Emergence of Prevocalic Stop Consonants in Children With Repaired Cleft Palate

Abstract

Purpose

This study determined the time course of the emergence of prevocalic stop consonants in young children with cleft palate following surgical repair.

Method

A total of 120 children in four cohorts from three institutions were followed from 12 to 24 months of age: (a) 24 with repaired cleft lip and palate (CLP), (b) 36 with repaired cleft palate only (CP), (c) 33 without clefts but with histories of frequent otitis media and ventilation tubes (OM), and (d) 27 typically developing (TD) children without clefts or OM. Emergence of prevocalic stops and symbolic language skills were determined during administration of the Communication and Symbolic Behavioral Scales Developmental Profile. Parametric survival models were fitted with and without covariates—recruitment site, gender, maternal education level, middle ear status, language ability, and age at surgery for children with clefts—to describe the time course of the emergence of prevocalic stops.

Results

The estimated age at which 80% of children demonstrated prevocalic stop emergence was 15.0, 15.3, 18.9, and 21.8 months for TD, OM, CP, and CLP groups, respectively (p < .001, unadjusted model). Both CP and CLP cohorts had a significantly longer time to stop emergence than either the TD or OM cohorts, even after adjusting for covariates. Abnormal middle ear status, lower symbolic language ability, and older age at palatal surgery were significantly associated with delayed stop emergence.

Conclusions

Survival model estimates show that four out of five children with repaired cleft palate will achieve emergence of prevocalic stop consonants by 19–22 months of age, corresponding to 9–12 months following palate repair. Clinical implications are discussed.

Children born with cleft palate are at a significant risk for delays in development of early speech sounds, especially stop consonants. Prior to repair of palatal clefts, infants typically show limited production of oral stops and increased production of nasal and glottal articulations compared to noncleft peers (e.g., Chapman, 1991; Chapman et al., 2001; O'Gara & Logemann, 1988). This is not surprising given that an unrepaired cleft of the secondary palate will effectively reduce, if not eliminate, the ability of an infant to generate adequate levels of oral air pressure required for the production of stop consonants. As further highlighted by Chapman (2009), however, even following palate repair, the “acquisition of age-appropriate speech is not achieved immediately or quickly for many children” (p. 255). Indeed, clinicians often encounter parents who are worried because their child is not using stop consonants such as /b/ or /d/ 4–6 months following palate repair at chronological ages of 14–16 months.

Although clinicians reassure parents that children with repaired cleft palate will catch up if velopharyngeal function and hearing are adequate, little empirical evidence exists to inform parents and guide clinical decisions. Jones et al. (2003) investigated 14 infants with cleft palate at 12 months of age before surgery and at 17 months of age, following surgery. Jones et al. reported modest gains for the children at 5 months post surgery relative to size of consonant inventories, canonical babbling ratios, and production of the stop consonant /b/. The production of /b/ is considered an important predictor of the success of palatal surgery given this is an early developing phoneme that requires velopharyngeal closure. Compared to a control group of 17-month-old infants without cleft palate, however, Jones et al. also reported that the infants with clefts were lagging behind in production of (a) total number of consonants, (b) overall frequency of stop consonants, and (c) the accuracy of stop consonants. In a longitudinal study, Chapman et al. (2003) followed children with cleft palate from 9 months of age prior to surgery to 21 months of age following surgery. They reported that, although the children with clefts made gains in speech acquisition following surgery, they remained behind noncleft peers in production of pressure consonants. In another longitudinal study, Vallino et al. (2019) followed 53 children with repaired cleft palate from 12 to 18 months of age. These children were a subset of children followed in this study. Vallino et al. reported that only 55% of the children at 18 months of age produced prevocalic /b/ compared to 82% of noncleft peers. These studies underscore that children with repaired cleft palate continue to lag behind their noncleft peers well into the second year of life relative to the production of stop consonants.

The reasons for delayed emergence of oral stop consonants following palate repair are not entirely clear. Obviously, if surgery does not achieve its goal of palatal reconstruction and levator muscle function, then velopharyngeal insufficiency may be present and will hinder the emergence of oral stops. In general, it has been estimated that approximately 20% of initial palatal surgeries fail resulting in velopharyngeal insufficiency (Kuehn & Moller, 2000). Children with cleft palate also experience high rates of otitis media (OM) with effusion (Flynn et al., 2009; Goudy et al., 2006; Muntz, 1993; Paradise et al., 1969). Thus, frequent and/or prolonged episodes of conductive hearing loss may also be a factor. Shriberg et al. (2000), for example, reported that children without cleft palate who had average hearing thresholds greater than 20 dB at 12–18 months of age experienced a significantly higher risk of clinical or subclinical speech delay at 36 months of age compared to children who had less than 20 dB average thresholds at 12–18 months of age. Although there are mixed results regarding the effects of OM on early speech development of children without cleft palate (see Lim, 2007, for a review), the combined effects of fluctuating conductive hearing loss, varying degrees of velopharyngeal insufficiency, and other oral cavity anomalies such as maxillary arch defects may place young children with cleft palate at a greater risk for delayed emergence of oral stops.

