This multi-site randomized controlled clinical trial, single-blind, parallel group design was prospectively registered in the Clinical Trials Registry (NCT02593825). This study presents a pre-planned secondary analysis from the START-Play clinical trial.9 Intervention and assessments were performed across five sites in the United States [Duquesne University, Pittsburgh, PA; University of Delaware (UD), Newark, DE; University of Nebraska Medical Center (UNMC), Omaha, NE; University of Washington (UW), Seattle, WA; Virginia Commonwealth University (VCU), Richmond, VA]. The study was completed in accordance with regulations set by the Duquesne University (protocol #2015/06/5) and VCU (protocol # HM20005431) Institutional Review Boards. Parents provided informed consent. Data were analyzed at the University of Nebraska-Lincoln (UNL), Lincoln, NE.

Children with motor delays were recruited during 2016–2019 across five sites. Eligibility criteria were: 1) 7–16 months of corrected age at baseline; 2) ability to sit with or without arm support for at least 3 seconds; 3) inability to transition in or out of sitting; 4) presence of spontaneous arm movements; and 5) gross motor delays evidenced by a scaled score of more than 1SD below the mean for the gross motor subscale of Bayley Scales of Infant and Toddler Development, 3rd Edition (Bayley-III).13 Exclusion criteria were significant visual disorder, ASD, Down syndrome, spinal cord injury, uncontrolled seizure disorder, or neurodegenerative disorders.

An a priori power analysis was performed for the primary outcomes (e.g., reaching, sitting, problem-solving, and global development)9 and a sample size of 152 children was suggested to detect intervention effects (two-tailed, α=0.05, power≥.80, attrition rate 8%).13,14 Initially, 155 potential participants were assessed for eligibility; 134 children were tested at baseline; 22 children were excluded from analysis (Supplementary material – Fig. 1S). The remaining 112 participants were classified as having mild (n = 62) or significant (n = 50) motor delays based on the Bayley-III motor composite score.9 Group comparisons are underpowered, and implications are discussed in the limitations section.

After the baseline assessment, concealed random assignment was performed as predetermined by an investigator not involved in data collection. Sequentially numbered, sealed, and opaque envelopes were used. All participants were randomly assigned in blocks, with stratification into groups of movement ability9 to either usual care early intervention (UC-EI, n = 55) or START-Play (START-Play, n = 57) group to achieve equivalent groups.

The START-Play intervention was provided twice weekly (40–60 minutes/session) throughout the first 12 weeks of the study by trained physical therapists at each site in collaboration with parents/caregivers, using key ingredients of START-Play.9,13,15 Children assigned to START-Play continued receiving their usual care (i.e., early intervention and/or outpatient therapy services). Those in UC-EI received only their usualcare services.

Children were assessed in their homes by trained examiners blind to children's intervention group at baseline and 1.5, 3-, 6-, and 12-months post-baseline (mean=4.5, SD=1.1 visits per child; 11.3% missing visits). At each visit, children were seated in a portable highchair (Fig. 1) and engaged in a MEPS task assessing their ability to pull a towel (means object) to obtain a toy (end object).3,4 Children were provided three 30-second trials with a small, interesting toy placed out of reach on the far end of the towel. Trials ended when the child touched the toy, moved any of the task objects out of reach, or 30 seconds elapsed.1,16

Fig. 1.

Experimental setup for the means-end towel-pulling task.

(0.23MB).

