3. Results

3.1. Study Population and Audiometric Evaluation

The final study population included 73 participants (Table 1), of whom 41 had COL1A1 variants (female/male = 20/21; OI Type III/IV = 22/19), and 32 had COL1A2 variants (female/male = 21/11; OI Type III/IV = 12/20). Audiological evaluations were conducted on individuals aged 1–45 years (mean age [years ± SD]: OI Type III 13.5 ± 7.6; OI Type IV 15.7 ± 9.5; COL1A1 15.1 ± 8.8; COL1A2 14.1 ± 8.6). Participants underwent between 1 (n = 3) and 20 (n = 1) hearing tests, with means of 7.8 ± 4.7 and 7.2 ± 4.5 tests per individuals for participants with COL1A1 and COL1A2 variants, respectively.

3.2. Audiometric Findings by OI Type

3.2.1. Incidence of HL

A total of 53% (n = 39) of the 73 total participants with OI Type III or IV exhibited HL (Table 2). For OI Type III, 71% (n = 24/34) of individuals had HL, compared with 38% (n = 15/39) in OI Type IV. Among the individuals with HL, the proportions of unilateral and bilateral HL were comparable between OI Type III (n = 6/24, 25% and n = 18/24, 75%, respectively) and OI Type IV (n = 6/15, 40% and n = 9/15, 60%, respectively) (Figure S1).

3.2.2. Degree of HL and Progression

Most affected participants had mild HL (75% OI Type III and 93% OI Type IV). Six participants with OI Type III experienced greater HL, ranging from moderate (n = 2, 8%) to severe (n = 3, 13%) to profound HL (n = 1, 4%). For OI Type IV, only one participant had HL that progressed, developing severe loss in one ear (Figure 1).

Figure 1: Degree of hearing loss and progression by OI type. All ears (individual lines) with hearing loss (horizontal dashed line represents degree of hearing loss designation; normal hearing = gray circle) and type (CHL = blue triangle, MHL = yellow inverted triangle, and SNHL = red square) indicated as a function of age (years). Each row represents LF‐, 4F‐, and HF‐PTA with columns noting stable or progressive hearing loss, by OI Type III or IV. Bone conduction not measured indicated by “x,” and unfilled circles represent cases unable to be categorized in terms of hearing loss type. Refer to Figure S1 for representation of all ears in the study (including normal hearing participants).

3.2.3. Type of HL and Progression

Among individuals with HL, 25% (n = 6/24) of those with OI Type III experienced a progressive decline, compared with 7% (n = 1/15) of those with OI Type IV. In OI Type III, progressive HL either remained conductive (33%) or transitioned from conductive to MHL (67%). The single progressive HL in OI Type IV remained conductive over a period of 37 years. Notably, the individual with OI Type IV and one individual with OI Type III demonstrated bilateral HL that was stable in one ear and progressive in the other. For all other individuals with stable HL, 50% of those with OI Type III and 39% of those with OI Type IV experienced mild CHL. A smaller subset, 22% in OI Type III and 23% in OI Type IV, transitioned from mild CHL to mild SNHL. Six percent of individuals with OI Type III and 8% of those with OI Type IV had SNHL with no history of CHL or MHL. Among those with stable HL, the type of loss could not be classified in 22% of OI Type III and 31% of OI Type IV (Figure 1).

3.2.4. Age of Onset of HL

For both OI Types III and IV, the age of HL onset was earlier than previously reported in most studies, with the majority occurring in the first decade of life (n = 18/24 [75%] OI Type III; n = 10/15 [67%] OI Type IV) (Figure 2A). First decade onset occurred among individuals with both stable and progressive HL. Among individuals with stable HL, 78% (n = 14/18) of those with OI Type III and 87% (n = 13/15) of those with OI Type IV experienced onset within the first decade. Among individuals with progressive HL, 67% (n = 4/6) of individuals with OI Type III and the one individual with OI Type IV had onset within the first decade. In the first 5 years of the first decade, HL onset was 46% (n = 11/24) in OI Type III and 33% (n = 5/15) in OI Type IV. In later decades, HL onset occurred in 21% (n = 5/24) and 13% (n = 2/15) during the second decade, 0% and 7% (n = 1/15) during the third decade, and 4% (n = 1/24) and 13% (n = 2/15) in the fourth and fifth decades for individuals with OI Types III and IV, respectively. This may be limited by an examination of fewer patients past their third decade of life (Table S3).

