June 6, 2016

Mediterranean diet, fat levels and body weight

The effect of a high-fat Mediterranean diet on body weight is examined in a paper published in the Lancet Diabetes & Endocrinology journal in which the authors report an unrestricted-calorie, high-vegetable-fat Mediterranean diet was associated with decreases in bodyweight in older people at risk of cardiovascular disease. Roundup comments accompanied accompanied this analysis.

Title, Date of Publication & Journal

‘Effect of a high-fat Mediterranean diet on bodyweight and waist circumference: a prespecified secondary outcomes analysis of the PREDIMED randomised controlled trial’

Lancet Diabetes & Endocrinology

Published on Monday 6 June 2016

Study’s main claims – and are they supported by the data

The study does suggest that some, but perhaps not all, Mediterranean diets may reduce weight compared to low fat diets but I agree with the headline emphasis that there is no weight gain, strictly speaking of course no evidence of any weight gain relative to a low fat diet in this particular cohort of people (older Spanish people with a risk of heart disease).

Strengths/Limitations

The title of the paper refers to ‘a prespecified outcomes analysis of the PREDIMED randomised controlled trial’, whereas the press release talks about a randomised trial (without ‘controlled’). The latter is correct because this was not a controlled trial in the normal sense – the only control used here was to recruit older, unhealthy white people living in Spain who were randomly allocated to 3 groups, each group receiving different dietary advice dispensed at regular intervals by trained dietitians. Note the word “control” in this sense does not refer to a control group (in this trial, one group was given standard low fat dietary advice as currently recommended by health authorities and could be regarded as the control group) but to the way the trial was designed, the criteria used to recruit the participants and the way their diets were controlled.

There are a number of other factors that could have influenced the result but they were not controlled for in the trial, so not all confounding factors were controlled for. I have always felt with dietary studies that ‘control’ is hard to achieve. Dietary effects are felt over long periods of time (as opposed to medical treatments for specific conditions) and by definition, you cannot really control for all the potential events that can occur in that time (this study lasted for five years) that could also affect weight and waist. So any dietary trial that seeks to be a true RCT under specialised conditions will always have a problem with its ability to be extrapolated onto the general population under everyday conditions.  My feeling is that this trial achieved the right balance between control and extrapolation.

At times in the paper, I was confused as sometimes they mentioned an increase in waist and other times they mentioned a decrease. The charts shown on page 8 though are consistent with the figures in the press release.

Dietary consumption was measured through self completed questionnaires. That sounds like an imprecise way to measure diet but the authors anticipated this issue and also took blood and urine samples at regular intervals and looked for key markers given by certain foods. That allowed them some measure of crosscheck with the questionnaire results.

I thought the paper was fair in terms of identifying possible weaknesses with the trial.

Comments in the press release regarding implications are generally reasonable but I think they should place greater emphasis on the fact that the trial was specifically for older unhealthy people. In other words, do they really think their findings could be extrapolated to younger healthy people? If so, on what basis do they equate these two populations? I could not see anything in the paper that justified such extrapolation to the wider population. It is possible that other trials have shown that such extrapolations are justified but no such references were given in the paper or press release.

Any specific expertise relevant to studied paper (beyond statistical)?

None

Before The Headlines is a service provided to the SMC by volunteer statisticians: members of the Royal Statistical Society (RSS), Statisticians in the Pharmaceutical Industry (PSI) and experienced statisticians in academia and research. A list of contributors, including affiliations, is available here.