.
Class IV- IF ‘Interoception’ – this class was not obese, had poor sleep, moderate to high muscle and joint pain, moderate abdominal pain but in contrast to the CFS dominated groups, only moderate post-exertional fatigue and good concentration and little shortness of breath. It also had low immune activation (low CRP, low IL-6) and normal cortisol but low insulin and high progesterone.
The authors inability to account for this groups symptoms using laboratory data apparently lead them to label this group ‘interoceptive’. This suggested these patients problems derived from an over sensitization of the brain to incoming stimuli. This group was dominated by idiopathic chronic fatigue patients (65%), had a substantial number of healthy controls (30%), and almost no CFS patients.
A look at the symptoms
The authors did not analyse the symptom findings. Perhaps not surprisingly given the limited extent of the laboratory data, it was symptoms not lab measures that played the biggest role in differentiating the groups. That the top eleven differentiating factors were all symptoms suggests that this data set did little to uncover CFS physiology.
This could not have been entirely unexpected. While many tests results in CFS are normal some (cortisol, NK cell numbers and function, RNase L fragmentation, increased apoptosis, serotonin levels, oxidative stress markers, reduced brain blood flows, NMR choline spikes in the brain, low HRV, unique hemodynamic instability index, increased TGF-b levels, prolonged acetylcholine activity) have been more or less consistent.
The HHV-6 Foundation makes a compelling case for increased HHV-6 activation in a subset of CFS patients as well. Only two of these (heart rate variability, cortisol) were used and both, not surprisingly showed up in the subsets. The CDC, of course, was focused on neuroendocrine markers and had to stop somewhere.
Some of these tests are very expensive. Still, one wonders how much more effective this study, in particular, would have been if more or different measures had been used. Contrast these results with reports that Dr. De Meirleir in a blind test was able to accurately identify almost all CFS patients without any symptom or clinical data simply by using six blood tests!
To their credit these researchers included symptoms such as post-exertional fatigue, shortened breath, photophobia and abdominal pains that are common in CFS but are not included in the CDC criteria.
The most important symptoms in differentiating the different groups were.
- Post-exertional fatigue – was the first and third most important differentiating variables in the PCA and Latent Class Analyses. Its discriminatory prowess was highlighted by the fact that it and concentration difficulties were the only variables not found at all in the Well Group (class 2). The very high levels of post exertional fatigue in the three classes dominated by CFS patients (78-91%) and the low to moderate levels of it in the classes dominated by idiopathic fatigue patients (33-41%) indicate that it plays, as Dr. Jason has suggested, a special role in CFS. CFS is often described as being an amalgam of very common symptoms but this study indicates that post-exertional fatigue is not common in the population nor in other unexplained fatiguing diseases. It was remarkable, at least to me, that these researchers did not draw attention to what an important differentiating factor post-exertional fatigue turned out to be.
- Sore throat – was found from moderate to high levels in the CFS dominated groups (29, 59, 73%) but was rarely found in the IF and well classes (8-17%). This perhaps reflects an important infectious subset in CFS (???).
- Shortness of breath – was also found mostly in intermediate amounts in the CFS dominated groups (32, 45, 54%) but was rarely found in the others (4, 5, 8%).
- Higher rates of concentration problems in the CFS groups (22%, 45, 86%) vs lower rates of concentration problems in the IF groups (4, 21%), and no concentration problems in the well group suggest that concentration problems fairly well differentiate CFS, IF and healthy controls.
- Unrefreshing sleep/sleep problems – The CFS patients had lots of sleep problems (90-100%), as did the IF patients (75-80%) but not the controls (13-30%). Thus poor sleep is good at differentiating people with medically unexplained fatigue from healthy people but differentiate well between CFS and IF patients.
In summary post–exertional fatigue, in particular, plus sore throat, shortness of breath and concentration problems appear to much more increased in CFS patients than in fatigued patients who do not meet the critieria for CFS or in healthy controls
Conclusions
This study was largely able to differentiate CFS from other fatigued patients and from controls. It was also able to create three classes of CFS patients. The authors did not analyze the symptom findings but an examination of them indicated that post-exertional fatigue was the most effective factor in differentiating the groups and that a suite of other symptoms (sore throat, shortened breath, concentration problems) were far more common in the CFS dominated groups than in the others.
