Computer Gaming: Your waist doesn’t really go to your head
[This story is being bumped ahead of the scheduled piece….]
A press release went out on Tuesday to enable the media, in lockstep, to announce the results of a new study on the embargoed date — before it was published or even available to medical professionals who had paid subscriptions to the journal — ensuring the media release spin would pre-empt any professional critique. Accordingly, at least 638 (and counting) news stories have been breathlessly reporting that “big bellies increase risk for dementia.”
This was our first clue that what we’ve been hearing is marketing, not science. They are not remotely the same thing.
If it seems like a lot of studies linking fat to dementia have been in the news over recent years, you would be right. But repeated media stories do not mean there’s a growing body of evidence, or any new research. Reports of the same studies, on the same patients, and by the same researchers, are frequently published in multiple journals, each time often generating a press release and media coverage. Repetition can give us an illusion of something being real.
And the more fanatical the rhetoric — such as the lead researcher quoted in the Wall Street Journal yesterday calling tummies “toxic fat” — the more it seems to reinforce the need for us to feel worried and anxious… and the more we know that our emotions are being played. That’s a marketing tactic, too. Not impartial science.
Wordsmithing can also lead us to think that this study had shown large waistlines to be the cause for dementia. In actuality, through the magic of computer modeling, a correlation was found. Computer models have this dandy ability to cull through data and endlessly find all sorts of meaningless correlations. Especially when the data is tenuous to begin with.
The headlines should have said: “After eliminating 3/4ths of the records, a single waist measurement taken nearly 40 years earlier was linked in a computer model to mental impairment checked off on electronic billing records.” That would have been recognized as another meaningless correlation and not generated the sensational media stories.
When study results are so overstated and employ such intense marketing techniques, it pays to take a look closer with a critical eye.
How soon we forget
Three years ago, the very same story hit the news, reporting: “Study links middle-age obesity to dementia.” Media around the world reported that this new study provided “the most convincing research so far” that being fat raised one’s risk for dementia. It was held up as another reason to be concerned about the “obesity crisis” and the need for action. “For God's sake, we better get cracking," said Philip James, the head of the obesity lobbying group, International Obesity Task Force.
That 2005 study was actually the first to suggest a link between fat and dementia and its authors are the main source for this continued fear. The 2005 study was done on the very same study population, using the same techniques and led by the same researchers, as the study released this week online, ahead of the print edition of the journal Neurology.
As readers may remember, the 2005 study also used science by press release and was not only reported in at least two different medical journals, it was released to the media before being published in the British Medical Journal.
When medical professionals finally had access to the actual study, however, they found that it hadn’t been able to find a single tenable correlation between dementia and BMI (body mass index) or body fat as measured by skinfold thickness (subscapular and tricep). By then, the media had moved on to the next news story and the public never heard these facts.
In the study, the authors said they were unable to come up with a biologically plausible explanation for how fatness might make anyone senile, either. They said they had exhausted all of the possible explanations among the fat dementia patients they could come up with — and found no link with cardiovascular disease, diabetes, “metabolic syndrome,” inflammatory markers, etc. Few consumers heard that, either.
Despite finding no link to obesity or body fat, for this new paper, the authors reanalyzed the data using new permutations, looking for a link between waist size and dementia. Examining the 2005 and 2008 studies together, comparing and contrasting, provides a valuable opportunity to see for ourselves the machinations in data dredges and enable us decide for ourselves how much “weight” these types of studies deserve.
Both studies were led by Rachel A. Whitmer, Ph.D., an epidemiologist with Kaiser Permanente Northern California. Kaiser Permanente is one of the largest health-maintenance organizations in the country and since 2002 has taken a leading role, partnered with the Robert Wood Johnson Foundation, CDC, American Association of Health Plans, Health Partners, and the National Business Group, to develop the national anti-obesity agenda, including clinical guidelines for doctors and education for children and healthcare professionals.
These studies were both retrospective. This is significant because it is the weakest type of observational study design. The authors culled through data they specifically chose to consider that had been gathered on a group of people, one time nearly 40 years earlier, searching retrospectively (backwards in time) for correlations among those with and without dementia recorded on their charts today.
