This is the cleanest evidence chain in nutrition, and it is worth walking end to end because almost every popular account of it stops at the first number and skips the sentence that explains it.
The design that makes the argument
Te Morenga, Mallard and Mann, in the BMJ 2012, ran a systematic review and meta-analysis of randomised controlled trials and prospective cohort studies. OVID Medline, Embase, PubMed, CINAHL, Scopus and Web of Science to December 2011, with a minimum duration of two weeks for trials and one year for cohorts, selection, assessment and analysis done under Cochrane Collaboration and GRADE methods. Trials of weight loss, or confounded by additional medical or lifestyle interventions, were excluded.
Thirty of 7,895 trials and 38 of 9,445 cohort studies were eligible. The first number is the one to remember when a page tells you the evidence is weak.
The four numbers, in order of importance
In trials of adults eating freely — ad libitum, meaning no strict control of food intake — reducing dietary sugars was associated with a decrease in body weight of 0.80 kg (95% CI 0.39 to 1.21, P < 0.001). Increasing sugars was associated with a comparable increase of 0.75 kg (95% CI 0.30 to 1.19, P = 0.001).
Now the third number, which is the one almost nobody quotes. Isoenergetic exchange of dietary sugars with other carbohydrates showed no change in body weight: 0.04 kg (95% CI −0.04 to 0.13).
Read the three together and the mechanism writes itself. When you take sugar out of a free-living diet and replace it with nothing, people lose under a kilogram. When you swap sugar for another carbohydrate at equal energy, nothing happens at all. The first effect is energy intake; the second is sugar specifically, and it is absent.
The authors’ conclusion states this: among free-living people on ad libitum diets, intake of free sugars or sugar sweetened beverages is a determinant of body weight, and the change in body fatness that occurs with modifying intakes seems to be mediated via changes in energy intakes, since isoenergetic exchange of sugars with other carbohydrates was not associated with weight change.
The cohort arm, and its confidence intervals
The same paper reports the prospective side. For intake of sugar sweetened beverages after one year of follow-up, the odds ratio for being overweight or obese was 1.55 (95% CI 1.32 to 1.82) comparing the highest intake group with the lowest.
The children’s trials in the same review are the part most often left out: trials involving recommendations to reduce intake of sugar sweetened foods and beverages had low participant compliance with the dietary advice, and showed no overall change in body weight. That is a fact about adherence, not about sugar, and it belongs in the same sentence as the rest.
The authors flag significant heterogeneity in one meta-analysis and potential bias in some trials, but report that sensitivity analyses showed the trends were consistent and the associations remained after those studies were excluded.
The outcome that isn’t weight
Imamura and colleagues, in the BMJ 2015, asked a different question with the same design: prospective associations of sugar sweetened beverages, artificially sweetened beverages and fruit juice with incidence of type 2 diabetes, before and after adjustment for adiposity, plus the population attributable fraction. Seventeen cohorts, 38,253 cases, 10,126,754 person-years.
Sugar sweetened beverages: 18 per cent greater incidence per one serving per day (95% CI 9 to 28 per cent) before adjustment for adiposity, 13 per cent (6 to 21) after. Artificially sweetened beverages: 25 per cent (18 to 33) and 8 per cent (2 to 15). Fruit juice: 5 per cent (−1 to 11) and 7 per cent (1 to 14).
Two things are worth noticing. The heterogeneity is enormous — I² of 89 per cent for the main estimate, 79 per cent after adjustment — so these are pooled estimates across very different populations. And the fruit juice confidence interval crosses zero before adjustment.
The argument against, taken seriously
Kaiser, Shikany, Keating and Allison, in Obesity Reviews 2013, wrote a paper whose title is the honest summary of the field’s asymmetry: evidence supporting the conjecture is strong, but evidence when testing the effect is weak. They abstracted randomised controlled trials examining sugar sweetened beverages and body weight and composition indices, requiring human trials of at least three weeks with random assignment to conditions differing only in the beverage.
Their updated meta-analysis of seven studies that added the beverages to people’s diets showed weight increases that grew with the amount consumed. The meta-analysis of eight studies attempting to reduce consumption showed the weaker side.
Both directions are in the same paper. A supplement page that cites only the first is doing what supplement pages do.
What this evidence does not show
It does not show that sugar is uniquely fattening — that is precisely what the isoenergetic exchange comparison rules out at these doses and durations. It does not show that reducing sugar is useless: 0.80 kg in free-living adults across trials is a real effect, and it is an effect of energy displacement, not a metabolic quirk. And the cohort figures describe association in populations with heterogeneity this large, not a mechanism; Imamura’s own I² values say how much the pooled number should be trusted.
This article summarises published meta-analyses. It is not medical advice and it is not a recommendation for any individual.
