Modifying quantitative sensory testing to investigate behavioral reactivity in a pediatric global developmental delay sample: Relation to peripheral innervation and chronic pain outcomes

Abstract

Early tactile and nociceptive (pain) mechanisms in children with global developmental delay at risk for intellectual and developmental disability are not well understood. Sixteen children with global developmental delay (M age = 5.1 years, SD=1.4; 50% male) completed a modified quantitative sensory testing (mQST) protocol, an epidermal (skin) punch biopsy procedure, and parent-endorsed measures of pain. Children with reported chronic pain had significantly greater epidermal nerve fiber (ENF) density compared to children without chronic pain. Based on the mQST trials, ENF density values were associated with increased vocal reactivity overall and specifically during the light touch and cool thermal stimulus trials. The findings support the feasibility of an integrative bio-behavioral approach to test nociceptive and tactile peripheral innervation and behavioral reactivity during a standardized sensory test in a high-risk sample for which there is often sensory dysfunction and adaptive behavior impairments.

Our scientific and clinical understanding of pain in general and nociceptive processing in particular remains relatively limited among children with intellectual and developmental delay and disability (IDD). Although a great deal of progress has been made in understanding the developmental biology of the somatosensory system and its nociceptive capacities early in human development, these observations have not been readily translated into insights specific to children with significant developmental delays at risk for intellectual disability. Typically developing neonates have lower mechanical thresholds, exhibit faster and larger responses dependent on stimulus intensity, and sensitize rather than habituate to repeated mechanical stimulation (Andrews & Fitzgerald, 1999; Willer, 1985). Somatosensory reactivity typically approaches maturity by about 42 weeks postconceptional age, with a gradual decrease in excitability as well as reorganization of spinal cord connections at that time (Andrews & Fitzgerald, 1999; Jennings & Fitzgerald, 1998).

The neurobiological mechanisms underlying the normative tactile and nociceptive developmental processes appear to be, in part, experience-dependent processes at the spinal cord level (Koch & Fitzgerald, 2013). There is ample evidence showing that early adverse experience (e.g., invasive long-term neonatal intensive care) is developmentally disruptive for somatosensory maturation such that developmentally expected increases in cutaneous thresholds, for example, do not always occur or are otherwise affected (Andrews & Fitzgerald, 1999; Fitzgerald, Millard, & McIntosh, 1989). Less clear is what the range of normal variability might be with respect to early somatosensory function and whether there are early functional differences already present in very young children for whom development is delayed. Part of the investigational constraint is practical – it remains difficult to interrogate somatosensory function and sensory experience in clinical pediatric populations with developmental delays and any associated motor, communicative, and cognitive impairments.

As a preliminary step toward improving our understanding of somatosensory function in very young children with delayed development, we adopted a quantitative sensory testing (QST) approach and adapted it (described in detail below) to accommodate associated developmental impairments to then test the relation between sensory reactivity with cutaneous innervation and epidermal nociceptors. Current sensory assessment approaches used in developmental disabilities are most often based on proxy report or tests of tactile preference. Alternatively, in some developmental disability groups (e.g., cerebral palsy [CP]), sensory testing research has relied on sensory stimuli designed to assess impaired discriminative tactile abilities such as two-point discrimination, texture perception, and shape perception mostly specific to large fiber afferent function. There has been little work incorporating sensory testing approaches that simultaneously evaluate loss and gain of function reflecting both large and small fiber afferent function. QST approaches provide the opportunity to evaluate the expression of sensory signs in relation to gain or loss of function and reveal clues about underlying somatosensory pathophysiology.

Critical to our long-term interest in possible somatosensory pathophysiology associated with developmental delay is a specific focus on cutaneous innervation including both large (myelinated) and small (unmyelinated) fibers. Given the almost complete absence of integrative bio-behavioral QST applications in pediatric pain and IDD research, our approach was discovery-oriented and patient-focused. Using a cross-sectional design with a clinical sample of young children with global developmental delays, we explored three sets of hypotheses. Specifically, we hypothesized that the observed behavioral reactivity to the mQST would, in part, be associated with differences in epidermal nerve density (ENFd) as a measure of C-fiber innervation. Similarly, we hypothesized that cases with reported chronic pain problems would be characterized by differences in ENFd values compared with cases with no reported chronic pain problems. And last, we hypothesized that directly measured sensory reactivity profiles would be related to proxy-reported pain problems; specifically, that observed behavioral reactivity as partial evidence for sensory gain would be associated with proxy-reports of increased pain symptoms. Sex as a biological variable and developmental risk severity differences were tested for all comparisons.

Method

Participants and Settings

Following Institutional Review Board approval, 16 children aged 3 to 7 years of age (50% male; M age = 5.10 years, SD=1.35) with global developmental delay were recruited from Gillette Children’s Specialty Healthcare. Parents were given the opportunity to experience the sensory testing stimuli prior to giving consent for their child to participate. Children were Caucasian (n=11), African American (n=3), Asian (n=1), and Hispanic/Latino (n=1). Participant characteristics (developmental diagnosis, special education services, birth history, and pain experience) are reported in Table 1. Sensory testing was performed at Gillette Children’s Specialty Healthcare (Gillette) in a clinic room or in a family consult room. Parents and sometimes siblings were present during the sensory test. Overall, the testing space was quiet with minimal distraction.

Table 1.

Sample medical chart diagnoses

	Total (n= 16)
Diagnosis
Global developmental delay	16 (100%)
Autism spectrum disorder	3 (19%)
Genetic Syndrome	3 (19%)
Motor coordination disorder	1 (6%)
Mixed receptive and expressive language disorder	1 (6%)
Pervasive developmental disorder (NOS)	1 (6%)
Intellectual disability	1 (6%)
Oppositional defiant disorder	1 (6%)
CDI subscale scores −2SD (30%) below age (n=14)
Developmental score	10 (71%)
Expressive language	10 (71%)
Language comprehension	10 (71%)
Birth History
Gestation
Term (range = 37–40+ wks)	12 (75%)
Preterm (range = 32–36 wks)	3 (19%)
Very preterm (range = 28–31 wks)	1 (6%)
Extremely preterm (range = 23–27 wks)	0 (0%)
Birth weight
Low birth weight (less than 2500 grams)	2 (12.5%)
Not low birth weight (2500 grams or greater)	14(87.5%)
NICU admittance	4 (25%)
Mean length of stay in days (SD)	12.0 (12.83)
Range	1–30 days
Pain experience
Pain in 7 days prior to study
Yes	6 (40%)
No	10 (60%)
Chronic pain (n=15)
Yes	4 (26.7%)
No	11 (73.3%)

CDI = Child Developmental Inventory; NICU = neonatal intensive care unit.

A comparison group of typically developing children was enrolled to evaluate differences in sensory reactivity (see Barney et al., 2017 for thorough comparisons between groups on sensory reactivity). Typically developing children (n=16, aged 2 to 5 years; 56% male, M age = 3.90, SD = 1.14) were enrolled via a university-affiliated daycare and were Caucasian (n=14), Asian (n=1), and African American (n=1). Sensory testing was performed in a quiet room at the daycare. Parents were not present during the sensory test but had experienced the sensory stimuli at the time of providing consent and were given the opportunity to schedule the testing for a time when they were available to attend.