Figure 2: Hearing loss onset and progression. Stable (filled light blue bars) and progressive (filled dark purple bars) hearing loss represented by decade for age group in years. (A) Hearing loss onset in OI Types III and IV occurs most commonly in first decade of life (n = 18 OI Type III; n = 10 OI Type IV). (B) Hearing loss onset in COL1A1 and COL1A2 occurs most commonly in the first 5 years of age (n = 10; n = 16 first decade) in COL1A1 and in the first decade (n = 12; n = 6 first 5 years) for COL1A2. (C) COL1A1 variants (in green) showed increased hearing threshold indicating greater hearing loss and steeper progression over time in OI Type III (left panel) versus OI Type IV (right panel) and COL1A2 variants (in orange), and OI Type III showed a slight decrease in hearing threshold over time (visit). Visit slope: COL1A1, OI Type III, β = 0.038, p = <0.0001; COL1A2, OI Type III, β = −0.023, p = 0.0023; COL1A1, OI Type IV, β = 0.013, p = 0.0014; COL1A2, OI Type IV, β = 0.0048, p = 0.14.

3.3. HL and Genotype–Phenotype Correlation

3.3.1. HL and Correlation With Alpha Chain Position of Missense Variant

More than half of participants with COL1A1 (n = 21/41; 51%) or COL1A2 (n = 18/32; 56%) variants had HL (Table 2). Except for a few individuals, the participants had missense variants, mostly glycine substitutions. Several participants had exon splicing defects or a large deletion. There was no correlation of splicing defect versus glycine substitution with HL. Our population did not include individuals with COL1A1 haploinsufficiency defects causing Type I OI.

Figure 3 shows the collagen variant position of individual participants with and without HL along the pro_α_1(I) and pro_α_2(I) chains. HL is associated with heterozygous collagen protein sequence variants along the full length of the pro_α_ chains, including the C‐propeptides. In the helical region of the _α_1(I) chain, there are two substantial gaps on the HL map, from Gly352 to Gly589 and from Gly613 to Gly832, representing approximately one‐half the length of the chain. Some participants with missense variants in these regions have normal hearing. In the _α_2(I) chain map, there is only one individual with HL in the region from Gly370 to Gly703. In addition, for individuals with OI Type IV, we note a lower incidence of HL in the third quartile of the Type I collagen helix with either COL1A1 or COL1A2 variants (Figure S2). Although the approximate overlap of these midchain gaps is intriguing, they should be considered preliminary pending additional data from other studies, since our maps represent a modest number of patients and about half of our patients were not followed into their third or fourth decade.

Figure 3: Hearing loss map. Normal hearing compared with hearing loss (by degree on y‐axis) mapped onto Type I collagen amino acid chain for OI Type III (blue hashed fill) and OI Type IV (red solid fill) participants with COL1A1 (top panel) and COL1A2 (bottom panel) variants. Each filled bar represents a single amino acid substitution. Bars above the baseline indicate hearing loss by degree, and bars below the baseline indicate normal hearing. Each bar represents an individual patient unless there is a number next to the bar indicating multiple patients with the same variant and audiology finding. Splicing mutations or deletions are indicated with brackets spanning the amino acids affected by the exon splice. Below the HL map, there are two additional sets of brackets. First, in the regions indicated by dashed lines, OI variants are associated with normal hearing. In our OI cohort, no or few patients with variants in this region have HL. Second, brackets composed of solid lines indicate the helical and C‐terminal nonhelical portions of the collagen alpha chains. Quartiles along the helical region are indicated by vertical lines.

Our population also gave us the opportunity to examine intrafamilial and interfamilial variability of this secondary feature of OI. These comparisons are limited to some extent by the broad age range of the participants and familial age gaps. Interfamilial variability was found at the COL1A1 Gly589 residue. Four unrelated patients, two each with Types IV and III OI, had COL1A1 Gly589Ser substitutions; mild HL was detected in one Types III and one Types IV, with onset at the age of 5 years in both cases, whereas the other Types III and IV participants had normal hearing when examined at the age of 3 and 15 years, respectively.

In COL1A2, there were three glycine residues in the amino‐quarter of the chain at which more than one participant had missense variants, involving both related and unrelated individuals. In a mother and her two daughters with Type IV OI caused by a Gly106Ser substitution, the mother had mild HL with onset at the age of 32, whereas the daughters had normal hearing at the ages of 16 and 6 years, leaving open the possibility of fourth decade onset in the children. A Gly238Ser substitution occurred in five individuals, two genetic half‐siblings with Type IV OI as well as two Type III and one Type IV OI unrelated participants. One half‐sibling had mild HL, as did one of the Type III OI participants, with onset at the ages of 10 and 3 years, respectively, whereas the second half‐sibling and unrelated Types III and IV participants had normal hearing when examined at the ages of 32, 17, and 15 years, respectively. Finally, Gly247Ser missense was found in three participants, a father and son with Type IV OI and an unrelated male with severe Type III OI. Both the Type III OI individual and the Type IV OI father had mild HL with onset at the ages of 17 and 11 years, respectively, whereas the Type IV OI son had normal hearing at age 14.

