Relationships between black tea consumption and key health indicators in the world: an ecological study. BMJ Open 7 November 2012
The paper does not prove a link between high black tea consumption and low diabetes prevalence, but it does add weak support to this hypothesis.
This paper looks at the prevalence of type-2 diabetes in countries where there is a high consumption of black tea. It should be noted that black tea means tea made with black leaves (as we drink in the UK) – it has nothing to do with whether or not you add milk.
The study appears to show a linear statistical correlation between high black tea consumption and low diabetes prevalence in the world, although the authors are frank that ‘correlation does not imply causality’.
The linear correlation (fig 3) is driven by Turkey, UK and Ireland in the bottom right (with high tea consumption) and a country with very high diabetes and low tea consumption (in the top left).
This plot suggests that the data are not appropriately distributed for this regression/correlation method – rank correlation should have been used instead.
Ireland (with the highest black tea consumption) has a small population and so may be weighted too highly in this analysis.
In summary the significance of the results displayed here may not be as great as it first appears.
The authors are honest about limitations.
They recognise that “the quality of data collection can be expected to be heterogeneous around the world.” – e.g. in the UK, recording of diabetes prevalence is likely to be more thorough than in other countries.
The authors add “health indicators such as diabetes depend on diagnostic criteria, which can vary across countries”. – e.g. in UK testing is more thorough so if you have diabetes here you are likely to know about it. In other countries you may live your whole life with diabetes and never be diagnosed.
The methods appear appropriate overall but data availability is the main limiting factor.
There is an increased chance of false positives because of multiple testing (five key health indicators are used) – in other words the more things you look at the more likely you are to find something that appears significant.
The p-value of 0.003 comes from the same test done in two different ways (the paper appears to suggest that the second way supports the first). It would be more convincing if this p-value came out of two entirely separate tests.
The coefficient r2 of 0.199 indicates weak evidence for a relationship.
There is no discussion of the obvious driver for the association which is genetic susceptibility. Similarly there is no attempt to correct for covariates, such as latitude and longitude, which would serve as potential proxies for genetic susceptibility.
In summary: at face value, the results suggest that a high level of black tea consumption (as in the UK) would halve the prevalence of type-2 diabetes; but a lot more research would be required to prove this and this paper alone does not constitute proof.
‘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.