Inclusion/Exclusion Criteria

Participants were included if they were between 2 and 7 years of age and had been referred to Gillette, a tertiary center for diagnosis of a developmental disability, due to failure to meet developmental milestones. Participants were excluded from the study if they had serious accompanying acute health impairments considered to be painful (e.g. reflux or otitis as determined by subjects’ physician record review and/or examination if necessary).

Sensory Testing

Sensory testing materials.

The sensory testing was conducted by applying six calibrated stimuli to the back of the child’s left and then right calf while the child was in a seated position. A light touch was applied with a von Frey monofilament (2.0 g) pressed against the skin until the filament bent. The filament was touched lightly to the skin five times in five seconds. A light pin prick was applied for < 1 second with a plastic US Neurological pin made for use during neurological exams. A cool touch was applied lightly to the participant’s skin for five seconds using a Tip Therm cool thermal probe. The end of the cool probe is circular with a radius of ¾ inch. Because the tip maintains room temperature it is cooler than the participant’s body temperature. Deep pressure was applied using a manual algometer (Wagner model FDX) with a rubber tip with a 0.5-inch circular diameter. The pressure was applied as steady and quickly as possible to a 4lb reading and then held at that consistent pressure for the remainder of the 5 second application. A von Frey monofilament (60 g) was applied repeatedly to the skin at 1 Hz for 30 seconds. The monofilament was touched to the skin until it bent and then removed. A thermal heat probe was applied at a temperature of 50°C. The thermal probe was electronic with a metal circle approximately 3 mm in diameter that was applied to the skin. The tester gently held the participant’s ankle as needed for stability in applying the stimuli; however, the participant was able to withdraw from any of the stimuli at any time if they became uncomfortable.

For eight children with global developmental delay and five typically developing children a sham trial was also included prior to the application of any other stimuli. For the sham trial a von Frey monofilament was altered to remove the filament but otherwise looked comparable to the light touch and repeated touch applications. The sham trial lasted 5 seconds and was comparable in every way to the other applications with the exception of the missing filament. The sham von Frey was moved toward the child’s calf; however, nothing came in contact with the child’s skin. Because the sample was comprised entirely of pediatric participants and because of the range of cognitive and communicative impairments and delays, stimulus trials were modified to be time-limited (≤ 5 sec.) and were only conducted once per site per participant at a consistent intensity. This approach sacrifices our ability to measure cutaneous sensory thresholds of various modalities. Although recognizing that it does not establish threshold, it does appear promising by providing a reproducible standardized way of comparing a range of time-locked behavioral reactivity to calibrated sensory stimuli across individuals with high relevance to pain sensitivity as well as sensory function.

Sensory testing procedures.

Each participant was tested individually. The examiner brought the child into the room and spent a few minutes playing with the child to help the child become accustomed to the environment. Then the child sat on a chair so that their lower legs hung over the edge to provide access to the calf. The examiner was seated to the child’s left side and the six stimuli were arranged behind the child’s chair. The sensory stimuli were adapted in order to minimize fearfulness of the stimuli by covering with colorful toys. The pins were not adapted because each was small, disposable and applied quickly and from behind so the child did not see the application.

Participants did not observe movies, use electronic devices or other distracting toys or equipment during the test but were given a stuffed toy to hold if they chose. Children were prompted to watch a visual timer during the sensory test that signaled the duration of the test and to encourage the child to remain in the seat. The examiner announced “each of my fuzzy friends are going to touch you right here (touching left calf) and here (touching right calf)” and then the testing began. Each stimulus was audibly signaled by the tester saying a number (one through six) associated with each stimulus when the stimulus was applied to the calf and then audibly signaled by the tester saying “off” when the stimulus was removed. The sham trial was signaled as zero when it was applied. The stimuli were always applied in the following order: 0) sham, 1) light touch, 2) pin prick, 3) cool, 4) pressure, 5) repeated von Frey, and 6) heat. The array of stimuli were designed to assess A-delta, A-beta, C-fibers, and to conduct a proxy assessment of central sensitization (repetitive activation of nociceptive primary afferents leads to an increase in the excitability of neurons in the CNS that can contribute to chronic pain) using the repeated von Frey application. Stimuli were presented in order of increasing intensity while also attempting to group stimuli by the primary sensory afferent (A-delta, A-beta, etc.) being tested. The order of application from non-noxious to noxious was designed to reduce the likelihood of carryover effects in terms of behavioral reactivity.

Behavioral Measurement and Outcome Variables

Sensory reactivity.

To capture facial expression, vocalizations, and whole-body movements during the sensory testing a camera was set up approximately 2.5 meters away and focused on the participant’s whole body. Digital video was used and coded using Pro-Coder for Digital Video (PCDV) (Tapp, 2003). PCDV is a software program designed to facilitate the collection of observational data from digital media files. This system provides a keyboard-driven coding platform that enables coders to scroll through designated time windows with playback options available. Behavioral events are coded as either present or absent throughout the observational period based on specific operational definitions for each code.

The operational definitions for facial, vocal and whole-body behavioral reactivity codes were derived from the items on the Pain and Discomfort Scale (PADS) (Bodfish, 2006). The PADS was developed with items derived from the Non-Communicating Children’s Pain Checklist (NCCPC, now NCCPC-R) (Breau, McGrath, Camfield, Rosmus, & Finley, 2000; Breau, McGrath, Camfield, & Finley, 2002), and was used independently by Phan and colleagues to measure pain expression in adults with IDD during a dental scaling procedure (Phan, Edwards, & Robinson, 2005). PADS was used in the current study to measure pre-specified behavioral codes in relation to the application of calibrated stimuli - the majority of which were likely non-noxious (we address the scientific and ethical challenges in the Discussion).

The behaviors coded were selected from a non-verbal pain behavior checklist and were pain and sensory-relevant. Vocal reactivity was defined as moaning, whining, whimpering, crying, screaming, or yelling. Facial reactivity was defined as a cringe or grimace, furrowed brow, change in eyes, mouth open, tightly puckered lips, pout, quiver, clenching of teeth, grinding of teeth, or tongue thrusts. Whole-body reactivity was defined as protecting or favoring a specific part of the body, flinching, or being sensitive to touch. Each of these items was explicitly defined based on the PADS descriptions (available upon request). For example, ‘crying’ is further defined by the PADS as “louder vocalizations made with mouth open or closed, tears may or may not accompany the vocalization” and ‘furrowed brow’ is further defined by the PADS as “inner and/or central portion of the eyebrow lowers; may produce vertical wrinkles between eyebrows; or produces muscle bulge from middle of forehead above middle of eyebrow down to inner corner of the eyebrow”. In addition to the PADS vocal items, laughter was also coded as a vocal behavior because there is evidence that under some circumstances it may be considered a paradoxical expression in relation to experiencing pain or sensation (Collignon & Giusiano, 2001). Further, any child vocalizing pain or discomfort using words or sentences was coded as vocal reactivity.

