Meta-analysis of trial data may support a causal role of hyperglycaemia in cancer

Excerpt

To the Editor: We read with interest the article by Johnson and Bowker published in Diabetologia [1]. However, this study had major methodological flaws that render the conclusion premature and invalid. First, the sample size was not sufficient to support the negative conclusion that intensive glycaemic control did not reduce cancer risk. We used the published data in the meta-analysis and performed the sample size and power analysis that would be required to make such a conclusion. In the control arm, 380 cancer events were recorded during 45,009 person-years of follow-up with an incidence of 8.44 (95% CI 7.60–9.29) per 1,000 person-years. Given an incidence of 7.44 (95% CI 6.67–8.12) per 1,000 person-years in the intensive glycaemic control arm (357 cancer events during a total follow-up time of 47,974 person-years), the current sample size had zero power at a type I error rate of 5% to make a conclusion of equivalence. In other words, data from this meta-analysis were severely underpowered to justify the conclusion that ‘cancer risk’ was not reduced by improving glycaemic control. …