The complete sequencing of the human genome by the Human Genome Project was a remarkable accomplishment and a cause for celebration. Several companies including 23andMe, Navigenics, and deCODE have capitalized on that scientific achievement by offering genomic testing directly to the public. They promise more than they can deliver, and consumers don’t understand the limitations of the test results. The subject has been covered in several SBM articles.
One of the expected benefits of genomic testing is that if people knew they were at high risk of a disease, they would take preventive steps to reduce their risk. That seems plausible; but a recent study, a systematic review in the BMJ (formerly the British Medical Journal) calls that assumption into question. It found that communicating DNA-based disease risk estimates did not increase risk-reducing health behaviors or motivation to engage in such behaviors.
Futurist Ray Kurzweil predicted:
Genomics testing may soon be able to predict precisely what foods are best for us, prescribe individualized exercise and other lifestyle prescriptions, and recommend a personalized list of supplements, neutraceuticals [sic], and prescription drugs for maximum health and disease avoidance.
His prediction may be correct, but how soon is “soon?” We aren’t there yet, not by a long shot.
Current risk predictions are imprecise and are based on assumptions
We can’t be sure that the high-risk gene variants cause disease: all we really know is that a variant was statistically associated with a disease in the particular population tested, and correlation does not prove causation. It’s complicated. Many diseases are multifactorial. The expression of one gene may be affected by other genes and by environmental and lifestyle factors. Having a gene associated with a disease does not mean you will get that disease; the most it might do is increase the probability. And conversely, having variants associated with low risk is no guarantee that you won’t get the disease. And the different companies don’t even agree with each other; they assess risk using different combinations of “snips,” single nucleotide polymorphisms. To assess prostate cancer risk the different companies test for 5, 13, and 9 variants respectively, but no company tests for all 16 variants known to be correlated with increased risk.
Are we there yet?
For genomic testing in general and for assessing risks of various diseases, many experts have asked “are we there yet?” and have concluded that we are not. For example, the American Academy of Ophthalmology has advised physicians, outside of research studies, against genetic testing for eye disorders such as macular degeneration, “until treatment or surveillance strategies can be shown to be of benefit to individuals with specific, disease-associated genotypes.”
An article in the New England Journal of Medicine (NEJM) provides a good analysis of the pitfalls of genetic risk predictions; it is well worth reading. It says “We are still too early in the cycle of discovery for most tests that are based on newly discovered associations to provide stable estimates of genetic risk for many diseases.”
Another article in the NEJM just this month addressed genetic misdiagnoses of the risk of hypertrophic cardiomyopathy:
Multiple patients, all of whom were of African or unspecified ancestry, received positive reports, with variants misclassified as pathogenic on the basis of the understanding at the time of testing. Subsequently, all reported variants were recategorized as benign. The mutations that were most common in the general population were significantly more common among black Americans than among white Americans (P<0.001). Simulations showed that the inclusion of even small numbers of black Americans in control cohorts probably would have prevented these misclassifications.
These are just a few of the many concerns about the accuracy of genetic risk prediction.
The BMJ study
“The impact of communicating genetic risks of disease on risk-reducing health behaviour: systematic review with meta-analysis” by Hollands et al. was published in BMJ in March 2016. They found 18 randomized controlled studies that assessed behavioral changes after subjects were told they were at high risk of conditions where risk could be reduced by behavioral changes. Multiple studies showed no significant effects on smoking cessation (6 studies), diet (7 studies), or physical activity (6 studies). They also found no evidence of effects on other behaviors including alcohol use, medication use, sun protection, attendance at screening or behavioral support programs, or on motivation to change behavior. Specifically, when patients were told they were at high risk of lung cancer based on their genetic makeup, the knowledge did not induce them to stop smoking; patients told they were at high risk of melanoma did not use more sunscreen; patients told they were at high risk of diabetes, obesity, cardiovascular disease, or hypertension did not change their diet or physical activity; and patients told they were at high risk of alcoholism did not reduce their alcohol intake. Fortunately, they found no evidence of adverse effects of such knowledge, although one might expect an increased level of worry. They noted that many of the studies they found were of low quality with considerable risk of bias.
What does this mean?
This is only one systematic analysis of low quality studies, and it is certainly not the final word. It does, however, strongly suggest that if genetic risk predictions are effective in changing behaviors, the effects must be rather small.
This ties in with my recent article about prevention. It will be valuable to have more reliable estimates of risk. But it would be far more valuable to find effective ways of persuading all patients to take preventive action, not just the patients we think are at the highest risk.