Each video was coded in PCDV using four passes. On the first pass the research assistant coded the onset and offset of each sensory stimulus. On the second, third, and fourth pass the research assistant coded the presence (versus absence) of vocal, facial, and whole-body reactivity behaviors, respectively. This coding approach quantified the duration (in seconds) of each reactivity behavior (e.g., vocal) occurring in conjunction with each stimulus application (e.g., light touch). The research assistants were trained to a minimum 90% criterion across all observational codes using practice videos and demonstrated an inter-observer agreement on practice videos exceeding 90% prior to coding the videos obtained for this study. Inter-observer agreement was calculated on > 25% of all sensory tests conducted for this study. Inter-observer agreement was calculated by PCDV for each type of pain behavior (vocal, facial, and whole-body). To control for any differences between participants in the duration of the stimulus application, duration metrics were converted to time-based trial proportions. Specifically, the duration in seconds (s) of each behavior code (i.e., vocal, facial, body) occurring within a given stimulus trial (e.g., cool, pin prick) was divided by the duration of the associated stimulus trial to equal a proportion ranging from 0–1.0. When assessing for overall reactivity, proportion was calculated as [vocal (s) + facial (s) + gross motor (s)]/duration of stimulus trial (s) and proportion scores ranged from 0.0–3.0.

Severity of delay.

Severity of the developmental delay was measured using the Child Developmental Inventory (CDI) total behavioral score and expressive and receptive language scores, available for the investigation through the child’s medical record. The CDI includes 300 items that are completed by the parent or caregiver to measure the child’s development in the following areas: social, self-help, gross motor, fine motor, expressive language, receptive language, letters and numbers. The CDI also includes questions related to the child’s health, growth, development and behavior. The CDI developmental scales correlate closely with age (r = 0.84) and the results identified the 26 children who were enrolled in early childhood/special education within the normative sample (N = 568)(Ireton & Glascoe, 1995). The CDI scales correlated with reading and academic achievement in kindergarten (Ireton & Glascoe, 1995). The CDI has also demonstrated strong significant correlations with the Clinical Adaptive Test/Clinical Linguistic and Auditory Milestone Scale (CAT/CLAMS; r = .87, p <.001) and the Bayley Scales of Infant Development, 2^nd Edition (BSID-II; r = .86, p < .001), demonstrating that the CDI generates scores consistent with content and construct validity evidence specific to typical and delayed child development (Doig, Macias, Saylor, Craver, & Ingram, 1999).

Pain outcomes.

At the time of the sensory test the primary caregiver of each child completed the Dalhousie Pain Interview (DPI) (Breau, Camfield, McGrath, & Finley, 2003) and the Brief Pain Inventory (BPI) (Cleeland & Ryan, 1994). The DPI is presented in an interview/script format consisting of 10 close-ended questions that provide a measure of episodic pain in the previous 7 days as well as a description of chronic pain that has been ongoing for six months or more. This measure has been adapted from the methodology used in previous research for obtaining pain information via proxy report when self-report is not possible or otherwise difficult to obtain (Breau et al., 2003). Specific items are anchored to whether there has been pain in the past week, its general description, possible cause, duration (cumulative hours, minutes, and seconds), frequency (number of episodes), and intensity (0–10; zero means “no pain at all” and ten means the “worst pain ever”). All pain episodes reported are categorized as accidental, gastrointestinal, musculoskeletal, neurological, stretching, positioning, equipment, orthopedic, spasm, other, or unknown pain.

The BPI pain interference subscale was developed initially as a method of measuring cancer pain; however, the tool has since been revised and validated for use with individuals with developmental disability and cerebral palsy (Cleeland & Ryan, 1994; Tyler, Jensen, Engel, & Schwartz, 2002). The modified BPI pain interference items were adapted to efficiently measure the extent to which pain interfered with twelve different aspects of daily living such as communication, mobility, school, daily activities, self-care, sleep, and mood in the previous week. Each of the 12 items are scored from 0–10, 0 meaning pain “did not interfere” with that item and 10 meaning pain “completely interfered” with that item. Each item score is the 0–10 score assigned by the rater. There are no subscale scores. Each participant receives a total score based on the addition of the 0–10 scores for each of 12 items. Thus, individual total scores have the potential to range from 0 to 120. Pain interference scores on the modified BPI have shown significant correlations with pain intensity ratings and have shown excellent internal consistency (Barney et al., 2018; Osborne, Raichle, Jensen, Ehde, & Kraft, 2006; Watters et al., 2010).

In previous studies with children and adolescents with significant levels of IDD (i.e., cerebral palsy, Batten disease, and Rett syndrome) the DPI, the modified BPI pain interference, and the PADS have performed very well psychometrically. The modified BPI has demonstrated excellent internal consistency with coefficient alphas between 0.96 and 0.97. The modified BPI pain interference scores correlated significantly with parent-reported pain intensity ratings on the DPI (r= 0.76, p< 0.05) in children with cerebral palsy. In a study involving girls and women with Rett syndrome, the modified BPI pain interference scores significantly correlated with pain expression on the Non-Communicating Children’s Pain Scale-Revised (NCCPC-R total score; r=0.58, p<0.05). In the same study, both the PADS (scored by trained coders during a pain examination procedure) and the NCCPC-R (completed by parent proxies) demonstrated that girls and women with Rett syndrome are most likely to show pain using facial expression (p<0.01).

In the current study the modified BPI’s coefficient alpha was 0.90. PADS inter-observer agreement (IOA; 2^nd independent coder) for 26% of the randomly selected coded sample was as follows: IOA was 98% for vocal reactivity behavior codes, 91% for facial reactivity behavior codes, and 88% for whole-body reactivity behavior codes. The average implementation fidelity score for the mQST (i.e., % correct application for each stimulus) for the application of sensory stimuli was 98%.

Peri- & post-natal short history.

Gillette medical records were reviewed to determine 1) whether there were complications during pregnancy or delivery, including admittance to the NICU, 2) the child’s gestational age at birth (in weeks), 3) the child’s weight at birth.

Peripheral Innervation & Nociceptive Biomarkers & Measurement

We were able to take advantage of ‘standard of care’ procedures that involved sedation for clinical purposes (e.g., imaging) and obtain a small (3 mm) epidermal (skin) biopsy from the back of the calf. Skin biopsies were procured following the mQST. Because the child was sedated, the procedure was nonpainful and there were no issues related to compliance and/or procedure-related anxiety. Biopsies were made with a 3 mm biopsy tool (Acupunch; Acuderm; Fort Lauderdale, FL). From skin samples, immuno-localization procedures for confocal microscopy permitted the direct quantification of epidermal nerve fibers (ENFs) with antibody to protein gene product (PGP) 9.5 and evaluation of nerve fibers containing substance P (SP). Biopsies were fixed in Zamboni’s solution, cryoprotected, and sectioned with a freezing sliding microtome (Leica, Nussloch, Germany). Diluent and washing solutions were comprised of 1% normal donkey serum(NDS) (Jackson ImmunoResearch, West Grove, PA) in 0.1 M PBS with 0.3% Triton X-100 (Sigma, St. Louis, MO). Floating sections were blocked with 5% normal donkey serum in the diluent solution. Nerve and tissue antigens were localized using primary antibodies to PGP 9.5 (1:1000; AbD Serotec, Raleigh, NC) and SP (1:1000; Immunostar, Hudson, WI) with each diluted in PBS/Triton X-100/NDS. Secondary antibodies specific to the IgG species used as a primary antibody and labeled with Cy dye fluorophores 2, 3 and 5 (Jackson ImmunoResearch) were used to locate two antigens in each section. Secondary antibodies alone were used as immunostaining negative controls. After immunohistological processing, sections were adhered to coverslips with agar, dehydrated via alcohol, cleared with methyl salicylate, and mounted in DPX (Fluka BioChemika, Ronkonkoma, NY). Stained sections were imaged with a CARV II spinning disk confocal microscope (Becton Dickinson) at 2micron z-steps for 16 sections with a 20X objective, and the number of fibers per millimeter of epidermis were counted using Neurolucida software (MBF Bioscience, Williston, VT). The primary dependent measure for the analysis was the number of fibers per millimeter of epidermis (density) for each child. We also examined substance P-immunopositive fiber counts (as a biomarker for nociceptors) in dermal areas of the biopsies.