The dominant role that symptoms played in differentiating these groups suggested that the suite of symptoms found in CFS is unique but that the data base accumulated was mostly unequal to explaining its pathophysiology. This was highlighted by the inability of any variable to track the levels of the hallmark symptom in CFS, post-exertional fatigue, as it fluctuated in the classes.
Nevertheless some intriguing clues were found. Low cortisol levels did differentiate a subset of CFS patients and immune markers appeared to differentiate at least one and perhaps two others. The authors felt obesity played a prominent role in producing the symptoms in one of the three classes of CFS patients created. A closer look at this issue, however, indicated that obesity by itself plays little if any role in producing the hallmark symptoms of CFS.
The authors at times appeared to favor psychological interpretations over biological ones.
Carmel, L., Effron, S., White, P., Vollmer-Conna and Rajeevan. 2006. Gene expression profile of empirically delineated classes of unexplained chronic fatigue. Pharmacogenomics 7, 375-386.
This same set of researchers determined if unique patterns of gene expression were associated with these groups. The statistical tests originally created three different solutions, one with four, five and six class sets. Only the six class set solution was used.
This time, however, they analyzed the gene expression data from both the five and six class sets and came up with set of 39 abnormally expressed genes, 19 of which were found in both class solutions. Of these 19 genes four were involved in immune functioning, four in gene transcription, four in ubiquitination, two in signal production and one in amino acid transportation.
Three of the genes that were upregulated in most of the five classes suggested that glutamate transport, the regulation of gene transcription and ubiquitin dependent protein catabolism.
Glutamate is the chief excitatory neurotransmitter in the brain. The function of thie gene upregulated in this study is to reduce extracellular glutamate levels before they become ‘excitotoxic’ i.e. start to damage or kill neurons. The presence of this gene, of course, suggests that increased glutamate levels are found in idiopathic fatigue and CFS.
This is an intriguing finding in many ways. We saw last year that CFS patients have reduced grey matter volume in their brains. Peter’s Selfish Brain theory suggests increased glutamate production may be occurring in CFS (see Do Chronic Fatigue Syndrome (ME/CFS) Patients Have a Selfish Brain?).
Dr. Pall has posited that high glutamate levels in the brains of CFS patients play an important role in nitric oxide up regulation and ultimately production of the dangerous free radical peroxynitrite. Several studies also suggest that increased glutamate levels produce mental fatigue.
The ubiquitin protease pathway is a pathway in which ubiquitin couples with proteins to catalyze their destruction by proteases.
Summary
The attempt to differentiate the CFS subgroups using gene expression data was not a success. The only group that was fairly well differentiated was the ‘interoception’group. This group, however, contained almost no CFS patients.
It was surprising then to find the authors end on a high note. In the ‘Outlook’ section they stated that in 5 or 10 years ‘we will have replicated or refined the heterogeneity in CFS using gene expression as an external validator’.
Overall conclusions
This was perhaps the most important but least successful of the four Pharmacogenomics groups. Given the amount of interpretation these types of studies are open to, it is perhaps unfortunate that these studies, in particular, were under the aegis of the one group of researchers with a history of interpreting CFS in a psychological manner.
The authors overall appeared confident that their studies validated the idea that different kinds of CFS are present. While CFS patients and their advocates cannot get excited about groups called ‘interoception-depression’ or statements that obesity contributes significantly to many of the symptoms of CFS, the ability of these researchers to differentiate different fatigue states was a significant step forward. It illustrated there are significant differences between idiopathic fatigue and CFS and further cemented the idea that as filled subsets as CFS presumably is it is still distinguishable from other types of fatigue.
Hopefully this study will lead, as the tenor of the authors remarks suggests, to much larger studies composed solely of CFS patients and using more and different kinds of laboratory data.
- The CDC’s Pharmacogenomics Studies on Chronic Fatigue Syndrome (ME/CFS) I: An Introduction
- The CDC’s Pharmacogenomic’s Studies II: The Allostatic Stress In CFS
- The Pharmacogenomics Studies on Chronic Fatigue Syndrome (ME/CFS) III: The Gene Expression Studies
- The CDC’s Pharmacogenomic’s Studies IV: Heredity and Chronic Fatigue Syndrome (ME/CFS)
- A Guide to the 2006 Critical Assessment of Microarray Data Analysis (CAMDA) Conference (Integrating Laboratory, Gene Expression, Gene Mutation and Proteome Data)