Appearing today in the New England Journal of Medicine is a stealthily encouraging study for the use of genetic testing to improve the assessment of the risk of the common forms of breast cancer. Stealthily, I say, because the authors seem oddly determined to provide a gloomy interpretation of their own data. The study, entitled ‘Performance of Common genetic Variants in Breast-Cancer (sic) Risk Models,’ by Wacholder et al, uses data from several major breast cancer studies to answer an interesting question: does adding the measurement of common SNPs linked to risk of breast cancer add to the risk assessment provided by the traditional ‘Gail score’ criteria – age, family history, age at menarche, age at first live birth and the number of previous breast biopsies?
The answer is clearly yes, though the authors of the paper seem not to want you to know that. Most importantly, the authors define as elevated risk those women between the ages of 50 and 79 who are at a greater than 0.575% chance of developing breast cancer in any given year. Using the Gail criteria alone, 18.9% of study participants were considered to be at elevated risk. But with the addition of the genetic risk factors – which are ten of the twelve risk factors tested for by deCODE Breast Cancer test – another 9% of participants could be identified as being in the higher risk category. A 50% improvement.
Similarly, using an Area Under the Curve calculation (customarily used to evaluate the accuracy of methods for diagnosing disease) the Gail model yielded an AUC of 58%, and the Gail-plus-genetics model yeilded an AUC of 61.8%. In an AUC model, the amount over 50% (the baseline of a test that is no better than random) is the measure of relative discriminatory power. So an increase from 8 to 11.8 is, yes, a small number, but also an improvement of something in the neighborhood of 45%. The study also shows that compared to each other, the set of genetic risk factors were more accurate predictors of breast cancer than were the Gail factors that are the current mainstay of risk assessment.
So I can see why the authors wouldn’t want to celebrating these results too loudly – because we need to do better. But what this study shows is that genetics is already taking us in the right direction, and that the addition of genetic risk to current clinical practice can – right now, today – provide a substantial improvement in the crucial task: to better risk stratify the population, focus screening on those who should have it, pick up more cancers earlier and save lives. I can’t see anythig but good news in that. Our task is to keep discovering new risk factors that will continue to increase the power of these tests, and we are committed to doing so.
Dr. Kari Stefansson