Predicting Stuttering Onset by the Age of 3 Years: A Prospective, Community Cohort Study

Transcription

1 ARTICLE Predicting Stuttering Onset by the Age of 3 Years: A Prospective, Community Cohort Study Sheena Reilly, PhD a,b,c, Mark Onslow, PhD d, Ann Packman, PhD d, Melissa Wake, MD a,b,c, Edith L. Bavin, PhD e, Margot Prior, PhD b,f, Patricia Eadie, PhD c, Eileen Cini, BAppSc c, Catherine Bolzonello, MSpPath c, Obioha C. Ukoumunne, PhD a,c a Department of Paediatrics, f School of Behavioural Sciences, University of Melbourne, Melbourne, Australia; b Royal Children s Hospital, Melbourne, Victoria, Australia; c Murdoch Childrens Research Institute, Melbourne, Victoria, Australia; d Australian Stuttering Research Centre, University of Sydney, Sydney, Australia; e School of Psychological Science, La Trobe University, Melbourne, Victoria, Australia The authors have indicated they have no financial relationships relevant to this article to disclose. What s Known on This Subject Stuttering involves repeated movements and fixed postures of the speech mechanism. Stuttering may severely impair communication, leading to speech-related social anxiety and social phobia and failure to attain occupational potential. Although stuttering is a common disorder, the cause remains unknown. What This Study Adds The cumulative incidence of stuttering onset by age 3 years was higher than previously reported. Risk factors for stuttering onset explained little of the variation in stuttering onset. Language delay and shyness were not associated with stuttering onset. ABSTRACT OBJECTIVES. Our goals were to document (1) the onset of stuttering and (2) whether specific child, family, or environmental factors predict stuttering onset in children aged up to 3 years. METHODS. Participants included a community-ascertained cohort of year-old Australian children recruited at 8 months of age to study the longitudinal development of early language. The main outcome measure was parental telephone report of stuttering onset, verified by face-to-face expert diagnosis. Preonset continuous measures of the child s temperament (approach/withdrawal) and language development were available. Information on a range of predictor measures hypothesized to be associated with stuttering onset was obtained (maternal mental health and education levels, gender, premature birth status, birth weight, birth order, twinning, socioeconomic status, family history of stuttering). RESULTS. By 3 years of age, the cumulative incidence of stuttering onset was 8.5%. Onset often occurred suddenly over 1 to 3 days (49.6%) and involved the use of word combinations (97.1%). Children who stuttered were not more shy or withdrawn. Male gender, twin birth status, higher vocabulary scores at 2 years of age, and high maternal education were associated with stuttering onset. The multivariable model, however, had low predictive strength; just 3.7% of the total variation in stuttering onset was accounted for. CONCLUSIONS. The cumulative incidence of stuttering onset was much higher than reported previously. The hypothesized risk factors for stuttering onset together explained little of the variation in stuttering onset up to 3 years of age. Early onset was not associated with language delay, social and environmental factors, or preonset shyness/withdrawal. Health professionals can reassure parents that onset is not unusual up to 3 years of age and seems to be associated with rapid growth in language development. Pediatrics 2009;123: peds doi: /peds The project was initiated by Drs Reilly, Onslow, Packman, Wake, Bavin, Prior, and Eadie; together with Ms Bolzonello they were responsible for managing the project including data collection and analysis. Dr Ukoumunne provided statistical advice and conducted the analyses in conjunction with Ms Cini. Dr Reilly wrote the article in conjunction with Drs Onslow and Packman, and all authors contributed to planning, reviewing, and editing the manuscript. Dr Reilly had full access to all of the data in the study, takes responsibility for the integrity of the data and the accuracy of the data analysis, and is the article s guarantor. Key Words stuttering, longitudinal study, risk factors, child, preschool, epidemiological studies Abbreviations ELVS Early Language in Victoria Study LGA local government area SEIFA Socio-economic Indexes for Areas RA research assistant CDI MacArthur-Bates Communicative Development Inventories CSBS Communication and Symbolic Behaviour Scales CI confidence interval IQR interquartile range OR odds ratio S Accepted for publication Apr 14, 2008 TUTTERING INVOLVES REPEATED movements and fixed postures of the speech Address correspondence to Sheena Reilly, mechanism and superfluous verbal and nonverbal behaviors. 1 PhD, Royal Children s Hospital, Department of In severe cases, Speech Pathology, Flemington Road, Parkville, fixed postures can arrest verbal communication, and associated grimacing can be Victoria 3086, Australia. sheena.reilly@ disfiguring. School-aged children who stutter suffer bullying and teasing. In later life, mcri.edu.au stuttering can lead to speech-related social anxiety and social phobia, failure to attain PEDIATRICS (ISSN Numbers: Print, ; occupational potential, and severely impaired communication. 2 Online, ). Copyright 2009 by the American Academy of Pediatrics The cause is unknown, although it is commonly believed to be a disorder of the neural processing underpinning spoken language, with genetic and environmental influences. 3 It is generally accepted that stuttering typically begins between 2 and 4 years of age and coincides with 270 REILLY et al

