Relationship of Genetic Variants With Procedural Pain, Anxiety, and Distress in Children

Abstract

Objective:

This study used a candidate gene approach to examine genomic variation associated with pain, anxiety, and distress in children undergoing a medical procedure.

Study Design:

Children aged 4–10 years having an IV catheter insertion were recruited from three Midwestern children’s hospitals. Self-report measures of pain, anxiety, and distress were obtained as well as an observed measure of distress. Samples were collected from children and biological parents for analysis of genomic variation. Genotyped variants had known or suspected association with phenotypes of interest. Analyses included child-only association and family-based transmission disequilibrium tests.

Results:

Genotype and phenotype data were available from 828 children and 376 family trios. Children were 50% male, had a mean age of 7.2 years, and were 84% White/non-Hispanic. In family-based analysis, one single-nucleotide polymorphism (SNP; rs1143629, interleukin (IL1B) 1β) was associated with observed child distress at Bonferroni-corrected levels of significance (p = .00013), while two approached significance for association with high state anxiety (rs6330 Nerve Growth Factor, Beta Subunit, [NGFB]) and high trait anxiety (rs6265 brain-derived neurotrophic factor [BDNF]). In the child-only analysis, multiple SNPs showed nominal evidence of relationships with phenotypes of interest. rs6265 BDNF and rs2941026 cholecystokinin B receptor had possible relationships with trait anxiety in child-only and family-based analyses.

Conclusions:

Exploring genomic variation furthers our understanding of pain, anxiety, and distress and facilitates genomic screening to identify children at high risk of procedural pain, anxiety, and distress. Combined with clinical observations and knowledge, such explorations could help guide tailoring of interventions to limit procedure-related distress and identify genes and pathways of interest for future genotype–phenotype studies.

Humans vary significantly in their mental, emotional, and behavioral responses to health and illness due to genomic and environmental factors (Champagne, 2012; James, 2013; McEwen, Eiland, Hunter, & Miller, 2012; Mogil, 2012). A growing number of studies are examining these relationships; however, most studies of behavioral outcomes have focused on pathological diagnoses. Studies of nonpathological behavioral outcomes, particularly in children, are limited (Donner, 2012; Erhardt & Spoormaker, 2013; Sakolsky, McCracken, & Nurmi, 2012; Trzaskowski et al., 2013; Trzaskowski, Zavos, Haworth, Plomin, & Eley, 2012).

Of particular interest to health professionals who work with children is medical procedural distress, which may affect procedure success and possible long-term sequelae of negative experiences (Du, Jaaniste, Champion, & Yap, 2008; Kennedy, Luhmann, & Zempsky, 2008; Matthews, 2011; Noel, Chambers, McGrath, Klein, & Stewart, 2012). Children vary widely in their responses to procedures, and their age and developmental stage may place them at higher risk for procedural pain, anxiety, and distress compared to adults (Du et al., 2008; Kennedy et al., 2008). As with many behavioral responses, this variation is likely due to genomic and environmental factors (McCarthy et al., 2010a). Human and animal studies have identified genetic factors that contribute to variable pain responses (James, 2013; Kleiber, Schutte, et al., 2007; Mogil, 2009, 2012; Shi et al., 2010) as well as pathological forms of anxiety (e.g., anxiety disorders; Gregory & Eley, 2007; Smoller & Faraone, 2008). However, few genotype–phenotype studies have examined nonpathological distress outcomes (e.g., everyday and procedure-related anxiety), including response to painful medical procedures.

Many children have painful medical procedures every year, which have short- and long-term impacts (Du et al., 2008). One common procedure is IV catheter insertion, which may cause needle-related pain and distress during and after the procedure. We previously reported a significant association between children’s self-reported pain during IV insertion and single-nucleotide polymorphisms (SNPs) in endothelin receptor type A (EDNRA; Kleiber, Schutte, et al., 2007). Recent evidence supports the involvement of genes in other biological pathways (e.g., the hypothalamic–pituitary–adrenal [HPA] axis, immune response) that likely affect children’s responses to medical procedures (Kim, Ramsay, Lee, Wahl, & Dionne, 2009; Ruau et al., 2012; Young, Lariviere, & Belfer, 2012). Analyzing SNPs in these pathways will provide a more comprehensive examination of genomic variants hypothesized to be associated with pain, anxiety, and distress in response to painful medical procedures.

