Associations of Sleep Disordered Breathing and Neurodevelopment with Pre-Speech Function in 2-month-old Infants with Oral Cleft: A Feasibility Study

Abstract

Objective:

Evaluate feasibility of a study to measure emergent language and pre-speech skills in infants with and without oral cleft. Functioning was assessed in relation to nighttime oxygenation and neural structure.

Design:

Observational, case/control study.

Setting:

Data was obtained at two clinical settings (University of Iowa and University of Pittsburgh Children’s Hospital).

Participants:

Twenty infants with and without oral cleft were enrolled and evaluated at 2 months of age. Seven additional infants the same age with a subset of measures from an earlier study were also included. The combined sample included 27 participants: 5 with cleft lip only (iCL), 8 with cleft lip and palate (iCLP), and 14 unaffected (UA).

Main Outcome Measures:

Parent ratings of general functioning and emergent language (Bayley-4 SEABQ), acoustic measures of vocalizations, overnight pulse oximetry, and structural MRI scans.

Results:

Success rates for measures were high for emergent language and vocal acoustics (100% and 89%, respectively) and moderate for oximetry and MRI scans (63% and 70%, respectively). Participants with iCLP had the lowest outcomes for emergent language and vocal acoustics, as well as the highest number of desaturations and lowest myelin intensity.

Conclusions:

The results of this study a) support feasibility of obtaining pre-speech, oxygenation, and neural measures in early infancy, b) identify patterns of higher risk for participants with iCLP, c) and suggest that some neural differences may be present prior to exposure to anesthesia.

Introduction

Children born with isolated cleft of the lip and/or palate (iCL/P), despite not having an identified syndrome or other congenital anomalies, are at increased risk for several language and learning disabilities. These include speech/articulation deficits,¹ delays in emergent language,² and later risk for reading and other learning disabilities.^3,4 Delays are associated with high use of Special Education,⁴ lower school matriculation,⁵ and increased psychosocial concerns.^6,7

The potential causes of speech and language deficits are multifaceted. In general, it has been felt that the deficits in iCL/P are secondary to factors such as early disruption of speech and hearing,^8,9 oral cavity abnormalities (with palatal involvement having higher risk),^10–15 and structural differences in the velopharyngeal mechanism.^16–19 More recently, researchers have been exploring etiologies that may be connected to speech and language outcomes through impacts on neural development. It is hypothesized that exposure to anesthesia²⁰ and increased rates of sleep disordered breathing²¹ during early, critical developmental periods may impact neural development or cause neuronal injury, further impacting outcomes. A third theory proposes there is disrupted neural migration during development (aka. Cortical dysplasia),^22,23 making neural differences a primary, rather than secondary factor affecting outcomes.

Exposure to Anesthesia

Patients with oral cleft undergo several procedures requiring sedation over the first few years of life. Because animal studies have demonstrated a relationship between increased exposure to anesthesia and cell death,²⁴ it is hypothesized that early surgeries may negatively impact brain development and subsequently, speech and language acquisition. However, the relationship between exposure and neural development in humans is less clear. While there is not a strong association to broad cognitive functioning, some have found delayed white matter development and lower performance on tasks of processing speed and working memory.^25,26 Research specific to children with cleft has been predominately retrospective with mixed findings. One study found that increased exposure was correlated to greater frontal lobe volume and decreased verbal IQ²⁷ while another found no relation to academic functioning.²⁸

Sleep Disordered Breathing

Patients with cleft palate are a higher risk for sleep disordered breathing, including obstruction, central apnea and hypoventilation (with resulting lowered oxygenation).^29,30 It is hypothesized that low oxygen saturation may disrupt neural development in the pre-frontal cortex and contribute to later cognitive deficits.^21,31 While the connection to neural structure and function has not been studied in patients with oral cleft, one study on children with cleft found a correlation between respiratory elements of sleep disordered breathing in infancy and lower language skills at 3 years of age.³²

Disrupted Neural Migration

An oral cleft occurs when there is disruption in the migration of cells that form the lips and palate.³³ These cells originate from the same cells that later develop into the brain and other components of the central nervous system. It is hypothesized that there may be concurrent abnormal migration in the developing brain cells, resulting in slightly abnormal neuronal structure and connections, and consequently, disrupted brain functioning.^22,23

Magnetic Resonance Imaging (MRI) research on patients with iCL/P has identified neuronal differences that start in infancy and continue through adulthood. Preliminary work in infancy suggests early decreases in tissue in the left temporal lobe (auditory cortex)³⁴ and global delays in white matter maturation throughout the cerebrum.³⁵ While there may be some catch up in white matter volume by childhood and adolescence,³⁶ developmental trajectories suggest delays in pruning of gray matter that result in higher volume in the anterior cortex (frontal and parietal lobes) but reduced volume in posterior regions (temporal and occipital lobes) by adulthood.³⁷ Additionally, volumes of the cerebellum and sub-cortical regions appear to be consistently lower.^36–38 A recent meta-analysis of studies through 2020 found reduced cerebellar, occipital, temporal, and total gray matter volumes in iCL/P compared to unaffected controls.³⁹

Another way to evaluate brain structure is through evaluation of white matter integrity. Diffusion tensor imaging (DTI) is a special type of MRI that maps the movement of water molecules in the brain. One measure of this movement is fractional anisotropy (FA); it provides information on how water molecules move more easily in one direction versus the other. Higher FA values (i.e., more ease of movement in a single direction) suggest an increased integrity of myelin, the protective sheath around nerve fibers that help neural signals travel more quickly. Recent work has found decreased FA in children and adolescents with iCL/P in tracts within the cerebellum and brainstem.⁴⁰

These structural differences have further shown sex and cleft-type effects, where males with palatal involvement typically have the most aberrant findings.³⁶ Finally, neural differences have been associated with cognitive,⁴¹ academic,⁴² behavioral,⁴³ social,⁴⁴ and speech outcomes.⁴⁵ Associations remain after controlling for measures of hearing and speech, suggesting that disrupted neural development may play a unique role in these outcomes.