2 advances in language development, particularly the development of 2- to 3-word phrases. There are many causal theories, but for the most part they remain untested. 4 To date, there have been 2 prospective, community studies of incidence and natural recovery. Stuttering was examined in the 1000-family study conducted in Newcastle-Upon-Tyne, England, 5,6 with all children born in that city in May and June 1947 assessed over a number of years. Cumulative incidence up to 5 years old (ie, the proportion of children experiencing onset by age 5) was 3.5%, and 18 (42%) of 43 children recovered naturally by 6 years. 6 Shortcomings of these data 7 include the fact that stuttering was identified by health visitors rather than by speech pathologists, and it was not clear how long after stuttering onset the children were assessed. The cumulative incidence of stuttering in 1042 preschoolers up to 3 years old born between 1990 and 1991 on the Danish island of Bonholm was 5%. 8 By 5 years of age, 71% were reported to have naturally recovered. Limitations of this study included the fact that the children were not assessed until their third birthday, and the presence of stuttering was not established before the age of 3 years. Many antecedents and risk factors have been thought to be implicated in stuttering onset. There is ongoing debate about the relationship between language and stuttering. 9 Whereas some studies revealed that language learning difficulties were not associated with stuttering, 10 others demonstrated that children who stutter have lower language proficiency than children who do not stutter. 11 In addition, they are reported to differ on various temperament characteristics, 12 including shyness, 13 sensitivity, 14 adaptability, 14 and vulnerability. 15 Some have proposed that these temperament characteristics play a role in the etiology of stuttering. 13 To date, however, no study we are aware of has measured either language or temperament before stuttering onset. Therefore, it is not clear whether these factors are associated with stuttering onset and, if they are, whether they precede or develop as a result of the stuttering. Our review of the literature also identified several other potential variables thought to be implicated in stuttering onset. Included were prenatal and perinatal factors, birth order, socioeconomic status, parent-child relationships, traumatic experiences in early childhood, and growth and neurologic development. 5,16 Again, few of these factors have been measured before stuttering onset, and the reported associations are therefore not well supported. We identified 1 exception, a study of 93 children in the Netherlands 17 in which language development measured before any of the children started to stutter did not predict stuttering onset. The only predictors associated with onset were increased child speech rate and reduced maternal language complexity. Children in the Dutch study were selected on the basis that at least 1 parent stuttered, and the results may not therefore generalize to the large numbers of children who stutter without an overt family history. Existing knowledge about stuttering onset has been used to inform the development of causal pathway theories and to underpin stuttering interventions during the preschool years. 18,19 Prospective information about stuttering incidence, antecedents and predictors, and natural recovery rates is urgently required to inform treatment practices with preschoolers. Particularly pressing is the need to identify children in whom stuttering is and is not likely to persist, so that differing advice (watchful waiting versus recommending treatment) can be correctly targeted. This report aims to begin to address some of these methodologic issues by prospectively documenting the onset of stuttering in children up to 3 years of age in a large, community-ascertained cohort of Australian children. Specifically, we aimed to describe the cumulative incidence, age, and characteristics of stuttering at onset, and to determine any child, family, or environmental factors that might predict stuttering onset. METHOD Overview of the Early Language in Victoria Study The children in this study were participants in a larger study of early language development, the Early Language in Victoria Study (ELVS), conducted in Melbourne, Australia (population: 3.6 million), and described in detail elsewhere. 