Lastly, assuming that palatal surgery is successful, do young children with cleft palate simply require some amount of time to practice and master production of stop consonants? As emphasized by Stoel-Gammon (1998), speech is a skilled motor activity that requires practice and feedback to develop. Young children are also still developing internal phonological representations during this time, which requires a combination of caregiver modeling, imitation, and feedback (Stoel-Gammon, 2011). Because children with cleft palate are at a higher risk for language and phonological delays (e.g., Eshghi et al., 2019; Scherer, 1999), the delayed emergence of oral stops may be driven by a combination of both language and phonological deficits as well as delayed motor practice with an adequate velopharyngeal mechanism following palate repair. In essence, perhaps “palatal age” is a better indicator of oral stop emergence than chronological age.

The purpose of this prospective multisite study was to determine the age when prevocalic stop consonants emerge (as a phoneme class) in children with repaired palatal clefts, with and without cleft lip, compared to noncleft peers. Two groups of children without clefts were studied for comparison: (a) children with early histories of frequent OM with effusion that required ventilation tubes and (b) typically developing (TD) children without early histories of frequent OM. In addition, we also explored the effects of gender, middle ear status, maternal education level, symbolic language ability, recruitment site, and age at palatal surgery for children with clefts on the emergence of stop consonants.

Method

Participants

Participants consisted of 120 children, 12–14 months of age at the time of enrollment, in four cohorts: (a) 24 children with repaired cleft lip and palate (CLP), (b) 36 children with repaired cleft palate only (CP), (c) 33 children with OM, and (d) 27 TD children. The children were enrolled at three different sites in the United States—the University of North Carolina at Chapel Hill (UNC); Nationwide Children's Hospital, Columbus, OH; and Alfred I. duPont Hospital for Children, Wilmington, DE—and followed up to 24 months of age. Children with CLP and CP were recruited through each institution's cleft/craniofacial clinic; children with OM were recruited through the institutions' ear, nose, and throat (ENT) clinics. TD participants were recruited through a combination of methods including flyers posted in the community, e-mail, and word of mouth. Across the three recruitment sites, approximately 60% of parents who had eligible children with repaired CLP or CP agreed to enroll their child in the study.

Inclusion criteria for all children were as follows: (a) at least 36 weeks of gestation, (b) from monolingual American English–speaking families, (c) good general health, (d) no documented sensorineural hearing loss, and (e) no documented global developmental delays. Additional inclusion criteria for children with CLP and CP were (a) single-stage palate repair before 14 months of age, (b) no oronasal fistulae, and (c) no known syndromes or Pierre Robin sequence. The average age of palate repair was 10.5 months across all sites: 9.3 months (range: 7.2–13.1 months) at UNC, 10.7 months (range: 9.1–12.6 months) at Nationwide Children's Hospital, and 11.6 months (range: 10.7–13.8 months) at Alfred I. duPont Hospital for Children. Although all sites adhered to a protocol of early palate repair, usually by 12 months of age, some children had later surgeries due to various factors including poor weight gain or social challenges scheduling families (e.g., transportation, child care). All but two of the 60 children with palatal clefts had undergone bilateral myringotomies with insertion of ventilation tubes at the time of palate surgery. All OM participants had three or more episodes of OM with effusion during the first year of life that required bilateral myringotomies with insertion of ventilation tubes. All TD participants had typical speech and language at 12 months of age as determined by the Communication and Symbolic Behavior Scales Developmental Profile (CSBS DP; Wetherby & Prizant, 2002; described below). The study was approved by the institutional review boards at the respective sites, and parents of all participants provided signed informed consent. Table 1 shows demographic characteristics of the four cohorts relative to recruitment site, gender, and race.

Table 1.

Demographic characteristics of the four cohort groups (N = 120 children).

