Post a new comment on this article
Post Your Discussion Comment
Please follow our guidelines for comments and review our competing interests policy. Comments that do not conform to our guidelines will be promptly removed and the user account disabled. The following must be avoided:
- Remarks that could be interpreted as allegations of misconduct
- Unsupported assertions or statements
- Inflammatory or insulting language
Why should this posting be reviewed?
See also Guidelines for Comments and Corrections.
Thank you for taking the time to flag this posting; we review flagged postings on a regular basis.close
Correspondence to PloS Medicine: Reply to Meir Stampfer, December 3, 2005.
Posted by plosmedicine on 30 Mar 2009 at 23:49 GMT
Author: Thorkild I.A. Sorensen
Position: Professor at the Institute of Preventive Medicine
Institution: Copenhagen University Hospital
Additional Authors: David F. Williamson, Jaakko Kaprio
Submitted Date: December 23, 2005
Published Date: January 5, 2006
This comment was originally posted as a “Reader Response” on the publication date indicated above. All Reader Responses are now available as comments.
In his correspondence , Stampfer raises two issues regarding excess mortality observed in the study reported by Sorensen et al. of presumed healthy overweight subjects who intended to lose weight and did lose weight . The first issue is whether the results could be due to inadequate control for smoking; the second issue is whether the study failed to measure the effects of intentional weight loss because no difference in mean weight loss was observed between those intending and not intending to lose weight. Below we address these two issues.
For Stampfer's smoking argument to be valid, the group who intended to lose weight and did lose weight should have more smokers and/or have smoked more at baseline, or during the period of time when the weight loss was observed, than the relevant comparison groups. Depending on the question, there are three relevant comparison groups: those intending, but failing to lose weight, those not intending to lose weight but who did lose weight, or those not intending to lose weight and keeping weight stable. There were no such skewed distributions of smoking among these groups; in fact, those who intended to lose weight and succeeded in weight loss actually smoked somewhat less than the other groups. As a consequence, and as already mentioned in the paper , adjustment for smoking increased the hazard ratio associated with intentional weight loss. Furthermore, the analysis was also conducted with finer smoking categories, but the difference in mortality between the groups was virtually unaffected; this was also mentioned in the paper and was the main reason results were presented for the cruder categories, which allowed inclusion of more covariates without overloading the statistical model. In addition to the control for baseline smoking, changes in smoking during the observation period were also controlled for, which also failed to alter the results.
In a supplemental analysis, which stratified by never, former, or current smoking either at the beginning or the end of the weight loss period, the excess mortality associated with intentional weight loss remained in all strata. Although Stampfer tries to dismiss the results of the simulation study  that examined the exclusion of smokers from analyses of obesity and mortality, most studies of the relationship between body weight, smoking, and mortality do not find a major impact on the relationship between obesity and mortality by excluding smokers [4, 5]. The results reported by Sorensen et al  are in accordance with these findings.
Stampfer provides a table purporting to demonstrate that there is no practical difference in weight change between those intending to lose weight and those not intending to lose weight in the study by Sorensen et al . Apparently, Stampfer does not appreciate that the 87% increase in all-cause mortality rate reported in Table 3 of Sorensen et al was calculated by dividing the mortality rate in those who reported intentionally losing weight and who lost weight, by the mortality rate in those who did not report intentional weight loss and whose weight was stable. Table 2 of Sorensen et al shows the weight changes experienced by these two groups over a 6 years period. Those who reported intentional weight loss and who lost weight, lost 1.21 kg/m2 (about 2.7 kg for a person 1.5 m tall), while those who did not report intentional weight loss and whose weight was considered stable, gained 0.36 kg/m2 (about 0.8 kg for a person 1.5 m tall). Thus, the net difference in weight change between these two groups was about 3.5 kg for a person 1.5 m tall, which is quite similar to the net difference in weight change observed in long-term randomized controlled trials of dietary/lifestyle interventions for weight loss .
Although it is interesting and relevant to draw inferences from an observational study, as if it had been a randomized trial, this requires a more careful analysis than Stampfer has presented . If we assume we have a group of overweight subjects who are randomised to a program supposed to induce intention to lose weight versus no program, then the weight loss observed in the total group intending to lose weight should be compared to the group not intending to lose weight. In the study by Sorensen et al.  there is no difference in weight loss between the two groups (Table 2); in fact both groups increased similarly in BMI (0.33 and 0.31, respectively), and there was no difference in mortality between the two groups (adjusted hazard ratio, 1.0, CI 0.75-1.32).
However, if we carry out another trial consisting entirely of subjects intending to lose weight and one group were randomised to a weight loss program and another group were not, then we could perhaps assume that those randomised to weight loss might exhibit a weight loss as observed in the study, i.e. -1.21 kg, while the other group would either remain stable or gain a small amount of weight, also as observed in the study . Then the relevant comparison of mortality rates would be as depicted in Figure 2 of the paper by Sorensen et al , and the hazard ratio would be 2.13 (CI 1.22-3.71).
On the other hand, it may be argued that if the alternatives relevant for public health is either to encourage people to maintain their weight or to lose weight, and we then want to assess what the effects of this policy might be on total mortality among those who succeeded in complying with these goals, then the relevant comparison would be between those intending to lose weight who lost weight and those who did not intend to lose weight and whose weight remained stable, which resulted in the hazard ratio of 1.87 (CI 1.22-2.87)
There is convincing evidence for a short-term beneficial physiologic effect of weight loss on risk factors for diabetes and cardiovascular diseases in overweight or obese subjects, but there is a serious lack of studies addressing the long-term effects of intentional weight loss. The expectation would of course be that those who lost weight would benefit from the weigh loss in the long run. The study by Sorensen et al.  clearly came out against these expectations for intended weight loss. Even though the findings are statistically significant, they may of course still be chance findings. However, it should be noted that the p-values are referring to testing the null hypothesis of no differences; given the expectation of the opposite result, the likelihood of obtaining the observed result is much lower than the reported p-values.
The studies that have examined the long-term effects of intentional weight loss have so far provided a confusing picture, raising much doubt about the long-term benefits of weight loss. In our opinion it would - from a public health point of view - be inappropriate to disregard results like those of Sorensen et al. that do not fit with conventional wisdom based on the short-term physiologic effects of weight loss.
Thorkild I.A. Sorensen
Institute of Preventive Medicine,
David F Williamson
Centers for Disease Control and Prevention
Atlanta, Georgia, United States of America
University of Helsinki
1. Stampfer M (2005) Author's reply. PloS Med 2(9): e326. DOI: 10.1371/journal.pmed.0020326
2. Sorensen TIA, Rissanen A, Korkeila M, Kaprio J (2005) Intention to lose weight, weight changes, and 18-y mortality in overweight individuals without co-morbidities. PloS Med 2: e171. DOI: 10.1371/journal.pmed.0020171
3. BMI in Diverse Population Collaborative Group (1999) Effect of smoking on the body mass index - mortality relation: Empirical evidence from 15 studies. Am J Epidemiol 150: 1297-1308
4. Troiano RP, Frongillo EA Jr, Sobal J, Levitsky DA (1996) The relationship between body weight and mortality: a quantitative analysis of combined information from existing studies. Int J Obesity 20: 63-75
5. Calle EE, Thun M, Petrelli JM, Rodriguez C, Heath CW (1999) Body-mass index and mortality in a prospective cohort of U.S. adults. N Engl J Med 341: 1097-1105
6. Douketis J, Macie C, Thabane L, Williamson DF (2005) Systematic review of long-term weight loss studies in obese adults: Clinical significance and applicability to clinical practice. Int J Obesity 29: 1153-1167