20,21 Briefly, ELVS is a prospective observational longitudinal study commencing at 8 months, followed by repeated measures at each birthday throughout the preschool and into the primary school years. Infants were recruited from 6 of metropolitan Melbourne s 31 local government areas (LGAs). The Disadvantage Index derived from the census-based Socio-economic Indexes for Areas (SEIFA), 22 represents attributes such as low income, low educational attainment, and high unemployment. The index was used to stratify the 31 LGAs into 3 tiers, after which 2 noncontiguous LGAs were selected from each tier to ensure the study sampled from geographic areas across the spectrum of disadvantage/advantage. Between September 2003 and April 2004, maternal and child health nurses were asked to consecutively approach parents of all infants at their universal 8-month visit (attended by over 80% of all Melbourne infants), followed by a formal recruitment process. Children were excluded if they had an obvious congenital (eg, Down Syndrome) or developmental (eg, cerebral palsy) problem or other serious intellectual or physical disability diagnosed by 8 months of age. Participants were also excluded if their parents did not speak and understand English sufficiently for recruitment and completion of questionnaires. The final sample was 1911 children. Parents reported on early communication and language development at each annual wave, and a broad range of other child, family, and environmental measures was also collected. 20,21 Procedure ELVS participants were eligible for inclusion in this stuttering study once they had returned their 2-year-old questionnaires. Parents were then sent a letter of invitation explaining the aims and procedures of the ELVS PEDIATRICS Volume 123, Number 1, January

3 Stuttering Study, an opt-out letter, and a refridgerator magnet defining stuttering with examples (see Appendix). Reminder letters were then sent every 4 months for 12 months to all families who had not opted out. All letters asked parents to telephone the research team if their child displayed any of the stuttering behaviors described (see Appendix). When a parent telephoned, the nature of the speech disruption was first clarified on the telephone by a research assistant (RA). If the RA verified that the speech characteristics reported resembled stuttering or was unsure, a 45-minute home visit was arranged as soon as possible. All RAs were qualified speech pathologists and consulted regularly with 2 of the authors (Drs Onslow and Packman). At the home visit, parents were interviewed to clarify the onset and characteristics of stuttering and to give a detailed family history of stuttering. This was followed by a videotaped 25-minute play session. Parents were asked to play with their child as they normally would, avoid asking too many questions, and reduce 1-word utterances. A standard set of toys, pictures, and questions were used to prompt verbal output. Outcome Measure The main outcome measure was the presence of RAconfirmed stuttering by 3 years of age. The Lidcombe 10-point stuttering severity scale was completed, with a score of 1 corresponding to no stuttering and 10 to very severe stuttering. 23 Where possible, home visits were conducted within 2 weeks of the parent report. A consensus panel considered all cases for whom there was uncertainty about the presence or absence of stuttering. The panel, comprising 2 speech pathologists experienced in the diagnosis and rating of stuttering (Drs Onslow and Packman), independently reviewed the videorecorded play sessions and completed the Lidcombe stuttering severity scale. Stuttering was determined to be present if the parent reported stuttering and the RA and the 2 experienced raters judged stuttering to be present (ie, scores 2 on the Lidcombe severity scale). Predictor Measures Predictor measures were largely drawn from the questionnaire completed by parents as children turned 2 years of age. Raw (quantitative) vocabulary scores were calculated from the MacArthur-Bates Communicative Development Inventories (CDI), with children below the 10th percentile for vocabulary production identified as late talkers. 24,25 The development of communicative behaviors up to 24 months of age was summarized using the Communication and Symbolic Behavior Scales (CSBS) Infant-Toddler Checklist 26 ; standardized Total Scores (normative mean: 100, SD: 15) were calculated according to the manual. Parents completed the 5-item Approach/Withdrawal scale, 1 of the 5 temperament dimensions that comprise the Australian Temperament Project-Short form suitable for toddlers aged 1 to 3 years (Toddler version). 27 Each item was rated on a scale from 1 (almost never) to 6 (almost always) and summed to provide a total score (maximum score: 30). 28 Shy/withdrawn children receive higher, and outgoing children lower, scores. Maternal mental health was measured using the Nonspecific Psychological Distress Scale (Kessler-6, or K-6), 29 which yields scores from a possible 0 (no distress) to 24 (maximum distress). The remaining parent-reported predictor variables were child gender, premature birth status, twin birth status, birth weight, existence of an older sibling, mother s education level, and family history of speech, language, reading, and stuttering problems. Finally, the continuous SEIFA Index of Relative Disadvantage based on the participant s postcode was used as the indicator of socioeconomic status, with lower scores representing greater disadvantage (Australian normative mean: 1000; SD: 100). 