Variable	CLP (n = 24)	CP (n = 36)	OM (n = 33)	TD (n = 27)	Overall (N = 120)
Site, n (%)
UNC	10 (42)	12 (33)	7 (21)	9 (33)	38 (32)
AID	3 (12)	12 (33)	17 (52)	10 (37)	42 (35)
NWC	11 (46)	12 (33)	9 (27)	8 (30)	40 (33)
Gender, n (%)*
Male	19 (79)	12 (33)	24 (73)	13 (48)	68 (57)
Female	5 (21)	24 (67)	9 (27)	14 (52)	52 (43)
Race, n (%)
Black/African American	1 (4)	3 (8)	1 (3)	2 (7)	7 (6)
White	23 (96)	25 (69)	29 (88)	22 (82)	99 (82)
More than one race	0 (0)	8 (22)	3 (9)	3 (11)	14 (12)

Note. CLP = cleft lip and palate; CP = cleft palate; OM = otitis media; TD = typically developing; UNC = University of North Carolina at Chapel Hill; AID = Alfred I. duPont Hospital for Children; NWC = Nationwide Children's Hospital.

^*Pearson chi-squared test: p < .05.

Procedure

Participants were seen up to 7 times between the ages of 12 and 24 months at 2-month intervals (i.e., 12, 14, 16, 18, 20, 22, and 24 months). If a child with cleft palate did not have surgery before 12 months of age, the first visit started at 14 months of age. Most participants were seen for at least four study visits at 12, 14, 18, and 24 months of age. Per the study protocol, some participants, mostly children with CLP and CP, were seen for additional visits at 16, 20, and/or 22 months of age depending on the emergence of stops (described below). The following assessments were performed at the visits.

Oral Examinations

All children with CLP and CP had oral examinations at each study visit to rule out the presence of oronasal fistulae and/or surgical dehiscence of the repaired palate. Oral examinations were conducted on OM and TD children at the time of enrollment to rule out bifid uvula and/or submucous cleft palate. Experienced craniofacial speech-language pathologists (SLPs) at each site conducted the oral examinations.

Tympanometry

Bilateral tympanograms were obtained at each study visit to assess middle ear function of all children. Normal middle ear function was defined as either a Type A tympanogram (normal compliance and pressure) or a flat Type B tympanogram with a large volume (greater than 1.0 ml when a ventilation tube was present) for both ears. Abnormal middle ear function was defined as either a Type A_s reduced compliance tympanogram (peak less than 0.2 ml), a flat Type B tympanogram with a small volume (less than 1.0 ml), or a Type C tympanogram with negative pressure (less than −250 daPa) for either ear. Middle ear status as documented at the first study visit (either 12 or 14 months of age) was used as a covariate in the study.

Hearing Assessments

Audiometric evaluations were attempted for all children at the time of the first study visit (either 12 or 14 months of age) and at 24 months of age. The assessments consisted of either sound-field testing or screenings at 20 dB HL across the frequencies of 0.5, 1.0, 2.0, and 4.0 kHz. Certified audiologists at the child's respective institution conducted all assessments. Because hearing was only screened in some cases and assessments were not obtained for all children (reported below), hearing was not included as a covariate in the study.

CSBS DP

The CSBS DP was conducted at each study visit for all participants to determine the emergence of prevocalic stops and symbolic language ability. The CSBS DP takes approximately 30 min and consists of six semistructured communication opportunities that involve the child, parent, and examiner. It assesses social communication; use of gestures, sounds, and words; and symbolic language of children up to 24 months of age. A standardized set of toys was used to elicit vocalizations, words, and conversational interactions during the CSBS DP. The toys targeted each of the stop consonants /p b t d k g/ in the initial and final word positions.

Trained study examiners at each site administered the CSBS DP. These were either experienced craniofacial SLPs or master's-level developmental specialists. Training consisted of watching videos provided with the CSBS DP manual followed by two practice administrations with children not in the study to ensure test administration fidelity. All administrations of the CSBS DP were video-recorded for later scoring at the lead site (UNC).

Six trained “stop consonant trackers” at UNC independently viewed the CSBS DP videos to document the occurrence of prevocalic stops. Two of the trackers were SLPs, and four were graduate students (either second year master's or doctoral level) in speech and hearing sciences. A relatively large number of trackers were used due to the need to expedite documentation of the emergence of stops and determine subsequent study visits (described below). All trackers completed training that consisted of independently viewing a 6-min sample of a CSBS DP video and identifying all clearly produced prevocalic oral stops. All trackers had to achieve at least 80% agreement with the first author, an experienced craniofacial SLP, on the total number of stops. Only prevocalic stops were documented for two reasons. First, consonant–vowel syllables are among the earliest and most frequent syllable type produced by infants and young children (Oller, 1986; Stoel-Gammon, 1985). Second, as part of training, we found it difficult, at times, to reliably determine the occurrence of word-final (postvocalic) stops if the child did not release the oral occlusion. We considered the child to display evidence of the emergence of stops if at least three prevocalic stops were produced during the 30-min sample. This could consist of the same prevocalic stop produced at 3 different times or any combination of prevocalic stops that totaled three productions. Study visits continued until a child achieved two consecutive visits with demonstrated emergence of prevocalic stops. Regardless of when stops emerged, all children returned for study visits at 18 and 24 months of age as part of the study's protocol.