The purpose of the present study was to expand our understanding of genomic variants associated with children’s pain, anxiety, and distress in response to a painful medical procedure. We examined the associations of SNPs in 34 candidate genes from multiple biological pathways with pain, anxiety, and distress in children having peripheral IVs inserted. Results will facilitate the identification of children at greater risk of procedural distress and the development of appropriate interventions and will identify biological pathways and genes with possible relationships to behavioral phenotypes.

Method

Study Sample and Setting

The primary study examined the effectiveness of parent distraction coaching for children having peripheral IV catheters inserted. The model guiding the parent study (Kleiber & McCarthy, 2006; McCarthy et al., 2010a) identified child, parent, and procedural factors associated with the child’s response to medical procedures. The basis for the analysis reported here is the contribution of child genotype to this response. Eligible children met the following criteria: (1) 4–10 years old, (2) scheduled for IV insertion for diagnosis or treatment in one of the three Midwestern children’s hospitals, (3) at least one parent present for the procedure, and (4) English speaking. Children were ineligible to participate if they had significant developmental delay, did not speak English, or did not have a parent present for the procedure.

One parent answered study questions and supported the child during IV insertion. Children had a variety of medical conditions, but none had significant developmental delay. All participants spoke English. Each family received US$30.00 compensation. The institutional review boards at each data collection site approved the study.

Procedures

Primary study procedures were as follows. Before IV insertion, children rated their state and trait anxiety. Clinic personnel followed clinic-specific IV insertion protocols, which included application of topical anesthetics to possible IV insertion sites. These personnel were blind to treatment condition and we asked them not to interfere with parent behavior during the IV insertion. We videotaped the child and parent from the time the child was placed on the exam table until the IV was secured. Following IV insertion, children rated their pain. We collected whole blood or buccal swabs from children, and parents provided saliva samples for DNA analysis. We gave families supplies and instructions for obtaining samples from biological parents not present. Further details on primary study procedures appear in prior publications (McCarthy et al., 2010a, 2010b).

Phenotypes

The phenotypes of interest for this analysis were procedural pain, state and trait anxiety, and distress, which were identified as factors that could affect child response to IV insertion in the parent study (McCarthy et al., 2010a, 2010b). We obtained self-rated pain (Oucher), state anxiety (Children’s Anxiety Meter [CAM]), and trait anxiety via child self-report, while we measured observed distress using the Observational Scale of Behavioral Distress–Revised (OSBD-R), a behavioral measure (Aradine, Beyer, & Tompkins, 1988; Ersig, Kleiber, McCarthy, & Hanrahan, 2013; Jay & Elliott, 1984; Jay, Elliott, & Varni, 1986). We derived an additional self-rated distress phenotype from self-rated pain and state anxiety scores: Children with high state anxiety and high self-rated pain were categorized as high self-rated distress. Each child’s parent completed a brief questionnaire to document demographic factors including child’s age, gender, grade level, and relevant diagnoses (e.g., attention deficit hyperactivity disorder, anxiety disorder). We describe the phenotypes in detail below.

Pain

The child’s perception of pain at the IV insertion site was measured with the Oucher (Aradine et al., 1988), a validated self-report scale for pain intensity for children aged 3–12 years. Depending on age and ability to order items by size, children point to either a face in a series of six photographs showing a child in varying degrees of discomfort or a vertical numerical scale (0–10) positioned next to the faces. Higher scores indicate greater perceived pain. We used Oucher scales with culturally representative pictures for White/non-Hispanic, African American, and Hispanic children.

State and trait anxiety

We used the CAM, a brief, validated self-report measure of anxiety for children 4–10 years of age, to measure state and trait anxiety (Ersig et al., 2013). Drawn to resemble a thermometer, the CAM has a bulb at the bottom with a horizontal line at each whole number. Children are asked to color in the amount of their worried or nervous feelings at the current time (CAM–state anxiety [CAM-S]) and the amount for how they usually feel at home (CAM–trait anxiety [CAM-T]; Kleiber & McCarthy, 2006). Children unable to rank order by size or who did not understand the instructions did not complete the measure. Two team members used an overlay marked in ½-in. increments to measure the amount of the thermometer that was colored; the resulting number functioned as a child’s CAM score.