22 Almost all of the putative risk factors identified in our review of the literature were collected within ELVS and in most cases before the onset of stuttering. ELVS was designed as a study of early language development. Therefore, data on some variables hypothesized to be implicated in stuttering onset were not measured (eg, traumatic experience before stuttering onset, physical growth and neurologic development, and parent childrearing skills). Analysis Logistic regression was used to calculate odds ratios (ORs) for the relationship between stuttering onset by 3 years old and the predictor variables specified earlier. Information sandwich estimates of standard error 30 that allow for the correlation between responses from twins were obtained. Crude ORs from bivariable models, in which each risk factor is used as a predictor on its own, and adjusted ORs from multivariable models, in which all the risk factors are used simultaneously as predictors, were obtained. The squared Pearson correlation measure was used to quantify explained variation (R 2 ). 31 Scores on the variables of interest for children born prematurely, defined as 36 weeks gestation, were age-corrected before analysis. All analyses were implemented by using Stata RESULTS The Sample Figure 1 summarizes the participant flow between ELVS and the Stuttering Study. At 2 years of age, 57 of the original 1911 ELVS participants were unavailable (withdrawn, lost contact, or moved out of Victoria), and an additional 235 parents contacted us and opted not to participate in the study, leaving 1619 participants (85% of the original sample). Table 1 compares the study participants with the nonparticipants, with larger proportions of children whose mothers had a degree or postgraduate qualification among the participants than the nonparticipants. A total of 158 (9.8%) of 1619 parents called to report the onset of stuttering by 3 years of age. The presence of stuttering was then confirmed in 137 children (8.5% [95% confidence interval (CI): 7.2% 9.9%]) by a trained research assistant during the initial home visit. 272 REILLY et al

4 ELVS cohort 1911 subjects FIGURE 1 Flowchart of ELVS participants and relationship to the stuttering study. Lost to follow-up 57 subjects Stuttering-study participants 1619 subjects Opted out 235 subjects Confirmed stutterers 137 subjects Nonstutterers 1482 subjects TABLE 1 Characteristics of Participants Versus Nonparticipants Variables Participants Nonparticipants (N 1619) a (N 292) b Gender Male, n (%) 824 (50.9) 142 (48.6) Female, n (%) 795 (49.1) 150 (51.4) Twin birth, n (%) 42 (2.6) 11 (3.8) Premature birth, n (%) 47 (2.9) 12 (4.1) Birth weight, mean (SD), kg 3.4 (0.5) 3.4 (0.6) Birth order First child, n (%) 820 (50.8) 134 (46.5) Second child or more, n (%) 795 (49.2) 154 (53.5) Family history No problem, n (%) 1213 (74.9) 222 (76.0) Speech/language/reading problems 312 (19.3) 60 (20.5) only, n (%) Stuttering problem, n (%) c 94 (5.8) 10 (3.4) SEIFA disadvantage score, mean (SD) 1037 (61) 1033 (60) Mother s education level Did not complete year 12, n (%) 359 (22.4) 72 (25.2) Completed year 12, n (%) 628 (39.2) 130 (45.5) Degree/postgraduate qualification, n (%) 615 (38.4) 84 (29.4) Maternal mental health score, mean (SD) 3.2 (2.9) 3.1 (2.9) a The number of participants ranged from 1469 to b The number of nonparticipants ranged from 242 to 292. c Children with a family history of stuttering may also have reported family histories of speech/ language and/or reading problems. The panel agreed that there were 21 borderline children in whom stuttering reports were ambiguous, and these children were classified as nonstuttering for the main analyses. The median age of onset was 29.9 months (interquartile range [IQR]: ; range: ) and of parent telephone to report stuttering onset was 31.8 months (IQR: ; range: ). The median period from reported onset to telephone contact was 1.4 months (IQR: ; range: ), and between reported onset and the home visit was 2.1 months (IQR: ; range: ). Figure 2 plots the proportion of the ELVS cohort who had stuttered by any age up to 3 years. The constant slope from 25 months emphasizes that there was no obvious shorter age window during which stuttering characteristically commenced. Eleven parents reported that their children started to stutter before 2 years of age. Most parents reported that they first noticed stuttering Proportion Age, mo FIGURE 2 Proportion of study participants who stuttered by a given age. Data shown for children up to 3 years of age. when the child was stringing 3 or more words together (133 of 137 [97.1%]). In more than half the children, stuttering started suddenly; 51 parents (37.2%) reported stuttering commenced during a period of 1 day, and 17 parents (12.4%) over 2 to 3 days. In some children, stuttering was reported to emerge more slowly: 1 to 2 weeks (37 [27.0%]), 3 to 4 weeks (19 [13.9%]), and during a period of 5 weeks in 8 children (5.8%). Four parents (2.9%) were unsure about the time it took for stuttering to emerge. For 1 case, the data were