Development of a Therapeutic Device to Improve Speech Sound Differentiation in Preterm Infants

The goal of this clinical study is to test a new, novel medical device designed to improve speech sound differentiation among hospitalized preterm infants. The device is designed to be used at an age equivalent to 32 weeks of gestation or older and to integrate readily into clinical practice for use by nurses and therapists staffing Level II to Level IV NICUs. Preterm born infants are at high risk for neurosensory impairments and developmental delays. In the NICU, infants are often deprived of infant-directed parental speech because of numerous challenges to parental visitation, resulting in reduced differentiation of speech sounds, altered brain structure and poor language outcomes. The study will explore the effectiveness of a novel medical device designed for infant learning through contingent sucking on a pacifier equipped with a sensor for suck pressure/timing, connected to a speaker that delivers mother's voice. The study will test the hypothesis that there will be a greater response difference between speech sounds on EEG, for infants receiving the suck-contingent mother's voice intervention than for infants hearing the same amount of non-contingent mother's voice from a speaker device.

Every year in the United States, approximately 10% of infants are born preterm, impacting nearly 400,000 families in the US alone. These infants often experience significant cognitive and language delays in early childhood, with negative implications on social and academic function at school age and beyond. Complex higher-order functions, such as language, build on early, more basic neural processes that lay the foundation for later connections and systems. A fundamental neural processing function necessary for the development of cognition and receptive language is the early ability to differentiate speech sounds from each other (above hearing, a simpler perception of sound). In preterm infants, speech sound processing is frequently deficient relative to full-term peers at discharge from the neonatal intensive care unit (NICU) and is associated with worse outcomes at two years of age. Time spent in the NICU after birth does not compensate for atypical speech sound processing, and by early childhood, many preterm infants require targeted therapies. Currently, no technological intervention addresses the problem of preterm infants' poor speech sound differentiation during NICU hospitalization, when the greatest potential for improving this basic neural process exists. Root of the problem: Contributing to poor speech sound processing of preterm infants is an atypical NICU auditory environment, experienced during a time of neural vulnerability and critical developmental phases. The NICU environment is rich in machine sounds or human voices which, unfiltered by the uterine environment, are often above recommended volume levels. Conversely, the NICU is poor in sounds that would normally wire developing connections in the newborn brain, such as human voices in a natural environment. The optimal auditory stimulus for infants after birth would be their parents' voices, at low volume and addressed in infant-directed and responsive patterns. However, despite supportive NICU policies, few parents can consistently be present to provide this voice intervention, due to social, geographic, and economic imperatives. Social and psychological determinants of health further aggravate the problem. Solutions: Providing infants with their parents' recorded voice when they are awake, physiologically stable, and receptive can help their developing auditory system receive appropriate experiences. In addition, if the voice is: 1) infant-directed; 2) at volumes and rhythms calibrated against naturally occurring stimuli; and 3) contingently provided upon infant request (consistently delivered after the same easy action, within a very brief timeframe), the conditions of optimal learning of speech sound in developing humans are met. Challenges to the design of effective interventions: Safety concerns exist in placing most speaker devices in incubators and cribs, namely infection risks from devices that cannot be fully sanitized and provision of loud/unfiltered sounds at times when an infant is not receptive, causing physiologic instability. Efficacy concerns exist regarding the provision of uncalibrated content when the neural circuitry of auditory processing is immature and inefficient. In addition, active learning has been demonstrated in infants and young children as far superior for brain plasticity and development compared to passive exposure to recorded material. Finally, measuring intervention effects in preverbal infants is inaccurate using standard behavioral assessments. In this investigation, the clinical study team will leverage strong evidence that shows reading and singing (calibrated against natural environment occurrences) are beneficial for cognitive and language development in preterm infants when mother's voice is directed to the infant upon request. The technology investigated is designed to support infants in the NICU while engaging their mothers. The technology is designed to optimize learning experiences by controlling the level and duration of the intervention; The clinical study team partners with NICU nursing teams to schedule delivery of mother's voice during stable and receptive states. The intervention device is designed to ensure learning through contingent sucking on a pacifier equipped with a sensor for suck pressure/timing, connected to a speaker that delivers mother's voice. The technology is proven sanitizable, to meet NICU infection-control standards. Through this design, parents are empowered to help their children even when they do have the means, ability, or time to stay in the hospital every day - this technology supports NICUs in providing both excellence and justice in care. Because this study includes a population with higher diversity of social determinants of health compared to national standards, and because of a family support process already in place at the enrolling study sites, the clinical study team has made efforts to ensure participation in research for all, equally, improving the generalizability of the resulting findings. Finally, the clinical study team measures outcomes objectively, using brain-based (EEG) measures of speech sound processing. These measures have high construct validity in preterm infants for the study purposes because they: (1) are elicited consistently across the spectrum of prematurity and settings (2) reflect degree of maturity of language processing (3) are highly predictive of receptive language scores in the first two years, and (4) are responsive to treatment with a 3-week contingent-voice paradigm. The study is designed to demonstrate the efficacy of the technology in a randomized controlled trial using rigorous protocols; the study design respects standard care that most often incorporates infant's passive listening to parent-recorded voice (providing a more conservative measurement of the intervention's effect than if it were compared it to no exposure at all); the study design investigates potential effects of the NICU environment such as room types (single vs. open bay) and ambient noise. The study aims follow. The study will test the hypothesis that there will be a greater response difference between speech sounds on EEG, for infants receiving the suck-contingent mother's voice intervention than for infants hearing the same amount of non-contingent mother's voice. The study will test whether the intervention effect size increases when NICU background noise is lower. The study will explore whether the intervention effect size varies by NICU room type.

