Statistical Methods in Genetic Epidemiology

Correction: no evidence of an association between the T16189C mtDNA variant and late onset dementia (Gibson et al) We believe that the title of Chinnery et al's paper should be corrected because the data the authors present do not include an analysis of the 16189 variant of mtDNA (Table 1).We defined the 16189 variant as the DNA sequence associated with a polydC tract, 2 resulting from a T16189C transition that may generate heteroplasmic length variation, table 1. Heteroplasmic length variation does not occur when the polymeric tract is interrupted by a cRt transition, which occurs at several different sites but commonly at nucleotice 16186 or 16192.Individuals with these additional polymorphisms are excluded from our definition of the 16189 variant because they no longer have a long homopolymeric c tract.The variant does not alter any coding sequences yet lies near to mtDNA control sequences, which can explain its effects on mitochondrial function.In studies of disease associations with variants in this region we chose to investigate the 16189 variant rather than any other sequence change, because of the likely functional effects of the homopolymeric C tract and heteroplasmic length variation.Gibson et al 1 have shown that the overall prevalence of the T16189C allele in their population is 12.6%, which is substantially higher than the 6.4-8.8 % prevalence of the 16189 variant reported in other studies. 2 3 This is because they have quantfied the prevalence of the T16189C transition per se rather than the variant.Including these additional polymorphisms may dilute out a real association with the 16189 variant.The authors have shown that the T16189C transition per se is not a risk factor for late onset dementia, 1 but to our knowledge this has not, in any case, been implicated with any disease phenotypes.However, they found that the heteroplasmic length variation, which implies the presence of the 16189 variant, was associated with a 2.2 fold increased risk.They did not, however, quantify the relative risk for the 16189 variant per se, which could well be significant, in direct contradiction of their title.From their data, it is possible that the variant might in fact predispose to late onset dementia.The 16189 variant is a risk factor for type 2 diabetes, 4 thinness at birth, 5 and aged 20 years 6 iron loading in haemochromato- sis, 7 dilated cardiomyopathy, 8 endometrial cancer, 9 and other multifactorial disorders. 10This variant may be mildly detrimental. 11Unlike many other mtDNA polymorphisms implicated in type 2 diabetes, this variant probably has bona fide functional consequences because it has arisen many times independently in the various populations studied, 2 4 12 excluding a founder effect.Because the authors did not perform mitochondrial haplotyping 1 to exclude a founder effect, their results may reflect the consequences of other co-segregating genes.