Observed distress

We measured children’s observed behavioral distress during IV insertion using the OSBD-R (Jay & Elliott, 1984; Jay et al., 1986). The OSBD-R is a validated rating scale made up of eight operationally defined behavioral categories indicative of distress in children during painful procedures. We videotaped time samplings of the child’s behavior, then scored and weighted them to indicate intensity. To calculate a total distress score, we added together the weighted values of each behavior at each interval. Higher scores reflect more distress. Jay and Elliot (1984) reported an internal consistency of 0.72; internal consistency for this study was 0.76. Interrater agreement for this study was 99.3%.

Self-rated distress

Of particular interest in this analysis was psychological distress, which reflects an individual’s response to a particular stressor, such as a medical procedure (Ridner, 2004). However, the parent study did not include a measure of psychological distress. To provide preliminary information on what genetic variants might be associated with this phenotype, we generated a measure based on measures of the two key components of psychological distress: pain and anxiety. Specifically, we created the self-rated distress phenotype by combining children’s self-rated pain and anxiety. Children with high scores on both (Oucher score ≥ 4; CAM-S score ≥ 5) were categorized as high self-rated distress. All other score combinations were defined as low self-rated distress.

Covariates

We identified covariates of interest with the potential to affect response to IV insertion from previous analyses (McCarthy et al., 2010a, 2010b). These included several child variables as well as the level of distraction provided.

Child variables

We obtained age, gender, diagnosis of anxiety disorder or attention deficit hyperactivity disorder (ADHD), previous IV insertion, and experiences with previous IV insertion via parent report. Nurses rated the difficulty of IV insertion using a 4-point Likert-type scale (1 = easy to 4 = unable to place IV). All nurses who inserted IVs during this study were pediatric nurses who worked regularly in these clinics, where IV insertion is a frequent nursing activity. While no formal evaluation was made, nurses evaluated their expertise in the context of other patient factors. If nurses deemed the IV insertion as being beyond their self-assessed abilities, they obtained assistance from more experienced nursing staff for IV insertion.

Treatment effect

In the primary study, children received varying levels of distraction coaching. Treatment effect was quantified using the distraction coaching index (DCI), a measure of the quality and frequency of distraction coaching. Content and construct validity have been previously reported (Kleiber, McCarthy, Hanrahan, Myers, & Weathers, 2007).

DNA Extraction

DNA was extracted from whole blood using QuickGene-610L (Autogen, Holliston, MA). Saliva samples were collected using Oragene kits (DNA Genotek Inc., Kanata, Ontario, Canada). DNA was extracted from saliva and buccal swab samples using prepIT (DNA Genotek Inc., Kanata, Ontario, Canada). Analyses were completed in 3–12 years following the collection of DNA samples. DNA quantification was performed using the Qubit 2.0 Fluorometer (Life Technologies Corporation, NY), with quantities ranging from 0 to 800 ng/µl; genotyping was attempted on all available samples. In total, 2,672 samples were available from children (1,040), and their biological parents (1,632 total: 974 mothers and 658 fathers; 626 complete trios).

Selection of Candidate Genes

We selected 34 genes based on known or suspected roles in pathways of interest, including the endothelin pathway, HPA axis, and immune response (Online Table S1; Kleiber, Schutte, et al., 2007) as well as evidence of association with study phenotypes. We selected 91 SNPs in these genes based on previous association or classification as a tagging SNP. SNPs selected for analysis had a minor-allele frequency (MAF) of at least .2 according to dbSNP builds 136 and 137 (Sherry et al., 2001). We selected tagging SNPs based on the results in Haploview. We made exceptions for SNPs supported by strong evidence in the literature, key novel variants of interest, and SNPs in haplotypes with other SNPs analyzed.

