On 2014 Jan 06, Tom Kindlon commented:

Caution required when making numerical comparisons between Wessely (1997) and the current study

In his editorial[1], Prof. White says: "Comorbid psychiatric conditions may have inflated the prevalence. A previous study found an equally high point prevalence of CFS (2.6%), by surveying United Kingdom primary care patients [10]. However, when those patients who also had a comorbid psychiatric disorder were excluded, the prevalence fell to 0.5%."

Reference to this paper[2] is also made in the editorial's concluding paragraph and in the accompanying Reeves paper[3].

A close inspection of table 2 of the referenced paper[2] reveals some strange figures (with regard to the estimates for the CDC '94 criteria mentioned above): (i) The Oxford criteria for CFS were found to have a lower prevalence, of 2.2%. Given that the CDC 94 criteria would be seen as more restrictive than the Oxford criteria (e.g. requiring symptoms as well as fatigue lasting six months), this suggests an error with one or both of the figures? (ii) the mean and 95% confidence intervals given for the prevalence rates without co-morbid psychological disorders for CFS (CDC 94) are given as 0.5 (0.1, 0.3) which makes no sense (the confidence intervals should be above and below the mean).

So these two observations mean that I'm not sure how much faith should be placed with some of the figures given in that study.

The methodology of the Wessely study was also different, using attendance at primary care physicians to screen for patients, which could lead to skewed data. The random number methodology in the Reeves study seems stronger.

It should also be remembered that the authors of the Reeves study[3] did exclude many patients with psychological disorders before giving the diagnosis of CFS. So even if one accepts the curious data presented in Table 2 in Wessely et al[2], it seems unlikely we can extrapolate from the drop in the figures found the Wessely study to produce a similar drop in figures found in the current study[3].

References:

[1] How common is chronic fatigue syndrome; how long is a piece of string? Peter D White Population Health Metrics 2007, 5:6 doi:10.1186/1478-7954-5-6

[2] Wessely S, Chalder T, Hirsch S, Wallace P, Wright D. The prevalence and morbidity of chronic fatigue and chronic fatigue syndrome: a prospective primary care study. Am J Pub Health 1997, 87:1449-1455.Available online at:http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1380968

[3] Prevalence of chronic fatigue syndrome in metropolitan, urban, and rural Georgia. William C Reeves, James F Jones, Elizabeth Maloney, Christine Heim, David C Hoaglin, Roumiana S Boneva , Marjorie Morrissey and Rebecca Devlin. Population Health Metrics 2007, 5:5 doi:10.1186/1478-7954-5-5

On 2014 Jan 06, Tom Kindlon commented:

Obesity and arbitrary criteria

Firstly, I thought I would clarify that I did not make my point about obesity rates based simply on the one study, the Wichita study[1]: the Chicago Study[2] found a prevalence of 0.422% using the same (or very similar) methodology and method of operationalizing the criteria as the Wichita study, producing a much higher score than the 0.235% score found in the Wichita study.

However this correspondence has caused to me to reflect on the issue: I still remain to be convinced that because the residents of Wichita were more obese than the general population, the prevalence figure for CFS (as defined then) of 0.235% was artificially increased; however perhaps if the new broadened criteria lack sensitivity and specificity, the figures in the latest studies could be artificially inflated because of a higher background obesity rate?

I think there is an important issue of a lack of sensitivity and specificity with the new method of operationalizing the criteria. As Peter White says, the current criteria are "arbitrary". Whether they are being used by a "jobbing physician", an epidemiologist or a researcher, one of the aims of criteria should be that they have good sensitivity and specificity rates. Perhaps a direction for discourse and research in the future should be trying to arrive at CFS criteria that reach that aim?

If necessary, having different criteria for different circumstances: for example, have one set of criteria when looking for expensive biological work but perhaps less stringent criteria for use in some clinical settings?