Following determination of stop emergence, three trained SLPs at UNC independently scored the videos to obtain a total CSBS DP score and three composite scores—social communication, speech, and symbolic language. The symbolic language score (SLS) reflects a combination of receptive understanding and object use. It is a standard score with a mean of 10 and an SD of 3. SLSs obtained at the first study visit (either 12 or 14 months of age) were used as a covariate in the study.

Agreement of Scorers

Emergence of Stops

Interscorer and intrascorer agreements of the six stop consonant trackers were assessed by randomly selecting 20% of study visits and repeating the scoring. All six of the stop trackers demonstrated 100% intrascorer agreements relative to the occurrence of at least three prevocalic stops. Five of the six stop trackers also demonstrated 100% interscorer agreements; one stop tracker demonstrated 97% interscorer agreement.

SLSs

Interscorer and intrascorer agreements of the three SLPs were also assessed by randomly selecting 20% of study visits and repeating the symbolic language scoring. Percentages of agreements were calculated by dividing the smaller score (either original or repeated) by the larger score and multiplying by 100. All three SLPs demonstrated at least 83% interscorer and intrascorer agreements.

Statistical Analysis

The primary analysis involved the fitting of parametric survival models to the time-to-event variable of emergence of stops after two consecutive visits that is interval censored (Allison, 2010, Chapter 4); in other words, the day that stops emerged was not observed, but it is assumed to have occurred within the interval defined by the two consecutive visits. Because we set the time origin at 12 months of age in our analysis, stop emergence was assumed to have occurred for each child after 12 months of age. The general form of the accelerated failure time model is

$$log{T}_i={\mathrm{\beta }}_0+{\mathrm{\beta }}_1{x}_{i1}+\dots {\mathrm{\beta }}_{\mathit{ik}}{x}_{\mathit{ik}}+\mathrm{\sigma }{\mathrm{\epsilon }}_{\mathrm{i,}}$$ (1)

where T_i is the time of stop emergence in months (following 12 months of age) assumed to have a continuous parametric distribution for all subjects; x_ij , j = 1, …, k, are covariates for the ith individual, ε _i is a random disturbance term, and σ is the scale parameter. Two general models were considered, one where T_i has a Weibull distribution and the other where T_i has a log-logistic distribution. These two were chosen because their survival functions can be inverted to solve for the age at which a given percentage of children achieved stop emergence. The better model between the two distributions was chosen based upon Akaike information criterion (see Results section). To compare “survival” times (i.e., time to stop emergence) between cohorts, models with and without covariates were fitted, where covariates were (a) study site; (b) SLS during the first visit following surgery (age of 12 or 14 months) centered at the mean score of TD children; (c) an indicator variable for female child gender; (d) an indicator variable for abnormal middle ear status during the first visit (1 = one or two ears affected vs. 0 = no ears affected, based on bilateral tympanometry); (e) an indicator variable for low maternal education level (1 = less than bachelor's degree vs. 0 = at least bachelor's degree); and (f) age at surgery for children with clefts, which was centered at the mean (10.5 months), or in the case of noncleft children set to 0. Exponentiated regression coefficients were generated as estimates of the relative mean survival times between groups (for categorical predictors) or the multiplicative effect of a unit increase in the covariates for SLS and age at surgery. The survival function for each cohort was inverted to estimate the expected age at which a fixed percentage of children achieved stop emergence. Covariate-adjusted models utilized “effect coding,” that is, ANOVA-like coding such as 1 = female and −1 = male, to obtain estimates for an “average” child. Specifically, we estimated the age at which 50%, 60%, 70%, and 80% of children achieved stop emergence. The latter criterion of 80% provided the primary end point specified a priori in the study protocol. We also computed the 95% confidence intervals (CIs) of these estimates based on standard application of the delta method. Statistical significance was defined as a p value < .05. All models were fit with Proc Lifereg in SAS Version 9.4 (TS1M1 SAS Institute Inc.).

