Characterizing Autism Traits in Toddlers with Down Syndrome: Preliminary Associations with Language, Executive Functioning, and Other Developmental Domains

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Characterizing Autism Traits in Toddlers With Down Syndrome: Associations With Language, Executive Functioning, and Other Developmental Domains

I have read the manuscript with interest, and I agree that autism in DS is a rare condition that was underestimated and still not investigated. that said, you can find here a series of recommendations, suggestions, and concerns, section by section.

In the introduction, the description of the genetics of DS needs to be added to the text. I know that is a well-known syndrome, but it is important for the general audience. Please, add more info about. Moreover, despite the principal focus being on toddlers, I suggest adding a brief description of the general or global characteristics of DS. Please, add.

Moreover, I suggest using Autism spectrum disorders instead of Autism, but I agree that a diagnosis of autism in DS is challenging.

The language development in DS and autism is well described and informative. However, I suggest distinguishing in a clearer way the social interaction alterations and alterations in language development. The definition of EFs needs to be improved, since it is quite vague.

Despite the trajectory of EFs during childhood and early childhood are still matter of investigation, and less is still known, you could make a parallelism between non-DS or non-autistic toddlers and DS or autistic toddlers, highlighting the EFs that are more affected during the development in these specific populations.

The research questions at the end of the introduction are clear and stated in a good way.

Methods: You recruited 38 children, but the mean age seems to be preschool-aged (4 y.o). Indeed, the range indicated that you collected data from children ~5 y.o. Please, clarify.

The age range in Table 1 is different from that reported in the text. Word production is reported to be ~113, but the range is 0-78. The measures used are described in a very detailed way, allowing the replication of the study in a different country with a similar population. I suggest adding more information, if possible, about the socioeconomic status and education of the parents. This variable was found to play a relevant role in the cognitive dysfunctions in children with neurological disorders.

The statistical analyses were described in a good and transparent way. I suggest being clearer about the way in which CARS-2 was used, since it is not clear if it was used for adjusting or weighting. Despite this, the data analysis is well-reported. Since you have missing data (I suppose at different levels/variables), maybe for a future study, you could consider the application of Bayesian statistics, but I appreciate that you reported this.

The results are interesting. I do not ask the fit index for the regression, since I know that it is redundant.

 The results are a bit different from the research questions (introduction). In this way, I suggest adding more information to the research questions in order to facilitate the reading.  

In the discussion, you are able to integrate your results in the light of the previously published evidence. I have also appreciated that the results obtained in the study were discussed critically. Despite this, you have acknowledged a series of limitations. In this way, I suggest adding to the title and the abstract the fact that the study can be considered as preliminary.

Author Response

Reviewer #1:

Reviewer #1: In the introduction, the description of the genetics of DS needs to be added to the text. I know that is a well-known syndrome, but it is important for the general audience. Please, add more info about. Moreover, despite the principal focus being on toddlers, I suggest adding a brief description of the general or global characteristics of DS. Please, add.

Response: We revised the introduction's opening paragraph to describe the genetic etiology of DS explicitly and to include a concise overview of common developmental and medical characteristics associated with DS. Specifically, the following text has been added: “Down syndrome (DS) is a genetically based neurodevelopmental condition caused by the presence of a full or partial extra copy of chromosome 21 (Bull, 2020). It is the leading genetic cause of intellectual disability with a prevalence of 1 in every 707 births (Mai et al., 2019). Individuals with DS share a broad developmental phenotype that often includes hypotonia, characteristic craniofacial features, congenital heart conditions, and delays across cognitive, motor, language, and adaptive domains, though substantial variability exists in how these features present (Daunhauer, 2011; Martin et al., 2009).” Please see highlighted text on pg. 3, lines 48-54.

Reviewer #1: Moreover, I suggest using Autism spectrum disorders instead of Autism, but I agree that a diagnosis of autism in DS is challenging.

Response: Though “autism spectrum disorder” is the formal diagnostic terminology, we intentionally used “autism” in several places to refer more broadly to autism traits or the behavioral phenotype, consistent with current research conventions. We reviewed the manuscript for clarity and now use autism spectrum disorder when referring specifically to the clinical diagnosis and autism when discussing traits or symptomatology more generally. Please see highlighted text on pg. 3, line 59, pg. 4, lines 67, 70, 74, 79, and pg. 5, lines 100-101.