References:

[1] Prevalence and Incidence of Chronic Fatigue Syndrome in Wichita, Kansas Michele Reyes, PhD; Rosane Nisenbaum, PhD; David C. Hoaglin, PhD; Elizabeth R. Unger, PhD, MD; Carol Emmons, PhD; Bonnie Randall, MCP; John A. Stewart, MD; Susan Abbey, MD; James F. Jones, MD; Nelson Gantz, MD; Sarah Minden, MD; William C. Reeves, MD, MSPH Arch Intern Med. 2003;163:1530-1536.

[2]. Jason LA, Richman JA, Rademaker AW, Jordan KM, Plioplys AV, Taylor RR, McCready W, Huang CF, Plioplys S: A community-based study of chronic fatigue syndrome. Arch Int Med 1999, 159:2129-2137.

On 2014 Jan 06, Tom Kindlon commented:

Reference to obesity a red herring?

In his editorial, Prof White says: "Georgia may not be representative of the USA as a whole. For instance, we do not know the body mass index (BMI) of the Georgian sample. The Wichita sample of CFS cases contained 43% of subjects with a BMI of 30 or over, representing significant obesity [9]. This compares with 20% in the USA as a whole [13]. Since obesity is associated with fatigue [14], a similar proportion in Georgia might inflate the prevalence of CFS."

Firstly, just because obesity can cause fatigue is quite a different from obesity causing the syndrome CFS. Using this logic, perhaps we should be saying that prevalence studies on any condition which can involve disabling fatigue (for example multiple sclerosis) may be questionable if there is a higher rate of obesity within the sample population. It is important to consider cause and effect i.e. just because people with a condition may be more obese when they are sampled years after having the illness is not the same as saying they were more obese before getting the illness and this caused them to develop the condition.

Also the Wichita study[1], to which Prof. White refers, found a relatively low prevalence rate, of 0.235%, for CFS compared to other random-number studies including the one under review. So it seems curious to refer to this study to try to justify a hypothesis that the obesity rate in the Georgia study artificially increased the prevalence rate.

References:

[1] Prevalence and Incidence of Chronic Fatigue Syndrome in Wichita, Kansas Michele Reyes, PhD; Rosane Nisenbaum, PhD; David C. Hoaglin, PhD; Elizabeth R. Unger, PhD, MD; Carol Emmons, PhD; Bonnie Randall, MCP; John A. Stewart, MD; Susan Abbey, MD; James F. Jones, MD; Nelson Gantz, MD; Sarah Minden, MD; William C. Reeves, MD, MSPH Arch Intern Med. 2003;163:1530-1536.

On 2014 Jan 06, Tom Kindlon commented:

Caution required when extrapolating prevalence rates to the full population

This editorial [1] says, with regard to the CDC study[2]: "The CDC has now repeated and extended the Wichita study in Georgia, and found a prevalence of between six and ten times greater, with 2.5% of the population suffering from CFS. If this prevalence was both accurate and representative of the USA as a whole, this would suggest that some 7.5 million Americans were sufferers, compared to the previous estimates of 0.7 to 1.2 million."

Before the 7.5 million figure is quoted, it might be useful to point out that the figure makes a number of assumptions, including that the prevalence rate for those under 18 and over 60 would be similar. However previous studies have suggested this is unlikely to be the case, with prevalence rates for young children in particular being much lower.

The round figure of 7.5 million would be equivalent to a population of 295,275,591. Using this data the population estimate for 2005 was 296,410,404 (i.e. a similar figure). Using the same data: The population under 18 years was 73,469,580, the population over 60 was 49,791,976 and population aged 18-60 was 173,148,444.

For a population of those aged over 18 and under 60 of this size, a back of the envelope calculation for CFS prevalence using the prevalence rate of 2.64%[2] would give: (173,148,444*0.0264)= 4,397,971.

References:

[1] How common is chronic fatigue syndrome; how long is a piece of string? Peter D White Population Health Metrics 2007, 5:6 doi:10.1186/1478-7954-5-6

[2] Prevalence of chronic fatigue syndrome in metropolitan, urban, and rural GeorgiaWilliam C Reeves , James F Jones , Elizabeth Maloney , Christine Heim , David C Hoaglin , Roumiana S Boneva , Marjorie Morrissey and Rebecca Devlin. Population Health Metrics 2007, 5:5 doi:10.1186/1478-7954-5-5

On 2014 Jan 06, Tom Kindlon commented:

Caution required when extrapolating prevalence rates to the full population

This editorial [1] says, with regard to the CDC study[2]: "The CDC has now repeated and extended the Wichita study in Georgia, and found a prevalence of between six and ten times greater, with 2.5% of the population suffering from CFS. If this prevalence was both accurate and representative of the USA as a whole, this would suggest that some 7.5 million Americans were sufferers, compared to the previous estimates of 0.7 to 1.2 million."