Research Questions

Research to date on these etiological theories of speech and language outcomes in oral cleft and the association to neural structure and function has been minimal and is limited by sample size, lack of concurrent measures, and use of retrospective chart reviews. To evaluate these theories, research leveraging neural imaging in infants early in development, before exposure to anesthesia and with information on pre-speech/language development and sleep disordered breathing collected in tandem is imperative. The current study assessed the feasibility of collecting data on a sample of infants with isolated cleft lip and/or palate (iCL/P) and unaffected controls (UA) that underwent a testing protocol at two months of age, before exposure to anesthesia. Descriptive information on pre-speech/language functioning, overnight oxygenation, and neural structure are presented.

As a feasibility study, inferential statistics will not be presented. Interpretation of findings focus on patterns of performance between groups and across different measures to guide the development of protocols for larger studies with greater power to evaluate these complex relationships. In general, it is hypothesized that emergent language skills will not show differences at this young age. However, refined measures of pre-speech vocal acoustics and neural structure will show differences. It is uncertain if pre-speech and neural differences will be associated with measures of oxygenation at this early of age. Patterns of differences prior to exposure to anesthesia will reduce support for the anesthesia exposure theory and increase support for the disrupted neural development theory – and the need to include MRI. Strong differences in sleep disordered breathing would provide support for the theory of sleep disordered breathing – and the need to include oximetry.

Materials and Methods

Participants

Infants with and without oral cleft (n = 20), were seen at 2 months of age (6 – 12 weeks old) between November of 2022 and April of 2024. Recruitment and assessment efforts took place at two clinical sites (University of Iowa [UI] and University of Pittsburgh Medical Center [UPMC]). Infants with iCL/P were identified through clinic schedules and advertisements posted on social media support groups for families impacted by oral clefts. Unaffected infants were identified through local advertisements posted on University-wide platforms, including ResearchMatch and READGreen.

Interested participants were screened prior to scheduling. Given the nature of this study, enrollment was limited to those with English as the primary language. Because of the higher risk for oral cleft among patients of preterm birth,⁴⁶ only those with gestational age over 36 weeks were included. Finally, participants were excluded if they had a significant medical condition that could impact development or functioning (e.g., known syndrome, cardiac or neurological disorder), had metal in their body that would prohibit participation in the MRI scan, or had any exposure to general anesthesia before the study date.

To provide more robust information across different cleft types, data from an additional 7 participants (3 iCL/P and 4 UA) seen between 2014 and 2016 at UI³⁵ were included in analyses. This data only included parental ratings of general functioning, vocal recordings, and MRI structural scan. Enrollment efforts and inclusion/exclusion criteria were the same as the current protocol.

Measures

All study procedures took place on site at each institution and on the same day for most participants. When possible, families returned within 1 week for second attempts on the MRI scanner if not successful in the first visit. The protocol lasted an average of 3 hours and included proxy questionnaires related to demographics, family functioning, and emergent language; collection of a pre-speech vocal recording; measurement of sleep disordered breathing; brain MRI; and structured medical chart review (See Figure 1). All procedures were reviewed and approved by the Iowa Institutional Review Board (IRB), which acted as the IRB of Record for all sites. Written informed consent was obtained from one legal guardian. Families were compensated $50 for their efforts and reimbursed for travel costs.

Figure 1. Study Protocol.

The full protocol for participants included proxy ratings of family functioning and emergent language, pre-speech vocal recordings, measuring sleep disrupted breathing using an oxygen monitor, structural brain MRI, and medical chart review for information on cleft type, audiology test results, and documentation of any completed procedures.

*Protocol components not completed by the n = 7 earlier pilot participants.

Demographics and Family Functioning.

Parents were asked to complete a questionnaire that provided information on sex (as defined at birth), race, ethnicity, and socioeconomic status using the Barratt Simplified Measure of Social Status⁴⁷ (BSMSS). They also completed the McMaster Family Assessment Device⁴⁸ (McMaster FAD), a 53-item, Likert-Scale questionnaire that provides information on family functioning. In addition to the Total Score, 7 subscales of Problem Solving, Communication, Roles, Affective Responsiveness, Affective Involvement, Behavioral Control, and General Functioning are calculated. Means for Non-Stressed and Stressed populations are provided for each subscale. Completion time is roughly 5 minutes and higher scores reflect more issues.

Emergent Language.

Parents completed the Bayley-4 Social-Emotional & Adaptive Behavior Questionnaire⁴⁹ (Bayley SEABQ). This 120-item questionnaire measures adaptive behaviors and monitors the development of independence. Participants from the earlier feasibility study completed the Bayley-3 SEABQ. General functioning was measured with the Total Adaptive Functioning (TAF) and Socio-Emotional (SE) Indices were used as measures of general functioning. The Receptive and Expressive Language subtests (RL and EL) from the Communication Index were used as measures of emergent language. Index scores have a mean of 100 and standard deviation of 15; RL and EL subtest standard scores were transposed to this scale for easier comparison. Completion time was roughly 15 minutes.

Vocal Recordings.

Audio recordings of spontaneous crying were obtained in a location free of extraneous noise. Throughout the visit, staff waited for moments of spontaneous cries (e.g., the infant was hungry, uncomfortable, or needed a diaper change) and obtained the recording using a TASCAM (DR-40x; 48 kHz, 16 bit) recorder placed near the infant’s mouth. No pain conditions or other methods of cry stimulation were utilized. All recordings were completed in the presence of the caregiver while the infant was supine. Parents were permitted to calm the infant as soon as the required 1 minute of recording time was obtained.

