2. Materials and Methods

2.1. Study Population

Longitudinal audiometric evaluations obtained between 1980 and 2023 were analyzed on 75 participants diagnosed with OI Type III or IV with heterozygous variants in COL1A1 or COL1A2 (Table 1). All participants had confirmed heterozygous variants in COL1A1 or COL1A2 resulting in small changes in the primary structure of the collagen alpha chains. The majority were missense substitutions for glycine residues; two were single exon skipping, and one was a small deletion. Mutations were determined by targeted gene panel sequencing (CTGT [now HNL Genomics], Allentown, Pennsylvania) with a few exceptions. Causative mutations for ID 208, 125, and 121 were reported by the Marini lab in 1993, 1996, and 2002, respectively [21–23]. Their cDNA level mutations and any intronic changes were sequenced by the dideoxy chain‐termination method (Sanger sequencing) using appropriate PCR amplification products.

Variables | COL1A1 | COL1A2 | Total
 | (n = 41) | (n = 32) | (n = 73)
Sex, n (%)
Female | 20 (48.8) | 21 (65.6) | 41 (56.2)
Male | 21 (51.2) | 11 (34.4) | 32 (43.8)
OI type, n (%)
OI Type III | 22 (53.7) | 12 (37.5) | 34 (45.2)
OI Type IV | 19 (46.3) | 20 (62.5) | 39 (52.1)
Age at assessment, (years)
Minimum | 1 | 2 | 1
Maximum | 40 | 45 | 45
Mean ± SD | 15.1 ± 8.8 | 14.1 ± 8.6 | 14.7 ± 8.7
Visits
Minimum | 1 | 2 | 1
Maximum | 18 | 20 | 20
Mean ± SD | 7.8 ± 4.7 | 7.2 ± 4.5 | 7.5 ± 4.6

The legacy notation for collagen protein variants, in which the first Gly of the collagen helical region is designated as Amino Acid 1 [18], is utilized in descriptions throughout this text; the current ACMG notation for their variants is also presented in Table 2. Children with OI from throughout the United States enrolled in a natural history study (Clinical Trials NCT03575221; Natural History of the Collagen‐Related Disorder OI and Genotype–Phenotype Correlation and NCT00001594; Evaluation and Intervention for the Effects of OI) approved by the institutional review board. All participants or their guardians gave informed consent. Participants were seen throughout childhood, with periodic examination of the primary skeletal and secondary audiological, dental, pulmonary, and neurological features of OI. Of the 75 participants, one was excluded from analysis because ear‐specific responses could not be obtained, and one was excluded because no pure tone averages (PTAs) could be calculated from the limited number of frequencies tested.