Before the 7.5 million figure is quoted, it might be useful to point out that the figure makes a number of assumptions, including that the prevalence rate for those under 18 and over 60 would be similar. However previous studies have suggested this is unlikely to be the case, with prevalence rates for young children in particular being much lower.

The round figure of 7.5 million would be equivalent to a population of 295,275,591. Using this data the population estimate for 2005 was 296,410,404 (i.e. a similar figure). Using the same data: The population under 18 years was 73,469,580, the population over 60 was 49,791,976 and population aged 18-60 was 173,148,444.

For a population of those aged over 18 and under 60 of this size, a back of the envelope calculation for CFS prevalence using the prevalence rate of 2.64%[2] would give: (173,148,444*0.0264)= 4,397,971.

References:

[1] How common is chronic fatigue syndrome; how long is a piece of string? Peter D White Population Health Metrics 2007, 5:6 doi:10.1186/1478-7954-5-6

[2] Prevalence of chronic fatigue syndrome in metropolitan, urban, and rural GeorgiaWilliam C Reeves , James F Jones , Elizabeth Maloney , Christine Heim , David C Hoaglin , Roumiana S Boneva , Marjorie Morrissey and Rebecca Devlin. Population Health Metrics 2007, 5:5 doi:10.1186/1478-7954-5-5

On 2014 Jan 06, Tom Kindlon commented:

Reference to obesity a red herring?

In his editorial, Prof White says: "Georgia may not be representative of the USA as a whole. For instance, we do not know the body mass index (BMI) of the Georgian sample. The Wichita sample of CFS cases contained 43% of subjects with a BMI of 30 or over, representing significant obesity [9]. This compares with 20% in the USA as a whole [13]. Since obesity is associated with fatigue [14], a similar proportion in Georgia might inflate the prevalence of CFS."

Firstly, just because obesity can cause fatigue is quite a different from obesity causing the syndrome CFS. Using this logic, perhaps we should be saying that prevalence studies on any condition which can involve disabling fatigue (for example multiple sclerosis) may be questionable if there is a higher rate of obesity within the sample population. It is important to consider cause and effect i.e. just because people with a condition may be more obese when they are sampled years after having the illness is not the same as saying they were more obese before getting the illness and this caused them to develop the condition.

Also the Wichita study[1], to which Prof. White refers, found a relatively low prevalence rate, of 0.235%, for CFS compared to other random-number studies including the one under review. So it seems curious to refer to this study to try to justify a hypothesis that the obesity rate in the Georgia study artificially increased the prevalence rate.

References:

[1] Prevalence and Incidence of Chronic Fatigue Syndrome in Wichita, Kansas Michele Reyes, PhD; Rosane Nisenbaum, PhD; David C. Hoaglin, PhD; Elizabeth R. Unger, PhD, MD; Carol Emmons, PhD; Bonnie Randall, MCP; John A. Stewart, MD; Susan Abbey, MD; James F. Jones, MD; Nelson Gantz, MD; Sarah Minden, MD; William C. Reeves, MD, MSPH Arch Intern Med. 2003;163:1530-1536.

On 2014 Jan 06, Tom Kindlon commented:

Obesity and arbitrary criteria

Firstly, I thought I would clarify that I did not make my point about obesity rates based simply on the one study, the Wichita study[1]: the Chicago Study[2] found a prevalence of 0.422% using the same (or very similar) methodology and method of operationalizing the criteria as the Wichita study, producing a much higher score than the 0.235% score found in the Wichita study.