Thus, of the eight participants who share a collagen variant with at least one other participant and have normal hearing at last exam, four have normal hearing in the decade older than the corresponding participant with HL (normal hearing in second and fourth decades of life vs. first decade HL), whereas four participants with normal hearing are not yet old enough to be confident that HL will not develop at the older age of onset in their corresponding HL participant.

3.3.2. Degree of HL, Type of HL, and Progression

For participants with variants in COL1A1 and HL, the majority were mild and stable 67% (n = 14/21) but HL progressed in 33% to moderate (n = 2/7), severe (n = 4/7), or profound (n = 1/7). In the group with progression, all HL began as CHL but developed into MHL in five individuals (Figure 1). In the subset with stable mild HL, the majority of participants (n = 10/14) had CHL, with 4/10 individuals exhibiting SNHL at discreet visits. SNHL with no history of CHL was seen in 2/14 individuals, and HL type could not be determined in 2/14 (Table 2; Figure S3).

For participants with variants in COL1A2 and HL, the degree of HL remained mild for all participants except one (n = 17/18), who had stable moderate HL in one ear. HL was conductive for the majority of participants (n = 13/18), with 4/13 individuals showing SNHL at discreet visits. One individual had SNHL with no history of CHL, and HL type could not be determined in 4/18 (Figure S3).

3.3.3. Onset of HL

For both COL1A1 and COL1A2 variants, onset of HL was most common in the first decade (n = 16/21, 76%; n = 12/18, 67%, respectively). Further, a substantial proportion of HL onset occurred by 5 years of age, with early loss in 48% (n = 10/21) for COL1A1 and 33% (n = 6/18) for COL1A2 variants (Figure 2B).

3.4. HL Progression Is Greatest for Type III OI Participants With COL1A1 Variants

PTA was modeled as a function of OI type, mutated gene, and visit to explore interactions between these factors. For 4F‐PTA, this revealed a significant interaction between visit and Type III OI (β = −0.028, 95% CI: [–0.044, –0.012], p = 0.0007), as well as a three‐way interaction between visit, genotype (COL1A1), and Type III OI (β = 0.053, 95% CI: [0.034, 0.072], p = 5.2_E_ − 08), indicating that the rate of HL progression over time varied significantly across genotype‐OI type groups (Table S4).

Post hoc contrasts revealed that individuals with COL1A1 variants exhibited significantly greater HL (4F‐PTA) than those with COL1A2, both in OI Type III (β = 0.35, 95% CI: [0.18, 0.52], p = 0.0001) and Type IV (β = 0.20, 95% CI: [0.0052, 0.40], p = 0.044). The genotype effect was more pronounced in Type III. In contrast, no significant differences in 4F‐PTA were observed between OI types within either genotype background (all p ≥ 0.38). Similar patterns of interaction and post hoc contrasts were observed in both LF‐PTA and HF‐PTA models, reinforcing the genotype– and OI Type–specific differences in hearing trajectory identified in 4F‐PTA analysis (Table S2).

Hearing trajectory revealed a significant increase in 4F‐PTA over time for individuals with COL1A1 variants, with stronger HL progression in OI Type III (β = 0.038, 95% CI: [0.032, 0.044], p < 0.0001) compared with OI Type IV (β = 0.013, 95% CI: [0.0050, 0.021], p = 0.0014). Notably, individuals with COL1A2 variants and OI Type III exhibited a significant negative slope (β = −0.023, 95% CI: [−0.038, −0.0083], p = 0.0023), suggesting potential stabilization or slight improvement in this subgroup (Figure 2C).

Breakpoints, defined as the age at which HL trajectory changes or begins to accelerate, were estimated separately for 4F‐PTA, LF‐PTA, and HF‐PTA. The mean estimated breakpoint for HL progression was age 13.5 specifically, for 4F‐PTA it was 13.0 ± 4.9 years, for LF‐PTA it was 13.7 ± 4.1 years, and for HF‐PTA it was 13.7 ± 2.7 years. The earliest breakpoints observed across all ears were 6.2 years for 4F‐PTA, 8.5 years for LF‐PTA, and 8.7 years for HF‐PTA, indicating that HL progression may begin as early as the first decade.