Title, Date of Publication & Journal

Changes in coffee intake and subsequent risk of type 2 diabetes: three large cohorts of US men and women. Bhupathiraju, Diabetologia, Apr 2014

Claim supported by evidence?

The paper does not provide new support for the claim that “increasing coffee consumption by on average one and half cups per day (approx 360ml) over a four-year period reduces the risk of type 2 diabetes by 11%.” 

Summary

• Investigation relates “change of coffee intake” instead of absolute amounts of coffee (or caffeine) intake to risk of developing diabetes
• Records were updated every 4 years, but update only included last year (no account for the coffee intake in the other 3 years)
• The study only evaluates potential short term effects (4 years); it does not provide evidence that long term coffee intake had benefits for the risk of diabetes.
• Therefore, no recommendation for coffee intake can be derived from this paper.

Study Conclusions

• Participants were stratified into 5 groups with regard to “change of coffee intake” compared to 4 years earlier. In these 5 groups, there appears to be a numerical advantage (risk reduction) of the participants who increased their coffee intake by more than one cup per day.
• The results are in line with other clinical trials; however the variable “change of coffee intake” might be misleading, so no new evidence is generated with this investigation (see Limitations below)

Strengths/Limitations

Strengths

+    Inclusion of three large, long term studies which focused on nutrition (coffee intake was not their primary objective)

+    Findings are in line with previous randomised trials of caffeine intake that reduces biological markers of diabetes risk

Limitations

–    Investigation does not provide new evidence, because the data were secondary in the trials, and the statistical analysis is potentially misleading

–    It is unclear why “Change of coffee intake” was used as a primary variable of interest, and it is not clear whether they had prospectively defined this variable.

–    Typically the best analysis variable would be the absolute “coffee intake” (which can be updated as “moving average”) – the authors have not commented why they have not used it (maybe just because it was not statistically significant).

–    Only the last year in every 4 year period was used in the analysis – and the risk of diabetes was obtained during the next 4 years, but not later.

–    The different groups (of coffee intake) are not equal at baseline – e.g. the group with larger reduction of coffee intake started with a relatively high coffee intake at baseline.

–    If is not clear from the description of the analysis whether the differences at baseline (in particular for coffee intake) are accounted for correctly

–    It is not clear how the “person-years” (the period in which the participants are at risk in this trial) have been calculated, in particular as someone might have increased coffee intake in one 4-year-period and decreased it in the next 4 years.