Genotyping

We used Fluidigm 192.24 Dynamic Array Integrated Fluid Circuits (Fluidigm Corporation, South San Francisco, CA) to complete genotyping. DNA samples and Taqman® SNP genotyping assays (Life Technologies Corporation, NY) were loaded into reaction chambers by the IFC Controller RX and a fast polymerase chain reaction (PCR) protocol was run on the FC1 Cycler. The chip was read using fluorescence detection by the BioMark HD Reader, and we used Fluidigm analysis software Version 3.1.2 to view and interpret our results.

We genotyped a total of 2,672 individuals (1,040 children and 1,632 biological parents; 626 complete trios) for at least one SNP each. We uploaded all genotype data to Progeny Version 8 (Progeny Software, LLC, Delray Beach, FL) to analyze sample quality. First, we removed individual samples with a no-call rate of >5% (n = 395) from the data set. Then, we removed individuals in families with a Mendelian error rate of >2% (n = 119). We then removed genotypes from five SNPs from the data set due to assay failure (rs737865, rs25531, and rs3811939) or not being in Hardy–Weinberg equilibrium (p value ≤ .05; rs1876831 and rs9574124). One final quality-control check examined concordance of SNP genotyping for rs1143629 with sequencing. Analysis indicated 94.8% concordance between methods for this particular SNP. Following quality-control procedures, we analyzed data from 2,160 individuals (n = 1,037 children, 1,123 parents) and 86 SNPs in 34 genes.

Derivation of Phenotypes for Statistical Analyses

We entered phenotype data into a Microsoft Office Access® database (Microsoft Corporation, Redmond, WA) and checked it using standard double-entry techniques. We generated descriptive statistics (e.g., mean, median, range) for all measures. The phenotypes of self-rated pain, state and trait anxiety, and self-rated distress had skewed distributions: The majority of children scored on the low end of these scales. Based on these data, we dichotomized results for statistical analysis. We identified cut points based on previous analyses for each phenotype variable (McCarthy et al., 2010a, 2010b). For self-rated pain, we considered Oucher scores of 4–10 indicate high pain intensity and those from 0 to 3 low pain intensity. For state and trait anxiety, we categorized CAM-S and CAM-T scores of 5–10 as high anxiety and 0–4 as low anxiety. For self-rated distress, we categorized children as high if they had high state anxiety and high self-rated pain. We analyzed the OSBD-R, which measures observed distress, as a continuous measure and computed scores as the weighted sum of multiple behavioral components totaled over 24 time intervals (Jay & Elliott, 1984). Higher scores reflect higher observed distress.

Statistical Analyses

We completed regression analyses of child genotype only for the phenotypes of interest using all child genotype data available. A subset of families also provided genotype data on both biological parents. We analyzed data from these family trios using transmission disequilibrium tests (TDTs), which allowed us to generate greater power to detect genotype–phenotype associations and to remove the potential influence of population admixture on results (Abecasis, Cookson, & Cardon, 2000; Spielman & Ewens, 1996).

For analyses of child genotype, we used linear regression for observed distress and logistic regression for all other phenotypes, with genotypes coded 0, 1, and 2 using an additive model. Covariates of interest included the child’s age, gender, and education, parent report of child diagnosis of ADHD or anxiety disorder, child’s experience with previous IV insertions, nurse perception of IV insertion difficulty, and effect of parent distraction (the treatment effect). Bivariate associations of each potential covariate and outcome identified covariates controlled for outcome-specific analyses. Analyses for self-rated distress, self-rated pain, and observed distress were controlled for DCI, previous IV experience, and difficulty of IV insertion. Analyses for self-rated distress also controlled for age, and those for observed distress also controlled for age and gender. Analyses for trait anxiety were adjusted for age, while those for state anxiety were adjusted for difficulty of IV insertion. Results for logistic regression analyses provide the odds ratio for increase or decrease in risk per rare allele. Linear regression analysis results indicate the percentage increase or decrease in the ratio of the mean OSBD-R for each additional rare allele.

We used TDTs to examine differential transmission of parental alleles to children exhibiting phenotypes of interest. We performed single-marker and multimarker (haplotype) tests in PLINK (V1.07; Purcell et al., 2007).

