Presbycusis phenotypes form a heterogeneous continuum when ordered by degree and configuration of hearing loss

Abstract

Many reports have documented age-by-frequency increases in average auditory thresholds in various human populations. Despite this, the prevalence of different patterns of hearing loss in presbycusis remains uncertain. We examined 'presbycusis phenotypes' in a database of 960 subjects (552 female, 408 male, 18-92 years) that each had 30 measures of peripheral hearing sensitivity: pure tone audiograms for left and right ears from 0.25 to 8 kHz and DPOAE for each ear with F(mean)=1-6.4 kHz. Surprisingly, the hearing phenotypes did not naturally separate into discrete classes of presbycusis. Principal component (PC) analysis revealed that two principal components account for 74% of the variance among the 30 measures of hearing. The two components represent the overall degree (PC1) and configuration of loss (Flat vs. Sloping; PC2) and the phenotypes form a continuum when plotted against them. A heuristic partitioning of this continuum produced classes of presbycusis that vary in their degree of Sloping or Flat hearing loss, suggesting that the previously reported sub-types of presbycusis arise from the categorical segregation of a continuous and heterogeneous distribution. Further, most phenotypes lie intermediate to the extremes of either Flat or Sloping loss, indicating that if audiometric configuration does predict presbycusis etiology, then a mixed origin is the most prevalent.

Figure 1

Mean Right Ear Pure Tone Thresholds (top) and DPOAE amplitudes (bottom) for Males (Left) and Females (Right), pooled by decade bin (+/− 5 years around nominal age listed in legend). Both sexes show the classic progression of increasing high-frequency hearing loss and loss of DPOAE amplitude with age, males tending to have greater high-frequency loss, as well as a modest notch loss near 4 kHz.

Figure 2

Average Auditory Age (AAA) vs. calendar age, for Males (crosses) and Females (filled circles). The dashed lines represent a conservative estimate of the precision limit of the audiogram, measured in 5 dB steps. The upper line is the AAA for the age-median HL (Robinson & Sutton, 1979) with 5dB added to each PTT, while the lower line was calculated from subtracting 5 dB for each PTT from the age-median HL, with HL scores less than zero being set to zero.

Figure 3

Weighting functions of the first two principal components for Right ear PTT (upper) and DPOAE (lower), plotted as a function of stimulus frequency. The first component weights all PTT positively and all DPOAE negatively, with a bias for higher frequencies: it represents a measure of overall hearing loss. The second component positively weights low frequencies and negatively weights high frequencies: it represents differential high-frequency vs. low-frequency hearing loss.

Figure 4

Representation of the PTT and DPOAE data for each of the 960 subjects in the two-dimensional space of the first and second principal components. While they are continuously distributed in this space, the data are categorized and labeled according to the results of a K-means cluster analysis.

Figure 5

Right ear PTT and DPOAE for the centroids of the eight clusters determined using K-means analysis. The clusters represent No Hearing Loss (1), Mild Hearing Loss (2), Steeply Sloping Loss of increasing degree (3, 5, 7), and Flat to Gently Sloping Loss of increasing degree (4, 6, 8).

Figure 6

Distributions of Overall and Differential Hearing Loss. Histograms of number of ears based on A) Pure tone average of 0.5, 1, 2, and 4 kHz; B) Difference between the averages of 4 & 8 kHz and 0.5 & 1 kHz; C) PC1 weighted 0.5 to 8kHz pure tone average; D) PC2 weighted 0.5 to 8kHz pure tone average. All four distributions appear bimodal and are well-fit by two Gaussian distributions. The gray shaded areas show data for subjects older than 65 years.