Results

Descriptive Statistics

Table 2 shows the characteristics of the four cohorts relative to the covariates. The table also shows the social and speech composite scores from the CSBS DP; these are presented as additional information but were not included in the models. Maternal education level (at least bachelor's vs. less than bachelor's degree) and middle ear status (normal vs. abnormal) were categorical variables; symbolic language, social, and speech composite scores were continuous variables. Most mothers of the children (76/120 or 63%) had a 4-year college degree (p < .001); however, there were differences across cohorts. While 50% (12/24) and 42% (15/36) of mothers of children with CLP and CP, respectively, had a 4-year college degree, 76% (25/33) and 89% (24/27) of mothers of OM and TD children, respectively, had a 4-year college degree. Regarding middle ear status, 51 children (42%) had abnormal tympanograms in one or both ears at the first visit (12 or 14 months of age); at 24 months of age, 31 children (30%) had abnormal tympanograms in one or both ears. As shown in Table 2, at the first study visit, children in the CLP, CP, and OM cohorts had similarly modest rates of abnormal tympanograms (28%–37%) while TD children had the highest rate (70%). At 24 months of age, all cohorts had generally similar rates of abnormal tympanograms (21%–38%). Although SLSs at the first visit were slightly higher for the TD and OM cohorts, there were no significant group differences. In addition, as indicated by the standard deviations in Table 2, most children in all cohorts scored within normal limits on symbolic language. As further shown in Table 2, SLSs continued to be similar, although higher, across the cohorts at 24 months of age (Visit 7). The social and speech composite scores had patterns similar to symbolic language, with the TD and OM cohorts having slightly higher scores and all cohorts having higher scores at 24 months of age.

Table 2.

Maternal education, middle ear status (MES), and Communication and Symbolic Behavior Scales Developmental Profile (CSBS DP) composite scores of the four cohort groups (N = 120 children).

Variable	CLP (n = 24)	CP (n = 36)	OM (n = 33)	TD (n = 27)	Overall (N = 120)
Maternal education, n (%)*
At least bachelor's	12 (50)	15 (42)	25 (76)	24 (89)	76 (63)
Less than bachelor's	12 (50)	21 (58)	8 (24)	3 (11)	44 (37)
MES at the first visit, a n (%)**
Normal (0 ears aff.)	15 (63)	26 (72)	20 (61)	8 (30)	69 (58)
Abnormal (1–2 ears)	9 (37)	10 (28)	13 (39)	19 (70)	51 (42)
MES at 24 months of age, b n (%)**
Normal (0 ears aff.)	15 (71)	21 (70)	22 (79)	16 (62)	74 (70)
Abnormal (1–2 ears)	6 (29)	9 (30)	6 (21)	10 (38)	31 (30)
CSBS DP standard scores at the first visit, a M (SD)
Symbolic language	7.8 (2.5)	7.7 (2.7)	8.2 (2.1)	8.1 (2.7)	7.9 (2.5)
Social	9.6 (2.4)	8.4 (2.2)	9.6 (2.0)	10.0 (2.5)	9.3 (2.3)
Speech	9.2 (2.1)	9.3 (2.1)	10.8 (2.1)	10.2 (2.4)	9.9 (2.3)
CSBS DP standard scores at 24 months of age , c M (SD)
Symbolic language	10.3 (2.7)	8.6 (2.8)	8.9 (3.2)	10.6 (3.2)	9.5 (3.1)
Social	11.9 (3.3)	10.5 (3.5)	10.1 (2.1)	10.3 (3.4)	10.6 (3.1)
Speech	10.2 (2.2)	10.7 (2.4)	11.0 (3.0)	12.3 (2.4)	11.1 (2.6)

Note. CLP = cleft lip and palate; CP = cleft palate; OM = otitis media; TD = typically developing; aff. = affected.

^a The first visit occurred at 12 or 14 months of age.

^b Fifteen children have missing tympanograms at 24 months of age.

^c Twelve children have missing symbolic language scores at 24 months of age.

^*Pearson chi-squared test: p < .001.

^**Pearson chi-squared test: p < .01.

The high rate of abnormal middle ear function at the first visit for TD children was driven by the frequent occurrence of Type A_s tympanograms in this cohort. Of the 19 TD children with abnormal middle ear status shown in Table 2, nine (47%) presented with Type A_s tympanograms. Although the cause of this is not entirely clear, impacted cerumen may have been a factor. As noted by Katz (1978), impacted cerumen “will give misleading information regarding the status of the middle ear” (p. 10). We did not perform otoscopic examinations on children to detect cerumen. However, some parents reported problems with excessive earwax when told their child had a Type A_s tympanogram. We also encouraged these parents to seek ENT evaluations if they had concerns with hearing. Based on the finding of substantially reduced abnormal tympanograms at 24 months of age for TD children, it is likely that some parents followed up on this recommendation. Of note, relatively few children with cleft palate or OM presented with Type A_s tympanograms, quite likely due to the presence of ventilation tubes and regular ENT evaluations. To be sure, the overall modest rates of abnormal middle ear findings among the children with cleft palate are testimony to early ventilation tubes as standard of care.