However this correspondence has caused to me to reflect on the issue: I still remain to be convinced that because the residents of Wichita were more obese than the general population, the prevalence figure for CFS (as defined then) of 0.235% was artificially increased; however perhaps if the new broadened criteria lack sensitivity and specificity, the figures in the latest studies could be artificially inflated because of a higher background obesity rate?

I think there is an important issue of a lack of sensitivity and specificity with the new method of operationalizing the criteria. As Peter White says, the current criteria are "arbitrary". Whether they are being used by a "jobbing physician", an epidemiologist or a researcher, one of the aims of criteria should be that they have good sensitivity and specificity rates. Perhaps a direction for discourse and research in the future should be trying to arrive at CFS criteria that reach that aim?

If necessary, having different criteria for different circumstances: for example, have one set of criteria when looking for expensive biological work but perhaps less stringent criteria for use in some clinical settings?

References:

[1] Prevalence and Incidence of Chronic Fatigue Syndrome in Wichita, Kansas Michele Reyes, PhD; Rosane Nisenbaum, PhD; David C. Hoaglin, PhD; Elizabeth R. Unger, PhD, MD; Carol Emmons, PhD; Bonnie Randall, MCP; John A. Stewart, MD; Susan Abbey, MD; James F. Jones, MD; Nelson Gantz, MD; Sarah Minden, MD; William C. Reeves, MD, MSPH Arch Intern Med. 2003;163:1530-1536.

[2]. Jason LA, Richman JA, Rademaker AW, Jordan KM, Plioplys AV, Taylor RR, McCready W, Huang CF, Plioplys S: A community-based study of chronic fatigue syndrome. Arch Int Med 1999, 159:2129-2137.

On 2014 Jan 06, Tom Kindlon commented:

Caution required when making numerical comparisons between Wessely (1997) and the current study

In his editorial[1], Prof. White says: "Comorbid psychiatric conditions may have inflated the prevalence. A previous study found an equally high point prevalence of CFS (2.6%), by surveying United Kingdom primary care patients [10]. However, when those patients who also had a comorbid psychiatric disorder were excluded, the prevalence fell to 0.5%."

Reference to this paper[2] is also made in the editorial's concluding paragraph and in the accompanying Reeves paper[3].

A close inspection of table 2 of the referenced paper[2] reveals some strange figures (with regard to the estimates for the CDC '94 criteria mentioned above): (i) The Oxford criteria for CFS were found to have a lower prevalence, of 2.2%. Given that the CDC 94 criteria would be seen as more restrictive than the Oxford criteria (e.g. requiring symptoms as well as fatigue lasting six months), this suggests an error with one or both of the figures? (ii) the mean and 95% confidence intervals given for the prevalence rates without co-morbid psychological disorders for CFS (CDC 94) are given as 0.5 (0.1, 0.3) which makes no sense (the confidence intervals should be above and below the mean).

So these two observations mean that I'm not sure how much faith should be placed with some of the figures given in that study.

The methodology of the Wessely study was also different, using attendance at primary care physicians to screen for patients, which could lead to skewed data. The random number methodology in the Reeves study seems stronger.

It should also be remembered that the authors of the Reeves study[3] did exclude many patients with psychological disorders before giving the diagnosis of CFS. So even if one accepts the curious data presented in Table 2 in Wessely et al[2], it seems unlikely we can extrapolate from the drop in the figures found the Wessely study to produce a similar drop in figures found in the current study[3].

References:

[1] How common is chronic fatigue syndrome; how long is a piece of string? Peter D White Population Health Metrics 2007, 5:6 doi:10.1186/1478-7954-5-6

[2] Wessely S, Chalder T, Hirsch S, Wallace P, Wright D. The prevalence and morbidity of chronic fatigue and chronic fatigue syndrome: a prospective primary care study. Am J Pub Health 1997, 87:1449-1455.Available online at:http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1380968

[3] Prevalence of chronic fatigue syndrome in metropolitan, urban, and rural Georgia. William C Reeves, James F Jones, Elizabeth Maloney, Christine Heim, David C Hoaglin, Roumiana S Boneva , Marjorie Morrissey and Rebecca Devlin. Population Health Metrics 2007, 5:5 doi:10.1186/1478-7954-5-5