Given the exploratory nature of our analyses, we used liberal p-value thresholds to identify SNPs with possible associations with phenotypes of interest; these associations should guide researchers’ decision-making when choosing SNPs to consider for future studies. For regression analyses, we highlight p values < .10, while for TDT analysis, we used p values < .05. For TDT analysis, we also calculated a conservative Bonferroni p value (5 phenotypes, 86 SNPs), with values of .00012 or smaller used to identify association with phenotypes of interest.

Results

Sample Characteristics

The total study sample was 1,116 children, of whom 828 had genotype and phenotype data. For child-only regression analyses, the number of child genotypes was 83–88, with most having 87 (n = 386) or 88 (n = 298). Sample size for each regression analysis varies according to the data available. For the family-level TDT analysis, data were available from 626 trios. We removed trios with one or more individuals with no calls ≥5% (n = 219) and those with a Mendelian error rate of >2% or more than one Mendelian error in the trio (n = 31), resulting in a final sample of 376 trios. Children were about half female (50.3%), had an average age of 7.2 ± 1.9 years, and were 84% White/non-Hispanic.

Phenotypes

Table 1 contains descriptive statistics for sample characteristics, potential covariates, and outcome measures of interest. Table 2 provides results for the child-only regression analysis with p values <.05. TDT results with p values <.05 are reported in Table 3. SNPs of particular interest included rs1143629 (interleukin 1β [IL1B]), which demonstrated possible association with high observed distress in the TDT analysis (p = .00013, equal to the Bonferroni corrected p value). We also identified possible relationships between this SNP and self-rated pain (p = .036) and state anxiety (p = .047). Also of interest were two SNPs associated with trait anxiety in child-only and family-based analyses: rs6265 (brain-derived neurotrophic factor [BDNF]) and rs2941026 (cholecystokinin B receptor [CCKBR]).

Table 1.

Descriptive Statistics.

Variables	Descriptive Statistics (n = 828, Unless Otherwise Indicated)
Sample characteristics
Gender (male), n (%)	417 (50.4)
Age (years), mean ± SD (range)	7.2 ± 1.9 (4–10)
Race/ethnicity
White/non-Hispanic, n (%)	692 (83.6)
Hispanic, n (%)	18 (2.2)
African American, n (%)	35 (4.2)
Asian, n (%)	11 (1.3)
Native American, n (%)	3 (0.4)
Mixed, n (%)	68 (8.2)
Unknown, n (%)	1 (0.1)
Potential covariates
Difficult IV insertion, n (%)	250 (30.2)
Topical anesthetic within expected dwell time, n (%)	671 (81.0)
ADHD diagnosis, n (%)	98 (11.8)
Anxiety disorder diagnosis, n (%)	33 (3.9)
Has child ever had an IV? Yes, n (%)	611 (73.8)
Outcome variables
High pain (Oucher ≥ 4), n (%)	299 (36.1)
State anxiety (CAM-S), n = 683, mean ± SD (range)	3.86 ± 3.27 (0–10)
Trait anxiety (CAM-T), n = 679, median (25–75%; range)	0 (0–2%; 0–10)
Self-rated pain (Oucher), n = 797, mean ± SD (range)	3.25 ± 3.34 (0–10)
Observed distress (OSBD-R), n = 795, mean ± SD (range)	2.90 ± 3.91 (0–26)

Note. CAM-S = Children’s Anxiety Meter–state anxiety; CAM-T = Children’s Anxiety Meter–trait anxiety; OSBD-R = Observational Scale of Behavioral Distress–Revised; SD = standard deviation; ADHD = attention deficit hyperactivity disorder.

Table 2.

Suggestive Results of Child-Only Regression Analysis (p < .05).