Parametric Survival Models

Unadjusted Models

The Weibull model had a better fit than the log-logistic model based on smaller values of Akaike information criterion, so the Weibull model results are reported. There were significant differences in time to stop emergence among cohort groups in the unadjusted model (p = .002). In particular, both the CLP and CP groups had a statistically significantly longer time to stop emergence than the TD and OM groups, whereas there was no significant difference between the TD and OM groups or between the CLP and CP groups. Moreover, the estimated age at which 80% of children had stop emergence was 15.0, 15.3, 18.9, and 21.8 months for TD, OM, CP, and CLP children, respectively (see Table 3); the table also shows the 95% CIs for the estimated ages.

Table 3.

Estimated age in months at which a fixed percentage of children in a cohort attained stop emergence (95% confidence intervals), based on the Weibull model unadjusted (first row) and adjusted (second row) for covariates.

Cohort	50%	60%	70%	80%
CLP	15.6 (13.8, 19.2)	17.0 (14.5, 22.2)	18.9 (15.3, 26.4)	21.8 (16.6, 32.9)
	16.6 (14.2, 21.8)	18.1 (14.8, 25.0)	19.9 (15.6, 29.3)	22.5 (16.7, 35.5)
CP	14.5 (13.3, 16.8)	15.5 (13.9, 18.8)	16.9 (14.5, 21.5)	18.9 (15.5, 25.8)
	14.8 (13.5, 17.5)	15.7 (13.9, 19.3)	16.9 (14.4, 21.7)	18.5 (15.2, 25.2)
OM	13.2 (12.6, 14.3)	13.7 (12.9, 15.2)	14.3 (13.2, 16.5)	15.3 (13.7, 18.4)
	13.4 (12.7, 14.7)	13.9 (13.0, 15.5)	14.4 (13.3, 16.6)	15.2 (13.7, 18.2)
TD	13.1 (12.3, 15.8)	13.5 (12.4, 17.3)	14.1 (12.6, 19.4)	15.0 (12.9, 22.6)
	12.9 (12.2, 15.6)	13.2 (12.3, 16.7)	13.6 (12.4, 18.2)	14.1 (12.6, 20.3)

Note. CLP = cleft lip and palate; CP = cleft palate; OM = otitis media; TD = typically developing.

Adjusted Models

Results from the covariate-adjusted Weibull model confirmed results from the unadjusted model, namely, that there were statistically significant differences in the time to stop emergence between the cleft groups (CLP, CP) versus the noncleft groups (TD, OM; see Table 4). For example, the time to stop emergence for a child with CLP beyond 12 months of age was estimated to be 4.90 times as long compared to a TD child (95% CI [2.26, 10.58], p < .001); for a child with CP, it was 3.01 times as long compared to a TD child (95% CI [1.52, 5.96], p = .002). Furthermore, there were no statistically significant differences in the time to stop emergence between the OM and TD groups or between the CLP and CP groups, adjusting for covariates. Importantly, the comparisons involving children with a cleft are made assuming age at surgery of 10.5 months. Table 4 shows that time to stop emergence among children with a cleft was estimated to be 1.36 times longer for an increase of 1 month in age at surgery. Therefore, multiplicative effects such as those cited immediately above for comparing the cleft versus noncleft groups are even larger when age at surgery is greater than 10.5 months.

Table 4.

Estimated ratios of expected time to stop emergence and 95% confidence intervals (CIs) based on the covariate-adjusted Weibull model (N = 120 children).

Variable	Mean survival ratio	95% CI	p value
Study site (reference = UNC)			.157
AID	1.79	[0.98, 3.24]	.056
NWC	1.32	[0.77, 2.28]	.311
Female gender	1.30	[0.81, 2.11]	.281
Low maternal education	0.95	[0.60, 1.49]	.809
Abnormal MES	1.87	[1.18, 2.96]	.008
Symbolic language score	0.91	[0.84, 0.99]	.037
Age at surgery (months) a	1.36	[1.07, 1.73]	.012
Cohort (reference = TD)			< .001 b
CLP	4.90	[2.26, 10.58]	< .001
CP	3.01	[1.52, 5.96]	.002
OM	1.50	[0.80, 2.84]	.208
Additional contrasts
CLP vs. CP	1.62	[0.88, 3.01]	.124
CLP vs. OM	3.25	[1.71, 6.18]	< .001
CP vs. OM	2.00	[1.08, 3.72]	.027

Note. UNC = University of North Carolina at Chapel Hill; AID = Alfred I. duPont Hospital for Children; NWC = Nationwide Children's Hospital; MES = middle ear status; TD = typically developing; CLP = cleft lip and palate; CP = cleft palate; OM = otitis media.

^a Applies only to children with clefts.

^b Overall 2 df test.