Statistical Analyses

Initial analyses indicated that age (r= −0.32, p=0.23) and sex (t(14)=0.28, p=0.78) were not significantly related to pain reactivity during sensory testing. Age (r=0.19, p=0.48) and sex (t(14)<0.001, p=1.00) were also unrelated to ENF density; thus, these variables were not included as covariates in further data analyses. ENF density was moderately associated with birth weight (r=−0.46, p=0.07); thus, birthweight was controlled for whenever possible during ENF related analyses. Birth weight did not correlate with height (r=0.22, p=0.46), weight (r=0.42, p=0.16), or body surface area (r=0.36, p=0.23) at the time the skin biopsy was procured. Data were tested for normality using the Shapiro-Wilk statistic. ENF density (W=0.96, p=0.73) and total duration of pain-behavior reactivity (W=0.90, p=0.19) met criteria for assumed normality. Data for parent-reported pain outcome measures did not meet criteria for assumed normality; including, pain intensity (W=0.60, p<0.001), duration (W=0.31, p<0.001), frequency (W=0.73, p<0.001) of pain episodes and pain interference (W=0.70, p<0.001). Analyses including the pain outcome measures were conducted using non-parametric tests. When checking for outliers (M +/− 3SD) it was determined that one behavioral reactivity score for children with global developmental delay (during application of heat) and six scores of typically developing children (during application of light touch, cool, and pressure) would be considered outliers. Outliers were likely more frequent in the control group because the majority of participants demonstrated little to no behavioral reactivity. Given the small sample size, the exploratory nature of the data, and the concern that we do not know the typical range of observed behavioral reactivity in typically developing children or children with global developmental delay, we opted to report on all available data.

Results

Feasibility outcomes

The sensory test was terminated prior to completion for three children. Reasons for terminating included parent request (n=1) and pronounced pain/discomfort behaviors; specifically, visible tears, repeatedly moving away from the tester, and loud vocalizations (n=2). This resulted in missing data for pressure (n=1), repeated von Frey (n=2), and heat (n=3) stimuli. In addition, the heat probe was broken during the sensory test for one child. After data collection had already begun, it was decided to include the sham stimulus trial. Thus, the sham trial was only instituted for the last 8 participants. This resulted in missing data. Sensory testing was terminated prior to completion for one typically developing child due to visible tears, resulting in missing data for repeated von Frey and heat stimulus trials. The sham trial was included for five typically developing participants.

Primary peripheral biomarkers for the sample of children with global developmental delay

The overall mean ENF density was 92.53 fibers/mm (SD=27.73; range=49–141). The mean SP positive fiber count was 9.88 (SD=5.08; range 3–20). Representative confocal images and sample quantification are reported in Figure 1 and Table 2, respectively.

Figure 1:

Confocal images of immunostained calf skin biopsies from two participants representative of epidermal nerve fiber density (ENFd) range (55.3 ENFs/mm [A] 141 ENFs/mm [B]). A superficial avascular epidermal layer of keratinocytes (blue layer labeled ‘epidermis’) with dispersed innervation by isolated (ENFs and, in some samples, Langerhans cells (LC) are also labeled in the epidermis. Below the dermal-epidermal junction (DEJ), lies the adjacent dermis which contains the capillaries and dermal nerves. The subepidermal neural plexus (SNP) is composed of dermal nerves that branch and ascend to form the ENFs of the epidermis or remain in the dermis to innervate capillaries and Merkel cells (not shown). In these confocal images, epidermal nerve fibers are indicated with small arrows. A and B are representative examples of calf skin shown as a projected image of 69 individual 0.75 microns/z-step images from the epidermis of 60 micron thick sections of the biopsy (using a 60X objective). In C, an image representing a projection of two adjacent image stacks (50 individual 1 microns/z-step images) from the same epidermal biopsy shown in B above (B is from the upper central portion of C) shows the abundant innervation over a larger scale view of the calf skin. In both the epidermis and dermis, PGP9.5 (green) stained all neural fibers and lightly stained fibroblasts and LCs. Collagen IV (red) stained basement membrane at the edge of the epidermis and dermis and on the capillaries and glial Schwann cells. Ulex (blue) stained a cell-surface glycoprotein on keratinocytes and endothelial cells. Colocalized immunostaining with PGP9.5 and Collagen IV is yellow for glial cell basement membrane in dermis. Apparent colocalized immunostaining with Ulex and Collagen IV is pink for capillaries in projected confocal images. Scale bars in A and B = 25 microns and in C = 100 microns.

Table 2:

Quantification of epidermal nerve fiber density [ENFd] and substance P-positive fiber count (SP+) in this sample.

Sample	Sex	Age (years)	ENFd	SP+
Participant 1	Female	5.8	129.7	15
Participant 2	Male	6.4	141.23	7
Participant 3	Female	3.5	78.55	15
Participant 4	Female	3.2	83.41	15
Participant 5	Male	4.8	82.96	20
Participant 6	Male	3.2	108.49	16
Participant 7	Male	5.7	87.02	8
Participant 8	Male	3.9	55.32	3
Participant 9	Female	7.1	97.89	9
Participant 10	Male	5.3	49.17	5
Participant 11	Female	4.1	125.74	10
Participant 12	Female	7.9	93.5	6
Participant 13	Male	5.3	127.83	6
Participant 14	Female	5.7	61.66	8
Participant 15	Male	4.8	88.2	3
Participant 16	Female	5.1	69.8	12

Behavioral reactivity – descriptive statistics

Means, standard deviations, and ranges for vocal, facial, and whole-body behavioral reactivity exhibited during each type of stimulus application are provided in Table 3 and displayed in Supporting Figure 1. All children in the sample exhibited behavioral reactivity during each active sensory testing trial suggesting that A-beta, A-delta, and C-fibers were functional (see Figure 2). On average, each child exhibited 14.69 (SD= 7.63; range=8–30) distinct episodes of codable behavioral reactivity that totaled a combined duration of 123.25 (SD= 97.49) seconds during all sensory testing applications combined. Specifically, there was an average of 4.44 (SD=4.16) episodes of vocal pain behaviors resulting in an average combined duration of 44.00 (SD=56.86) seconds; 3.56 episodes of facial pain behaviors resulting in an average combined duration of 32.75 (SD=34.58) seconds; and 6.69 (SD=3.20) episodes of body pain behaviors resulting in an average combined duration of 46.50 (SD=28.17) seconds.