For processing, a cry utterance was defined as the onset and the offset of identifiable acoustic energy in the waveform that occurred on the expiratory phase of a single respiratory cycle. PRAATv 6.1.35,⁵⁰ an open-source software, identified and measured cry utterances within each audio file and distinguished between expiratory and inspiratory phases as well as background noise. All automatic segmentations were double checked by an audio-visual analysis. Any errors in identifying segmentations were manually corrected by modifying cursor positions for start and end points for cry utterances. The number of utterances for each infant was calculated, then each utterance was evaluated to determine if it contained one or more melody-breaking instances (aka. segmentation error). A minimum of 25 utterances were required for processing.

Two key acoustic measures were calculated. The percentage of cry utterances containing a segmentation error for each infant was their resulting Segmentation Index (SI). A high SI is hypothesized to represent immature vocal control, disruption to neural networks associated with rhythmicity (cerebellum, brain stem), and is a risk factor for subsequent language disorders.⁵¹ The Melody Complexity Index-II (MCI-II) is the percentage of complex melodies (utterance with a contour of more than one ascending-then-descending melodic arch) that are unsegmented. A lower MCI-II is hypothesized to represent immature vocal control and right hemisphere dysfunction.

Sleep Disordered Breathing.

Overnight measures of blood oxygen were obtained within a 1-week window of the research visit. Parents were given a NellcorTM Bedside SpO2 Patient Monitoring System (PM100N) with an infant sensor and were trained in set up. They took the device and either conducted the recording at their own home or in a hotel the night of the study. Trend data was downloaded and processed using Nellcor Analytics Tool V1.6. Baseline threshold for desaturation detection was defined as a drop of 3% below baseline oxygen saturation (SpO2); calculated as previous peak SpO2 value. To be considered an event, this condition must have been met for 5 seconds. Instances of interference and desaturation outliers (>3SD) were removed and recordings with at least 2 hours of data were included. Key variables of percent time at or above SpO2 90, desaturation rate, and longest desaturation event were calculated.

Magnetic Resonance Imaging.

Scans were completed using the feed and wrap technique⁵² during the infant’s natural nap time. For all participants, the mother took the infant to a quiet nursery near the scanner and fed and rocked the infant. After the infant fell asleep, they were wrapped in an MRI-safe swaddling device and protective coverings were placed on their ears to minimize noise. A research assistant remained in the scanner room maintaining visual contact with the baby and notified the technicians to end the scan if signs of distress were noted.

All MRI data were acquired on a 3T scanner (GE Premier/UI and Siemens PRISMA/UPMC) using a 48-channel or 32-channel head coil, respectively. The imaging protocol was aligned with the NIH ABCD Study⁵³ and included: 1. Localizer; 2. T1 MP-RAGE Sagittal: TE=2.88ms, TI=1060ms, TR=2500ms, FOV=256×256×176, Matrix=256×256×176, flip angle = 8, iPAT=2; 3. T2 SPACE Sagittal: TE=565ms, TR=3200ms, FOV=256×256×176, Matrix=256×256×176, iPAT=2; 4. DTI Axial: TE=88ms, TR=4100ms, FOV=240×240, Matrix=140×140, Slice Thickness=2.0, b-value=1000, # directions=30, MB=3, iPAT=1, 4 x b=0; 5. Reverse Polarity B0 Image: Same parameters as DTI but reversed phase encoding direction (4 x b 0); and 6. fMRI (resting state): TE=30ms, TR=800ms, FOV=216×216, Matrix=90×90, flip=52, iPat=1, MB=6 (8 minutes). Total scanner time was 25 minutes.

Anatomical MR images, T2-weighted (T2w) and T1-weighted (T1w), were processed using a multi-stage, multi-atlas approach similar to a family of methods that combines multiple templates and label fusion, e.g., (BRAINSAutoworkup⁵⁴ and MAGeT⁵⁵). For these infant brain images, we utilized the M-CRIB 2.0^56,57 atlas with 10 expertly segmented exemplars consistent with the common Desikan-Killiany-Tourville⁵⁸ atlas. In the first stage, anatomical images were cleaned and a multi-atlas coregistration was leveraged to generate accurate parcellation of the intracranial volume (i.e., skull stripping). Processing steps included, i) z-plane clipping, to limit images to the head region and increase similarity to template images; ii) denoising using a spatially-varying Rician noise model;^59,60 iii) intensity non-uniformity correction, using the N4 method,⁶¹ iv) coregistering T1w images to the T2w image (as available), and v) a preliminary mask of the intracranial volume (ICV) was generated by coregistering atlas exemplar images to the T2w image and combining ICV masks for each exemplar using correlation voting (where voting strength is modulated by similarity to the participant’s anatomical image).

In the second stage, images were normalized to atlas space using non-linear registration to multiple atlas exemplars. Processing steps in stage 2 included: i) applying the ICV mask from stage 1 to remove gross differences in skull and face in the image that otherwise draw the registration algorithms away from the brain and its edges. ii) Images were aligned to atlas space using a rigid coregistration, this pushed each brain to roughly ACPC alignment. iii) Aligned images were then normalized to multiple atlases using rigid, affine, and symmetric normalization implemented in Advanced Normalization Tools.^59,62 iv) Once normalized, atlas exemplars and labels were pushed to participant’s native brain space and combined using joint label fusion.⁶³ v) Next, anatomical labels underwent a rigorous visual inspection to ensure accuracy; the most common source of error was extension of labels to non-brain tissues, which was corrected by manually editing the ICV mask to exclude the offending regions from the pipeline.

