Discussion

In this discussion section, we begin by reiterating key methodological conclusions. We follow this with a discussion of what we believe the currently available data on relative body weight and mortality show and what the currently available data on body weight and mortality mean. We can point out that these are not necessarily the same thing.

Based on the information reviewed above, we reach the following conclusions:

1. While controlling for smoking either by stratification or statistical adjustment is a sound process and smoking is a plausible confounder of the BMI-mortality relationship, in actual data sets, adjusting for smoking has very little impact on the results of the analysis. This does not imply that one should not control for smoking. It only implies that smoking appears to be an unlikely explanation for the U- and J-shaped relationships frequently observed between BMI and mortality.

2. Excluding subjects who die during the first few years of follow-up is not a reliable way of controlling for confounding due to occult disease. In the presence of confounding due to occult disease such exclusions can either increase or decrease the bias, although in practice such exclusions appear to make little difference. Because such exclusions can actually increase the bias under some circumstances and result in an overall reduction of sample size, we do not recommend that subjects dying during the first few years be excluded from the analyses.

3. There is no a priori reason to assume, if a quadratic model is fitted to describe the relationship between BMI and mortality and the minimum of this quadratic equation solved for the resulting estimated BMI associated with minimum mortality, that the estimate will systematically overestimate the true BMI associated with minimum mortality. However, other methods for es timating the BMI associated with minimum mortality are available and may be superior. These methods do not require that BMI be categorized into quantiles but can be applied to BMI treated as a continuous variable.

4. BMI is a continuous variable and, as with other continuous variables, there is little advantage to categorizing BMI in the final analysis. It is certainly useful to treat BMI categorically in an exploratory manner. However, it is possible to treat BMI continuously in the final analysis and there are a number of advantages to doing so.

5. Though highly correlated with body fatness, BMI is not a true measure of body fatness and it cannot be assumed that BMI will have the same relationship with mortality in either direction or form as will a valid measure of body composition. Therefore, it is strongly suggested that future research consider including measures of body composition rather than just BMI.

6. There is substantial variation in results from study to study, some of which is probably due solely to random sampling variations. Because of this, selective review of the data can be used to support virtually any conclusion. Therefore, it is essential that reviews of the literature, if they are intended to be objective, evaluate the entire body of the literature to the greatest extent possible. This approach is exemplified in the recent papers by Allison et al. (32), Troiano et al. (9), and The BMI in Diverse Populations Collaborative Group (47).

7. The relationship between BMI and mortality appears to vary substantially by age, sex, and race. Other variables yet to be fully explored may also moderate this relationship. Therefore, it is ill advised to generalize from studies in the one population (e.g. white middle-aged females) to other populations (e.g. young black males or elderly Asian females). Moreover, investigators who wish to make broad statements about the overall 'average' relationship between BMI and mortality for the US population will need to rely on samples that are representative of the US population. Finally, this implies that it is wise for investigators to attempt to stratify by or fit interaction terms with their demographic variables and other possible moderators when analyzing the relationship between BMI and mortality.

Psychology Of Weight Loss And Management

Psychology Of Weight Loss And Management

Get All The Support And Guidance You Need To Be A Success At The Psychology Of Weight Loss And Management. This Book Is One Of The Most Valuable Resources In The World When It Comes To Exploring How Your Brain Plays A Role In Weight Loss And Management.

Get My Free Ebook


Post a comment