The estimated ages at which 50%, 60%, 70%, and 80% of children had stop emergence based on the Weibull model adjusted for covariates are also shown in Table 3 (see rows below unadjusted estimates). Similar to the unadjusted model, 80% of the children with CLP and CP had stop emergence at older ages (22.5 and 18.5 months, respectively) compared to the children with OM and TD children (15.2 and 14.1 months, respectively).

The covariate-adjusted Weibull model (see Table 4) further shows that there were no statistically significant differences in time to stop emergence according to gender, maternal education, or study site. On the other hand, poorer symbolic language ability corresponded to a delay in emergence of stops. By inverting the mean ratio and its lower and upper confidence limits, the time to stop emergence corresponding to a unit decrement in SLS increased by a factor of 1/0.91 = 1.10 (95% CI [1.01, 1.19], p = .037). Additionally, the time to stop emergence for children with abnormal middle ear status was estimated to be 1.87 times as long as the time to stop emergence for children with normal middle ear status (95% CI [1.18, 2.96], p = .008). Adjusting for the relative disadvantage of the TD children in the study owing to a higher frequency of abnormal tympanograms at the first visit had the apparent effect to decrease their time to stop emergence, especially at the 70% and 80% quantiles, thereby increasing the gap in time to stop emergence when compared to children with CLP and CP.

Discussion

The results of this study further confirm that children with cleft palate lag behind their noncleft peers in the emergence of oral stop consonants following palate repair. More importantly, the study shows that the emergence of stops is a protracted process that may take up until 19–22 months of age for some children, depending upon various factors.

The first, and most obvious, factor is status of the velopharyngeal mechanism. If surgery has not successfully reconstructed palatal structures—which may occur due to numerous factors including different surgeons and reconstruction techniques (i.e., intravelar veloplasty vs. double-opposing z-plasty) and differences in cleft severity among children—then oral stops are likely not to emerge in a timely fashion. In this study, there were only three children (5%) with palatal clefts who did not achieve stop emergence by 24 months of age. These children are likely to have some degree of velopharyngeal insufficiency and are currently being followed clinically. In addition, there were 10 children (17%) with palatal clefts who did not achieve stop emergence until 20 months of age or later. Physiologic information relative to velopharyngeal function was available for six of these 10 children. This consisted of nasal ram pressure (NRP) data obtained during production of oral stops at the time of emergence. NRP monitoring is a procedure that determines the binary status of the velopharyngeal port as open or closed by detecting pressure at the exit of the nose (Bunton & Hoit, 2018; Thom et al., 2006). It is a fairly well-tolerated procedure by young children. The six children in this study who had late stop emergence successfully tolerated NRP monitoring (see Eshghi et al., 2017, for details of the procedure). Four of the six children (67%) showed velopharyngeal closure on less than 50% of the oral stops that they produced; one child had closure on 33% of the stops, while the other three had 0% closure. It should be noted that NRP does not provide information on the size of the velopharyngeal gap and these four children may have had relatively small openings. Nonetheless, these limited data suggest that some degree of velopharyngeal insufficiency may have prolonged, but not prevented, the emergence of oral stops in these children.

A second factor may be middle ear status and hearing. In this study, the survival model suggests that it may take a child with or without cleft palate who has abnormal middle ear function at baseline 1.87 times longer to achieve stop emergence compared to a child with normal middle ear function. If, indeed, the emergence of stops requires phonological learning and motor practice with adequate feedback as suggested by Stoel-Gammon (1998) and others (e.g., Guenther, 2006), then degraded auditory feedback during critical periods of learning and practice may delay the emergence of stops for some children. Relative to infants with cleft palate, Hardin-Jones et al. (2002) reported some evidence to support this. They investigated the effects of palatal obturation prior to surgery on consonant development in infants with cleft palate. Hardin-Jones et al. reported no statistically significant difference between groups of infants who did and did not undergo obturation. However, when all infants were categorized according to tympanometry screening results, those who failed the screening produced significantly more glottal stops than those who passed the screening. Hardin-Jones et al. suggested that this implicated the role of degraded auditory feedback on the development of oral consonants of the infants.

As previously noted, the effect of OM with effusion on the speech development of children without clefts is controversial. As emphasized by Lim (2007), it is surprising that many studies that investigate the effects of OM with effusion on speech development lack audiometric findings. In this study, hearing assessments were attempted at the time of enrollment (either 12 or 14 months of age) and again at 24 months of age. At the time of enrollment, assessments were successfully obtained for 113 of the participants, 57 with palatal clefts (CLP, CP) and 56 without clefts (OM, TD). There was no difference between the number of children with (n = 8) and without (n = 8) palatal clefts who had abnormal hearing, typically mild-to-moderate conductive losses, as reported by an audiologist. At 24 months of age, assessments were successfully completed for 101 of the children, 49 with palatal clefts and 52 without clefts. Eleven (22%) of the children with palatal clefts had abnormal hearing; five (10%) of the children without clefts also had abnormal hearing (p = .067). Although these findings downplay the role of hearing loss in children with palatal clefts, it must be noted that episodes of conductive hearing loss associated with OM with effusion fluctuate over time. Therefore, the infrequent hearing assessments may have missed episodes of hearing loss that occurred at other times.