Two measures of brain structure were obtained for analysis. The volume (mm³) of each atlas labelled region was calculated within participant’s native brain space. Volumes were segregated into gray and white matter within the left and right hemisphere of the cerebrum and cerebellum, as well as the four cerebral lobes and limbic tissue. Lastly, myelin integrity was quantified by taking the ratio of T1w to T2w images after scaling their intensity using histogram matching to a common target brain image of the same modality.^64,65 Normalizing within-modality to a common template effectively placed individual estimates of myelin intensity on a common scale across participants (i.e., higher values = higher myelin intensity).^56,57 Values were obtained for white matter tracts across both hemispheres of the cerebrum and cerebellum/brain stem.

Medical Chart Review.

Following each visit, research staff reviewed medical charts for information on cleft type (LAHSHAL), and results of audiology tests (Otoacoustic emissions [OAE] and Auditory Brainstem Response [ABR]). Staff also confirmed that no procedures involving anesthesia were conducted prior to the participation date.

Statistical Methods

Categorical data were analyzed by counts and percentages. For continuous data, means and standard deviations were calculated. As not all participants had complete data across all measures, sample size was determined on a pair-wise basis across variables.

Demographics by Study and Cleft Type.

First, demographic variables (i.e., gestational age, birth weight, age at visit, and parental socioeconomic status), family functioning (i.e., McMaster FAD Total score and subscale scores), and general functioning (Bayley-4 SEABQ TAF and SE Indices) were presented separately for the two samples (i.e, the current protocol and earlier data). After verifying similarity between the two samples, they were combined and descriptive statistics were repeated separately for each of the cleft types on measures of demographics and family functioning.

Language, Vocal Acoustics, Sleep Disordered Breathing, and Neural Structure.

Descriptive statistics, split by cleft type, were conducted for all measures of interest. This included emergent language (Bayley-4 SEAQ RL and EL subscales), vocal acoustics (total utterances, SI, and MCI-II), and sleep disordered breathing (percent time at/above SpO2 90, desaturation rate, and longest desaturation event).

For measures of neural structure (i.e., brain volume and myelin intensity), measures were presented separately for left vs. right hemisphere and gray vs. white matter when possible. For brain volume, dependent variables included the cerebrum, cerebellum, and the four cerebral lobes (i.e., frontal, parietal, temporal, and occipital). For measures of myelin intensity, dependent variables included 10 association fibers (connecting regions within a hemisphere), 8 projection fibers (connecting regions from inferior to superior regions), 8 commissural fibers (connecting regions across hemispheres), and 5 tracts within the cerebellum and brainstem. In addition to means and standard deviations, z-scores were calculated for participants with iCL and iCLP using data from UA participants as reference.

Results

Participants and Demographics

A total of 27 participants were included in this study. Twenty were enrolled with the feasibility protocol (10 UA, 4 iCL, and 6 iCLP) and an additional 7 participants (4 UA, 1 iCL, and 2 iCLP) were included from the earlier study (See Supplemental Table 1).

The combined sample included 14 UA participants (9 male), 5 participants with iCL (3 male), and 8 participants with iCLP (6 male). Most participants (93%) were White and non-Hispanic/Latinx (93%). Gestational age ranged from 36^4/7 to 41^6/7 (mean [SD] = 39.59 [1.4] UA, 38.00 [1.4] iCL, and 39.00 [1.5] iCLP) and birthweight ranged from 2.47 to 4.25 kg (mean [SD] = 3.43 [0.5] UA, 2.90 [0.1] iCL, and 3.39 [0.5] iCLP). Participants ranged in age from 1.64 to 2.86 months of age at the time of testing (mean [SD] = 2.03 [0.3] UA, 2.49 [0.3] iCL, and 2.14 [0.4] iCLP). Finally, the socioeconomic status of parents of participants ranged from 33 to 61 (mean [SD] = 52.14 [7.8] UA, 47.20 [5.2] iCL, and 43.81 [8.8] iCLP).

Results of newborn hearing screening were available for 20 participants. For the 7 who did not have information on newborn hearing (4 UA, 2 iCLP, and 1 iCL), parents did not mark concerns with hearing on their record forms and no issues with hearing were noted within the medical record. Of the 20 participants with results of newborn hearing screening, 18 passed and 2 (1 UA and 1 iCLP) were referred. Parental ratings of receptive and expressive language were reviewed for the 2 participants who were referred. Ratings for these participants were above the 25^th percentile for the entire sample. Because ratings of emergent language were not discrepant for these participants, they were included in subsequent analyses.

Scores on the McMaster FAD were evaluated separately in relation to non-stressed and stressed population means for each group. For UA families, Behavioral Control was over the mean for the stressed population and Affective Involvement was above the non-stressed population mean. For families of patients with iCL, Roles and Behavioral Control were above the non-stressed population mean. Finally, families of those with iCLP had 5 out of 7 subscales above the non-stressed population mean: Communication, Roles, Affective Involvement, Behavioral Control, and General Family Functioning. The total score for participants with iCLP (mean = 91.01) was higher (reflecting more issues) than the score for those with iCL (mean = 85.99) and UA participants (86.82). General functioning (TAF and SE) was in the average range for UA participants (mean = 101.57 & 103.57, respectively). Scores for participants with iCL ranged from average (TAF mean = 95.20) to above average (SE mean = 121.00). For participants with iCLP, scores ranged from low average (TAF mean = 92.50) to average (SE mean = 96.88; See Table 1).

Table 1.

Demographics, Family Functioning, and General Functioning [Mean (SD)].