ID | Gene | Legacy notation | ACMG DNA notation | ACMG protein notation | OI type | Sex | Min age | Max age | Max degree HL | HL types | LF‐PTA degree HL | 4F‐PTA degree HL | HF‐PTA degree HL
128 | COL1A1 | G76E | c.761G>A | p.G254E | III | F | 8 | 19 | Profound∗ | CHL/MHL | B(profound) | B(profound) | B(profound)
131 | COL1A1 | G88E | c.797G>A | p.G266E | III | F | 9 | 29 | Severe∗ | CHL/MHL | B(severe) | B(mod) | B(mod)
111 | COL1A1 | G148D | c.977G>A | p.G326D | IV | M | 7 | 22 | Mild | CHL/SNHL | B(mild) | L(mild) | L(mild)
142 | COL1A1 | G154R | c.994G>A | p.G332R | III | M | 4 | 26 | Mild | CHL/SNHL | L(mild) | L(mild) | L(mild)
124 | COL1A1 | G187A | c.1094C>A | p.G365A | III | F | 6 | 27 | Mild | CHL/SNHL | B(mild) | — | —
130 | COL1A1 | G193S | c.1111G>A | p.G371S | III | F | 4 | 15 | Mild | CHL | B(mild) | — | B(mild)
102 | COL1A1 | G217S | c.1183G>A | p.G395S | III | F | 11 | 31 | Mild | SNHL | R(mild) | — | —
107 | COL1A1 | G286R | c.1390G>A | p.G464R | III | M | 10 | 32 | Severe∗ | CHL/MHL | R(severe) | R(severe) | B(severe)
 |  |  |  |  |  |  |  |  |  |  | L(mod) | L(mod) | 
133 | COL1A1 | G352S | c.1588G>A | p.G530S | IV | M | 7 | 39 | Mild | CHL/SNHL | B(mild) | — | —
134 | COL1A1 | G352S | c.1588G>A | p.G530S | IV | M | 4 | 36 | Mild | CND | R(mild) | — | L(mild)
139 | COL1A1 | G352S | c.1588G>A | p.G530S | IV | F | 3 | 22 | Mild | CHL | B(mild) | R(mild) | —
125 | COL1A1 | Splice out exon 33‐36 | c.2337_2451+62del g.50190267_50190825del |  | III | F | 3 | 23 | Mild | CHL | B(mild) | B(mild) | —
119 | COL1A1 | G589S | c.2299G>A | p.G767S | III | F | 2 | 17 | Mild | CHL | B(mild) | — | —
136 | COL1A1 | G589S | c.2299G>A | p.G767S | III | M | 3 | 30 | Mild | CHL/SNHL | B(mild) | B(mild) | B(mild)
141 | COL1A1 | G613A | c.2371G>A | p.G791A | III | M | 6 | 10 | Mild | CHL | R(mild) | — | —
121 | COL1A1 | Splice out exon 41 | g.12743A>C | IVS41+4A>C | III | M | 3 | 15 | Mild | CND | L(mild) | — | —
135 | COL1A1 | G832S | c.3028G>A | p.G1010S | IV | F | 2 | 40 | Severe(R)∗ | CHL/MHL(R) | R(severe) | R(mod) | R(mod)
 |  |  |  |  |  |  |  |  | Mild(L) | CHL/SNHL(L) | L(mild) | L(mild) | L(mild)
110 | COL1A1 | G898S | c.3226G>A | p.G1076S | III | F | 4 | 22 | Severe∗ | CHL/MHL | R(severe) | B(mod) | R(mod)
 |  |  |  |  |  |  |  |  |  |  | L(mod) |  | L(severe)
118 | COL1A1 | G898S | c.3226G>A | p.G1076S | III | M | 5 | 10 | Mod∗ | CHL | B(mod) | R(mod) | R(mild)
 |  |  |  |  |  |  |  |  |  |  |  | L(mild) | 
127 | COL1A1 | G997S | c.3523G>A | p.G1175S | III | M | 4 | 28 | Mod(R)∗ | CHL | R(mod) | R(mod) | R(mod)
 |  |  |  |  |  |  |  |  | Mild(L) |  | L(mild) | L(mild) | L(mild)
132 | COL1A1 | P1266H CPRO peptide | c.4331C>A | p.P1444H | IV | F | 3 | 9 | Mild | CND | B(mild) | — | —
207 | COL1A2 | G106V | c.589G>T | p.G196V | IV | F | 2 | 36 | Mild | CHL/SNHL | L(mild) | L(mild) | L(mild)
208 | COL1A2 | Splice out exon 16 | g.16194G>A | IVS16+1G>A | IV | F | 4 | 33 | Mod(L) | CHL | L(mild) | L(mod) | R(mild)
 |  |  |  |  |  |  |  |  | Mild(R) |  |  |  | L(mod)
203 | COL1A2 | G190V | c.839G>T | p.G280V | III | F | 4 | 22 | Mild | CHL | R(mild) | R(mild) | —
212 | COL1A2 | G238S | c.982G>A | p.G328S | IV | F | 2 | 35 | Mild | CHL | L(mild) | — | —
224 | COL1A2 | G238S | c.982G>A | p.G328S | III | F | 3 | 6 | Mild | CND | B(mild) | — | —
223 | COL1A2 | G247C | c.1009G>T | p.G337C | III | M | 17 | 26 | Mild | CHL/CND | B(mild) | R(mild) | —
202 | COL1A2 | G247S | c.1009G>A | p.G337S | IV | M | 8 | 45 | Mild | CND | — | — | B(mild)
219 | COL1A2 | G250S | c.1018G>A | p.G340S | III | F | 6 | 16 | Mild | CHL | B(mild) | — | R(mild)
211 | COL1A2 | G268S | c.1072G>A | p.G358S | IV | F | 6 | 26 | Mild | CHL | R(mild) | — | —
200 | COL1A2 | G337S | c.1279G>A | p.G427S | III | F | 7 | 18 | Mild | CHL/SNHL | B(mild) | — | —
222 | COL1A2 | G370S | c.1378G>A | p.G460S | III | M | 4 | 12 | Mild | CND | — | — | L(mild)
213 | COL1A2 | G511S | c.1801G>A | p.G611S | IV | F | 14 | 35 | Mild | CND | — | — | B(mild)
218 | COL1A2 | G703R | c.2377G>A | p.G793R | III | F | 3 | 14 | Mild | CHL/CND | B(mild) | — | —
221 | COL1A2 | G706S | c.2386G>A | p.G796S | III | F | 4 | 5 | Mild | CND | B(mild) | — | —
220 | COL1A2 | G922S | c.3034G>A | p.G1012S | III | F | 5 | 17 | Mild | CHL | B(mild) | R(mild) | B(mild)
228 | COL1A2 | G922S | c.3034G>A | p.G1012S | IV | F | 25 | 28 | Mild | CHL/SNHL | R(mild) | B(mild) | B(mild)
201 | COL1A2 | G940D | c.3089G>A | p.G1030D | IV | M | 8 | 22 | Mild | CHL | R(mild) | — | —
120 | COL1A2 | P1011L | c.3299C>T | p.P1101L | IV | M | 3 | 18 | Mild | CND | — | L(mild) | —

Primary outcomes included HL type, degree, age of onset, and progression. Secondary outcomes included “breakpoints” in trajectory of progressive HL. Sex was defined as sex documented in the medical record and collected to determine prevalence of sex‐specific audiologic manifestations.