Reviewer #1: The language development in DS and autism is well described and informative. However, I suggest distinguishing in a clearer way the social interaction alterations and alterations in language development. The definition of EFs needs to be improved, since it is quite vague.

Response: We added several sentences highlighting social-interaction characteristics associated with ASD (e.g., reduced reciprocity, differences in the social use of gestures) and clarified how these differ from language-specific delays. We also modified the discussion of gesture use to emphasize its dual role in both language development and socially directed communication. These revisions appear on pages 6-7, lines 106-127, and improve the conceptual differentiation requested. Additionally, we have expanded the definition of EFs. Please see pg. 6-7 lines 131 – 133.

 

Reviewer #1: Despite the trajectory of EFs during childhood and early childhood are still matter of investigation, and less is still known, you could make a parallelism between non-DS or non-autistic toddlers and DS or autistic toddlers, highlighting the EFs that are more affected during the development in these specific populations.

Response: We revised the Executive Functions section to incorporate a clearer developmental comparison across neurotypical toddlers, toddlers with DS, and autistic toddlers. We added text describing typical early EF emergence and highlighted how specific EF components (e.g., inhibitory control, working memory, flexibility, planning) differ in DS and autism relative to neurotypical developmental trajectories. We also clarified which EF domains appear to be areas of shared vulnerability and which are more characteristic of autism. These additions improve conceptual clarity and strengthen the developmental framing requested. Revisions can be found on page 6-8, lines 133-138, 143- 148, and 155- 160.

Reviewer #1: You recruited 38 children, but the mean age seems to be preschool-aged (4 y.o). Indeed, the range indicated that you collected data from children ~5 y.o. Please, clarify.

Response: We have revised the participant description to explain that although children ages 2;0–5;11 were eligible for the larger project, the enrolled sample for this phase of data collection skewed slightly older, resulting in a mean age of approximately 4 years. Specifically, the following text has been added: “Although children ages 2;0–5;11 were eligible, the enrolled sample skewed slightly older, yielding a mean age of 4.19 years.” Please see the highlighted text on pg. 11, lines 233-234

Reviewer #1: The age range in Table 1 is different from that reported in the text. Word production is reported to be ~113, but the range is 0-78. The measures used are described in a very detailed way, allowing the replication of the study in a different country with a similar population. I suggest adding more information, if possible, about the socioeconomic status and education of the parents. This variable was found to play a relevant role in cognitive dysfunction in children with neurological disorders.

Response: We have corrected both in the manuscript to ensure consistency between the text and Table 1. Additionally, we have added descriptive information about caregiver education and socioeconomic status to the Participants section to provide greater contextual detail for the sample (pgs. 12-13).

Reviewer #1: The statistical analyses were described in a good and transparent way. I suggest being clearer about the way in which CARS-2 was used, since it is not clear if it was used for adjusting or weighting. Despite this, the data analysis is well-reported. Since you have missing data (I suppose at different levels/variables), maybe for a future study, you could consider the application of Bayesian statistics, but I appreciate that you reported this.

Response: The CARS-2 was used as the outcome variable in all analyses as the measure of autistic traits. We have clarified this on page 17, lines 361-362. “CARS-2 scores were used as the outcome variable in all models.”

We also did not have missing data for any measures used in our regression models; we only had some missing data for some demographic variables. We have revised the missing data section on page 18, line 369-370 to clarify this. “We did not have any missing data across the measures used in our regression analyses. However, we did have some missing data related to participant demographics. Specifically, two participants did not report sex, race, and ethnicity. For these variables, we have a “not-reported” category.”

We appreciate your suggestion for incorporating Bayesian statistics into future studies, this is definitely an underutilized approach in clinical research.

Reviewer #1: The results are interesting. I do not ask the fit index in the regression, since I know it is redundant.

Response: In our regression tables (Tables 2-4) we report R² and adjusted R² to aid in both interpreting a single model and across models. We have removed the reporting of these from the narrative section of the results to avoid redundancy with the table. 

Reviewer #1: The results are a bit different from the research questions (introduction). In this way, I suggest adding more information to the research questions in order to facilitate the reading.  

Response: To improve clarity and better guide the reader, we have expanded and refined the research questions to more directly reflect the domains included in our analyses. Specifically, we now state that our goal was to examine how (a) early communication skills (words understood, words produced, and gesture use), (b) components of executive functioning (inhibit, shift, emotional control, working memory, plan/organize), and (c) broader developmental domains (fine motor and visual reception) are associated with autistic traits in toddlers with DS.

