Evaluation of a Novel Intervention for Infants At Risk for Neurodevelopmental Disorders

Rationale:

Providing intervention during infancy, before the full emergence of the symptoms that would
lead to a diagnosis of ND based on a behavioral phenotype (e.g., ASD, language disorder, or
attention-deficit/hyperactivity disorder) is supported by 4 premises: (1) The first two years
of life are an especially active period of neural development. Due to rapid synaptic
proliferation and experientially-influenced shaping of functional connectivity, interventions
initiated in infancy may be powerful in promoting more typical neural connectivity (2)
Biologically-based differences in infants at-risk for NDs lead to observable differences in
sensory reactivity and communication behaviors in most infants by 9-15 months, prior to the
full emergence of diagnostic symptoms. (3) Differences in infant behaviors influence the
quantity and quality of parent responses. (4) Parent-child transactional processes begin
early in infancy and impact long-term child outcomes. Based on these premises, the
investigators propose a "proof of concept" evaluation of a novel intervention, Parents and
Infants Engaged (PIE), for prodromal infants at-risk for NDs. PIE is designed to directly
impact parent responses to behaviors commonly observed in infants at-risk for NDs. Without
intervention, these behaviors may fail to elicit parent responses that efficiently scaffold
child communication development. Extensive research shows positive associations between
caregiver responsiveness and child communication outcomes. Responsiveness is defined by
multiple dimensions (i.e., sensitivity, contingency, encouragement, matching
interests/activity level, physical affection, quality of language input [e.g., verbal
scaffolding], reciprocity, and shared control), which vary within and across caregivers.
Children play an active role in eliciting responses from caregivers, emphasizing the
co-regulatory or transactional nature of these interactions. This includes biobehavioral
co-regulation of arousal levels. Whereas much research comes from studies of typical
development, similar transactions occur with young children with NDs. Parent responses vary
depending on the preceding behaviors of the child. For example, parents are more likely to
respond, and to give a verbal response, to their one-year-olds' gestures than nongestural
communicative bids (vocalizations, gaze, actions); also, adults are more likely to respond to
infants' speech-like than nonspeech-like vocalizations. Parent responsiveness, in turn,
predicts communication outcomes of children with varied NDs.

Aims:

Specific Aim 1: Evaluate the differential changes in attuned parent responsiveness following
coaching on two PIE domains - responses to variable infant (a) sensory reactivity (SR) or (b)
prelinguistic communication (PC) - as well as cumulative changes in attuned parent responses
following coaching on both PIE domains.

Specific Aim 2: Estimate the separate and combined effects of PIE intervention domains on
parent-infant engagement and infant-initiated communication with parents.

Specific Aim 3: Determine the extent to which autonomic indices of infant self-regulation
change over the course of the PIE intervention.

Recruitment:

To identify infants at-risk for neurodevelopmental disorders, the investigators will use a
population-screening method based on birth records in North Carolina, supplemented with
distribution of postcards/flyers through physician's offices and public health clinics and
email and listserv announcements. Completed FYIs will be scored and screened for risk status.
Infants who score at-risk will be flagged, and those families will receive a phone call
informing them of the results of the screening (by a qualified/trained project coordinator),
and they will be invited for a more comprehensive developmental assessment.

Summary of Measures to be Completed at Each Assessment Time point:

Pretest:

- Full MSEL

- SPA

- BOSCC

- RSA/SCL Protocol

- Parent-Child Interaction

- Attention Following Protocol

- SEQ

- MB-CDI

- Parent Stress Scale

Posttest 1 (6-8 weeks after pretest):

- Parent- Child Interaction

- SEQ

Posttest 2 (13-16 weeks after pretest):

- MSEL Receptive and Expressive Language

- SPA

- BOSCC

- RSA/SCL Protocol

- Parent-Child Interaction

- Attention Following Protocol

- SEQ

- Intervention Rating Profile (intervention group only)

The intervention phases of the study will use a randomized comparative trial design, with two
phases. For Intervention Study Phase 1, dyads will participate in the Pretest assessment
battery; then the project methodologist will randomize families of eligible infants,
stratifying randomization by age (<13 months, 30 days or ≥ 14 months). Families will be
randomized to one of two treatment arms: Arm 1 families will participate in initial coaching
on the SR domain of PIE, and Arm 2 families will participate in initial coaching on the PC
domain of PIE. Families will participate in 6 coaching sessions in their respective treatment
arms, and then return for Posttest-1 (to test the separate impacts of the PIE content domains
on parent responses and infant outcomes). For Intervention Study Phase 2, dyads in Arm 1,
will receive 6 coaching sessions on the PC domain, and dyads in Arm 2 will receive 6 coaching
sessions on the SR domain; coaching for each group will also review their respective content
domain coached in Intervention Study Phase 1. Then families will return for Posttest-2 (to
evaluate the effects of the full PIE intervention).