missing. Family histories were obtained from all the children in the ELVS cohort (Table 1). A more detailed family history of stuttering and related communication problems was elicited after onset from the families of children who stuttered, which revealed a specific family history of stuttering in 71 (51.8%) parents of children who started to stutter. Between 20% and 30% reported family histories of other related conditions including speech problems, problems at school, difficulty with math, and difficulty with reading and/or writing. Less than 20% reported problems in a variety of domains including epilepsy, intellectual impairment, language problems, motor difficulties, and attention and concentration problems. The majority of parents described the nature of their child s stuttering as episodic (125 [91.2%]) rather than continuous. The most commonly reported stuttering behavior was whole word repetition (97 [71%]). Table 2 PEDIATRICS Volume 123, Number 1, January

5 TABLE 2 Comparison Between Stutterers and Nonstutterers Variables Stutterers Nonstutterers (N 137) a (N 1482) b Male gender, n (%) 84 (61.3) 740 (49.9) Premature birth ( 36 wk), n (%) 2 (1.5) 45 (3.0) Twin birth, n (%) 7 (5.1) 35 (2.4) Has older siblings, n (%) 62 (45.3) 733 (49.6) Family history No problem, n (%) 105 (76.6) 1108 (74.8) Speech/language/reading 20 (14.6) 292 (19.7) problems only, n (%) Stuttering problem, n (%) c 12 (8.8) 82 (5.5) Birth weight, mean (SD), kg 3.5 (0.5) 3.4 (0.5) SEIFA disadvantage score, mean (SD) 1047 (50) 1036 (62) Mother s education level Did not complete year (13.3) 341 (23.2) Completed year (28.1) 590 (40.2) Degree/postgraduate 79 (58.5) 536 (36.5) Maternal mental health score at 2 y, 3 (1, 5) 2 (1, 4) median (IQR) Temperament score at 2 y, mean (SD) 15.5 (4.6) 15.7 (5.0) CDI raw vocabulary score at 2 y, mean 299 (157) 259 (162) (SD) CDI percentile rank at 2 y, mean (SD) 48.8 (27.6) 40.6 (28.3) Late talker (score 10th percentile on 11 (8.0) 268 (20.0) CDI) at 2 y CSBS total standardized score at 2 years, mean (SD) 108 (15) 104 (15) a Sample size ranged from 134 to 137 for stutterers. b Sample size ranged from 1308 to 1482 for nonstutterers. c Children with a family history of stuttering may also have reported family histories of speech/ language and/or reading problems. shows the stuttering group had higher proportions of children with each of the following characteristics than the nonstuttering group: boys, twins, and parents educated at or above degree level. The stuttering group had a lower proportion of late-talkers and higher total scores on the CDI and CSBS at 2 years than children who did not stutter, but mean shyness, birth weight, and maternal mental health scores were similar (Table 2). Table 3 summarizes the results of the logistic regression analyses of stuttering status on the risk factors of interest showing both crude (bivariable) and adjusted (multivariable) ORs. No evidence of nonlinearity in the relationships between the quantitative predictors and the log odds of stuttering onset by 3 years was detected using the fractional polynomials approach. 33 In the adjusted model male gender, twin birth status, a mother with a degree or postgraduate qualification, and having a high CDI raw vocabulary score at 2 years were all associated with stuttering. An increase of 100 words on the vocabulary score corresponds to an increase of 17% in the odds of beginning to stutter. The findings for twin birth should be interpreted cautiously; the OR estimate for this variable may be unstable, because the number of twins who started to stutter was small. There was weak evidence that children with a family history of stuttering were more likely to develop stuttering in comparison to those with no problems or those with only a history of general language/ speech/reading problems. Overall, the multivariable model had low predictive strength, accounting for only 3.7% of the total variation in stuttering onset. A sensitivity analysis in which the 21 borderline cases were classified as stutterers rather than nonstutterers provided similar results to the main adjusted analysis. DISCUSSION The cumulative incidence of stuttering onset by 3 years old was 8.5%, which is almost twice the percentage previously reported in other studies. Although we confirmed several of the hypothesized predictive associations, very little of the variation in stuttering onset (3.7%) could be explained by the 12 risk factors we investigated. TABLE 3 Logistic Regression of Stuttering Status (Stuttering Versus not Stuttering) Variable Crude Adjusted OR 95% CI P OR 95% CI P Male gender Premature birth ( 36 wk) Twin birth Birth weight, kg Has older siblings Family history No problem Reference Reference Speech/language/reading problems only Stuttering problem a SEIFA disadvantage score (per 100-unit increase) Mother s education level Did not complete year 12 Reference Reference Completed year Degree/postgraduate Maternal mental health score Temperament score of child at 2 y CSBS total score at 2 y (per 15-unit increase) CDI raw vocabulary score at 2 y (per 100-unit increase) The number ranges from 1442 to 1619 for unadjusted analyses and is 1335 for adjusted analysis. a Children with a family history of stuttering may also have reported family histories of speech/language and/or reading problems. 274 REILLY et al