Testing procedures were video-recorded, and behaviors were scored from videos by trained experimenters blind to intervention group allocation using the MEPSAT. Videos were recoded for intra- and inter-rater reliability. Strong intra-rater [ICC= 0.996 (0.992, 0.999)] and inter-rater [ICC=0.997 (0.995, 0.999)] reliability evidence of means-end performance scores was found.2

The MEPSAT outcomes were: 1) MEPS learning: the determination of successful MEPS was based on a rubric incorporating three outcomes measures: a) Toy contact: contacting the toy by pulling the towel (yes vs no); b) Looking at the toy: looking at the toy in the five seconds prior to contacting it (yes vs no), c) Intentionality: attempting to retrieve the toy with clear intention, meaning focus on the toy, interacting with the towel to retrieve the toy, and looking at and grasping the toy (yes: clear intention vs no: no evidence of intention or unclear/ambiguous intention). MEPS learning occurred when contact was made with the toy, visual attention to the toy occurred within the five seconds before contact, and intention to retrieve the toy was rated as “clear” for a trial; and 2) MEPS level of performance: rated on a scale of 0–9 using a decision tree with 9 marking the highest level of performance (e.g., 0: no towel contact, 9: toy contact, looking at the toy).2 All variables were averaged across the three 30-second trials for statistical analyses.

MEPS mastery achievement occurred at the first visit when the child scored at the highest level of means-end performance (score 9) in at least two out of three trials. Note that mastery was not synonymous with perfection in the task. Mastery was defined based on one MEPS variable (level of performance: toy contact, while looking at the toy; successful trial) and allowed room for unsuccessful performance on one trial. Previous research showed that a score of nine in two of three trials likely represents intentional performance for young children.2 Therefore, it was possible for participants to show improvements in any MEPS variables even after achieving mastery.

Piecewise linear mixed modeling was performed using Mplus version 8.5 (Muthén & Muthén, Los Angeles, CA). Linear mixed modeling accounted for repeated measures within each child. Piecewise modeling, using individually varying time points, allowed estimation of the intercept (baseline), slopes during the intervention (baseline to 3-month visits), and post-intervention phase (3–12-month visits) for the MEPS outcomes.

Independent variables included in the models were: 1) two-pieced individually varying time points to estimate two slopes (Time1: baseline to 3-months, and Time2: 3- to 12-months); 2) intervention effect (INT: 0=UC-EI; 1=START-Play); 3) child's motor delay severity (SEV: 0=mild; 1=significant); and 4) MEPS mastery timing (MST: Early mastery: mastery achieved at baseline or 1.5-month visit; Late mastery: mastery achieved after 1.5-month visit or not in this study). The 1.5-month visit was chosen as a cut-off point to dichotomize MEPS mastery timing based on our knowledge of typical and atypical developmental trajectories for MEPS.1,2,16

Two statistical models were performed. Model 1 evaluated severity and intervention effects across baseline, Time1, and Time2. To test whether the timing of MEPS mastery affected MEPS developmental trajectories, Model 2 evaluated the main effects of severity and intervention, as well as SEV*MST and INT*MST interactions across baseline, Time1, and Time2. Both models controlled for children's age at baseline and recruitment site.

Adhering to an intention-to-treat perspective, all cases were included in the analyses regardless of missing visits17 using full information maximum likelihood (FIML) estimation. Due to the complexity of the models, which included random effects and individually-varying times of assessment, a residual analysis could not be performed in Mplus. However, non-normality was suggested by univariate descriptive statistics and histograms, so robust FIML (denoted MLR in Mplus) was used to obtain standard errors and test statistics robust to non-normality. Statistical significance was based on α=0.05. Hedges g was calculated as a measure of effect size, with standardized differences of 0.20, 0.50, and 0.80 being interpreted as small, medium, and large effects, respectively.18 Positive g values signify the magnitude of the effect favoring children with significant motor delays and in the START-Play group, whereas negative g values denote the magnitude of the effect favoring children with mild motor delays and in the UC-EI group.

Estimated statistical parameters for Model 1 are summarized in Table 3, presented in Supplementary material - Table S1, and illustrated in Fig. 2A-B. No differences were found between the START-Play and UC-EI intervention groups for either MEPS outcome at baseline, during the intervention, or post-intervention.

Fig. 2.

