An extensive evaluation of the scientific basis for commercial genetic tests, published in the Americal Journal of Human Genetics, suggests that evidence of clinical validity is often minimal and sometimes entirely absent [Janssens ACJW et al. (2008), Am. J. Med. Genet. 82: 593-599
]. This article is extremely timely, following the recent media interest in over-the-counter genetic tests sold directly to the consumer (see previous news article
) and the two recent reports on diagnostic testing from the PHG Foundation and Sense about Science (see previous news article
The authors looked specifically at seven companies – Genelex, Genovations, Integrative Genomics, Salugen, Sciona and Suracell – that all offer predictive genetic testing using multiple susceptibility markers, for conditions such as asthma, various cancers, psoriasis and Alzheimer’s disease Many of these tests are available directly to the consumer over the internet. They found that, of the 56 genes tested amongst the seven companies, just under half (43%) were not reviewed in meta-analysis. Furthermore, of those single nucleotide polymorphisms (SNPs) that had undergone meta-analysis, only 38% were found to have a statistically significant, albeit generally modest, association with the disease in question. These results suggest that only a third of tests have a well proven association between the SNP tested and the disease.
However, whilst a real association between a particular gene or SNP and a disease is necessary for a test to be clinically valid, it is by no means sufficient. The authors of this study comment that, “because the predictive value of genetic testing depends upon disease risk, genotype frequencies and odds ratios for the association… the profiles should be evaluated in the target population.” Although not addressed in this article, preliminary research at the PHG Foundation suggests that evidence of the second facet of clinical validity – assessment of the clinical performance of a test in terms of sensitivity, specificity and positive predictive value – is absent from the majority of predictive genetic tests sold directly to the consumer.