This revision brings the introduction, research questions, and analytic approach into clearer alignment, thereby facilitating the reader’s understanding of how each developmental domain examined in the results connects to our overarching study aims. The revised research questions can be found on pg. 10, lines 211-212, 218-219, and 224-225.

Reviewer #1: In the discussion, you are able to integrate your results in the light of the previously published evidence. I have also appreciated that the results obtained in the study were discussed critically. Despite this, you have acknowledged a series of limitations. In this way, I suggest adding to the title and the abstract the fact that the study can be considered as preliminary.

Response: We have added the term “preliminary” to both the title and the abstract to reflect the exploratory scope of the work and the limitations noted in the Discussion. Please see highlighted text throughout the discussion.

Reviewer 2 Report

Comments and Suggestions for Authors

Please see the attached file 

Comments for author File: Comments.pdf

Author Response

 

Reviewer #2:

Reviewer #2: The study includes only 38 toddlers, yet the authors run ten separate robust regression models, each including covariates. This sample size is insufficient for stable parameter estimates and increases the likelihood of both Type I and Type II errors. The effect sizes should be interpreted cautiously, and conclusions should be tempered substantially. The authors should avoid using language implying strong predictive relationships or developmental mechanisms.

Response: We agree that our sample size is underpowered. Unfortunately, this is very common in the intellectual disability field, and our sample size is in line with or larger than other DS studies (e.g., De Falco et al., 2008; Schworer et al., 2022; Sterling & Warren, 2014). While this is common in our field and something we always consider when interpreting results, we agree that we can be more transparent about this and have revised the manuscript throughout to reflect this: pg. 17, lines 362-364: “Despite the use of robust regressions, our parameter estimates may be unstable, given our sample size, and results should be interpreted accordingly.” In our discussion section we have also tempered our conclusions, removing causal/developmental conclusions, and softening language (e.g., preliminary evidence, suggest). Please see highlighted text throughout the discussion.

 

Reviewer #2: Although the manuscript frequently uses terms such as predict, contribute, shape, these are not appropriate for a cross-sectional dataset. The design only supports associational interpretations. The language throughout the manuscript should be revised to reflect correlational findings rather than developmental prediction.

Response: We carefully reviewed the manuscript and replaced all language implying prediction, causality, or developmental influence with terminology that accurately reflects correlational and associational interpretations. Please see the highlighted throughout the results and discussion.

 

Reviewer #2: A central conceptual limitation is the absence of a DS-only vs DS+autism comparison. Thus, the study cannot meaningfully distinguish between: DS-related developmental variability vs. autism-related characteristics. This limits any conclusions about co-occurring autism or the developmental phenotype of DS+autism. Many of the findings may instead reflect the typical DS profile rather than autism-specific traits.

Response: We appreciate the reviewer’s point regarding the limitations of not including categorical DS-only and DS+autism groups. We agree that such comparisons would allow clearer differentiation between DS-related variability and autism-specific characteristics. In response, we have revised the Discussion and Limitations to explicitly acknowledge that the associations identified here may reflect overlapping features of the broader DS phenotype rather than autism-specific mechanisms. We also now emphasize that our findings are exploratory and hypothesis-generating and should not be interpreted as definitive evidence of DS+ASD phenotypes. These revisions appear on pg. 30 lines 591-599.

 

At the same time, we respectfully note that the goal of this study was not to categorize DS+autism, but rather to examine dimensional variation in autistic traits within DS. This dimensional approach has been increasingly adopted in recent DS and neurodevelopmental research, as it allows investigators to capture the full continuum of autistic traits, including subclinical or emerging traits, and to investigate how these traits relate to developmental domains without requiring dichotomous diagnostic groups (e.g., Bradbury et al., 2021; Lorang et al., 2017; Spinazzi et al., 2023). Dimensional models can offer complementary insights to categorical comparisons, particularly in early childhood when formal diagnosis may be delayed or more difficult to establish, and when traits may present heterogeneously across individuals with DS.

 

Nevertheless, we agree that future studies incorporating both categorical comparisons (DS-only vs DS+ASD) and dimensional trait measures, ideally using gold-standard diagnostic tools, will be essential to determining whether the patterns observed reflect autism-specific features or broader variability within DS. We have added this point to the revised Discussion and Limitations.

