We believe that there are limitations in the study by Hodgkinson, Millward, and Demaine (1) about the association of the NAD(P)H oxidase p22phox polymorphisms with susceptibility to diabetic nephropathy. The case-control association studies on the polymorphisms in p22phox have been performed by a number of researchers, and their results are still in conflict, especially with regard to coronary artery disease. On the other hand, the allele frequencies of the C242T polymorphism have been found to be consistent in various Caucasian populations (2,3). The C242T allele frequencies in the control samples by Hodgkinson, Millward, and Demaine differ from those in previous reports, and the genotype frequencies also deviated significantly from Hardy-Weinberg equilibrium (P = 0.025), suggesting a nonrandom sampling or some technical error.
The estimation of haplotype frequencies in case-control association studies is often performed on the basis of a maximum-likelihood method with an expectation-maximization algorithm. It is usually impossible to determine whether an individual with the genotype Aa-Bb has haplotypes A-B and a-b or A-b and a-B. Moreover, the χ2 test by contingency table is not appropriate for comparing the estimated haplotype frequencies. Hodgkinson, Millward, and Demaine did not describe how the haplotypes of their samples were determined. Tsai et al. (4) applied a proper method, referred to as permutation-based hypothesis testing, to evaluate the association between multilocus angiotensinogen gene polymorphisms and hypertension.
Although Hodgkinson, Millward, and Demaine reported a very low and striking P value, we believe the association of diabetic nephropathy with p22phox remains unsolved.
