I posted about the cancer and CFS study last week based on the abstract only. Now I have a copy of the full paper, and there is much more to discuss. Dr. Suzanne Vernon also wrote about the study, so you might want to check out her analysis too. I expressed concerns about two aspects of the study: whether the subjects actually had CFS and whether the odds ratio represented a significant risk of cancer. Reading the full paper, my concerns were justified.
Case Selection
The study began with the selection of 1.2 million cancer cases from the SEER cancer registry. These cases were all first cancers in people aged 66 to 99 between 1992 and 2005. Cases had to have Medicare coverage for at least 13 months prior to diagnosis with cancer, and people with HMO coverage were excluded. Controls (100,000) were selected at random from the Medicare database and matched by gender, age, and location to the cancer cases. Both the cancer cases and controls were then evaluated for CFS in the Medicare database.
The presence of CFS was evaluated based on the database only. No confirmation of the diagnosis was attempted. More importantly, two codes were counted as CFS. The CFS code (780.1) was introduced in 1998. For older cases, this study used the code for neurasthenia (300.5). Neurasthenia is a topic unto itself, but its definition should give you pause:
a condition that is characterized especially by physical and mental exhaustion usually with accompanying symptoms (as headaches, insomnia, and irritability), is believed to result from psychological factors (as depression or emotional stress or conflict), and is sometimes considered similar to or identical with chronic fatigue syndrome
In this study, the authors used two definitions of CFS. The first group was diagnosed with CFS or neurasthenia beginning in 1991, and the second group was diagnosed only with CFS after 1998. Neurasthenia is not the same as CFS, and is much more likely to be applied to people with chronic fatigue and other vague symptoms. Such a diagnosis is not the same as the 1994 Fukuda definition of CFS. And it should be quite obvious to readers that neither the neurasthenia nor the CFS diagnoses were made using anything like the Canadian Consensus Criteria.
The authors did attempt to apply stringent requirements to the CFS groups. The diagnosis had to be documented in at least 1 hospital claim or by at least two outpatient claims that were 30 days apart. They also did not consider CFS diagnoses made during the year prior to the diagnosis of cancer to avoid the possibility that undiagnosed cancer was the source of fatigue. Despite these conditions, the authors acknowledge, “We were unable to assess whether CFS was correctly diagnosed by physicians and reported in Medicare claims.” This is a major weakness of the study. The authors point to their overall prevalence rate of 0.5% as being consistent with community-based studies like Dr. Leonard Jason’s. But Dr. Jason’s study was not limited to diagnosed cases. If we assume that only 25% (a generous number) of CFS cases are diagnosed, then the prevalence rate of this population would be 2%. That’s a prevalence number approaching that of the CDC’s empiric definition, and we know that definition is highly likely to capture patients with depression and other sources of fatigue.
Based on all of this, I have little confidence that the prevalence rate and CFS case selection in this study is accurate. It would be very interesting to pull the CFS cases and use chart review to attempt to confirm the diagnosis.
Actual Cancer Risk
This study found higher odds that certain cancer cases had CFS (compared to controls) in non-Hodkin’s lymphoma, pancreatic and kidney cancers. The study found lower odds that cases of breast and oral cavity/pharynx cancers would have CFS compared to controls. However, the only association that held up after multiple comparison adjustment was non-Hodgkin’s lymphoma. Even when the neurasthenia cases are excluded, the odds ratio holds up (Table 2 of paper). This study was also able to examine sub-types of non-Hodgkin’s lymphoma, and three subtypes were significantly associated with CFS/neurasthenia: diffuse large B cell lymphoma, marginal zone lymphoma, and B cell non-Hodgkin’s lymphoma not otherwise specified.
There are more than a dozen subtypes of non-Hodgkin’s lymphoma, but interestingly the first two subtypes associated with CFS have a median age at diagnosis of 65 to 70 years of age. The authors note the significant limitation of their study, “because our study was limited to people aged 66 years and older, our results may not be generalizable to younger (nonelderly) populations.” The authors also note that neither CFS nor neurasthenia was associated with cancer overall. Only the odds ratio with non-Hodgkin’s lymphoma held up in the study.
The actual numbers should be noted. According to Table 2 in the paper, 500 controls had CFS/neurasthenia (.5% of 100,000). Of the almost 1.2 million cancer cases, 5,885 had CFS/neurasthenia (.5% of 1,176,950). Of the 57,632 cases with non-Hodgkin’s lymphoma, 403 people had CFS/neurasthenia. While this is probably the biggest statistical study of CFS and cancer, the actual numbers of cases are still quite small. Even so, the hypothesis that chronic immune activation or infection could play a role in the association of CFS and non-Hodgkin’s lymphoma is supported by this study. As the authors note, “Our study results support continued efforts to understand the biology of CFS.”
Abstract vs. Paper
The hype in the CFS community when this study was released can be summarized as “Ack! People with CFS have a higher risk of non-Hodgkin’s lymphoma!” Here is yet another illustration of the absolute peril of relying upon paper abstracts and media spin. The paper explicitly refuses to make such a claim:
We would also caution further against any direct interpretation or application of our results in a clinical setting. We could not estimate the absolute risk of NHL associated with CFS, but the risk is likely too small to affect the clinical management of patients with CFS.
Everyone got that? This paper does not say that people with CFS have a higher absolute risk of non-Hodgkin’s lymphoma. This paper did find some interesting statistical associations that should be investigated further, especially given the role of immune activation in CFS. But that’s all the paper found.
Thanks!
The reason why several types of cancer have been associated with CFS/ME (see e.g. The Clinical and Scientific Basis of ME/CFS, which discusses this in several articles) is, of course, not immune activation, but immunosuppression. Viruses, hormonal deficiencies and other factors may also play a role. I’ve written an article about the subject: http://suite101.com/article/cfsme-and-cancer-risk-a226291
The hypothesis in this paper was based on immune activation leading to cancer. Regarding NK cell activity, the paper states, “Because NKT cells may play a role in suppressing autoimmunity, a disturbance in NKT activity could lead to increased immune activation.” The immune system is so complex, with so many pathways and interconnections, that I’m not sure it makes sense to speak in terms of activation or suppression without more explanation. For example, elevated cytokines also suggest immune activation, but both elevated cytokines and reduced NK cell activity have been reported. I’ve also heard it discussed that some patents have activation and some suppression, and that this may be a way to subtype patients or at least target treatments. This is the first large scale study to examine the association of CFS with cancers, and as the authors state, this is an area that warrants a great deal more study.
A nice summary and critique of this study. I was alarmed to see that Dr. Vernon only had positive things to say about the study and did not even note that the use of the Neurasthenia diagnosis renders the study invalid. I think her former employment at CDC clouds her judgment of HHS’ science.