NEJM obesity study is junk science:

Study doesn’t link being overweight with heart failure

 

By Steven Milloy

July 31, 2002

 

Summary

 

Obesity is associated with a variety of health problems. However, there are conflicting results and genuine controversy over whether just being moderately overweight is a health risk.

 

This study by Kenchaiah et al. in the New England Journal of Medicine (Aug. 1, 2002) doesn’t resolve this controversy, despite the claims of the researchers and the accompanying editorial.

 

Taken in the best light, the results are statistically weak. Even so, the reported results for men and women inexplicably conflict.

 

Finally, there was no medical determination that overweight caused the observed cases of heart failure - not in one case. The researchers attempted to make the link with statistics alone. Family history of heart problems and physical activity were not considered as confounding risk factors for heart failure.

 

The researchers used statistical trickery and outright dishonesty to make their claims.

 

The bottom line is that this study does not link overweight with heart failure.

 

Analysis

 

The study authors acknowledge at the outset of their article that the studies to date linking overweight with obesity are inconclusive:

 

Whereas extreme obesity has been associated with heart failure, data are limited regarding the influence of overweight and lesser degrees of obesity on the risk of heart failure.

The researchers grouped their study subjects by body mass index (BMI) into three categories: Normal (BMIs 18.5-24.9); Overweight (BMIs 25.0-29.9); and Obese (BMIs greater than 30.0).

 

Confounding risk factors for heart failure were considered in two ways: (1) base line - defined at the beginning of the study; and (2) time dependent - updated every four years.

 

For these groupings, the authors report weak statistical associations among overweight women and heart disease for base line and time dependent confounding factors: - (1) Base line confounders, Relative Risk = 1.50, 95% Confidence Interval (1.12 – 2.02); (2) Time dependent confounders, Relative Risk = 1.68, 95% Confidence Interval (1.25- 2.27).

 

These results are too weak to be meaningful. As the National Cancer Institute states,

 

"In epidemiologic research, relative risks of less than 2.0 are considered small and usually difficult to interpret. Such increases may be due to chance, statistical bias or effects of confounding factors that are sometimes not evident."

 

For men, the reported results are also weak and - worse - not statistically significant, meaning that they could be flukes: (1) Base line confounders, Relative Risk = 1.20, 95% Confidence Interval (0.87 – 1.64); and (2) Time dependent confounders, Relative Risk = 1.17, 95% Confidence Intervals (0.86-1.61). Because the lower bounds of the confidence intervals are less than 1.0, the relative risks aren’t statistically significant.

 

These results are problematic for the researchers not only because are they weak, but because they conflict even if the weakness is ignored. That is, the result for men isn’t statistically significant and so cannot be viewed as linking overweight with heart failure. If overweight was a real risk factor for heart failure, why in women, but not men? They offer no explanation and this problem is fatal to the link.

 

So the researchers tried statistical trickery to eliminate this problem. They looked at BMI as a “continuous variable” - i.e., BMI categories in increments of 1 such as 25.0-25.9, 26.0-26.9, etc.

 

But even this failed as the (again) weak associations still conflicted.

 

For women, the reported association was: Relative Risk = 1.07, 95% Confidence Interval (1.04 – 1.10). For men, the reported association was: - Relative Risk = 1.04, 95% Confidence Interval (1.00 – 1.07).

 

The men’s result is again not statistically significant as the confidence interval includes the no-effect level of 1.0. This result does not link overweight with heart disease and inexplicably conflicts with the women’s result.

 

The authors acknowledge this problem, but ignore it:

 

Although the hazard ratio for heart failure in overweight men did not achieve statistical significance, analyses evaluating body mass index as a continuous variable and findings of trend models support the existence of a continuous gradient of heart-failure risk with increasing body mass index in both sexes.

 

Their statement about the continuous variable supporting an association between BMI and heart failure is obviously false - the men’s confidence interval includes the value 1.0 and so the relative risk is not statistically significant.

 

Their claim that “trend models” support the association is also incorrect.

 

The trend model relies on the existence of the reported men’s relative risk which, as discussed above, is not statistically significant. If the relative risk is not statistically significant, then it cannot be a data point. So there is no “trend.”

 

Finally, the accompanying editorial, written by a consultant to drug manufacturers Bristol-Myers Squibb and Merck, promotes the spurious link between overweight and heart failure apparently to sell drugs to the population of overweight people:

 

On the basis of the findings… overweight and, particularly obesity should now be added to the list of risk factors for the development of heart failure…. Certain therapies for hypertension and diabetes, such as inhibitors of the rennin-angiotensin system and metformin, appear to be particularly effective for treating or preventing some of the consequences of overweight and obesity and may thereby reduce the risk of heart failure. Other therapies, such as antiplatelet drugs for dyslipidemia, may accomplish the same goal…

 

Is this yet another embarrassing conflict of interest for the New England Journal of Medicine?

 

Steven Milloy is the publisher of JunkScience.com , an adjunct scholar at the Cato Institute and the author of Junk Science Judo: Self-defense Against Health Scares and Scams (Cato Institute, 2001).

1