			Rare Allele (MAF)	Phenotype
Gene		SNP		Self-Rated Distressa p-value, OR (95% CI)	Self-Rated Painb p-value, OR (95% CI)	Trait Anxietyc p-value, OR (95% CI)	State Anxietyd p-value, OR (95% CI)	Observed Distresse p-value, OR (95% CI)
BDNF	Brain-derived neurotrophic factor	rs6265	A (.197)			.039, 0.573 (0.337, 0.973)
CCKBR	Cholecystokinin B Receptor	rs2941026	T (.455)			.048, 0.698 (0.489, 0.996)
CNR1	Cannabinoid Receptor 1	rs2023239	C (.146)				.006, 1.470 (1.120, 1.930)
COMT	Catechol-O-methyltransferase	rs4818	G (.439)	.046, 0.730 (0.530, 0.990)			.001, 0.670 (0.530, 0.850)
COMT	Catechol-O-methyltransferase	rs6269	G (.439)				.004, 0.710 (0.570, 0.900)
CRH	Corticotropin-releasing hormone	rs2446432	G (.490)			.033, 0.667 (0.459, 0.968)
CRH	Corticotropin-releasing hormone	rs4613981	G (.136)			.029, 1.540, (1.044, 2.272)
CRH	Corticotropin-releasing hormone	rs12721510	A (.050)		.039, 0.570 (0.334, 0.973)
CRHR1	Corticotropin Releasing Hormone Receptor 1	rs110402	T (.439)					.045, 1.131 (1.003, 1.276)
EDNRA	Endothelin Receptor Type A	rs10003447	T (.386)	.046, 0.710 (0.500, 0.990)	.016, 0.741 (0.581, 0.946)			.016, 0.851 (0.746, 0.971)
EDNRA	Endothelin Receptor Type A	rs4110	A (.313)	.032, 1.400 (1.030, 1.900)	.043, 1.273 (1.008, 1.608)
EDNRA	Endothelin Receptor Type A	rs4563479	T (.101)	.018, 0.490 (0.280, 0.890)
EDNRA	Endothelin Receptor Type A	rs5333	C (.253)		.020, 0.748 (0.586, 0.955)			.047, 0.874 (0.764, 0.998)
EDNRA	Endothelin Receptor Type A	rs6812093	T (.242)			.030, 1.480 (1.039, 2.108)
EDNRA	Endothelin Receptor Type A	rs6827096	T (.152)	.022, 0.600 (0.390, 0.930)
EDNRB	Endothelin Receptor Type B	rs1041619	T (.320)					.035, 1.135 (1.009, 1.277)
NFKB2		rs7897947	G (.192)		.045, 0.752 (0.570, 0.993)			.033, 0.849 (0.730, 0.987)
NR3C1	Nuclear Receptor Subfamily 3 Group C Member 1	rs12656106	C (.490)			.016, 0.629 (0.431, 0.918)
OPRM1	Opioid Receptor Mu 1	rs634479	C (.167)					.042, 0.858 (0.741, 0.994)
SERPINA6	Serpin Family A Member 6	rs941601	A (.152)				.019, 0.690 (0.510, 0.940)
SERPINA6	Serpin Family A Member 6	rs11627241	T (.237)					.030, 1.158 (1.014, 1.322)
SERPINA6	Serpin Family A Member 6	rs1998056	C (.409)					.029, 1.144 (1.014, 1.292)

Note. MAF = minor-allele frequency; SNP = single-nucleotide polymorphism; OR= odds ratio; CI = confidence interval; DCI = distraction coaching index; CAM = Children’s Anxiety Meter.

^aHigh self-rated distress (n = 670; Oucher > 4 and state anxiety > 5), adjusted for DCI, age, previous IV, and difficulty of IV insertion.

^bHigh self-rated pain (n = 797; Oucher > 4), adjusted for DCI, age, previous IV, and difficulty of IV insertion.

^cHigh trait anxiety (n = 683; CAM–trait anxiety > 5), adjusted for age.

^dHigh state anxiety (n = 679; CAM–state anxiety > 5), adjusted for difficulty of IV insertion.

^eLinear regression analysis (n = 795; Observational Scale of Behavioral Distress–Revised), adjusted for DCI, age, gender, previous IV, and difficulty of IV insertion. For observed distress only, results report mean (95% CI) for +1 difference in rare allele.

Table 3.

Significant and Suggestive Results of Family-Based Genetic Association Test.