	Unaffected	iCL	iCLP
	n = 14	n = 5	n = 8
Sex (defined at birth)
Male	9	3	6
Race
White	12	5	8
Multiracial	2	0	0
Ethnicity
Hispanic/Latinx	1	0	1
Hearing
Pass in at Least 1 Ear	13	5	7
Refer	1	0	1
Gestational Age (w)	39.59 (1.44)	38.00 (1.42)	39.00 (1.39)
Birth Weight (kg)	3.43 (0.45)	2.90 (0.13)	3.39 (0.47)
Age at Visit (m)	2.03 (0.30)	2.49 (0.30)	2.14 (0.40)
Parent SES	52.14 (7.75)	47.20 (5.22)	43.81 (8.80)
McMaster FAD Total Score	86.82 (13.01)	85.99 (35.75)	91.01 (13.96)
Problem Solving  [NSt. M = 2.22 & St. M = 2.44]	1.57 (0.31)	1.60 (0.71)	1.65 (0.41)
Communication^  [NSt. M = 1.90 & St. M = 2.14]	1.79 (0.45)	1.83 (0.81)	1.94 (0.34)
Roles  [NSt. M = 1.96 & St. M = 2.31]	1.93 (0.38)	2.03 (0.81)	2.03 (0.28)
Affective Responsiveness^  [NSt. M = 2.16 & St. M = 2.42]	1.48 (0.30)	1.53 (0.68)	1.75 (0.37)
Affective Involvement  [NSt. M = 1.74 & St. M = 2.04]	1.86 (0.31)	1.60 (0.76)	1.84 (0.45)
Behavior Control  [NSt. M = 1.43 & St. M = 1.52]	1.58 (0.30)	1.49 (0.62)	1.47 (0.31)
General Family Functioning  [NSt. M = 1.49 & St. M = 1.68]	1.39 (0.30)	1.42 (0.67)	1.52 (0.40)
Bayley-4 SEABQ
Total Adaptive Functioning	101.57 (7.63)	95.20 (9.26)	92.50 (7.91)
Socio-Emotional	103.57 (12.00)	121.00 (8.22)	96.88 (17.31)

Note. iCL = isolated cleft lip only. iCLP = isolated cleft lip and cleft palate. SES = socioeconomic status. FAD = Family Assessment Device. NSt. = Non-Stressed Population. St. = Stressed Population.

Emergent Language, Vocal Acoustics, Sleep Disordered Breathing, and Neural Structure across Groups

Emergent Language.

Twenty participants (14 UA, 4 iCL, and 6 iCLP) had complete data for the Bayley 4 SEABQ Receptive and Expressive Language subscales (RL and EL, respectively). Ratings were within the average range for UA participants (RL mean = 98.5 and EL mean = 105.0), in the low average to average range for participants with iCL (RL mean = 97.5 and EL mean = 92.5), and in the lower average range for participants with iCLP (RL mean = 94.2 and EL mean = 92.5; See Table 2).

Table 2.

Emergent Language, Pre-Speech Vocal Acoustic Measures, and Sleep Disordered Breathing Across Cleft Types.

	Unaffected	iCL	iCLP
Bayley 4 SEABQ	n = 14	n = 4	n = 6
Receptive Language	98.50 (7.47)	97.50 (6.45)	94.17 (6.65)
Expressive Language	105.00 (5.77)	92.50 (5.00)	92.50 (13.69)
Vocal Acoustics	n = 13	n = 4	n = 7
Total Utterances	57.08 (37.28)	52.75 (19.00)	47.86 (14.24)
Segmentation Index	.38 (0.25)	.33 (0.34)	.49 (0.18)
Melody Complexity Index-II	.39 (0.24)	.49 (0.37)	.32 (0.23)
Sleep Disordered Breathing	n = 8	n = 4	n = 5
Oximetry Recording Duration (h)	7:46 (4:19)	8:14 (2:59)	6:22 (3:27)
Percent SpO2 at/above 90	98.54 (1.05)	98.90 (1.87)	99.26 (0.44)
Desaturations per hour	2.86 (3.84)	2.15 (2.24)	9.66 (7.28)
Longest Desaturation (min)	0.51 (0.33)	1.27 (1.88)	1.57 (1.46)

Note. iCL = isolated cleft lip only. iCLP = isolated cleft lip and palate. SEABQ = Social-Emotional & Adaptive Behavior Questionnaire. SpO2 = Percent saturation of oxygen in the blood.

Vocal Acoustics.

Adequate vocal recordings were obtained for 24 of the 27 participants (13 UA, 4 iCL, and 7 iCLP); 3 had too few utterances for analysis. UA Participants had an average of 57.08 utterances while those with iCL and iCLP were slightly lower (mean = 52.75 and 47.86, respectively). Vocal control for participants with iCLP was the most immature, with the highest percent of segmentation errors (SI = 49%; UA = 38% and iCL = 33%) and the lowest percent of complex melodies (MCI-II = 32%; UA = 39% and iCL = 49%; See Table 2).

Sleep Disordered Breathing.

Adequate recordings were obtained for 17 participants (8 UA, 4 iCL, and 5 iCLP); 3 were excluded because they were under 2 hours in duration and the 7 earlier feasibility participants did not complete this part of the protocol. Participants with iCLP had had the most percent time at/above SpO2 90 (mean = 99.26%), followed by those with iCL and then UA participants (mean = 98.90% and 98.54%, respectively). However, participants with iCLP had over three times the desaturations per hour than UA infants (mean = 9.66 and 2.86, respectively) and over four times those with iCL (mean = 2.15). Participants with iCLP also had the longest desaturation duration (mean = 1.57 minutes) followed by those with iCL and then UA participants (mean = 1.27 and 0.51 minutes, respectively; See Table 2).

Neural Structure.

