Does Intentional Weight Loss Reduce Mortality?
Wednesday, April 1, 2015
Given that virtually every risk-factor guideline (from hypertension to diabetes) recommends “weight-loss” as “first-line” treatment, to many this may seem a rather stupid question, but in reality, there is in fact almost no high-qulaity evidence to show that intentional (non-surgical) weight loss actually reduces mortality.
This may well be simply because randomised controlled studies on intentional weight loss have either never been long enough, big enough, or able to sustain large enough differences in body weight between the interventions arms.
Nevertheless, now a meta-analysis by Stephen Kritchevsky and colleagues from the Wake Forest School of Medicine, published in PLOS one, suggests that being randomised to the intervention arm in a weight-loss study may reduce mortality by 15%.
The researchers identified 15 randomised controlled trials of behavioural interventions for weight loss of at least 18 months duration, of which 12 reported at least one death in either of the intervention arms.
These studies included 17,186 participants (53% female, mean age 52 years, mean BMI range 30–46 kg/m2 ) with follow-up times ranging from 18 months to 12.6 years (mean: 27 months), and an average weight loss of 5.5±4.0 kg.
Based on a total of 264 deaths in weight loss groups compared to 310 in non-weight loss groups, the authors calculated a 15% lower all-cause mortality risk (RR = 0.85; 95% CI: 0.73–1.00) in the weight-loss group.
Thus, the authors conclude that being randomized to the weight-loss arm in a behavioural weight-loss study may indeed reduce mortality risk.
However, as readers may realise, this study certainly does not “prove” that it is the actual weight loss that mediates these effects. After all, to achieve and sustain weight-loss through behavioural interventions, participants would have had to change their diet and activity levels to a greater extent that those in the control group. Thus it is very possible that the difference in mortality between the groups could well have been due to changes in health behaviours rather than due to the actual weight loss.
Nevertheless, the findings are reassuring in that they at least do not show an increase in mortality, something that people have feared may happen with intentional weight loss, especially in older individuals.
I guess the most we can conclude from this study is that being lucky enough to be randomised into the “weight-loss arm” of a behavioural weight-loss RCT may just help reduce your mortality risk.
Edmonton, AB
Thursday, April 2, 2015
I am glad you are blogging about this study. I was looking at it just yesterday and it left me scratching my head. Hopefully, I am just being paranoid, but there seems to be something fishy about their data.
I am trying to understand how they came up with the 95% confidence intervals in Figure 3. That’s the basis of the whole analysis. I don’t have access to most of the full-text articles (paywall) but some of the abstract report 95% confidence intervals on total mortality. But they don’t match those reported in the meta-analysis.
For instance, the Shea 2010 study reports an HR of 0.5 (0.3 to 1.0) in the abstract. The meta-analysis reports 0.5 (0.28 to 0.89) instead.
Shea 2011 reports an HR of 0.82 (0.55 to 1.22) in the abstract. The meta-analysis reports 0.93 (0.65 to 1.33) instead.
The Wing 2013 study (full text) reports death HR of 0.85 (0.69 to 1.04). The meta-analysis reports 0.86 (0.71 to 1.05) instead.
I don’t have access to the other ones. The three results I could check were “adjusted” in unknown ways. I don’t know what to think.
Monday, April 6, 2015
That is always the problem with these type of analyses – we need to take this study for what it is worth – I’d much rather have a prospective RCT that shows a reduction in mortality. Unfortunately we don’t even have that for surgery (and never may).
Monday, April 6, 2015
Sorry, I don’t understand what you mean by “That is always the problem with these type of analyses.”
Do you think there are mistakes in the paper?
Is the data subjected to some statistically magic that is not explained in the paper?
Is it outright fraud?
I know that meta-analyses can be tricky because the authors can choose selection criteria to get just the results they want. But I was expecting to see the 95% confidence intervals reported exactly from the references. I don’t understand why that is not the case.
Monday, April 6, 2015
Sorry, don’t have an explanation for you – perhaps you should contact the authors?