Reader Comments

Post a new comment on this article

Questioning the use of weighted GRS

Posted by agonzalez1 on 23 Sep 2011 at 11:09 GMT

weighted genetic risk score (GRS)
http://plosone.org/article/info:doi/10.1371/journal.pone.0024380#article1.front1.article-meta1.abstract1.sec2.p1

I understand that weighted genetic risk scores are needed to account for the different effect size of each risk allele when doing case-only analysis. However, its use in case-control comparisons is inappropriate because of double inclusion of the same type of comparison: the odds ratios used for weighting come from previous case-control studies. Therefore, the results are not coming from the current comparison between cases and controls, but from an exaggeration of current differences based on past studies if risks alleles are showing the same direction of effect and a decrease of the difference if they are of contrary direction. These changes will distort the genetic risk estimates.
Weighting by O.R. from past studies has the added distortion of the possible different distributions of each of the subtypes of RA patients between past and current studies and of different distributions of cases and patients in the variables used for adjustment.
It is very likely that weighting genetic risk scores for case-control comparisons in this report has had no effect in the overall conclusions, but it will have affected the size of the differences between patient subgroups.

No competing interests declared.