“With few exceptions, retrospective study designs are inferior to prospective study designs,” said Dean R Hess, Ph.D., RRT FAARC, at Massachusetts General Hospital, and Harvard Medical School in Boston, Massachusetts. As he explained in “Retrospective Studies and Chart Reviews,” published in a 2004 issue of Respiratory Care, this study design should never be used when a prospective design can be done.
Retrospectives are rife with disadvantages that cloud their results, said Dr. Hess, such as “garbage in—garbage out” as they rely on the accuracy of records not generated for the purpose of the study; they are subject to bias and the actual important data and confounders may not be considered or available; there is no randomization and no blinding; and you cannot establish cause and effect. The “results are, at best, hypothesis-generating,” he emphasized.
Each of his points are evident in these studies.
These studies both used the identical database and exact same criteria to identify the people to include in their analyses — but the numbers and demographics of those included were very different, with no explanation given for the discrepancies.
The studies stated they included the records of members of the Kaiser Permanente Medical Care Programme of Northern California who participated in Multiphasic Health Checkups (MHC) in San Francisco and Oakland, California, between 1964 and 1973. Both of these papers used people aged 40-45 during those years and who were still members of Kaiser Permanente in 1994. These studies used the information recorded at the first checkup. As part of the MHC, comprehensive physical exams were conducted by healthcare professionals, which included blood tests for cholesterol and glucose, vital signs, physical measurements (height, weight, skinfold thicknesses, waist circumferences, etc.), demographic and medical history information.
The 2005 paper reported that 25,290 Kaiser members met this criteria. According to this paper, of the original cohort, 13,014 people were lost in follow-up (including 2,598 who died and 10,407 who were no longer members by 1994). The 2005 paper conducted its analysis on 10,276 people.
The 2008 paper said there were only “8,664 continuous members” who met the very same criteria, and they excluded 2,081 more for missing data. This paper conducted its analysis on 6,583 members.
So, the 2008 analysis was done on a mere 26% of the original cohort, but nowhere did this paper disclose that fact or that there were actually 25,290 people in the original cohort and most had been excluded. Nor did it provide any explanation why their figures for continuous members differed from 2005. Nor do we know how the people they chose to exclude differed (physically, demographically, or medically) from those they used.
Neither paper presented the medical, demographic and physical characteristics of members they lost in follow-up or excluded, compared with those that they used. We might be tempted to trust there were no differences, except, that doesn’t appear to be the case. While detailed information on the study participants wasn’t provided, what is available even shows notable differences between the 2005 and 2008 papers. For example, only 19.5% had less than a high school education in the 2005 paper, compared to 38.3% in the 2008 study. In the 2005 paper, 10% of the people were obese in mid-life, but only 7% were obese among the people used in the 2008 paper.
No explanation is given for how they managed, using the same criteria, to create an incarnation of members for this 2008 study that differed so much from the earlier one.
How dementia was identified may surprise you. The cases of dementia were not determined by actual clinical examinations or confirmed by medical evaluations. The authors used electronic billing records and noted when the ICD code for “dementia” was checked off. The codes they tallied included the code for any “memory impairment.” These codes are taken from the International Classification of Diseases, an enormous, complicated and continually changing system which assigns a number to every disease and medical procedure, and currently has about 12,000 codes. The medical literature is filled with documentations of their inaccuracies in reflecting actual patient disease rates. Over recent years, healthcare providers are being encouraged to check these codes in order to receive reimbursements.
This is especially problematic for a difficult diagnosis like dementia. Dementia is a deterioration in brain function and cognitive skills that comes with aging. Today, it’s known to have many different manifestations and causes, including brain abnormalities, vascular problems and infections. Dementia doubles in prevalence every 5 years after age 60, so that by 85 years old, 30 to 50% of seniors are affected. Diagnosing dementia is extremely imprecise, according to the Merck Manual of Geriatrics, and requires a very thorough physical evaluation to differentiate it from benign age-related memory loss. It also needs to be differentiated from a lot of treatable conditions that mimic dementia among elderly, such as the effects of illness; prescription medications; hypothyroidism; vitamin B12 deficiency; depression and isolation; and poor oxygenation due to lung, heart or circulation problems. A busy practitioner might understandably check-off the ICD code for memory impairment to ensure receiving compensation for the extra time spent with a confused, elderly patient, without having done a full work-up for an accurate diagnosis of true dementia, as these authors assume.