Adequate MRI scans were obtained on 19 of the 27 participants (12 UA, 2 iCL, and 5 iCLP). Unsuccessful scans were due to the infant not falling asleep, waking during the scan, or having too much motion for processing. Motion artifacts for participants with iCL precluded obtaining their regional gray matter volumes. Additionally, the limited number of adequate scans for this subgroup created high variability in measures of white matter volume. Participants with iCL consistently had higher white matter volumes than controls, with the greatest differences in the right temporal and limbic regions. For participants with iCLP and those who were unaffected, tissue volume (mm³) within the cerebrum was generally consistent. Within the cerebrum, gray matter was generally lower and white matter was generally higher for those with iCLP. Regionally, the largest difference was bilateral white matter in the frontal lobe, where the volume for participants with iCLP was nearly 1 standard deviation larger than UA participants. Cerebellar volume was roughly half a standard deviation smaller for participants with iCLP. (See Figure 2 and Table 3).

Figure 2. Regional Gray and White Matter Z-Scores.

A) Gray Matter and B) White Matter volume was converted to z-scores using the mean and standard deviation of unaffected participants as reference (y = 0). iCL = isolated cleft lip only (movement artifacts precluded calculation of regional gray matter). iCLP = isolated cleft lip and palate.

Table 3.

Brain Structure: Gray and White Matter Volume.

	Unaffected (n = 12)		iCL ( n = 2)		iCLP ( n = 5)
	Left Hemisphere	Right Hemisphere	Left Hemisphere	Right Hemisphere	Left Hemisphere	Right Hemisphere
Volume (mm3)
Cerebrum
Gray M	132,789 (11,575)	134,665 (11,483)	14,662 (1,500)	14,628 (1,769)	131,523 (10,341)	134,386 (11,935)
White M	77,943 (7,961)	78,111 (7,919)	90,805 (13,569)	140,168 (19,713)	80,822 (4,844)	80,252 (6,646)
Frontal
Gray M	40,322 (3,932)	40,334 (4,105)			40,195 (2,302)	40,709 (3,617)
White M	23,120 (3,010)	31,199 (3,713)	42,183 (4,426)	51,145 (7,009)	25,279 (1,925)	33,499 (2,499)
Parietal
Gray M	31,749 (2,807)	30,430 (2,884)			29,746 (3,199)	29,719 (3,496)
White M	18,348 (1,880)	19,776 (1,946)	25,344 (4,800)	24,470 (5,855)	17,748 (1,156)	19,534 (1,191)
Temporal
Gray M	19,968 (2,166)	20,984 (2,242)			21,286 (2,303)	21,723 (2,064)
White M	11,408 (1,341)	11,750 (1,277)	14,424 (2,604)	21,796 (1,565)	11,998 (1,216)	11,822 (1,398)
Occipital
Gray M	16,329 (1,749)	17,426 (1,317)			16,069 (2,172)	17,549 (2,534)
White M	8,318 (1,158)	8,488 (994)	9,541 (1,742)	9,848 (1,525)	8,871 (565)	8,850 (927)
Limbic
Gray M	16,238 (1,353)	11,293 (1,178)			16,570 (1,091)	10,893 (933)
White M	8,554 (847)	8,796 (779)	9,576 (973)	19,786 (2,670)	8,469 (662)	8,570 (568)
Cerebellum	15,749 (2,403)	15,560 (2,252)	15,810 (2,052)	15,638 (2,010)	14,650 (1,290)	14,578 (1,648)

Note. iCL = isolated cleft lip only (movement precluded regional measures of gray matter volume for both participants). iCLP = isolated cleft lip and palate. H = Hemisphere. M = Matter. Separate measures of gray and white matter for the cerebellum were not available.

Measures of myelin intensity could not be calculated for 1 participant with iCL and 1 UA participant due to incomplete scans. For the single participant with iCL, myelin intensity was higher (reflecting more intensity) than UA participants in most association fibers and half the projection and commissural fibers. In contrast, all intensity ratios (except for the left sagittal striatum) were lower for participants with iCLP compared to those who were unaffected. Participants with iCL and iCLP had the strongest differences (iCL > iCLP by over 1 SD) in association, projection, and commissural fibers (See Figure 3 and Table 4).

Figure 3. Myelin Intensity Z-Scores.

The myelin intensity of A) Association, B) Projection, C) Commissural, and D) Cerebellar and Brainstem fibers were converted to z-scores using the mean and standard deviation of unaffected participants as reference (y = 0). iCL = isolated cleft lip only. iCLP = isolated cleft lip and palate.

Table 4.

Brain Structure: Myelin Intensity of White Matter Tracts.