A third factor that affects the emergence of stops for children with cleft palate is the age at palate surgery. Unfortunately, there is little consensus for this in the United States with different centers and surgeons having their preferred protocol. Although all three sites in this study adhered to a similar schedule of relatively early palate repair, the children's ages did differ across the three sites. Because most of the surgeries appeared to be successful as indicated by few children who did not produce stops at 24 months of age, this provided an opportunity to assess associations with age at surgery and age at stop emergence in cleft children. As indicated in Table 4, a delay in age at surgery of 1 month may result in a 36% increase in the time to stop emergence.

A fourth factor that may affect the emergence of stop consonants is general language ability. As previously noted, children with cleft palate are at an increased risk for language delay (e.g., Eshghi et al., 2019; Scherer, 1999). In this study, symbolic language ability (SLS) was measured during administration of the CSBS DP. This reflected a child's receptive understanding of basic concepts (e.g., body parts, common objects) and object use (e.g., play with a bottle, baby doll, or dishes). As noted by Rescorla and Ratner (1996) and others, limited language skills may afford limited opportunity to develop speech skills. It should be noted, however, that symbolic language had a relatively small effect in this study, with only a 10% increase in age of stop emergence per unit reduction in SLS.

Finally, we must reemphasize that the purpose of this study was to determine the age of children at the time of stop emergence, as a general class of phonemes. Therefore, we did not determine specific phonetic inventories of the children. However, it should be noted that voiced stops, particularly /b/ and /d/, were the stops that emerged for most children regardless of cohort. This is not surprising given that young children initially make no acoustic voicing distinction for stops and, when they do, both /p/ and /b/ fall into the voice onset range for /b/ (Macken & Barton, 1980).

Clinical Implications

Craniofacial SLPs often encounter parents who are concerned regarding limited speech production and lack of oral stop consonants in their child following palate repair. To our knowledge, this study is the first attempt to use parametric survival models to provide evidence-based information on stop emergence that can be used to counsel parents and guide clinical decisions. Based on the present findings, for example, parents can be reassured that it is not a red flag for velopharyngeal insufficiency if a child does not have stop emergence following palate repair by 16 months of age. Conversely, clinicians should be more concerned and proactive once a child reaches 20 months of age without the emergence of stops. In either case, clinicians may choose to promote parent-implemented speech-language stimulation activities such as providing frequent models of target words that begin with stop consonants (i.e., auditory bombardment), visual placement cues, positive reinforcement, and expansion of correctly produced words with oral stops (e.g., Scherer et al., 2008). Clinicians should also refer to early intervention services, especially for toddlers without stops at 20 months of age. In these cases, it may be important to provide specific stimulation recommendations (such as those given to parents) as many clinicians in early intervention do not see children with cleft palate on a regular basis. Furthermore, as always, clinicians should be cognizant of the possible effects of conductive hearing loss and encourage parents to seek medical care if their child shows any signs of middle ear infections.

Limitations

A limitation of this study was the lack of diagnostic hearing assessments at frequent intervals to potentially identify episodes of conductive hearing loss associated with OM. A related minor limitation was the lack of otoscopic evaluations to rule out the presence of impacted cerumen in all children. Finally, despite the strong evidence for differences in the time to stop emergence between children with clefts versus TD children or children with OM, the wide CIs reported in this study suggest that obtaining estimates of the effects of factors associated with time to stop emergence with greater precision would require larger sample sizes.

Summary and Conclusions

This study used parametric survival models to estimate the age of stop emergence in four cohorts of children with and without repaired cleft palate. Estimates from the unadjusted model showed that the age at which four of five children will achieve emergence of prevocalic stops was 15.0, 15.3, 18.9, and 21.8 months for TD children and children with OM, CP, and CLP, respectively. Sex of the child and maternal educational level did not significantly affect the estimates. Age at the time of palatal surgery was a significant factor, with later ages associated with increased age of stop emergence. These findings provide clinicians with evidence-based information to counsel parents and inform clinical decisions.

Funding Statement

Research reported in this publication was supported by the National Institute of Dental & Craniofacial Research of the National Institutes of Health under Award R01DE022566.