 

Reviewer #2: Conducting 10 separate regression analyses in a sample of 38 participants increases the risk of spurious findings even with Bonferroni correction. A more statistically sound approach would involve: dimension reduction (e.g., PCA or composite scores), hierarchical regression, or structural equation modeling with fewer parameters. At minimum, the authors should explicitly justify their analytic strategy and clearly communicate that estimates may be unstable.

Response: We agree that, while we tried to reduce the likelihood of spurious findings, our approach does not fully reduce the risk of spurious findings. Because our research questions focused on individual components within each developmental domain (e.g., gesture use, expressive vocabulary, receptive vocabulary; five distinct EF subdomains; fine motor and visual reception), dimension reduction approaches such as PCA were not appropriate. While PCA can be useful for simplifying data structure, it would collapse theoretically meaningful constructs and obscure dissociations that are well documented in DS and autism (e.g., Caselli et al., 2008; Daunhauer et al., 2014). Thus, we wanted to examine predictors separately. While hierarchical regression aligns with our research questions and study purpose, we did

think that hierarchical regression was the best approach, power wise, given that for RQ2, it would include a model with 7 predictors (i.e., 5 EF variables, along with site and age). As noted by Babyak (2004), models with many predictors and small samples also inflate both Type I and Type II errors and produce unstable estimates. Thus, we believed that, while not a perfect option, running separate regression models for each developmental domain was the best approach given our sample size and research questions.

We have justified this approach more clearly on pg. 17, lines 351-355: “Given the exploratory nature of our study, we aimed to examine how individual components within each developmental domain (e.g., RQ1 communication: receptive, verbal expressive, and nonverbal expressive language) were related to autistic traits; therefore, we conducted a series of regression analyses. To minimize the impact of underpowered and overfitted models given our sample size, we ran separate models for each study measure.”

We have also explicitly stated that our parameter estimates may be unstable in the methods, pg. 17, lines 362-364: “Despite the use of robust regressions, our parameter estimates may be unstable, given our sample size, and results should be interpreted accordingly.”

We have also added this explicitly as a limitation pg. 28, lines 559-561 “Similarly, while we opted to conduct more regression models with fewer predictors to reduce overfitting and underpowered models, given our sample size, parameter estimates may still be unstable, and our results should be interpreted accordingly.”

However, if after a second review, the reviewer(s) still thinks that another approach is better we are open to the idea of incorporating PCA into our study.

 

Reviewer #2:  The sample spans ages 2.2–6.0 years, but raw scores (especially in motor, visual reception, and language) change dramatically across this period. Although age is included as a covariate, raw scores are not developmentally linear. This raises concerns about whether associations simply reflect age-related developmental progress rather than autism traits.

Response: We agree that the use of raw scores can be problematic in neurotypical development across a large chronological age span because of the development that occurs during early childhood. However, this developmental growth is not as steep in intellectual disabilities and as a result, standard scores are not sensitive to developmental growth. For example, children with DS are delayed in producing a two-word utterance an average of 18 months compared to the neurotypical developmental milestone (Oliver & Buckley, 1994). Additionally, many children are at floor on standardized assessments, and thus the use of standardized scores can often mask true growth in a measure (Esbensen et al., 2017). Therefore, the use of raw scores is common within the field. For example, expressive and receptive MSEL raw scores were used in Hahn et al. 2019 which included children with DS ages 3-6 years. Expressive and receptive MSEL raw scores were also used in Barton-Hulsey et al. 2020 which included children with DS ages 22-63 months. Additionally, this age range is frequently used to examine development in in DS (e.g., De Falco et al., 2008; Schworer et al., 2022; Sterling & Warren, 2014)

 

Furthermore, as noted above, the use of standardized norm-referenced scores can be problematic since, in most studies, most participants are at or near the floor. We have included descriptives for our sample in the MSEL description in the methods to illustrate this pg. 16, lines 322-326: “Specifically, for visual reception, 31 participants were at floor, two were below floor and for two participants we were not able to calculate t-scores. Three participants scored above floor (M = 29, SD = 5.35, range: 22-35). For fine motor, 35 participants were at floor, for two participants we were not able to calculate t-scores. One participant scored above floor (t-score = 30).”

