For those who have been devastated by suicide, the news was encouraging: Researchers have discovered a simple blood test that can allegedly identify those who might consider taking their own life.
"With a test like ours, we may be able to stem suicide rates by identifying those people and intervening early enough to head off a catastrophe," says Zachary Kaminsky, an assistant professor at the Johns Hopkins University's School of Medicine and lead author of the study in a press release.
The researchers claim that a risk model based on the results of a blood test was able to predict with 80- to 97-percent accuracy which of the participants were having suicidal thoughts or had attempted suicide.
The test might best be used to predict future suicide attempts "in those who are ill," "in the military to test whether members have the gene mutation that makes them more vulnerable [so that] those at risk could be more closely monitored when they returned home after," "in a psychiatric emergency room" and "in all sorts of safety assessment decisions like the need for hospitalization and closeness or monitoring", says the press release.
The discovery was world news.
Regrettably, the scientific reality is far less promising. The news coverage and the press release are in sharp contrast with the scientific article, which was published last month in the American Journal of Psychiatry. The article repeatedly emphasizes that the study was very small and had notable shortcomings, that the results still need to be confirmed, and even that one of the main findings might be a statistical artifact.
When researchers claim to have discovered a new genetic test that predicts with high accuracy, they tend to make the same errors in the analysis and interpretation. Understanding a simple fact is enough to tackle most of the discoveries.
The fact is that accurate prediction based on a DNA test is only possible when the tested gene has a substantial role in the development of the disease. Well-known examples are the mutations in the BRCA1 or BRCA2 genes that markedly increase the risks of breast and ovarian cancer, and the mutations in the CTFR gene that cause cystic fibrosis when they are inherited from both parents.
When diseases are impacted by many risk factors that each have a minor to moderate impact on disease risks, testing a single risk factor can never accurately tell who will develop the disease. Smokers, for example, have a higher risk of cardiovascular disease, but many smokers will not get the disease, and many nonsmokers will. The impact of smoking is high enough to recommend smoking cessation, but smoking status on its own is not an accurate predictor of disease.
As there are numerous reasons that people commit suicide, it is extremely unlikely that, out of the blue, a single genetic predictor appears to be the leading cause.
This simple fact should be enough to be skeptical about the alleged discovery, but a simple check confirms that it's unlikely that the researchers have found a major suicide gene.
The scientific article reports that the SKA2 gene increases the risk of suicide by 1.15-fold.* In comparison, BRCA1 or BRCA2 mutations increase the risk of breast cancer by more than 4-fold: Female carriers of the BRCA mutations have a risk of 45 to 65 percent instead of 12 percent, which is the breast cancer risk in women who do not carry these mutations. A 1.15-fold increase would bring a risk of 12 percent to 13.8 percent but not higher. As the risk of suicide is much lower than 12 percent, around 13 per 100,000 people died from suicide in 2010, people who are genetically predisposed to risk variants of the SKA2 gene have only a minimally higher risk than others.
Accurate prediction requires that a risk factor substantially increases the risk of disease, and there are no exceptions to that rule of thumb. There are only instances in which application of the rule may be more complex, such as when many risk factors together increase the risk of disease. When that is the case, the rule of thumb is that individuals with the highest predicted risks, based on multiple factors, should have much higher risks than those with the lowest predicted risks.
How much higher the risk is is a value that is almost always given in scientific studies, because comparing the highest and lowest risks by definition gives the most favorable presentation of the value of the test. The study on suicide risk did not give this value, for obvious reasons: The two study populations in which the predictive ability was tested were too small to distinguish high- and low-risk groups.
A simple check of how much the risk is increased is often enough to debunk false claims about prediction accuracy of novel genetic tests. A few other questions may unmask the rest: How large was the study population? In which population was the study conducted? And has the study already been confirmed in an independent population? Prediction studies require hundreds of participants who are representative of the population in whom the test is going to be used, and their results need to be replicated in even larger populations before predictive accuracy can be claimed.
In the suicide study, the predictive ability was replicated in a population of 30 pregnant women with psychiatric diseases, and in a population of 22 children of patients with major depression. None of the claimed future uses of the test was investigated.
It is evident that this suicide blood test is entirely oversold in the media. A genetic test for suicide prevention is nowhere near, and the fact that two senior authors are co-inventors on a patent to evaluate risk of suicidal behavior using genetic and epigenetic variation at the SKA2 locus, as the disclosure paragraph of the scientific article says, does not change that.
This simple blood test for the prediction of suicide risks not only lacks a proper scientific basis but signifies unacceptable ignorance of the motives behind suicide thoughts and suicide attempts. Because of the complex nature of suicide, it is unlikely that a genetic test will ever be the key to prevention.
Need help? In the U.S., call 1-800-273-8255 for the National Suicide Prevention Lifeline.
*The number is calculated from Table 2 of the scientific article. The table reads that in the largest population (325 participants) the genetic effect was -0.14 for individuals who carry two copies of the C allele. Calculating e-0.14 translates this effect into an odds ratio of 0.89, which means a 0.89-fold decrease in risk. This in turn means that the reference group of individuals who carry two copies of the T allele had a 1/0.89=1.15-fold increased risk.