Reader Comments

Post a new comment on this article

scn9a and minor correction

Posted by alistair1 on 07 Jan 2013 at 16:00 GMT

Interesting paper and study design using the extremes of normal distribution. Would have been interesting to know i) which of the 6 statistical tools the authors recommend best and also whether SCN9A came up at all (see PMID: 20212137 and 17167479).
Also I think there is a minor error as the main text and fig2 are inconsistent; according to figure, the single pain sensitive subject with A95T variant in GZMM was homozygous so should count as 2 alleles for the reported Fisher's exact test.

No competing interests declared.

RE: scn9a and minor correction

HydeC01 replied to alistair1 on 17 Jan 2013 at 19:24 GMT

This analysis was done when gene-centric collapsing methods for analysis were still in early development, and indeed if we had felt there was a clear, single best method at the time, we’d have simply reported on that rather than combining six tests. Other methods published more recently may perform better. As for SCN9A, it was not at all covered by the capture array for TUK1, and not well captured even in TUK2, so not surprisingly it did not come up, but this was really mainly due to poor coverage.

Regarding Fisher’s exact test on GZMM, we regret that the discussion of allelic counts in the text was misleading: the p-values for Fisher’s exact test on GZMM were calculated based on the number of carriers (so, by subject), not by allelic counts, since there were so few homozygotes. Hence, the p=0.005 in TUK1 and p=0.0016 in TUK2 are correct for the carrier, non-carrier analysis. The commenter is correct in that the text incorrectly reports 1 rare allele instead of two for the heat sensitive subjects in TUK1, but this is of no consequence to the carrier/non-carrier analysis.

Competing interests declared: author, Pfizer employee