Table 3.

Duration of vocal, facial, and body pain behaviors (combined) exhibited by children with global developmental delay during the sensory test

	Behavioral Reactivity M(SD)
	Vocal	Facial	Body	Total
Total seconds of pain behavior during stimulus applications
Sham (n=8)	0.13(0.35)	0.00(0.00)	0.13(0.35)	0.25(0.71)
LT	2.44(4.65)	2.81(5.18)	3.25(4.93)	8.50(13.06)
Pin Prick	2.19(3.04)	1.13(2.73)	2.75(2.44)	6.06(6.01)
Cool	3.19(7.02)	2.13(6.48)	3.38(6.21)	8.69(18.51)
Pressure (n=15)	2.53(4.00)	2.73(4.37)	1.93(2.60)	7.20(9.35)
RVF (n=15)	12.40(12.81)	14.33(14.19)	25.00(16.92)	51.73(33.03)
Heat (n=14)	1.43(2.56)	0.64(1.39)	3.71(3.58)	5.79(6.42)
Proportions of pain behavior controlling for duration of stimulus applications (vocal behaviors [s] + facial behaviors [s] + body behaviors [s] / stimulus duration [s])
Sham	0.02(0.04)	0.00(0.00)	0.02(0.04)	0.03(0.08)
LT	0.15(0.30)	0.18(0.31)	0.29(0.33)	0.62(0.81)
Pin Prick	0.28(0.38)	0.14(0.32)	0.36(0.29)	0.79(0.68)
Cool	0.23(0.44)	0.11(0.26)	0.30(0.49)	0.65(0.97)
Pressure	0.24(0.33)	0.27(0.40)	0.18(0.26)	0.69(0.78)
RVF	0.35(0.39)	0.40(0.41)	0.59(0.33)	1.35(0.93)
Heat	0.13(0.26)	0.07(0.14)	0.33(0.30)	0.53(0.61)

Note. LT = light touch; RVF = repeated von Frey monofilament.

Figure 2:

Violin and box plots depict behavioral reactivity exhibited during each modified Quantitative Sensory Testing (mQST) stimulus application. Box plots show median reactivity values, interquartile ranges (IQR; boxes), 1.5 x IQR (whiskers), and >1.5x IQR (outliers - flagged with a diamond). Proportions have been reported [vocal (s) + facial (s) + gross motor (s)] / duration of stimulus application (s) to control for differences in duration of stimulus applications; thus, reactivity scores range from 0.0–3.0. Stimulus applications include sham (SH), light touch (LT), pin prick (PIN), cool (CO), pressure (PR), repeated von Frey (RVF), and heat (HT).

Behavioral reactivity – comparison to typically developing children

The comparison sample of typically developing children was, on average, one year younger compared to children with global developmental delay, which was statistically significant (p=.01). Descriptively, children with global developmental delay exhibited a greater proportion [vocal (s) + facial (s) + gross motor (s)]/duration of stimulus trial (s), proportion scores ranged from 0.0–3.0] of behavioral reactivity across all stimulus trails, with the exception of the sham trial. Children with global developmental delay (M=0.99, SD=0.67) were overall significantly more reactive to the sensory test compared to typically developing children (M=0.43, SD=0.42; p=0.008) and were specifically more reactive to pin prick (p=0.006) and repeated von Frey (p=0.007), with comparisons for light touch trending similarly (p=.06).

Pain outcome measures – descriptive statistics

Six children (40%) had experienced at least one episode of pain in the week prior to participating in the study. Among the seven children with pain, 5 types/sources of pain were reported including accidental pain (n=3), gastrointestinal pain (n=2), musculoskeletal pain (n=3), neurological pain (n=1), and pain of unknown origin (n=1). Pain parameters by pain type are reported in Table 4. On average, pain intensity was rated 5.5 out of 10 (SD=1.80; range 2–10) with the average frequency of pain episodes being 2.3 per week (SD=1.36; range 1–7). One participant was reported to be living with constant neurological pain while other parents reported pain lasting on average 66.69 minutes in the previous week (SD=68.88; range 5–300 minutes). Four children (27%) were reported to have ongoing, chronic pain conditions which may or may not have been a problem in the previous week but had occurred on a regular basis for more than 6 months. Chronic pain sources included headaches (n=1), neuro-irritability (n=1), stomach pain (n=1), and sinus pain (n=1) and all had lasted longer than one year.

Table 4:

Pain parameters by pain type reported in the previous 7 days M(SD).

Pain	MSK (n=3)	GI (n=2)	Accidental (n=3)	Neuro (n=1)	Other (n=1)
Intensity	5.00(2.00)	6.50(1.50)	6.00(2.67)	4.00(0.00)	5.00(0.00)
Range	2–7	5–8	3–10	-	-
Duration (mins)	53.33(44.44)	195.00(105.00)	6.72(5.52)	10080(0.00)	30(0.00)
Range (mins)	5–120	90–300	0.17–15	-	-
Frequency of episodes	3.33(2.44)	3.00(0.00)	1(0.00)	1.00(0.00)	3.00(0.00)
Range	1–7	3–3	1–1	-	-

M=mean; SD = Standard Deviation; MSK = Musculoskeletal pain; GI = Gastrointestinal related pain; Neuro = Neurological pain; Other = Mouth pain.

Birth history & ENF density

ENF density was moderately associated with birth weight (r=−0.46, p=0.07; Supporting Figure 2). There was no significant difference in ENF density related to gestational age (t(14)=−0.04, p=0.97). Participants born at term (≤ 38 weeks gestation; n=12) had similar ENF density (M=92.35, SD=29.27) compared to those born pre-term (≥37 weeks gestation, n=4; M=93.08, SD=26.54). ENF density did not differ significantly between participants with a remarkable birth history (i.e., complications noted during pregnancy or birth; M=85.53, SD=28.52) compared to those with an unremarkable birth history (M=95.71, SD=28.16; t(14)=0.67, p=0.52).

Behavioral reactivity & ENF density

Overall ENF density in the skin biopsies did not predict overall behavioral reactivity during the sensory test with (β=−0.26, t(14)=−0.94, p=0.37) or without controlling for birth weight (β=−0.46, t(14)=−0.85, p=.41; Figure 3). ENF density did not significantly correlate with pain behaviors exhibited during the sensory test (r=−0.22, p=0.41). However, when total proportion of pain behavior during the sensory test was dichotomized into high responders (cases in the upper 75^th percentile) and average responders (all other cases), an independent samples t-test indicated a significant mean difference in ENF density between high pain behavior responders (M=72.97, SD=13.96; n=4) and average pain behavior responders (M=99.05, SD=28.47; t(11.2)=2.42, p=0.03; n=12; Figure 4). ENF density significantly predicted vocal behavior during all sensory testing applications combined (β=−0.51, t(14)=−2.18, p=0.048). ENF density accounted for 25% of the variance in vocal pain behaviors (R²=0.25, F(1,15)= 4.75, p=0.048). Specifically, ENF density predicted vocal pain behaviors during light touch (β= −.58, t(14)=−2.66, p=0.04) and cool thermal stimulus applications (β=−.53, t(14)=−2.31, p=0.02; Figure 5). ENF density accounted for 34% of the variance in reactivity during light touch (R²=0.34, F(1,15)=7.05, p=0.02) and 28% of the variance in cool thermal (R² =0.28, F(1,15)=5.31, p=0.04).

