The short version
- Measuring hearing in decibels across 16,057 people found three gene regions tied to hearing loss, while a yes-or-no diagnosis across 61,469 people found none.
- Hearing loss looked about four times more heritable when it was measured with an audiogram than when it was read from a medical code.
- The genetic risk score that came out of it is weak, raising the odds of reported deafness by about 9%, and it was built only from people of European ancestry.
A hearing test that records how quiet a tone you can hear is a better tool for finding the genes behind hearing loss than a diagnosis in a medical record. That is the finding of a Vanderbilt University Medical Center study published October 8 in JAMA Otolaryngology–Head & Neck Surgery, and it comes from a neat piece of design: the team ran the same search twice in the same biobank and changed only how hearing loss was defined.
The design
Both searches were genome-wide association studies, which scan the whole genome for DNA variants that are more common in people with a trait. Both drew on BioVU, Vanderbilt’s DNA biobank linked to patient records.
In the first, hearing loss was a yes or no taken from diagnosis codes: 10,164 people coded with sensorineural hearing loss, the kind that arises in the inner ear or the nerve, against 51,305 people without the code.
In the second, the team used the audiograms of 16,057 people. For each person they took the pure-tone average, the mean of the quietest tones heard at 500, 1,000 and 2,000 Hz in the better ear, and treated it as a number on a continuous scale instead of a label.
What they found
The diagnosis-code search found nothing that cleared the statistical bar for a genome-wide result. The audiogram search, with about a quarter as many people, found three regions of the genome, mapped to four genes: EML6, SPTBN1, ARHGEF28 and EYA4.
EYA4 is a known quantity. It is involved in the organ of Corti, the strip of the cochlea where hair cells sit, and it has already been tied to inherited hearing loss. The other three work in the nervous system: the paper describes roles in the internal scaffolding of nerve and sensory cells, in brain development, and in how nerve fibers branch and form synapses.
The share of hearing variation that common DNA variants could explain was 11.78% with audiograms and 2.90% with codes, roughly a fourfold difference from the same biobank.
The team then tested both versions on 109,096 participants in the National Institutes of Health’s All of Us program. A genetic risk score built from the audiogram results tracked with people who reported being deaf (odds ratio 1.09, 95% confidence interval 1.07 to 1.12). The score built from diagnosis codes did not (odds ratio 1.00).
The limits
An odds ratio of 1.09 is small. The authors themselves place it in the weak range. Nobody could be told anything useful about their own hearing from this score.
Every genetic analysis was restricted to people of European genetic ancestry, so the results may not carry to anyone else.
The measure uses only three test frequencies, so it does not capture high-frequency hearing difficulty, and the study did not test how well people follow speech in noise. The All of Us check rested on a single self-reported question about hearing. And the study compared people at one point in time; it did not follow people with good hearing to see who lost it.
Why it matters
“Long term, we want to be able to identify our patients with hearing loss before it ever emerges,” lead author Andie DeFreese, an audiologist and doctoral candidate at Vanderbilt, said in comments reported by Hearing Health & Technology Matters.
That is a distant goal. The nearer lesson is about measurement. A diagnosis code puts mild and severe loss under one label. A threshold in decibels keeps the difference, and in this study that detail outperformed a sample nearly four times as large.
What to watch next: whether the three new genes hold up in biobanks that include other ancestries, and whether adding higher frequencies finds more.
