These GEITP pages have often discussed genome-wide association studies (GWAS), in which a phenotype (i.e. trait — such as height or cancer) is selected by a research team, and then the genomes of hundreds or many thousands of subjects are searched — to see if any DNA loci can be identified as statistically significantly associated with that trait (rather than a chance occurrence, i.e. random event leading to a false positive association). Note above the word, “searched,” because this is what GWAS generally are for: it is a fishing expedition, hoping to find some unexpected gene (especially when associated with a complex disease such as, e.g. type-2 diabetes, cancer or dementia).
Still, much clarification of studies involved in genetic risk assessment is needed, if one wishes one day to implement personalized medicine. Genetic risk predictions are largely made with the intention of “positive prediction,” albeit of a disease state [i.e. the goal is to ascertain which individuals (among any ‘at-risk’ population) have the highest likelihood of developing a condition, or of progressing to a more severe state]. Thus, genome-wide polygenic risk scores (PRSs) have been generated — for coronary artery disease, atrial fibrillation, Crohn disease, type-2 diabetes, and breast cancer — in each case identifying a threshold above which a small percentage (i.e. a subset) of the population has a “disease risk” of at least 3-fold higher than that of the general population.
Because single-gene mutations with such a magnitude-of-effect are sometimes regarded as clinically actionable — yet affect a much smaller proportion of people — it has been argued that PRSs are now at the point at which it might be appropriate to integrate them into clinical care. At the very least, this might lead to encouraging high-risk individuals to meet with an appropriate medical specialist or (personally) to initiate behavioral change. More commonly, however, PRSs might encourage a course of preventive medication. As the costs of healthcare continue to increase, the impact on both patient and the healthcare system comes into focus. The percentage of incidents (prevented by pre-emptive treatment) will reflect a function of the proportion of the population who are treated, and the rate of favorable response to treatment — which itself may vary, possibly as a function of disease risk.
It has previously been argued that, because negative prediction is almost always more accurate than positive prediction — owing to the low ratio of cases to controls — the potential for using PRSs to identify low-risk individuals should be given more attention than previously has been done to date. This is because, if only the highest-risk individuals are treated, then most cases are not prevented, yet treating everyone is both prohibitively expensive and (given the possibility of adverse drug reactions) potentially harmful. Both relative and absolute risk can be used to assess efficacy of medications: relative risk focuses on reducing the rate of incidents (e.g. from 5% to 4%); absolute risk focuses on reducing the number needed to treat (NNT; e.g. from 50 to 20 incidents prevented for each person taking the drug). Both of these numbers ought to be considered, when approaching the question of how to utilize PRSs in a manner that effectively focuses medical attention on the largest population with a high likelihood of effective response to therapy.
Four variables are critical in making this assessment: the prevalence of the condition, risk in each PRS-positive target group, proportion of the population in the group, and therapeutic response rate. The author [see attached article] reviews five conditions — across a broad spectrum of chronic disease (opioid pain medication, hypertension, type-2 diabetes, major depressive disorder, and osteoporotic bone fracture), considering — in each case — how genetic prediction might be used to target drug prescription. This concept leads to a call for more research, designed to evaluate genetic likelihood of response to therapy and a call for evaluation of PRS, not just in terms of sensitivity and specificity — but also with respect to potential clinical efficacy.
DwN
PLoS Genet Apr 2019; 15: e1008060