2.2. Defining Hearing Status

Comprehensive, age‐appropriate audiometric evaluations were completed in sound‐treated booths and consisted of behavioral audiometry for air conduction (AC) from 0.25 to 8 kHz and bone conduction (BC) from 0.25 to 4 kHz, speech reception thresholds, word recognition ability in quiet, 226‐Hz tympanometry, and acoustic reflex threshold and decay testing. A subset of these tests was used for analysis. Hearing status was evaluated using three PTAs: 4‐frequency (4F) PTA (0.5, 1, 2, and 4 kHz); low frequency (LF) PTA (0.25 and 0.5 kHz); and high frequency (HF) PTA (4 and 8 kHz) (see Table S1 for detailed definitions in this section).

HL was defined in individual ears by the identification of PTAs > 20 dBHL in any of the three PTAs at one or more visits. HL was characterized based on degree, type, age of onset, and progression. HL type was determined as: (1) CHL defined as the presence of an air‐bone gap (ABG) > 10 dB when AC > 20 dBHL and BC ≤ 20 dBHL; (2) MHL defined as AC and BC > 20 dBHL in combination with ABG > 10 dBHL; and (3) SNHL defined as AC > 20 dBHL with ABG ≤ 10 dBHL.

Degree of HL was defined as mild (> 20 and ≤ 40 dBHL); moderate (> 40 and ≤ 70 dBHL); severe (> 70 and ≤ 95 dBHL); or profound (> 95 dBHL). Hearing sensitivity was defined as normal if all three PTAs were ≤ 20 dBHL. This cutoff was chosen to establish a balance between the normal hearing thresholds for children (15 dBHL) and adults (25 dBHL) [24, 25]. HL was defined as stable unless there was progression. Progressive HL was defined as worsening beyond mild into a greater degree for the duration of the time the participant was followed. In audiometry, test–retest variability of ±5 dBHL is expected and does not demonstrate a change in hearing sensitivity [26].

2.3. Statistical Analysis

Statistical analyses were performed in R software (Version 4.3.3). LF‐PTA, 4F‐PTA, and HF‐PTA were log‐transformed to address nonnormality of residuals and heteroscedasticity [27] (i.e., nonconstant variance of residuals across levels of the fitted values). Linear mixed‐effects models were fitted using the lme4 package (v1.1.34) to examine association between log‐transformed PTA outcomes and fixed effects, including genotype (COL1A1 vs. COL1A2), OI type (III vs. IV), and Visit time (modeled continuously). Sex and age at onset of HL were included as covariates. All main effects and interaction terms among visit, genotype, and OI type were included to assess both independent and joint effects of genetic background, clinical subtype, and longitudinal progression on hearing thresholds. The model was specified as follows:

logPTA~sex+age+visit×genotype×OItype+1subject/ear.

Random intercepts were included for subjects and ears nested within subjects, to account for repeated measurements within individuals and between ears. Model fit was evaluated using Akaike information criterion and Bayesian information criterion [28], which balance model goodness‐of‐fit against model complexity, with lower values indicating a more parsimonious model. Model assumptions were assessed using diagnostic plots generated with the plot_model() function (type = “diag”) from the sjPlot package (v2.8.14).

Statistical significance of fixed effects and interaction terms was assessed using F‐tests with degree of freedom estimated via Satterthwaite′s approximation [29], which provides improved small‐sample inference for mixed‐effects models by approximating denominator degrees of freedom, implemented in the lmerTest package. Post hoc contrasts and visit‐related slopes were estimated from the linear mixed‐effects model adjusting for age and sex, with random intercepts for subject and ear. Positive slopes indicate increasing PTA values (i.e., worsening hearing thresholds) over time (Table S2). Post hoc pairwise comparisons and slope analyses were conducted using the emmeans package (v1.10.1), with Tukey adjustment applied for multiple comparisons.

To further characterize HL progression, the ages of inflection breakpoints were quantified. To identify breakpoints, we applied segmented linear regression to AC PTA data for the individual ears of the seven patients with progressive HL. This approach models the PTA trajectory as a series of linear segments separated by breakpoints, allowing detection of time points at which the rate of hearing change (slope) significantly shifts. Analyses were conducted using the segmented package (v2.2‐1) in R.