			Rare Allele (MAF)	Phenotype
Gene		SNP		Self-Rated Distressa p-value	Self-Rated Painb p-value	Trait Anxietyc p-value	State Anxietyd p-value	Observed Distresse p-value
BDNF	Brain-derived neurotrophic factor	rs6265	A (0.197)			0.005
CCKBR	Cholecystokinin B receptor	rs2941026	T (0.455)			0.046
EDNRA	Endothelin receptor type A	rs12647366	T (0.298)			0.018
EDNRA	Endothelin receptor type A	rs4110	A (0.313)			0.018
EDNRA	Endothelin receptor type A	rs6827096	T (0.152)					0.035
FKBP5	FK506-binding protein 5	rs1360780	T (0.278)				0.027
FKBP5	FK506-binding protein 5	rs9296158	A (0.273)				0.039
GRM3	Glutamate metabotropic receptor 3	rs6465084	G (0.253)				0.096
IL1B	Interleukin 1 beta	rs1143629	C (0.353)		0.036		0.047	0.0001
NFKB2	Nuclear factor kappa B subunit 2	rs7897947	G (0.192)				0.034
NGFB	Nerve growth factor	rs6330	T (0.424)	0.048			0.009
OPRM1	Opioid receptor mu 1	rs563649	A (0.051)		0.050
PENK	Proenkephalin	rs2576577	T (0.429)					0.023
SLC6A4	Solute carrier family 6 member 4	rs140700	A (0.106)					0.036

Note. Transmission disequilibrium test, p < .05. MAF = minor-allele frequency; SNP = single-nucleotide polymorphism; DCI = distraction coaching index; CI = confidence interval; CAM = Children’s Anxiety Meter.

^aHigh self-rated distress (n = 670; Oucher > 4 and state anxiety > 5), adjusted for DCI, age, previous IV, and difficulty of IV insertion.

^bHigh self-rated pain (n = 797; Oucher > 4), adjusted for DCI, age, previous IV, and difficulty of IV insertion.

^cHigh trait anxiety (n = 683; CAM–trait anxiety > 5), adjusted for age.

^dHigh state anxiety (n = 679; CAM–state anxiety > 5), adjusted for difficulty of IV insertion.

^eLinear regression analysis (n = 795; Observational Scale of Behavioral Distress–Revised), adjusted for DCI, age, gender, previous IV, and difficulty of IV insertion. For observed distress only, results report mean (95% CI) for +1 difference in rare allele.

Self-rated pain

Regression analyses identified SNPs in multiple genes (corticotropin-releasing hormone [CRH], EDNRA, nuclear factor κ B subunit 2 [NFKB2]) with possible relationships to self-rated pain. Of these SNPs, three were in EDNRA, with two associated with lower odds of high pain (rs10003447 and rs5333, p = .016 and .02, respectively) and one with higher odds (rs4110, p = .043). TDT analysis of rs5333, of interest given our prior results (Kleiber, Schutte, et al., 2007), showed no significant relationship (p = .089), although each rare allele lowered the odds of high pain by .252.

State anxiety

In regression analyses, three SNPs were associated with lower odds of high state anxiety for each rare allele: two in catechol-O-methyltransferase (COMT; rs4818 and rs6269, p = .001 and .004, respectively) and one in serpin family A member 6 (SERPINA6; rs941601, p = .019). We also found that one SNP in the cannabinoid receptor 1 gene (CNR1; rs2023239, p = .006) was associated with higher odds of high state anxiety for each additional rare allele.

TDT analysis identified one SNP (rs6330; nerve growth factor [NGF]) and two haplotypes related to high state anxiety (p = .009). For the haplotypes, windows of 2, 3, and 4 SNPs in the corticotropin-releasing hormone receptor 1 gene (CRHR1) included rs110402, rs173365, rs1876831, and rs878886 (CTAG haplotype; global: p = .005 and haplotype specific: p = .03). A two-SNP haplotype in the gene for FK506 binding protein 5 (FKBP5) of rs9296158 and rs3777747 was also related to high state anxiety (AG genotype; global: p = .02 and haplotype-specific: p = .004).

