NEW YORK – Genetic analyses of behavioral traits like drinking alcohol could be swayed by people who misreport their behavior, for example how much they drink, a new study has found.
Genetic studies of behavioral traits typically rely on questionnaire data, but due to lapses in memory or in an effort to present themselves in a more positive light, people often underreport how much they drink or smoke and overreport how much they exercise. In addition, how much someone drinks, smokes, or exercises can change over time.
Using GWAS summary statistics, a team of researchers from Australia and China has found that this can affect the results of genome-wide association studies and follow-up analyses, and possibly account for discrepancies between studies examining the effect of drinking on common diseases.
As they reported in Nature Communications on Tuesday, the researchers found that among UK Biobank participants, people with a higher disease burden were misreporting their alcohol consumption. After correcting for misreports or longitudinal changes in drinking behavior, the researchers found that most of the previously reported negative estimates of genetic correlation between alcohol use and common diseases became either positive or non-significant.
Senior author Jian Yang from Westlake University in China said in an email that such findings "thus need to be interpreted with great caution."
Within the UK Biobank, Yang and his colleagues first uncovered signs of misreports of or longitudinal changes in alcohol consumption. For instance, nearly 15,000 biobank participants said they were never drinkers, but follow-up studies or medical record data for about 3,600 of them indicated otherwise, such as through a diagnosis of alcoholic hepatitis or alcohol use disorder. Similarly, other participants who said they did drink reported a consumption level of zero, suggesting they underreported how much they drank.
Additionally, more than 60 percent of UK Biobank participants who said they did drink noted that their current consumption levels were different than they were 10 years ago, pointing to possible changes in drinking behavior over time.
Within the UK Biobank, individuals who said they drink but didn't say how much, as well as those suspected of underreporting the amount they drink, had a higher mean disease count of 18 common diseases than other drinkers, the researchers reported.
Further, 8.3 percent of current drinkers who said they drank more a decade ago indicated that they curbed their alcohol intake due to illness or their doctor's advice. That group had nearly twice the mean disease count of other drinkers and, in particular, 2.7 times the prevalence of cardiovascular disease.
Based on this, the researchers developed a procedure for partially mitigating the effect of misreports or longitudinal change in drinking behavior. It excludes people who might underreport alcohol consumption and those who reduced drinking due to doctors' advice.
In genome-wide association studies for alcohol consumption and disease performed with and without that correction, the researchers identified 16 loci that became non-significant following the correction, many of which were associated with metabolic or cardiovascular traits. They further showed that follow-up analyses based on summary statistics as well as Mendelian randomization studies would also be influenced by misreports or longitudinal changes in drinking behavior.
Expanding beyond alcohol consumption, the researchers noted similar effects of misreports or behavior changes over time in other self-reported behavioral traits such as smoking and physical activity.
They cautioned that scientists should be careful when studying and interpreting behavioral traits from biobank datasets that relied on self-reporting.