The 2005 study reported that among the 10,276 people seen between 1964-1973 and who were still plan members in 1994, 713 cases of dementia (about 7 percent) were identified between January 1994 and April 3, 2004, at an average age of 74.5 years. About 77 cases/year were identified during these follow-up years.
The 2008 study reported that among the 6,583 members in their analysis, there were 1,049 cases of dementia between January 1994 and June 16, 2006. They provided no explanation as to why in the two additional follow-up years, the annual rate of dementia rose by more than 112%, considerably beyond what would be expected.
In 2005, the researchers created three models for each measure of fatness (BMI and skinfold thicknesses), looking for a correlation to dementia. Their computer models were unable to find any tenable associations — above random change or a computer modeling error for these types of studies. Underweight was associated with a 24% higher odds ratio with dementia; “overweight” with a 35% risk, and “obese” a 74% risk.
Not only were these correlations untenable, but fatness was not what distinguished those with dementia. Increased risks for dementia were found in elderly women of all weight categories, but not in the men — although the men were fatter than the women! The relative risks for dementia associated with the greatest obesity in men was 0.84 to 1.20 — hugging either side of 1 and within random chance— a null finding. They tried to explain this by blaming it on not having the power to statistically detect an effect: “There were fewer obese and overweight men, thus the power to detect an effect was reduced.” But there were more obese and overweight men (2,741) than women (1,960).
This 2008 data dredge found the most variability in risks for dementia among the ‘normal’ weight people (odds ratios 0.94 to 3.81) compared with heavier people, negating a link to weight itself, but to other factors, such as genetic predispositions. Individual body shapes, including where fat accumulates, are also genetically determined, admitted lead author Dr. Whitmer. Yet, genetic or familial predispositions were not considered.
According to the authors in both papers, “those with a grade-school education were significantly more likely to be diagnosed with dementia,” suggesting a role for socioeconomic factors in memory function among elderly. They admitted, for example, that no information on dieting, nutrition, or actual cognitive function had been collected on the study participants at any time. Undernutrition and underweight has repeatedly been shown in other studies to have a key role in mental decline among elderly. Nor did they examine the use or overuse of certain prescription medications, even though it is one of the most common reasons for cognitive problems among elderly.
Once again, the association between fatness and dementia appears spurious. Even if the data used had been sound and a link to obesity tenable, countless confounding factors were not included or considered that could be the actual contributing causative factors, especially since they included all memory impairments in their term for dementia. Elderly women, for example, are more likely than men to have things known to exacerbate mental decline with advanced aging, increasing their risks of being misdiagnosed as having dementia in this study: be widowed, live alone with less social and mentally-stimulating interactions, live on lower fixed incomes and to have been smokers; and to have spent years “watching their figures,” undereating and restricting their food choices, and to have nutritional shortfalls.
Correlations are not causations. Correlations reported in epidemiological studies, as Dr. Hess noted, can only suggest a hypothesis for a role that then needs to be tested in clinical intervention trials and replicated. Here, there are no other studies. As the Kaiser Permanente author and colleagues wrote in 2005: “To date, this is the first study to determine the contribution of mid-life adiposity... on risk of dementia.”
And that study showed no link between BMI or obesity and dementia. With no link, weight loss makes no sense, especially since there is no known way to safely lose weight and keep it off. This new study is being cited as “another good reason to drop a few pounds” with no connection at all.
Even more nonsensical, this study is being used to recommend people “get rid of” their bellies. But everyone knows, spot reducing is a weight loss myth, too.
© 2008 Sandy Szwarc. All rights reserved.