To the Editor:

Dr. Stack correctly points out two limitations of genome-wide scans used for predicting common diseases: the clinical context is lacking, and there is limited utility in comparing individual risks with the average risk. To evaluate the utility of tests, defining the clinical context is crucial. A genetic test usually is intended for a specific population and useful only if it changes health decisions, for example, if test results lead to different recommendations or different interventions. In some instances, testing can be beneficial also in the absence of interventions because people may value the information gained from learning about their health risks. This benefit is proven for monogenic diseases, such as Huntington disease, but unclear for complex diseases.1

The question arises what is the clinical context in predictive genetic testing for type 2 diabetes, which we used as an example in our analysis?2 Currently, there are no guidelines on risk thresholds for type 2 diabetes prevention,3 similar to the thresholds of the Framingham risk score for cardiovascular disease.4 Furthermore, the only available preventive strategy is adoption of a healthy lifestyle, which is recommended to all and should not need a genetic test to justify it. However, many companies promote that genetic tests will motivate preventive behavior. They argue that motivation increases when people learn that they are at higher risk than average. Whether their tests can provide this benefit, without encouraging careless behavior among those at lower than average risk, remains to be proven.

We agree with Dr. Stack that it is doubtful to consider 20.8% as increased risk when the average is 20%. Although many companies initially did present the results in this way, several have changed their layout and now consider an additional average category. This evidently reduces the probability that people directly change from increased to decreased risk categories or vice versa. However, most likely, more people will move between risk categories, because the percentage of reclassification increases with the number of cutoff values.5 Individuals may move from the average risk category to increased or decreased risks and vice versa. An interesting question then would be to investigate whether individuals prefer to be “below average risk” or “not above average,” in other words, whether individuals put more value on one cut-off value than on the other.

Finally, Dr. Stack puts forward the question which average risk to consider. Although this is an important question when one is interested in the absolute risk of type 2 diabetes, it is not an issue for the mere fact of being higher or lower than average risk. All companies calculate an individual's risk starting from some average risk, which is then multiplied with the odds ratios of the genotypes of the variants carried by the individual. The companies then compare the individual's risk with the average risk they used as a starting point for the calculation. This deviation from average, i.e., whether someone has a higher or lower risk than average, is determined by the cumulative effect of the multiple variants, and this is the same whichever average is taken. Thus, whether a general average risk is taken, as we did in our study, or whether an age- and sex-specific risk is taken, the results with respect to the percentage of reclassification, our main measure, remain the same. The absolute risk of disease for an individual may be entirely wrong, as Dr. Stack points out, when other important risk factors are not included in the calculation as well. In our case, if body mass index is not taken into account, even age- and sex-specific average risks are incorrect. Given these observations and the fact that currently there are no preventive or therapeutic benefits associated with the results of these genome-wide DNA scans, these tests should only be bought to learn how one's DNA sequence compares with others but not for medical reasons.