Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2017 Apr;18(2):371-385.
doi: 10.1007/s10162-016-0596-2. Epub 2016 Nov 9.

Longitudinal Changes in Audiometric Phenotypes of Age-Related Hearing Loss

Affiliations
Free PMC article

Longitudinal Changes in Audiometric Phenotypes of Age-Related Hearing Loss

Kenneth I Vaden Jr et al. J Assoc Res Otolaryngol. .
Free PMC article

Abstract

Presbyacusis, or age-related hearing loss, can be characterized in humans as metabolic and sensory phenotypes, based on patterns of audiometric thresholds that were established in animal models. The metabolic phenotype is thought to result from deterioration of the cochlear lateral wall and reduced endocochlear potential that decreases cochlear amplification and produces a mild, flat hearing loss at lower frequencies coupled with a gradually sloping hearing loss at higher frequencies. The sensory phenotype, resulting from environmental exposures such as excessive noise or ototoxic drugs, involves damage to sensory and non-sensory cells and loss of the cochlear amplifier, which produces a 50-70 dB threshold shift at higher frequencies. The mixed metabolic + sensory phenotype exhibits a mix of lower frequency, sloping hearing loss similar to the metabolic phenotype, and steep, higher frequency hearing loss similar to the sensory phenotype. The current study examined audiograms collected longitudinally from 343 adults 50-93 years old (n = 686 ears) to test the hypothesis that metabolic phenotypes increase with increasing age, in contrast with the sensory phenotype. A Quadratic Discriminant Analysis (QDA) was used to classify audiograms from each of these ears as (1) Older-Normal, (2) Metabolic, (3) Sensory, or (4) Metabolic + Sensory phenotypes. Although hearing loss increased systematically with increasing age, audiometric phenotypes remained stable for the majority of ears (61.5 %) over an average of 5.5 years. Most of the participants with stable phenotypes demonstrated matching phenotypes for the left and right ears. Audiograms were collected over an average period of 8.2 years for ears with changing audiometric phenotypes, and the majority of those ears transitioned to a Metabolic or Metabolic + Sensory phenotype. These results are consistent with the conclusion that the likelihood of metabolic presbyacusis increases with increasing age in middle to older adulthood.

Keywords: animal models; audiogram classification; longitudinal; metabolic presbyacusis; sensory presbyacusis; supervised machine learning classifiers.

Conflict of interest statement

The authors state that they have no conflict of interest to declare.

Figures

FIG. 1
FIG. 1
QDA classification probability (CP) quantifies the similarity of an audiogram (points) to the distribution of expert-labeled training data (shaded regions: Older-Normal (gray), Metabolic (green), Sensory (red), Metabolic + Sensory (blue)). Pure-tone thresholds and fitted curves from an example audiogram for each category are shown in relation to the distribution. Higher CPs are typically obtained for correct classifications than for incorrect ones.
FIG. 2
FIG. 2
Most individual ears (lines) were classified as the same phenotype with increasing age (four shaded regions; classification probabilities on the ordinate), although these classification probabilities varied with age (abscissa, points). While the phenotype categories were stable longitudinally for these ears, classification probabilities were observed to vary with increasing age.
FIG. 3
FIG. 3
Audiometric patterns consistent with each phenotype were preserved despite increasing thresholds. The simulated pure-tone threshold values were produced with a GLMM linear regression model that estimated age-related changes separately for each frequency and phenotype for females (left column), males (open circles, middle column), and their combined data (right column).
FIG. 4
FIG. 4
Mean threshold changes as a function of frequency (dB/year, SEM bars), estimated with separate GLMM regressions for frequency and phenotype. Asterisks denote significant hearing sensitivity declines with increasing age: *P < 0.05 and **P < 0.001 (Bonferroni corrected). Changes were estimated and plotted separately where sex significantly interacted with age effects, which only occurred for the Metabolic + Sensory ears at 4 and 8 kHz (Bonferroni corrected P < 0.05; open circles depict mean threshold change for males).
FIG. 5
FIG. 5
Example audiogram data from individual ears that changed audiometric phenotype with increasing age. Ears that were initially classified as Older-Normal phenotype and later transitioned to Metabolic, Sensory, or Metabolic + Sensory phenotypes (top row). Ears initially classified as Metabolic or Sensory phenotype and later transitioned to Metabolic + Sensory phenotype (bottom row).
FIG. 6
FIG. 6
Most of the individual ears that changed phenotypes with increasing age (four shaded regions; classification probabilities on the ordinate) transitioned to a Metabolic or Metabolic + Sensory phenotype (O-N Older-Normal, MET Metabolic, SENS Sensory, MET + SENS Metabolic + Sensory). Classification probabilities (abscissa, points) varied with age. Each subplot displays a particular phenotype transition (e.g., Older-Normal to Metabolic (top row)). The Older-Normal ears that appeared to transition directly to Metabolic + Sensory changed phenotypes during a relatively long interval between clustered visits (8.74 ± 1.17 years), compared to those that changed to Metabolic phenotypes (5.98 ± 0.88 years) or Sensory phenotypes (5.03 ± 0.51 years).
FIG. 7
FIG. 7
Mean threshold changes as a function of frequency (dB/year, SEM bars) for ears with changing phenotypes, estimated with separate GLMM regressions for frequency and phenotype (same procedure as shown in Fig. 4). Asterisks denote significant hearing sensitivity declines with increasing age: *P < 0.05 and **P < 0.001 (Bonferroni corrected). There were no significant interactions between participant sex and age-related threshold changes.
FIG. 8
FIG. 8
The likelihood of phenotype change or no change (left). Filled circles depict the phenotype classification at the initial visit (Older-Normal (gray), Metabolic (green), Sensory (red), Metabolic + Sensory (blue)), and same-colored lines depict phenotype changes by the final visit (e.g., 30 % of Sensory ears became Metabolic + Sensory). Closed loops show ears that were classified with the same phenotype at the initial and final visit. The transitional probabilities sum to 100 % for each phenotype, although the total for the rounded percentages was 101 % for the Older-Normal phenotype. A frequency matrix shows the total number of ears that were classified with each phenotype at their initial visit (rows) and final visit (columns) (right). Ears with the same phenotype at the initial and final visit (diagonal) included 422 ears with stable phenotypes as well as 24 ears with a different intermediate classification, which could reflect an ambiguous audiometric configuration or classification error.

Similar articles

See all similar articles

Cited by 7 articles

See all "Cited by" articles

LinkOut - more resources

Feedback