6 To our knowledge, this is the largest epidemiological study of early stuttering onset to date, and the first to recruit a cohort and measure putative risk factors before onset and to incorporate reliable, expert diagnoses. Some of our findings confirm previous reports about early stuttering. For example, boys were more likely to stutter than girls. In addition, we confirmed the reported rapid onset of stuttering, its episodic nature, and the prominence of repeated words at onset. Consistent with existing retrospective evidence, the majority of parents reported that stuttering onset occurred during the 2- or 3-word stage of language development. Reports of a family history of stuttering varied according to when the information was obtained. Before stuttering onset, 12 (8.8%) of 137 families with stuttering children reported a family history of stuttering. When family history data were collected after stuttering onset, many more of these families recalled a family history (71 of 137 [51.8%]). Similar findings have been reported in studies involving screening for hearing impairment, 34 and the finding may simply reflect parents being unaware of the risk factor at the time of questioning (eg, before onset). The finding, however, does raise concern about using a potential risk factor such as family history to identify children at risk of stuttering onset. The high cumulative incidence of stuttering onset of 8.5% by 3 years old may seem surprising. We are confident about the estimate, because the 95% CIs were narrow (7.2% 9.9%). The response rate was high (85% of parents in the original ELVS cohort), and stuttering was rigorously verified by experts using strict criteria, so we can be reasonably confident that both the majority of children in the population who started to stutter before age 3 were identified and that the high rate was not because of false-positives. Finally, the high incidence is plausible: it is likely we have identified stuttering that would have been missed in previous studies, either because it was mild or because it was short-lived. In the Newcastle-Upon-Tyne study, 5 for example, stuttering sometimes seemed to last only a few months. Whenever a major study reports more extreme findings than previous research, the possibility of bias must be considered. We acknowledge that our cumulative incidence estimate may be somewhat inflated, because disadvantaged parents and infants were underrepresented in this cohort (a common limitation in longitudinal population studies), whereas children of highly educated mothers were both more likely to stutter and to be included in the study. However, this does not alter our conclusions. Australian census data on mothers of children born in 2001 indicate that 43.8% did not complete school, compared with only 22.4% in our stuttering study. When we applied inverse probability weights accordingly as a sensitivity check, the estimated cumulative incidence fell from 8.5% to 7.5% (95% CI: 6.2% 8.8%), which is still considerably higher than previously reported. Other contributors to selection bias seem unlikely: the ELVS cohort was not originally recruited to a study of stuttering, recruitment occurred well before any concern about stuttering onset might have arisen, and our findings indicate that the great majority of those with a family history of stuttering were unaware of it (ie, this could not have been the reason for self-selection into either ELVS or the stuttering study). Finally, the study successfully captured families from across the entire socioeconomic spectrum to quantify internal predictor-outcome relationships, for which a representative sample is not essential. This report includes onset only up to 3 years. The cumulative incidence of stuttering onset by 4 years will be higher. We may have missed some cases because parents elected not to call or stuttering duration was so brief that it did not elicit a parent report. As we continue our observation of this cohort into the preschool years, we will be able to ascertain any missed cases of persistent stuttering that were not notified by parents. Four factors were significantly associated with stuttering onset at the 5% level. The association with gender is the best-recognized factor in the literature, but the associations with both twinning and maternal education were surprising and, as noted earlier, should be interpreted cautiously. Mothers with a degree or postgraduate qualification were more likely to have children that stuttered by 3 years. It could be that well-educated mothers were more likely to be aware of (and therefore to report) stuttering onset. It is also possible that the association with maternal education may be an artifact; stuttering onset typically occurs with the development of 3-word combinations, and such combinations may simply occur earlier in children of more highly educated mothers. Against this, vocabulary scores at 2 years were significantly associated with onset of stuttering but not with maternal education, 21 and clearly this unexpected association requires additional elucidation as the children mature. Our results do not support the premise that language problems are associated with stuttering onset but that communication skills and vocabulary are more highly developed in children who start to stutter by 3 years of age. Of clinical and theoretical interest is our finding that the dimension of temperament we measured did not predict stuttering onset children who started to stutter were neither more shy nor more outgoing than their peers who did not stutter. Importantly, our measures were taken before stuttering onset by using a well-validated parent report instrument normed on Australian children. 