Title, Date of Publication & Journal

Sanjay Basu, Paula Yoffe, Nancy Hills, Robert H. Lustig

The Relationship of Sugar to Population-Level Diabetes

Prevalence: An Econometric Analysis of Repeated Cross-

Sectional Data. PLOS ONE 2013, Volume 8, Issue 2, e57873

Claim supported by evidence

Does it show that sugar has a causal role in diabetes? Is that because of obesity or does sugar have a special role?

• The major message from this paper is “Obesity is the major factor for getting diabetes, but independently of this, also the amount of sugar intake has its role”.
• There is an (independent) impact of the amount of sugar intake, but obesity is a much stronger predictor for the risk of diabetes.

How strong is the evidence for the claims?

• The results are epidemiological research, but do not prove the underlying mechanism.
• The effect of sugar intake is statistically significant in all models they had investigated.
• The claim is based on aggregate measures across countries, not on individual data – this lowers the evidence slightly.

Summary

• Overall, the study methodology is sound, as it thoroughly investigates a number of models that capture different environmental effects, and check their robustness.
• In all of these models, obesity is a stronger predictor of risk of diabetes than the intake of sugar, but sugar independently leads to some risk increase.

Study Conclusions

• The main conclusion of sugar attributing to diabetes is supported, but this should not downgrade the larger risk of obesity on getting diabetes.
• They found that “in the periods after a country lowered its sugar availability, diabetes prevalence reduced by 0.074%” after controlling for other things. This is a pretty small reduction which does not allow to conclude similar as Sir Austin Bradford (hence the reference to him is wrong).

Strengths/Limitations

+        Sound epidemiological research

+        Various models with different environmental effects were investigated, and the impact of sugar was present in all models

+        Numerous checks for statistical robustness were performed

• The results depend on the quality of the data source, which was not checked for this review
• Without access to the supplementary material, it might not be possible to fully judge the methodology

–  The impact of obesity in this analysis is statistically much larger than the impact of sugar intake

–  The use of aggregated measurements might potentially be misleading when drawing conclusions on individual level.
It does not imply that at an individual level, within countries, it’s the individual people who eat a lot of sugar that are more prone to get diabetes.
It could happen that, in a country where a lot of sugar is eaten (on average) and diabetes rates are high, one lot of people eat a lot of sugar, and an entirely separate lot of people get diabetes at high rates.

‘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.