Weight and Mortality I - the study

First the study that everyone is talking about, is available free online at the JAMA site.

I think it has yet to sink in what a big deal this is. It is almost as if all of a sudden smoking wasn't bad for you.

I want to explain what the article shows and then discuss a bit.

The study looks at mortality in 3 cohorts of US citizens designed to represent the population at large. The cohorts are part of a longitudinal (basically following people over time) study funded by the NIH called the National Health and NutritionExamination Survey (NHANES) for which various information like height,weight, smoking status.

The study looks at 3 NHANES cohorts recrutied in 1971-1975, 1976-1980 and 1988-1994. Mortality was followed until 1992 for the first 2 and until 2000 for the 3rd group. They divided their cohorts into groups based on Body Mass Index (BMI) which is defined as weight in kg divided by height in meters squared (kg/m^2). Based on federal guidelines they defines a BMI of <18.5 as underweight, 18.5 to 25 as normal, 25-30 as overweight and >30 as obese. They further subdivided the obese group into classes with BMI of 30-35 and those >35

To give you an idea of about what these mean, I'm 6'2" and I'd be underweight at 144.5 lbs or less, regular weight from 145 to 194.5, overweight from 195 to 234.5, obese I from 235-273.5, obese 2 from 274 up.

They then looked at mortality based on weight using a statistical technique called multivariate analysis to control for things like age (older people die more) and smoking. This is important because it allows the true effect of weight itself to come out.

Surprisingly they find much less of an effect of increase BMI than had been previously believed. In patients <60, the lowest level of death was seen in the "overweight" category, but this was NOT statistically signficiant. Underweight and obese with BMI <35 individuals seemed to have higher death rates, but again this didn't reach statistical significance. Those with BMI >35 did have a statistically significant increased mortality. For older patients the data looked the same, but there were some minor differences in what did and did not reach statistical significance.

They go on to calculate the excess (or decreased) deaths attributable to people not being in the "regular" category, based on the US population as a whole. They find 110,000 excess deaths in the obese and 34,000 excess deaths in the underweight (there are a lot more obese people than underweight). The overweight group has 86,000 fewer deaths. Most of the increased deaths in the underweight occurred in people >70 suggesting being young and thin is probably okay.

For reasons that I can't explain clearly but have to do with statistics, the changes in estimated deaths were statistically significant, even thought the mortality rates in the first part of the study were not.

Finally, they show that the increased risk associated with obesity seems to have declined from the first cohort to the second and third, suggesting fat people are healthier than they used to be.

So to summarize the results

1. BMI<35 seems to play a small role in determining overall mortality.
2. The lowest mortality is seen in the BMI range of 25-30.
3. Underweight is associated with increase mortality, but the majority of this risk is in the elderly
4. Over time, the US population is getting fatter, but the negative health effects of obesity are diminishing.

Remember this is only one study, but a big one with roughly 36,000 people and 9,000 deaths. There may be criticisms raised about the methods or analysis and there is always a chance the sample, although large, is anomalous.

Remember that differences in mortality in the underweight (higher) and overweight (lower) were NOT statistically significant.

Nonetheless, the results challenge the idea that carrying a few excess pounds is unhealthy, despite the clear association of obesity with cardiovascular disease and diabetes.

My #1 takehome message for now would be not to worry so much about one's weight but focus on healthy behaviors like good diet, regular exercise, not smoking and moderate alcohol consumption.