Figure 3:

Scatter plots depict the associations between epidermal nerve fiber (ENF) density and overall behavioral reactivity (vocal, facial, and body combined) exhibited during each type of stimulus application. Proportions have been reported [vocal (s) + facial (s) + gross motor (s)] / duration of stimulus application (s) to control for differences in duration of stimulus applications; thus, scores range from 0.0–3.0. Lines of best fit and associated 95% confidence intervals are shown.

Figure 4:

Violin and box plots depict the distribution on epidermal nerve fiber (ENF) density by participants who were high pain behavior responders (cases in the upper 75^th percentile) and those who were average responders (all other cases) across stimulus applications. Box plots show median ENF density values, interquartile ranges (IQR; boxes), and 1.5 x IQR (whiskers).

Figure 5:

Scatter plots depict the associations between epidermal nerve fiber (ENF) density and vocal behavioral reactivity exhibited during each type of stimulus application. Proportions (vocal reactivity [s]/ duration of application [s]) are reported to control for differences in the duration of stimulus applications. Lines of best fit and associated 95% confidence intervals are shown.

Pain outcomes & ENF density

Children with reported chronic pain had significantly greater ENF density (M=118.34, SD=27.40) compared to children without chronic pain (M=87.09; 21.02; t(13)=−2.36, p=0.03; Figure 6). ENF density did not correlate significantly with parent-reported pain intensity (r=0.36, p=0.18), frequency (r=0.003, p=0.99) or duration (r=−0.14. p=0.62) of pain episodes in the week prior to study participation. ENF density also did not correlate significantly with parent-reported pain interference with activities of daily living in the previous week (r=0.11, p=0.70).

Figure 6:

Violin and box plots depict the distribution on epidermal nerve fiber (ENF) density by participants who did and did not have chronic pain. Box plots show median reactivity values, interquartile ranges (IQR; boxes), 1.5 x IQR (whiskers), and >1.5x IQR (outliers - flagged with a diamond).

Substance P-positive fiber count (SP+)

SP+ fiber count was not significantly associated with birth history, behavioral reactivity, or any pain outcomes in this sample.

Behavioral reactivity & pain outcomes

Overall behavioral reactivity during the sensory test did not correlate with parent-reported pain intensity (r=−0.16, p=0.56), duration (r=−0.07, p=0.80), frequency of pain episodes (r=0.04, p=0.90), or pain interference with activities of daily living (r=−0.24, p=0.37) in the week prior to examination. Behavioral reactivity segmented by individual stimulus application did not correlate with parent-reported pain outcome measures (p>.05). Behavioral reactivity segmented by type of reactivity (face, whole body, vocal) also did not correlate with parent-reported pain outcome measures (p>.05).

Discussion

In this preliminary investigation, we took an integrative step toward establishing the rationale for a long-term developmental study of early peripheral innervation and sensory function in children with developmental delays at risk for lifelong developmental disability. Given the increasing importance placed on the construct of sensory function/dysfunction in relation to healthy development and neurodevelopmental disorders, we tested three exploratory integrative sensory-bio-behavioral relevant hypotheses. The first concerned the general relation between peripheral innervation biomarkers (epidermal nerve fiber density [ENFd] and substance P-positive fiber count (SP+) and objectively coded behavioral reactivity during a modified quantitative sensory test (mQST); the second concerned the general relation between those same biomarkers and proxy-reported (subjective) pain outcomes; and the third concerned the relation between sensory reactivity and pain outcomes.

Overall, a subgroup of children who were highly behaviorally reactive to the modified quantitative sensory test (mQST; i.e., ‘high pain-behavior responders’) had significantly reduced ENF densities compared to average behavioral reactivity responders. Reduced ENF density was associated with increased vocal reactivity overall and specifically during the light touch and cool stimulus trials. As this was a preliminary investigation, the evidence is not confirmatory of sensory subgroups.

The evidence from developmental biology suggests there are critical events important to the development of a fully functioning somatosensory system that depend on the integrity of the component parts, including those in the periphery. We suggest that aspects of sensory phenotypes associated with severe intellectual and neurodevelopmental disability in which individuals are described as hyper- and hypo-responsive to myriad stimuli including tactile and nociceptive may well be related to fundamental but overlooked functional impairments in peripheral components of the somatosensory nervous system. There are numerous pre-clinical and clinical observations across various species including human that abnormal sensory input is associated with developmental costs, but it is less appreciated that normal sensory input accompanied by abnormal underlying somatosensory circuitry can also be associated with significant developmental costs. This notion is gaining some traction, particularly as the phenotypic features of disorders like autism continue to be detailed, in which altered sensory function is a core feature.

In particular, a subgroup approach based on sensory response profiles generated by QST approaches could help delineate functional differences in somatosensory regulation that may ultimately relate to developmental differences (e.g., developmental delay, genetic syndromes, etc.). With that said, given the clinical population we are addressing (children with global developmental delay as part of a larger class of individuals with neurodevelopmental disorders and intellectual disability), it is important to acknowledge some issues specific to applying QST to vulnerable populations with cognitive, motor, and communication impairments. Overall, there have only been a handful of studies with clinical samples with developmental disorders. Collectively, results suggest the following: 1) QST procedures can be feasible but adaptations may be necessary; 2) reaction-time confounded by motor impairment is not trivial; 3) there are no definitive guidelines recommending which of the two dominant approaches (method of limits vs levels) should be used; 4) when language is impaired, establishing tactile/nociceptive thresholds is an extraordinary challenge using either method (Defrin, Pick, Peretz, & Carmeli, 2004; Defrin, Riabinin, Feingold, Schreiber, & Pick, 2015).

Because of the above issues with applying QST, we took an alternative approach for vulnerable clinical populations with complex communication needs by modifying QST such that behavioral reactivity is recorded (including any vocal responses) and time-locked to application of stimuli without requiring repeated applications. Such an approach sacrifices the ability to measure cutaneous sensory thresholds of various modalities. We have initiated such an approach in multiple clinical and field-based settings with multiple populations that have developmental disabilities and neurodegenerative diseases with evidence of excellent validity and very good psychometric properties in different applications, but we heretofore have not used it as a tool to integrate behavioral outcome measures with biomarkers specific to peripheral innervation (Barney, Feyma, Beisang, & Symons, 2015; Barney, Merbler, Simone, Walk, & Symons, 2020; Barney, Tervo, Wilcox, & Symons, 2017; Merbler et al., 2020; Symons, Harper, Shinde, Clary, & Bodfish, 2010).