	Unaffected (n = 11)		iCL (n = 1)		iCLP (n = 5)
	Left H.	Right H.	Left H.	Right H.	Left H.	Right H.
Association Fibers
Cingulum Cingulate	0.910 (0.071)	0.901 (0.065)	1.043 (.)	1.013 (.)	0.872 (0.072)	0.880 (0.077)
Cingulum Hippocampus	0.879 (0.064)	0.874 (0.056)	0.926 (.)	0.934 (.)	0.865 (0.078)	0.862 (0.099)
Sup. Long. Fasciculus	0.934 (0.071)	0.916 (0.067)	1.010 (.)	0.983 (.)	0.898 (0.055)	0.861 (0.074)
Sup. FOS	1.031 (0.059)	1.028 (0.052)	0.881 (.)	0.935 (.)	0.956 (0.109)	0.969 (0.079)
Inf. FOS	0.975 (0.046)	0.972 (0.037)	1.020 (.)	0.998 (.)	0.961 (0.044)	0.954 (0.051)
Uncinate Fasciculus	0.932 (0.057)	0.941 (0.055)	0.995 (.)	0.957 (.)	0.911 (0.079)	0.940 (0.085)
Sagittal Striatum	0.952 (0.042)	0.948 (0.043)	0.973 (.)	0.961 (.)	0.964 (0.076)	0.946 (0.066)
External Capsule	0.999 (0.051)	0.991 (0.037)	1.032 (.)	1.024 (.)	0.947 (0.074)	0.942 (0.091)
Fornix	1.025 (0.049)	0.995 (0.049)	0.756 (.)	0.706 (.)	1.023 (0.059)	0.983 (0.091)
Stria Terminalis	1.154 (0.028)	1.147 (0.029)	1.246 (.)	1.172 (.)	1.130 (0.047)	1.111 (0.058)
Projection Fibers
Internal Capsule: Ant. Limb	1.169 (0.054)	1.169 (0.047)	1.189 (.)	1.149 (.)	1.114 (0.099)	1.14 (0.071)
Internal Capsule: Post. Limb	1.351 (0.127)	1.35 (0.129)	1.324 (.)	1.345 (.)	1.288 (0.148)	1.294 (0.134)
Internal Capsule: RS	1.229 (0.053)	1.208 (0.066)	1.294 (.)	1.249 (.)	1.176 (0.106)	1.17 (0.108)
Corona Radiata Ant.	0.911 (0.066)	0.901 (0.071)	0.932 (.)	0.935 (.)	0.894 (0.075)	0.897 (0.085)
Corona Radiata Sup.	1.050 (0.049)	1.06 (0.044)	1.057 (.)	1.048 (.)	0.983 (0.078)	0.978 (0.08)
Corona Radiata Post.	0.919 (0.096)	0.893 (0.099)	1.001 (.)	0.950 (.)	0.869 (0.078)	0.856 (0.094)
Cerebral Peduncle	1.280 (0.125)	1.287 (0.131)	1.211 (.)	1.203 (.)	1.238 (0.126)	1.246 (0.134)
Post. Thalamic Radiation	1.005 (0.066)	0.989 (0.063)	1.028 (.)	1.026 (.)	0.981 (0.080)	0.948 (0.058)
Commissural Fibers
Potine Crossing	1.204 (0.162)		1.083 (.)		1.100 (0.148)
Corpus Callosum	0.989 (0.052)		1.003 (.)		0.959 (0.066)
Prefrontal	0.981 (0.050)		1.014 (.)		0.957 (0.049)
Pre-Motor	0.974 (0.050)		0.988 (.)		0.930 (0.075)
Motor	1.014 (0.056)		1.016 (.)		0.953 (0.102)
Sensory	1.039 (0.061)		1.025 (.)		0.996 (0.102)
Parietal-Temp.-Occ.	0.997 (0.062)		1.000 (.)		0.975 (0.074)
Tapetum	0.992 (0.105)	0.973 (0.092)	1.012 (.)	0.871 (.)	0.965 (0.086)	0.938 (0.071)
Cerebellum/Brainstem
Corticospinal Tract	1.198 (0.157)	1.205 (0.173)	1.059 (.)	1.014 (.)	1.097 (0.159)	1.124 (0.159)
Medial Lemniscus	1.321 (0.205)	1.320 (0.211)	1.221 (.)	1.222 (.)	1.235 (0.203)	1.233 (0.228)
Cerebellar Peduncle Sup.	1.352 (0.212)	1.352 (0.209)	1.195 (.)	1.151 (.)	1.317 (0.159)	1.298 (0.185)
Cerebellar Peduncle Mid.	1.327 (0.195)	1.334 (0.204)	1.162 (.)	1.170 (.)	1.235 (0.191)	1.231 (0.220)
Cerebellar Peduncle Inf.	1.448 (0.267)	1.425 (0.256)	1.362 (.)	1.253 (.)	1.376 (0.248)	1.362 (0.246)

Note. iCL = isolated cleft lip only (movement precluded regional measures of gray matter volume for both participants). iCLP = isolated cleft lip and palate. H = Hemisphere. Sup = Superior. Long = Longitudinal. FOS = Fronto-Occipital Fasciculus. Inf = Inferior. Ant = Anterior. Post = Posterior. RS = Retrolenticular Segment. Temp = Temporal. Occ = Occipital. Mid = Middle. Myelin intensity is a ratio of T1w:T2w images, where higher values reflect higher myelin intensity.

Discussion

Despite the high rate of speech and language disorders among children with oral cleft and growing evidence of an association to brain structure and function, very little is understood about the role that neural development plays in early language development nor the relationship to potential etiological factors such as exposure to anesthesia and sleep disordered breathing. The current study obtained multiple measures of language function in tandem with measures of oxygenation and structural brain imaging prior to exposure to anesthesia. This design permitted the earliest evaluation of neural development and the potential impact of sleep disordered breathing without the confound of surgery. While the sample size is small and results are interpreted with caution, patterns across measures: i) reflect the importance of measuring multiple aspects of language development, ii) suggest more frequent desaturations for participants with palatal involvement, iii) replicate patterns of neural differences at 2 months of age, prior to exposure to anesthesia (See Supplemental Table 2).

Emergent Language and Vocal Acoustics

Research has consistently demonstrated a high need for speech and language services among children with iCP±L.⁶⁶ While the need is lower for children without palatal involvement, emerging works suggests that rates of articulation disorders and language delays are still higher for those with iCL than unaffected peers.⁶⁷ The assessment of clinically-meaningful receptive and expressive language skills is limited until around 12 to 16 months of age. As the current sample was tested at 2 months of age, it is not surprising that proxy ratings were predominately in the average range. When evaluating cleft type patterns, ratings of expressive language showed the expected pattern of lower scores for participants with iCL/P.