Participants will receive 20 minutes of non-contingent recorded mother' voice during two 20-minute sessions, with a maximum of 2 sessions per day. Recordings are played through the smallTalk speaker device in passive mode, which limits play time to 20-minute exactly and volume to 45 dB in the A-weighted scale as per American Academy of Pediatrics recommendations. For these sessions, the therapist will remain at bedside with the infant and the device, as if they were administering the intervention. Nurses do not typically remain at bedside for the procedure due to workload issues and the low risk of the intervention.

Participants will receive 20 minutes of contingent recorded mother' voice also at 45 dBA, during two 20-minute sessions, with a maximum of 2 sessions per day. The smallTalk Active system integrates a wireless, lightweight and sealed sensor unit that securely fits into a Philips NICU Soothie pacifier. The speaker device is factory set to communicate constantly with the sensor unit, and to only deliver a predetermined 10 seconds of recorded parent's voice upon detection of a suck that meets a pressure threshold, which is automatically set by the speaker device.

The novel smallTalk NICU Active product design allows a disposable pacifier (equipped with the smallTalk sensor) to act as an infant-controlled mechanism for administration of developmentally appropriate parental voice, delivered by the NICU-safe speaker contingent upon the infant suck strength meeting an individually calibrated threshold.

A high-density array of 128 electrodes embedded in soft sponges (high-density 64-electrode waveguard™ net, ANT Neuro, Hengelo, Netherlands) will be used to record ERPs with a sampling rate of 1000 Hz, filters set to 0.1-400 Hz. Recording of brainwaves will be controlled by eego™mylab system (v. 9.2, ANT Neuro, Hengelo, Netherlands). E-Prime (v. 4.0, PST, Inc., Pittsburgh, PA) software will control stimulus delivery. The stimuli will be presented at 60 dBA. The ERP speech sound paradigm included a computer-generated woman's voicing one of six syllables (/ba/, /da/, /ga/, /bu/, /du/, /gu/) randomly and at random intertrial intervals for a total of 25 trials per syllables over 7-10 minutes. Inter-trial intervals will be varied between 1600ms and 2600ms. All stimuli are presented thus to prevent habituation.

Initial ERP Test is performed 2 days prior to intervention. Intervention timeframe is 20 sessions across 3 weeks. ERP Test is repeated within 24 hours of the last session.

We will test whether the effect size of the between-treatment-group, pretest-adjusted effect will vary by amount of background noise, favoring lower background noise. During each session, a sound level meter will record sound amplitude exposure at the baby's head. Sound levels will be logged onto a built-in SD card at 15 min intervals and downloaded after a 24-hour period. Measurements will occur every 3 days. Average sound exposure will be calculated from all measurements over the course of the intervention.

Within 3 months of the full enrollment of study, we will assess if the primary outcome measure is impacted by the level of NICU background noise.

This aim is only exploratory and will investigate the possibility that the between-treatment-group, pretest-adjusted effect size will vary by the room setting in which mother's voice exposure occurred.

Within 3 months of the full enrollment of study, we will assess if the primary outcome measure is impacted by the NICU room type

Inclusion Criteria: CGA 32 0/7-35 0/7 weeks at study start, and GA 35 0/7 weeks at birth Exclusion Criteria: Ventilation using an endotracheal tube, major congenital malformations, family history of genetic hearing loss, and use of sedative/seizure medications (medications potentially masking ERP measured sensory processing) and severe white matter injury as it increases the likelihood of hearing deficits.

All reasonable requests from qualified scientists for unique research resources (protocols and expertise) developed with NIH funds for research purposes will be honored.

All reasonable requests from qualified scientists for unique research resources (protocols and expertise) developed with NIH funds for research purposes will be honored. We will respond to requests in a timely manner. We will adhere to the NIH Grant Policy on Sharing of Unique Research Resources including the Sharing of Biomedical Research Resources Principle and Guidelines for Recipients of NIH Grants and Contracts. There is no unique biological information that could be made available to the scientific community. Data sharing will be made available by publishing articles in scientific publication, presenting research findings at scientific meetings, and by responding directly to data requests by other investigators. De-identified assessment data will be retained in REDCap and these data will be made available to other investigators who make specific inquiry for a reasonable purpose.

Institute of Medicine (US) Committee on Understanding Premature Birth and Assuring Healthy Outcomes; Behrman RE, Butler AS, editors. Preterm Birth: Causes, Consequences, and Prevention. Washington (DC): National Academies Press (US); 2007. Available from http://www.ncbi.nlm.nih.gov/books/NBK11362/