Data Analyses:

All data will be cleaned and inspected for outliers, missing data and distributional
irregularities. Where error distributions potentially deviate from normality, or
heteroscedasticity is suspected, the tests of the contrasts will be conducted using exact
(resampling-based) nonparametric methods. Proportional outcomes will be arc sin transformed
before entering them into the models. For the frequency count outcome, if counts are not
sufficiently large (e.g., ≥ 8) that a normal approximation to a Poisson error distribution is
appropriate, Poisson or negative binomial regression methods or nonparametric approaches may
be employed.

All primary analyses will be conducted to explore a priori contrasts of interest within a
repeated measures framework. Of primary interest will be the contrast of Pretest to
Posttest-1 scores and the contrast of Pretest to Posttest-2 scores (time effect). In
addition, the models will include terms for treatment arm effects and treatment arm-by-time
interactions.

At Posttest-1, the investigators anticipate statistically significant time, treatment arm,
and arm-by-time interactions, with parents showing differentially greater attuned responses
to infant SR or PC, consistent with their respective treatment arm. By Posttest-2, however,
the investigators predict group equivalence in responsiveness, with no differential time
effects. A follow-up test of the contrast between Posttest-1 and Posttest-2 will verify that
parents in Arm 1 maintained their attuned SR responsiveness gains, while parents in Arm 2
"caught up" on attuned SR responsiveness, and vice versa for PC attunement. For H2a, the
investigators expect increases in joint engagement from Pretest to Posttest-1, with
additional increases at Posttest-2 (time effects), and no differential treatment arm effects
or interactions at either Posttest. For H2b, the investigators anticipate minimal increases
in infant intentional communication at Posttest-1 in either treatment arm, with comparable
improvements (time effects) in each group at Posttest-2. Similarly, on the physiological
measures addressing H3a (measured only at Pretest and Posttest-2) the investigators
anticipate comparable improvements in RSA and SCL (time effect) in both arms.

Although the investigators are performing multiple statistical tests across outcomes and
hypotheses in addressing the specific aims, in this project it is more important to avoid
overlooking statistical signals of effectiveness of the innovative PIE intervention approach
(Type II errors), than avoiding false assertions of effectiveness (Type I errors). The
investigators believe it is premature, therefore, to employ conservative adjustments to the
Type I error rates in statistical tests, which would compromise the power of statistical
tests to detect such signals. Assuming recruitment of 44 infant-parent dyads and allowing for
a 9% (n=4) dropout rate, which is consistent with high retention rates (> 95%) in our prior
intervention studies, 40 dyads (20 per treatment arm) will have complete data for analyses.
Assuming a conventional Type I error rate of .05, and intercorrelations among the repeated
measures from .3 to .7, the magnitude of treatment group effects detectable with a .80
statistical power will range from f=.37 to .42, respectively, which are large standardized
effects. The magnitude of time effects and time-by-treatment arm interaction effects
detectable with a .80 power will range from .27 to .18, also respectively, which are
medium-sized standardized effects. Thus, the study is underpowered to detect any but large
differences between the treatment arms and medium-sized or larger effects on time and time-by
treatment arm interactions. More importantly than the statistical comparisons, though, the
analyses will yield key descriptive characterizations of the treatment-arm effects at the two
posttest points (i.e.,means, medians, proportions, and variances) on key outcomes, as well as
effect size estimates that can be used in planning a larger scaled efficacy trial of the PIE
intervention.

Inclusion Criteria:

- For the intervention trial: infant must meet risk criteria on the First Years
Inventory (Calculated based on data collected in another study where risk status on
the FYI was confirmed with follow up at 3 years. We empirically determined
combinations of the two FYI domain scores that would sort respondents into "at-risk"
and "not-at-risk." The resulting cut-points enabled us to sort children such that we
capture about 1/3 of those who would go on to be confirmed as at-risk while
misclassifying less than 5% of the TD sample as a-risk) AND must score at least one
s.d. below the mean on either the Receptive or Expressive subscale t scores AND meet
the following SPA inclusion criteria on HYPO or HYPER:

- "HYPO": Cut-point (equal or greater than) of 1.69 for Mean of the raw orienting
score across 7 items, each with the range of 1 to 4 possible points)

- "HYPER": Cut-point (equal or greater than) of .333 for Mean of the raw
approach/avoid novel toys score across 9 items, each with the range of 0 to 2
possible points) OR Any clear "defensive" response on orienting items or "Yes" to
covering ears to sound (in stereotypies checklist)

- Inclusion criteria for the low risk control group will be scoring below the 95th
percentile on the FYI between 11 and 16 months. Inclusion criteria for the high risk
but ineligible for intervention control group will be scoring at or above the at-risk
score on the FYI but not meeting criteria on the SPA and/or MSEL during their
behavioral assessment.

Exclusion Criteria:

- families who speak English < 50% of the time at home

- infants with previously identified genetic disorders (e.g., Down syndrome)

- infants with identified vision/hearing/physical impairments.