Investigation found they gave wildly different results and arrived at predictions that were no better than flipping a coin
Personalised health tests that screen thousands of genes for versions that influence disease are inaccurate and offer little, if any, benefit to consumers, scientists claimed on Monday.
An investigation into the services found they gave wildly different results, and in some cases arrived at medical predictions that were no better than flipping a coin.
The findings of the Dutch study will bolster calls for tighter regulations around personalised genetics tests which can cost more than £500. Critics claim the tests are a waste of money that could mislead people about their future health.
According to the group at Erasmus University medical centre in Rotterdam, tests from rival companies predicted conflicting risks for some diseases, often because they disagreed on how common the conditions were in the general population.
Another flaw was that the tests looked only at genetic factors, whereas many diseases are governed more by lifestyle and other environmental factors.
In the study, researchers used a computer to simulate genetic information for 100,000 typical people. They then used formulas from two of the largest genetic testing companies, deCODEme and 23andMe, to predict the risk of eight medical conditions, including heart attack, prostate cancer, coeliac disease, an eye disease known as age-related macular degeneration and diabetes.
"When we looked at individual risk of disease, we saw enormous differences between the two companies," said Cecile Janssens, an epidemiologist who led the research.
"A major problem with these tests is that they don't take other factors into account, such as your age, diet, exercise and whether you smoke, which in most cases have a greater impact on disease risk.
"For complex diseases, if you ignore these non-genetic factors, you are looking at only a small part of the picture and missing the main story," she said.
While the companies used similar techniques to predict an individual's risk of disease, the test results differed in some cases by more than 50 percentage points.
A number of tests based on deCODEme risk models appeared to suggest the odds of developing five disorders, including age-related macular degeneration, were greater than 100%, a result that in itself raised concerns over the validity of the tests, the researchers will tell the European Human Genetics Conference in Amsterdam .
But Kari Stefánsson, the chief executive of deCODEme, said the Dutch team had misinterpreted the computer models their findings were based. "We never report a lifetime risk over 90%. This is not how we use these models."
Brian Naughton, founding scientist at 23andMe said: "We are aware that most people will be at low risk for most diseases – this is what we expect, and is not an indication of inaccuracy.
"23andMe takes extensive measures to provide both accuracy and to highlight the context for relevance at all steps in the process of analyzing and presenting data. In addition, research undertaken by 23andMe seeks to expand relevant findings.
"Lastly, 23andMe cautions that conclusions regarding 23andMe's accuracy and relevance should not be based on review of deCode or others."