A recent meta-analysis⁶⁶ found that when comparing groups on norm-referenced tests, effect sizes of differences were higher than when using proxy-ratings. However, as production of speech has not yet started, normed-referenced tests are not available and vocal acoustics of pre-speech sounds are the only measures for infants with cleft at this age. The current study measured two acoustic characteristics reflecting maturity of vocal control (i.e., SI and MCI-II). Infants with iCLP made the most segmentation errors (SI) and had the fewest complex melodies (MCI-II), reflecting the least mature vocal control. In contrast, infants with iCL performed the best on both measures, reflecting the more mature vocal control than UA participants. This pattern is consistent with previous research using these measures⁶⁸ showing more vocal immaturity among infants with cleft when the palate is involved.

Sleep Disordered Breathing

Clinically, measures of sleep disordered breathing are usually obtained through sleep studies conducted after palatal surgery in the presence of symptoms of obstructive sleep apnea. Little is known about sleep disordered breathing among infants with cleft over the first year of life. While home monitoring systems are not as detailed as sleep studies, important information on oxygen saturation can be easily obtained. The current study focused on the frequency and degree of desaturation events. Participants with iCLP had the most time at/above SpO2, but the most desaturation events per hour and the longest durations. While the frequency of events (9.66 per hour) was nearly three times that of UA participants (2.86 per hour), it was under the upper limit for the normal range (15.10 per hour).⁶⁹ Previous work has reported associations between report habitual snoring (another symptom of sleep disordered breathing) and speech problems among children with iCL/P.⁷⁰ While the size of the current study precluded assessment of possible associations between increased desaturations and measures of language, findings demonstrate the importance of evaluating oxygenation, the potential usefulness of these metrics, and the necessity of collecting more data to better understand patterns of oxygenation and how they impact early development.

Neural Structure

Global measures of brain volume were similar between participants with iCLP and UA participants. The largest differences were increased bilateral frontal white matter and decreased cerebellar tissue for those with iCLP. Previous research on neural structure in patients with oral clefts has reported regional volumes or gray-to-white matter ratios without exploring hemisphere interactions. This makes comparison of findings difficult. However, the pattern is consistent with previous findings of higher cerebral volumes among children with iCL/P at school-age³⁶ and reduced cerebellar volumes across all ages.³⁷

Further, myelin intensity was lower for infants with iCLP. This is similar to a recent study finding lower fractional anisotropy among cerebellar and brainstem white matter tracts in children, adolescents, and young adults with cleft.⁴⁰ This suggests delayed white matter development and potential inefficiency of neural connectivity. While findings for participants with iCL should not be interpreted due to the limited sample size, including their metrics provides information for future researchers and possible meta-analyses.

It is uncertain what impact sleep disordered breathing may have on neural development and at what level disruption becomes clinically significant. The combined findings of increased desaturations and altered white matter and cerebellar volumes among participants with iCLP lends support to a potential connection between sleep disordered breathing and brain development. What is unclear is if the potential role of respiration could be driving some of the neural differences, or if there is an additional developmental etiology driving subtle differences in brain development and myelination that starts in-utero and is present before exposure to anesthesia.

Presence of Differences Prior to Anesthesia Exposure

A key finding from this study was the presence of neural differences before participants were exposed to anesthesia. Further research with pre- and post-surgery visits needs to be conducted to fully evaluate the potential role of anesthesia, but the current study suggests that it may not be a primary factor. There is also the possibility that later exposure to anesthesia may interact with other factors. For example, children with obstructed airways are at a higher risk for oxidative stress^71,72 and hypoxemia⁷³ when exposed to anesthesia. This could result in a “double hit” on neural development, where participants with sleep disordered breathing could be at higher risk during anesthesia exposure.

Limitations

As discussed, there are several aspects of the current study that limit generalizability and interpretation of results. In addition to the small sample size limiting power (especially evaluating subgroup differences), the sample was predominately White. Additionally, three infants born at 36^4/7-36^6/7 (late preterm) were included. This decision was made to make the sample more representative of infants with cleft in the general population, as there is a higher rate of oral cleft among infants born prematurely. However, this also introduces possible confound of premature birth. The use of a clinical-grade home monitor for measuring overnight oxygen levels is not as robust or reliable as a full sleep study would have been. However, it was more feasible to implement and provided helpful information and identified areas of potential concern. Finally, not all participants were able to complete the full battery of evaluations, resulting in differing sample sizes across analyses.

Implications for Clinical Practice and Future Research

Because of the nature of this study, any implications for clinical practice need to be supported by further research. While vocal acoustic measures are informative at this age for research purposes, the data do not have normative references or clinical relevance and should not be used to guide treatment decision-making. However, the increased desaturation events for participants with iCLP does highlight the importance of screening for respiratory issues early.

The findings from this study should be viewed as a framework for building future protocols. The pattern of findings supports the need to continue incorporating various measures of language and speech development – through proxy ratings, task-based tests, and experimental tasks – to obtain a clear picture of which skills are developing as expected, which are delayed, and how best to implement interventions. The timing of visits should happen before and after exposure to anesthesia and measures of sleep disordered breathing and neural structure and function should be included. Care will need to be taken in choosing measures that can be compared across ages and developmental stages. Finally, studies should include racially and ethnically diverse male and female participants with and without palatal involvement.

Conclusion

The current study demonstrated feasibility of a study design that collects measures of emergent language, pre-speech acoustics, sleep disordered breathing, and neural structure among infants with iCL/P prior to exposure to anesthesia. Initial patterns of lower language ratings and vocal acoustics with higher desaturations provide support for the need to evaluate the potential roles of early development and sleep disordered breathing on speech and language development among children with oral cleft. Understanding how early aspects of having a cleft palate and the role treatment decisions may play in later speech and language functioning will be essential in guiding clinician treatment decisions related to hypoxia, language development, and surgical planning.