Based on a limited but existing literature and a conceptual model based, in part, on peripheral/central sensitization, we hypothesized that children with developmental delay who had a greater ENF density would exhibit more behavioral reactivity during the sensory test. In particular, this hypothesis was based on the findings of Symons et al. (2009) who found that greater ENF density in the skin biopsies of adults with intellectual disabilities was correlated with greater sensory reactivity during a standardized sensory test (Symons, Wendelschafer-Crabb, Kennedy, Heeth, & Bodfish, 2009). A similar relationship was found in a study by Selim et al., (2010) for whom healthy adults with greater ENF density were better/faster at detecting sensation and pain (Selim et al., 2010). This relationship holds for adults with large-fiber diabetic neuropathy, for whom a more substantial reduction in ENF density is associated with reduced sensory and pain detection (Sommer & Lauria, 2007). However, the opposite relationship was found for individuals with diabetic or idiopathic small-fiber neuropathies affecting A-delta and C-fibers. For individuals with small-fiber neuropathy, reduced ENF density was associated with increased sensitivity to thermal stimuli (Sommer & Lauria, 2007). In addition, individuals with small-fiber neuropathy may experience allodynia (pain in relation to a stimulus that does not normally provoke pain), hyperalgesia (increased sensitivity to pain), and dysesthesia (abnormal and unpleasant sensations) (Sommer & Lauria, 2007).

Given the paucity of related evidence in childhood populations, at this point there is no reason to expect that children with global developmental delay would differ from healthy adults. Selim et al. (2010) found that healthy adults with greater ENF density reported more pin prick applications as painful and were more sensitive to detection of light touch (Selim et al., 2010). Based on this evidence, it was expected that children with greater ENF density would be more sensitive to pin prick and light touch in particular; however, given the current study did not test detection thresholds, this relationship may not have been measureable. It should be noted that several studies have not found reliable relations between ENF density and sensory testing outcomes (Chiang, Chen, Chien, & Hsieh, 2005; Periquet et al., 1999). Sommer and Lauria (2007) suggest that differences in findings may be related to methodological differences between labs and equipment as well as different populations tested (‘normal’, clinical, etc.) (Sommer & Lauria, 2007). Given the Selim et al. study was conducted within the same lab where the biopsies for the current study were processed, there was increased control over methodological differences between these two studies.

PGP staining for ENF density in epidermal skin biopsies quantifies predominantly C-fiber afferents and the majority of the stimuli tested were more likely to activate A-fibers. The general patterns of observed behavioral reactivity in relation to stimulus application suggests that A-fibers were functioning in the individuals making up the sample. Given what is known about the development of tactile and nociceptive spinal cord circuits, it is plausible to suggest that children with developmental delays in this sample with reduced C-fiber density and enhanced sensory reactivity may have had an immature sensory system possibly dominated by A-fiber afferent input (Koch & Fitzgerald, 2013). Neonatal spinal circuits are similarly highly responsive to tactile inputs that are predominantly transduced via A-fiber input. Within the first post-natal weeks C-fiber central synaptic inputs become stronger, which in turn drives the development of glycinergic inhibition. Maturation of glycinergic inhibition dampens A-fiber excitability and may contribute to reduced receptive field size and align inhibitory and excitatory receptive fields. Previous pre-clinical research has demonstrated that C-fiber destruction during the critical developmental period resulted in disorganized receptive fields and lack of A-fiber inhibition (Wall, 1982; Wall, Fitzgerald, Nussbaumer, Van der Loos, & Devor, 1982). This disruption in typical sensory circuit development occurred because the glycinergic interneurons failed to mature in these animals due to absent C-fiber input.

In our sample of children with developmental delay with reduced ENF density, there may, in turn, be relative reductions in adequate C-fiber input to produce maturation of glycinergic interneurons resulting in a persistent state of A-fiber dominance (i.e., inferred from heightened behavioral reactivity to some forms of tactile stimuli). Typical development of the tactile and nociceptive circuitry requires input from low-threshold A-fibers in very early development followed by C-fiber input during a later critical period. As discussed previously, some children with developmental delay may have markedly different sensory experiences potentially contributing to altered maturation of tactile and nociceptive circuitry.

The observation that children in this sample with reported chronic pain had significantly greater (not less) ENF density compared to children without chronic pain deserves comment. Four children with global developmental delay experienced chronic pain that had lasted longer than six months. This pain was reported in the form of headaches (n=1), neuro-irritability (n=1), stomach pain (n=1), and sinus pain (n=1). Of these four children, it would be reasonable that their reported chronic health problems (headache, gastrointestinal, and sinus) were associated with reported chronic pain. It is not clear in what way increased ENF density would be involved directly (or indirectly) as our design precludes any casual inference. The one case with reported neuro-irritability makes for interesting speculation about possible association between the irritable clinical presentation and the possibility of there being sensory or autonomic involvement as indicated, in part, by increased ENF values.

The most investigated direction of effect or relation between ENF density and pain is that reduction of the former (as in small fiber neuropathy) lead to the latter (chronic pain), although this is not always a clear relation (Sorensen, Molyneaux, & Yue, 2006). But in other cutaneous disorders (psoriatic itch, atopic dermatitis), intra-epidermal densities tend to be increased and related to pain (Huet & Misery, 2019). For our purposes, one of the unknowns is the lack of referent values providing a clear picture of increased/decreased relative to a normative criterion. In a recent study in children, among the few to examine ENF density in relation to chronic pain, the majority of the chronic pain subgroup had abnormal ENF density values – most commonly reduced; but, the referent values were all against adult (Gorlach et al., 2020). Thus, it seems we do not necessarily know for certain what to expect with regard to direction of effect for increased or decreased ENF density values in children in general, and certainly not for children with significant developmental delay. Of course, our current design precludes any causal inference – we simply do not know whether the increased innervation of these fiber types would be in the causal chronic pain pathway for this group.

It should be noted that there are other factors independent of ENF density that influence sensory reactivity in the periphery. For example, nerve growth factor (NGF) has been shown to play an important role in sensory reactivity independent of ENF density (Hirth, 2013) and mast cell degranulation has been linked with increased cutaneous reactivity during a sensory test in adults with intellectual disabilities (Symons et al., 2009). In this sample mast cells were also analyzed (not reported) and found to be predominantly intact and not degranulated. Mast cells also secrete exosomes containing miRNA and proteins. In addition to mast cells in the dermis, other potential sources of NGF in the skin include karatinocytes and Langerhans cells in the epidermis., Clearly, the transduction of sensory information is influenced by many factors external (e.g., environment, learning, behavior) and internal (e.g., genetics, ENF density, neuropeptides, mast cell degranulation, NGF, etc.) to each individual (Fitzgerald, 2005; Koch & Fitzgerald, 2013).

Beyond transduction, it is also important to situate our work within central mechanisms and the construct of ‘sensitization’. The chain of reasoning for our program of research is informed, in part, by detailed clinical observation of the phenomenon of repeated tissue damaging self-injury among individuals with intellectual and developmental disabilities. Although not the specific focus of this study, it – the phenomenon of repeated injuring oneself to the point of tissue damage, led to considering the notion of the sensitization of the nociceptive system occurring after repeated or particularly intense noxious stimuli, such that the threshold for its activation falls and responses to subsequent inputs are amplified (Latremoliere & Woolf, 2009). With this as initial starting point, our larger goal has been to consider approaches to testing for evidence of central and peripheral sensitization that may help understand individual differences in sensory/nociceptive function among nonverbal individuals living with intellectual disability but also as a possible risk factor for sensory/behavioral problems among children with global developmental delay.