 

Reviewer #2: The discussion often treats associations as evidence of developmental mechanisms. Given methodological limitations, interpretations should be reframed to emphasize hypothesis-generating exploratory findings.

Response: We carefully revised the Discussion to remove any language implying causality, prediction, or developmental influence. We reframed interpretations to highlight that these findings are exploratory and hypothesis-generating, and we clarified that associations observed in this sample should not be interpreted as evidence of underlying developmental pathways. Please see the highlighted text throughout discussion (e.g., pg. 23, line 429, 431-435, 437, 442; pg. 25, line 480, pg. 26 line 498)

 

Reviewer #2: Thirty-two of the 38 participants are White. This limits generalizability to more diverse populations, an important issue to address given racial and cultural differences in communication behaviors and parental reporting.

Response: We have revised the Limitation section to explicitly address that our sample constrains the generalizability of our findings. Specifically, the following text has been added: “Third, the racial and ethnic composition of our sample was predominantly White (32 of 38 participants), which limits the generalizability of these findings to more diverse populations. Communication behaviors, caregiver-report ratings, and the interpretation of autism-related traits can vary across cultural and linguistic contexts. Therefore, caution is warranted when applying these results to children from historically underrepresented communities. Future research that includes more racially and culturally diverse samples will be essential for improving the representativeness and applicability of this work.” Please see pg. 29, lines 570-576.

 

References:

 

Barton-Hulsey, A., Lorang, E., Renfus, K., & Sterling, A. (2020). Maternal input and child language comprehension during book reading in children with Down syndrome. American Journal of Speech-Language Pathology, 29(3), 1475–1488. https://doi.org/10.1044/2020_AJSLP-19-00156

de Falco, S., Esposito, G., Venuti, P., & Bornstein, M. H. (2008). Fathers' play with their Down Syndrome children. Journal of Intellectual Disability Research, 52(Pt 6), 490–502. https://doi.org/10.1111/j.1365-2788.2008.01052.x

Esbensen, A. J., Hooper, S. R., Fidler, D., Hartley, S. L., Edgin, J., d’Ardhuy, X. L., Capone, G., Conners, F. A., Mervis, C. B., Abbeduto, L., Rafii, M., Krinsky-McHale, S. J., Urv, T., & Group, O. M. W. (2017). Outcome measures for clinical trials in Down syndrome. American Journal on Intellectual and Developmental Disabilities, 122(3), 247–281. https://doi.org/10.1352/1944-7558-122.3.247

Hahn, L. J., Brady, N. C., & Versaci, T. (2019). Communicative use of triadic eye gaze in children with Down syndrome, autism spectrum disorder, and other intellectual and developmental disabilities. American Journal of Speech-Language Pathology, 28(4), 1509–1522. https://doi.org/10.1044/2019_AJSLP-18-0155

Oliver, B., & Buckley S. (1994). The language development of children with Down’s syndrome: First words to two-word phrases. Down Syndrome Research and Practice, 2, 71-75. https://doi.org/10.3104/reports.33

Schworer, E. K., Esbensen, A. J., Nguyen, V., Bullard, L., Fidler, D. J., Daunhauer, L. A., Mervis, C. B., Becerra, A. M., Abbeduto, L., & Thurman, A. J. (2022). Patterns and predictors of adaptive skills in 2- to 7-year-old children with Down syndrome. Journal of Neurodevelopmental Disorders, 14(1), 18. https://doi.org/10.1186/s11689-022-09430-4

Sterling, A., & Warren, S. F. (2014). Maternal responsivity in mothers of young children with Down syndrome. Developmental neurorehabilitation, 17(5), 306–317. https://doi.org/10.3109/17518423.2013.772671

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors addressed the concerns that I have raised in a satisfactory manner. 

I have no other questions.  

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have responded thoughtfully and thoroughly to my prior concerns. In particular, they now explicitly acknowledge the limitations imposed by the small sample size, multiple regression analyses, and cross-sectional design, and they appropriately temper their language and conclusions throughout the manuscript. The analytic strategy is more clearly justified, and the study is now framed transparently as exploratory and hypothesis-generating rather than predictive or mechanistic. Conceptual limitations related to the absence of DS-only versus DS+ASD group comparisons and limited sample diversity are clearly articulated. Overall, the revisions substantially improve the rigor, clarity, and interpretive caution of the manuscript, and my major concerns have been adequately addressed.