Estimated modeled developmental trajectories for means-end learning and performance by baseline motor severity (mild vs significant motor delays); first averaged across MEPS mastery timing levels (Model 1; A: MEPS learning; B: MEPS performance), and then presented according to the timing of MEPS mastery (early vs late mastery) (Model 2; C: MEPS learning; D: MEPS performance).

(0.41MB).

Statistically significant substantial clinical effects were observed between children with mild vs significant motor delays. At baseline, children with mild delays outperformed those with significant delays in MEPS learning (g=−1.08, large effect; p< 0.001) and performance (g=−1.21, large effect; p< 0.001). During the intervention phase, both motor delay severity groups improved in MEPS learning and performance at the same rate of change. However, during the post-intervention phase, statistically significant substantial clinical effects were detected between the two severity groups in their rates of change for MEPS learning (g = 0.50, medium effect; p = 0.033) and performance (g = 0.81, large effect; p< 0.001): while children with mild motor delays had plateaus in these two outcomes, their peers with significant delays continued improving their MEPS learning and performance.

Estimated statistical parameters for Model 2 are summarized in Table 4, presented in Supplementary materials - Tables S2 and S3, and illustrated in Fig. 2C-D. Model 2 showed no significant INT*MST interaction for either MEPS outcome at baseline, during intervention, or post-intervention. This indicates that the intervention effect did not significantly differ for infants who achieved MEPS mastery early vs. late. In addition, when stratifying by MEPS mastery (i.e., when looking at the intervention effect separately for infants who achieved MEPS mastery early vs. late), there were no significant intervention effects at baseline, during intervention, or post-intervention in either mastery group.

Model 2 also showed no significant SEV*MST interaction for either MEPS outcome at baseline, during intervention, or post-intervention, indicating that the severity effect did not significantly differ for infants who achieved MEPS mastery early vs. late. When stratifying by MEPS mastery, among children who achieved mastery late, there was a significant and clinically meaningful difference favoring children with mild motor delays at baseline in MEPS learning (g= −0.67, medium to large effect; p = 0.019) and performance (g= −0.82 – large effect, p = 0.003). Among children who achieved MEPS mastery early, the difference between children with mild vs. significant motor delays was not statistically significant and was smaller in magnitude but still clinically meaningful for MEPS learning (g= −0.40, small to medium effect; p = 0.299) and performance (g= −0.59, medium effect; p = 0.163).

During the intervention phase, no differences by severity were observed among children who achieved mastery early or among children who achieved mastery late. During the post-intervention phase, among children who achieved mastery late, children with significant delay demonstrated greater improvement in MEPS performance than children with mild delay; this difference was clinically meaningful but not statistically significant (g = 0.47, small to moderate effect; p = 0.090). No statistically or clinically meaningful severity effects during the post-intervention phase were observed for MEPS learning among children who achieved mastery late, or for either outcome among children who achieved mastery early.

There are potential limitations of this study. First, the number of participants included in the final analyses was less than the number suggested by an a priori power analysis, suggesting the study was underpowered. However, this study includes a secondary data analysis, whereas the power analysis was based on available data from preterm infants14 and estimated for the primary outcomes.9,13 Those data may not be representative of the children who ultimately participated in the study or for the MEPSAT outcomes. Regardless, the effects' clinical magnitude and statistical significance should be considered, and follow-up studies with larger sample sizes should be conducted to evaluate whether the findings replicate. Second, more than two-thirds of the children with mild delays achieved MEPS mastery early. Although it was possible for children to show improvements in MEPS outcomes even after achieving mastery, we cannot rule out the possibility of a ceiling effect for children with mild motor delays. Third, the intervention doses were not matched between the two groups. Future studies should investigate comparison with dose-matched intervention for this population and for other children with different motor and cognitive abilities. Lastly, our analyses controlled for children's age at baseline and recruitment site but did not control for socioeconomic status or dose of intervention. Future studies should evaluate the effect of these potential moderators.