27 Temperament is regarded as a relatively stable trait, but is nonetheless receptive to environmental experiences. 27,35,36 It is possible that when children begin to stutter they become more vigilant and more reticent, as described by Anderson et al. 12 Alternatively, it may be that parents, who are the ones that typically fill out temperament questionnaires, perceive their children to be so simply because they stutter. Regardless, our results do not suggest that innate shyness has a causal role in stuttering. Additional study of this cohort will enable estimation of the cumulative incidence of stuttering onset during the preschool years and identify whether stronger predictors emerge with stuttering onset after 3 years of age. We will also be able to estimate the natural recovery rate during the preschool years and provide a much-needed PEDIATRICS Volume 123, Number 1, January

7 evidence base regarding symptom clusters that are likely to persist and resolve with time. The study did not identify risk factors or antecedents that could be used by health professionals in the clinical case history to reliably predict children at risk of stuttering onset. The known prevalence of stuttering in adults is around 1%, 37 so one significant implication of our high incidence is that natural recovery could be much higher than previously thought. Parents presenting to health professionals with concerns about stuttering onset should be reassured about the following points: first, stuttering onset is not unusual in children up to 3 years of age. Second, stuttering onset was associated with rapid growth in language development, most noticeably the combining of words into phrases and short sentences. Third, children who start to stutter do not seem to be more shy or withdrawn compared with other children of the same age who do not stutter. Finally, parents should be reassured that there is no strong evidence that the social and environmental factors measured in this study are related to stuttering onset. APPENDIX: INFORMATION PROVIDED TO PARENTS ABOUT THE ONSET OF STUTTERING Children who stutter have trouble getting their words out. Stuttering is when children: Repeat words or syllables over and over (eg, can can can I go ) Make long prolonged sounds (eg, caaaaaaaaaaaaaan I go ) Have speech stoppages or blocks where no sounds come out Stuttering can start quite suddenly or it can begin gradually over days, weeks, or months. If you think your child is stuttering please contact the ELVS team as soon as you notice it. ACKNOWLEDGMENTS This study was funded by Australian Research Council grant DP The ELVS was funded by Australian National Health and Medical Research Council (NHMRC) grant The project was supplemented by an NHMRC program grant held by Professors Onslow and Packman (402763) and additional grants from the Murdoch Childrens Research Institute and the Faculty of Healthy Sciences at La Trobe University. Drs Onslow and Wake were funded by NHMRC Research Fellowship and NHMRC Population Health Career Development Award , respectively. Dr Okoumunne s postdoctoral position is funded by NHMRC Population Health Capacity Building grant Ethical approval was obtained from the Royal Children s Hospital Melbourne (23018) and La Trobe University Human Ethics Committee (03-32). We sincerely thank all the participating children and parents. REFERENCES 1. Teesson K, Packman A, Onslow M. The Lidcombe Behavioral Data Language of stuttering. J Speech Lang Hear Res. 2003;46(4): Messenger M, Onslow M, Packman A, Menzies R. Social anxiety in stuttering: measuring negative social expectancies. J Fluency Disord. 2004;29(3): Buchel C, Sommer M. What causes stuttering? PLoS Biol. 2004; 2(2):E46 4. Packman A, Attanasio JS. Theoretical Issues in Stuttering. London, England: Psychology Press; Andrews G. The epidemiology of stuttering. In: Curlee RF, Perkins WH, eds. Nature and Treatment of Stuttering: New Directions. San Diego, CA: College-Hill Press; 1985: Andrews G, Harris M. The Syndrome of Stuttering. London, England: Heinemann; Ingham RJ. Onset, prevalence, and recovery from stuttering: a reassessment of findings from the Andrews and Harris study. J Speech Hear Disord. 1976;41(2): Mansson H. Childhood stuttering: incidence and development. J Fluency Disord. 2000;25(1): Nippold MA. Concomitant speech and language disorders in stuttering children: a critique of the literature. J Speech Hear Disord. 