Our prior work produced indirect evidence consistent with this notion including studies documenting differences in peripheral autonomic markers (Symons, et al., 2001), nociceptive biochemistry (Symons, et al., 2003), and abnormal peripheral innervation (Symons, et al., 2007) among a subgroups of adult individuals with chronic self-injury and severe intellectual disability. One key pressing issue concerns the physiological mechanisms involved in pain perception. But, because access to the subjective experience of pain is not readily testable in persons with severe cognitive impairment using conventional means (i.e., self-report), our goals have been to examine some of the structural and bio-chemical features of sensory innervation that are known to be linked to peripheral and central sensory perception and experience. Our focus on the periphery, however, should not be interpreted as discounting central processes.

Central sensitization provides an organizing mechanistic explanation for several temporal, spatial, and threshold changes in pain sensation and perception in acute and chronic clinical pain conditions and highlights the fundamental contribution of changes in the CNS to the generation of abnormal pain sensitivity (Latremoliere & Woolf, 2009). Given we conduct our sensory testing and procure the skin biopsies from non-injured body sites/tissue, our observations are therefore not restricted to an injury site, and would be consistent with central regulatory processes. Moreover, while peripheral sensitization appears to play an important role in altered heat but less so in mechanical sensitivity, our sensory testing findings tend to involve differences as well in mechanical sensitivity which is a key feature of central sensitization (Woolf & Salter, 2000). In studies of autism spectrum disorder (ASD), there have been two studies explicitly extending the German Network’s QST protocol to adults with autism (verbal, normal IQs). In general, results indicated feasibility (as these were the first to do so) and some indication that central mechanisms underlying sensory stimulus integration may be more likely responsible for some of the observed somatosensory features of ASD rather than peripheral dysfunction, per se (Frundt et al., 2017; Vaughan, McGlone, Poole, & Moore, 2020).

There are several study limitations that should be summarized. First, the approach used in this study did not directly test pain thresholds and therefore this study cannot and does not address the pain thresholds of children with global developmental delays. Most of the sensory stimuli as applied were likely sub-threshold as noxious stimuli and while every step was taken to insure standardization it is likely there were some small variations in stimulus application. For example, for pressure although a 5-second timer count was routinely used, there may have been some unmeasured inconsistency (child’s arm was not stable, movement would mean inconsistent application) and therefore error that would create ‘noise’ in that not everyone experienced 4lbs of pressure for a full 5 second in exactly the same way. The results should, therefore, be interpreted with appropriate caution and the implications of the study should therefore be limited to discussion of the duration of behavioral reactivity exhibited during a standardized array of sensory stimuli. It is important to note that quantification of behavioral expression exhibited during sensory testing cannot be assumed to directly represent pain experience. Second, although the observational coders of behavioral reactivity established strong inter-rater reliability (88–99%) and were blind to the specific research questions, it was not possible to keep the coders blind to the fact that they were observing children with developmental delays. Finally, the sample was formed based on clinical convenience; thus, the results should be considered sample specific. Given the skin biopsy procedure was invasive with no benefit and the sample by definition was comprised of vulnerable participants (pediatric, global developmental delay), it is not ethically realistic or feasible to create a random sample from this population. With respect to ethics, it is worth noting and commenting on the thermal test temperature. Given the small 3 mm probe, 50°C was chosen as per a prior thermode tip size comparison manipulation in a heat pain sensation study by Khalili, et al. (Khalili, Wendelschafer-Crabb, Kennedy, & Simone, 2001). Their study was adult based. Given our pediatric sample, and given the very small thermode tip size, we settled on an approximate midpoint between 48°C and 53°C at 50°C. As this temperature for the thermal stimulus setting (50°C) would be considered high by many current protocols, and 3 participants did not complete the full mQST protocol specific to the thermal stimulus trial, it would be worth investigating, for scientific and ethical reasons, whether similar behavioral response profiles could be found using lower intensities. Consideration should also be given to the small thermode tip size (3mm). It may be that larger tip sizes combined with lower temperature intensity would also be a feasible approach. Finally, we cannot say with absolute certainty that these are all C fibers nor distinguish among nociceptive fibers or tactile afferents.

A different set of issues related to the analysis should be pointed out. Given the hypotheses tested in the study were exploratory, some data were analyzed without multiplicity adjustment (Bender & Lange, 2001). Exploratory studies, such as the current study, necessitate a flexible approach to design and statistical analyses such that simply controlling for multiple tests does not solve the problem of making valid statistical inferences using an exploratory data-driven approach (Bender & Lange, 2001). A common multiple comparison correction procedure – the Bonferroni adjustment - was designed to reduce Type I error rates for decision-making processes, but in doing so Type II error rates are inflated such that truly important differences between groups could be missed (Perneger, 1998; Rothman, 1990). We think that it might be the wrong time to miss possible important differences given the stage of research and state of scientific knowledge in this vulnerable population. Although there are differences of opinions on the issue, when it applies to reporting exploratory findings, techniques such as the Bonferroni are less useful and can actually be detrimental to revealing important effects (Perneger, 1998; Savitz & Olshan, 1998). Rothman noted that not making adjustments for multiple tests is preferred when the data are not random numbers but are based on observations occurring in nature – this will lead to fewer errors in the interpretation of results (Rothman, 1990). In one sense, the multiple sensory tests were quasi-independent of one another to the degree that the different stimulus modalities were engaging functionally related but independent aspects of sensory (touch, pain) and transduction physiology (A-beta, A-delta, and C-fibers). Thus, in some of the testing for the different research questions, multivariate approaches were used, but in other instances multiple comparisons were conducted without adjusting for multiple tests. For this reason, and because of the exploratory nature of the current study, the results are descriptive only and not for purposes of decision-making. Significant findings will need to be further tested in properly powered confirmatory studies.

Overall, the modified QST provided an approach to interrogate somatosensory function and defined/calibrated tactile sensory experience in clinical high-risk pediatric sample with global developmental delays. Reliable, objective observations of different patterns of tactile and acute nociceptive reactivity were associated for some children with a peripheral biomarker relevant for sensory and pain processing. Children with parent-reported chronic pain appeared to differ from the rest of the sample in their epidermal nerve fiber density. Future work seems warranted to more precisely identify whether there are phenotypic sensory subtypes and establish the nature of relation between clinical problems with chronic pain and underlying peripheral innervation variables for this vulnerable group of children; similarly, the nature of the relation between sensory function and behavioral phenotype, issues specific to restricted and repetitive behavior including self-injury in children with global developmental delay, remains to be conclusively elucidated.

Supplementary Material

fS1

Supporting Figure 1: Violin and box plots showing the distribution of vocal, facial, and body behavioral reactivity exhibited during all sensory testing applications combined. Box plots show median reactivity values, interquartile ranges (IQR; boxes), and 1.5 x IQR (whiskers). Proportions (behavioral reactivity [s]/ duration of application [s]) are reported to control for differences in the duration of stimulus applications.

fS2

Supporting Figure 2: Scatter plot describing the association between birth weight (in grams) and epidermal nerve fiber (ENF) density. Line of best fit and associated 95% confidence interval are shown.