October 29, 2015

A non-controlled intervention trial  was conducted by Lustig et al. among a small group (n=43) of Latino- and African-American youth (aged 8-18) who were obese, had high habitual sugars intakes, and had at least one other co-morbidity for metabolic syndrome (e.g. hypertension, hypertriglyceridemia, impaired fasting glucose, hyperinsulinemia). The study attempted to address an important question: does fructose influence major risk factors for diabetes or heart disease independent of Calories?

For 9 days, participants’ intakes of fructose were reduced through substitution with other carbohydrates in the form of fruits, bagels, cereal, pasta and bread (although the authors did not provide any detail on the macronutrient composition of the test meals or the actual food intake). Self-reported body weight was recorded daily; body composition and metabolic markers were measured at the beginning and the end of the intervention.

Due to flaws in the study design and statistical analyses of the results, conclusions must be drawn with caution. These are discussed in detail below:

1) There was significant weight loss due to Calorie deficits.

This study claimed to demonstrate that isocaloric substitution of fructose with other carbohydrate foods improved metabolic parameters independent of caloric/weight change. However, the findings do not support this conclusion as over 75% of participants lost weight—on average 2 pounds— within the short 9-day timeframe. For such a rapid weight loss to occur, the participants would have been in negative caloric balance. As reported in the paper, 33 participants were unable to consume all of the food provided for weight maintenance, resulting in reduced Caloric intake as compared to their usual eating patterns. It is also unclear which foods the participants were unable to consume and how it may have affected macronutrient or fructose intake.

With weight loss and Caloric deficits as confounding factors, the observed changes in metabolic outcomes cannot be attributed to isocaloric substitution of fructose for other carbohydrates.

2) This was not a randomized trial and there was no control group.

There was no external control group, where the same group of participants (or a separate group with similar demographics and baseline characteristics) consume their habitual diet or a control diet for the same period of time with all other variables staying the same. As a result, the “placebo effect” cannot be ruled out, the impact of which has been well documented in the literature. The participants spent 9 days thinking about “the special treatment” and weighed themselves daily; both of these factors may have subconsciously changed other dietary and behavioural habits. Furthermore, physical activity levels of participants were not recorded and participants’ actual food intakes were not reported in the paper. All of these unknowns make it difficult to interpret the results.

3) Flawed statistical analysis was used and if properly performed, there would have been no difference on measured outcomes in participants who did not lose weight.

Many of the outcomes measured in this study are not independent. For example, glucose and insulin are correlated; i.e. the change of one may lead to the change in the other. A similar relationship exists for blood triglycerides, LDL-cholesterol and HDL-cholesterol levels. Under such circumstances, the statistical significance level needs to be further adjusted (such as a ‘Bonferroni correction’) to account for such multiple testing of correlated metabolic markers (so that there is less probability the observed changes were due to chance) [1]. However, no adjustment was done in this study despite carrying out multiple testing. If proper statistical adjustment was applied, none of the measured parameters would be found to be different at the end of the intervention compared to baseline among the 10 participants who did not lose weight. Even with flawed analysis, only blood glucose and insulin levels differed whereas no change was observed for all other lipid and liver function parameters.

4) Wide participant demographics but small sample size further limit the data extrapolation to the general public.

There were only 43 participants with a wide range of ages (8-18 years old), different pubertal development stages, and different ethnic backgrounds (27 Latino and 16 African-American). The plausibility to extrapolate findings to the general public is very limited.

While the data and protocol described in this paper do not provide convincing data on fructose per se and contradict the study title and conclusion, what can be concluded is that weight loss (and possibly other undetected dietary and lifestyle changes) can lead to rapid improvements to blood glucose, insulin and lipid profiles. Health benefits of weight loss have been tested and proven by an overwhelming amount of clinical evidence. Further research is warranted to explore the underlying reasons for the marked improvement in measured outcomes (especially a 50% reduction in both blood triglycerides and insulin) within just 9 days, with a well-controlled study design.


Further reviews and critiques of this study from:

[1] Bender R, Lange S. Adjusting for multiple testing—when and how? Journal of Clinical Epidemiology. 2001;54:343-9.