1990;55(1): Watkins RV, Yairi E, Ambrose NG. Early childhood stuttering III: initial status of expressive language abilities. J Speech Lang Hear Res. 1999;42(5): Anderson JD, Wagovich S, Hall NE. Nonword repetition skills in young children who do and do not stutter. J Fluency Disord. 2006;31(3): Anderson JD, Pellowski MW, Conture EG, Kelly EM. Temperamental characteristics of young children who stutter. J Speech Lang Hear Res. 2003;46(5): Guitar B. Stuttering: An Integrated Approach to its Nature and Treatment, 2nd ed. Baltimore, MD: Williams and Wilkins; Riley G, Riley J. A revised component model for diagnosing and treating children who stutter. Contemp Issues Commun Sci Disord. 2000;27(2): Oyler M. Vulnerability in stuttering children. Diss Abstr Int. 1996;56 (UMI No ): Yairi E, Ambrose NG, Niermann R. The early months of stuttering: a developmental study. Journal of Speech and Hearing Research. 1993;36(3): Kloth S, Janssen P, Kraaimaat F, Brutten GJ. Child and mother variables in the development of stuttering among high-risk children: a longitudinal study. J Fluency Disord. 1998;23(4): Onslow M, Harrison L, Rousseau I, eds. The Lidcombe Program 2007: Clinical Tales and Clinical Trials. 3rd ed. New York, NY: Thieme; Jones M, Onslow M, Packman A, et al. A randomised controlled trial of the Lidcombe Program for early stuttering intervention. BMJ. 2005;331(7518): Reilly S, Eadie P, Bavin EL, et al. Growth of infant communication between 8 and 12 months: a population study. J Paediatr Child Health. 2006;42(12): Reilly S, Wake M, Bavin EL, et al. Predicting language at 2 years of age: A prospective community study. Pediatrics. 2007; 120(6). Available at: 6/e Australian Bureau of Statistics. Socio-economic Indexes for Areas. Canberra, Australia: Australian Bureau of Statistics; Onslow M, Packman A, Harrison E, eds. Lidcombe Program for Early Stuttering Intervention: Clinician s Guide. Austin, TX: Pro- Ed; Fenson L, Dale PS, Reznick JS. The MacArthur Communicative 276 REILLY et al

8 Development Inventories: User s Guide and Technical Manual. San Diego, CA: Singular Publishing Group; Fenson L, Pethick SJ, Renda C, Cox JL, Dale PS, Reznick JS. Short-form versions of the MacArthur Communicative Development Inventories. Appl Psycholinguistics. 2000;21(1): Wetherby A, Prizant B. Communication and Symbolic Behaviour Scales. Baltimore, MD: Paul H. Brookes Publishing Co; Prior M, Sanson A, Oberklaid F. The Australian Temperament Project. In: Kohnstamm GA, Bates JE, Rothbart MK, eds. Temperament in Childhood. Chichester, Sussex, United Kingdom: Wiley; 1989: Prior M, Sanson A, Smart D, Oberklaid F. Pathways From Infancy to Adolescence: Australian Temperament Project Melbourne, Australia: Australian Institute of Family Studies; Kessler R, Mroczek D. Final Version of Our Non-specific Psychological Distress Scale [memo]. Ann Arbor, MI: Institute for Social Research; Hanley JA, Negassa A, Edwardes MD, Forrester JE. Statistical analysis of correlated data using Generalized Estimating Equations: an orientation. Am J Epidemiol. 2003;157(4): Hosmer D, Lemeshow S. Applied Logistic Regression. New York, NY: Wiley; Stata Corporation. Stata Statistical Software [computer program]. Release 9.2. College Station, TX: Stata Corporation; Royston P, Ambler G, Sauerbrei W. The use of fractional polynomials to model continuous risk variables in epidemiology. Int J Epidemiol. 1999;28(5): Russ R, Rickards F, Poulaki Z, Barke M, Saunders K, Wake M. Six year effectiveness of a population based two tier infant hearing screening programme. Arch Dis Child. 2002;86(4): Sanson A, Hemphill S, Smart D. Connections between temperament and social development: a review. Soc Dev. 2004;13(1): van den Boom D. Do first-year intervention effects endure? Follow-up during toddlerhood of a sample of Dutch irritable infants. Child Dev. 1995;66(6): Craig A, Hancock K, Tran Y, Craig M, Peters K. Epidemiology of stuttering in the community across the entire life span. J Speech Lang Hear Res. 2002;45(6): HEALTH CLINICS INSIDE STORES LIKELY TO SLOW THEIR GROWTH The boom in walk-in health clinics located inside pharmacies, supermarkets and big-box retailers is showing signs of slowing. Hailed as an inexpensive option for treating minor health ailments like sore throats and rashes, the retail clinics have grown in number to 963 as of May 1 from just 125 three years ago. The clinics typically feature nurse practitioners who can prescribe basic drugs, and the price for a visit ranges from $50 to $75. But in recent months, retail health-clinic operators based in New York, Nevada, Indiana and Alabama have closed their doors, shuttering 69 clinics in 15 states. Now, the biggest retail-clinic operator, CVS Caremark Corp., says it is scaling back expansion plans for its MinuteClinic brand. The cost of setting up an in-store clinic runs about $ That is one reason why much of the future growth in walk-in health centers is expected to come from big companies with deep pockets and from hospital systems that are already well-known within a community and don t have to spend so much on marketing. Armstrong D. Wall Street Journal. May 7, 2008 Noted by JFL, MD PEDIATRICS Volume 123, Number 1, January