This paper by Byrnes et al describes a small study (total of 28 CFS patients and 28 controls) and concludes that gene expression in PBL is not useful as a biomarker for CFS. However, there are several weaknesses and some incongruous aspects in this study.

The patients were identified by symptom screen and subsequent clinical assessment. But it is not reported how many of the 28 CFS patients had previously been
diagnosed as having CFS. This recruitment strategy will identify a different population from those studies which focus on patients with a prior diagnosis of CFS who attend a CFS clinic.

In addition, the inclusion of patients with 'idiopathic chronic fatigue' is just padding (presumably as it was difficult to make up the numbers), as this diagnosis (ICF) is not one that is formally recognised. The title does not include the formally diagnosed 'Chronic Fatigue Syndrome', but 'chronic fatigue', for which no diagnostic criteria were provided in the paper.

It is stated that after array findings were negative, the authors tried to confirm 7 previously reported CFS-associated genes by real-time PCR with 18S-rRNA as an endogenous control. We have found that 18S-rRNA is differentially exressed between CFS patients and normal blood donors to a statistically significant degree (unpublished). So, it is not a useful endogenous control gene and will give misleading results in this context. We used GAPDH in our studies instead which does not vary across CFS and normal populations (Kaushik et al, 2005; Kerr et al, 2008); Powell et al, 2003 also used GAPDH.

The authors mention that the gene MSN (moesin) is the only one to date that has been found in common between CFS gene studies (this is incorrect - what about PRKARIA, EIF4G1, EGR1, CXCR4, etc); this gene, MSN, was first identified by Powell and confirmed by Kaushik and Kerr, using PCR, in a total of 62 CFS patients and >100 normal controls (combined subject numbers between these studies). So surely this would have been an ideal gene to test and disprove as a biomarker, having so much background, and PCR confirmation. In addition, we reported 88 human genes using microarrays and PCR (Kerr et al, 2008), 16 of which were previously identified using a separate study using microarrays and PCR (Kaushik et al, 2005). But instead the authors chose 7 genes identified by a small study comprising 8 CFS cases and 7 normals (Gow et al, 2009), using an array only without PCR confirmation, which identified 366 genes. Why did the authors not use the identical PCR assays and endogenous controls used in previous studies and disprove some confirmed CFS gene markers, such as MSN and any of 87 others? This is clear bias in gene selection.

The biggest weakness in the paper is that the SF36 scores for the patients versus controls were not particularly different. It seems that the patients did have low SF36 scores, but that the 'normal' controls exhibited scores approaching and even overlapping those of the patients! The SF36 physical function was 41.2 (range 28-48) for CFS and 48.2 (range 40-52) for Normals. To support their proper case selection, a t-test would have been useful, but I suspect that this would reveal these 2 populations were not statistically different, based on these scores. These SF36 results are very problematic and are not convincing of the fact that the phenotypes of these 2 populations differ significantly.

Why did the authors not publish the complete scoring for the SF36 and all its domains, and not just the physical function and mental function. The total score is the most useful and would have demonstrated clearly the global functioning of the subjects and would have facilitated a comparison of functionality between cases and controls. In our studies, mean SF36 total scores were <50 for CFs, and >80 for controls, reflecting distinct phenotypes. We also used neurocognifive function testing by the Cantab software, which also confirmed defects for the CFS patients (Kerr et al, 2008).

There are so many weaknesses with this study that the results cannot be interpreted with confidence.