Trait anxiety

Regression analyses identified six SNPs related to trait anxiety. Of these, four were associated with lower odds of high trait anxiety for each additional rare allele: rs6265 (BDNF, p = .039), rs2941026 (CCKBR, p = .048), rs2446432 (CRH, p = .033), and rs12656106 (nuclear receptor subfamily 3 group C member 1 [NR3C1], p = .016). The other two SNPs were associated with higher odds of high trait anxiety: r4613981 (CRH, p = .029) and rs6812093 (EDNRA, p = .03). Allelic TDT analysis also identified a possible association of rs6265 with high trait anxiety (BDNF, p = .0045) as well as rs2941026 (CCKBR, p = .046); these SNPs are of particular interest, given their identification in regression analysis as having a possible relationship to lower trait anxiety.

Observed distress

Regression analyses for observed distress identified nine SNPs of interest, with five of these associated with lower odds for high observed distress, while each additional rare allele in the other four was associated with higher odds for high observed distress. SNPs in EDNRA (rs10003447 and rs5333, p = .016 and .047, respectively), endothelin receptor type B (EDNRB; rs9574124, p = .006), NFKB2 (rs7897947, p = .033), and opioid receptor mu 1 (OPRM1; rs634479, p = .042) were associated with lower odds for high distress, while SNPs in CRHR1 (rs110402, p = .045), EDNRB (rs1041619, p = .035), and SERPINA6 (rs11627241 and rs1998056, p = .03 and .029, respectively) were associated with higher odds for high distress.

As noted above, allelic TDT analysis identified one SNP of interest for high observed distress: rs1143629 (IL1B, p = .00013). Regression analysis also determined that, for each rare allele in this SNP, children had lower odds of high distress (p = .061).

Self-rated distress

Regression analysis identified five SNPs that were related to high self-rated distress. For one SNP in COMT (rs4818, p = .046) and three in EDNRA (rs10003447, rs4563479, and rs6827096, p = .046, .018, and .022, respectively), each additional rare allele was associated with lower odds of high self-rated distress. Each additional rare allele in rs4110 (EDNRA, p = .032), in contrast, was associated with higher odds of high self-rated distress.

TDT analysis identified one two-SNP and one three-SNP haplotype in CRHR1 of interest for high self-rated distress. The two-SNP window included rs1876831 and rs878886 (AG; global: p = .007 and haplotype specific: p = .03), while the three-SNP window covered rs173365, rs1876831, and rs878886 (TAG; global: p = .02 and haplotype specific: p = .03).

Additional Analyses

Some SNPs were related to multiple phenotypes. rs1143629 (IL1B) showed a possible relationship with observed distress, high self-rated pain, and high state anxiety. SNPs in EDNRA and NFKB2 (rs7897947) were of interest for self-rated and observed distress and self-rated pain.

We also conducted all analyses on the subset of White/non-Hispanic participants, as nearly 85% of the study sample identified as this race. Results for this sample did not differ in direction of effect, but we did note differences in the magnitude of the effect due to the redistribution of the proportion of the sample with each genotype or phenotype.

Discussion

Analysis of this unique data set provides important information regarding the relationship between genomic variants and variability in children’s responses to painful medical procedures. Although few results reached statistical significance following the correction for multiple comparisons, findings highlight genes and, especially, biological pathways of interest for future studies of similar phenotypes in pediatric populations. Of particular interest was the association of rs1143629 (IL1B) with observed distress, indicating that the cytokine pathway may play a role in response to procedural pain. In addition, child-only and family-based analyses both identified two SNPs as possibly being related to trait anxiety, a relationship of interest for future studies. Multiple additional SNPs had possible relationships to pain, distress, and anxiety.

Part of the cytokine pathway, IL1B mediates the body’s inflammatory response, with involvement in cell proliferation, differentiation, and apoptosis. IL1B contributes to inflammatory pain hypersensitivity through induction of cyclooxygenase-2 (PTGS2/COX2; National Center for Biotechnology Information: http://www.ncbi.nlm.nih.gov/gene/3553). The SNP of interest from our analyses (rs1143629) is intronic, with no previously identified clinical associations. However, other studies have indicated that IL1B may play an important role in the inflammatory pain response. Painful conditions lead to elevations in proinflammatory cytokines, of which IL1B is one (DeVon, Piano, Rosenfeld, & Hoppensteadt, 2014). Oliveira and colleagues (2014) examined the association of genomic variation in the cytokine pathway with pain intensity in adult cancer patients. Of interest in Oliveira et al.’s and other studies is rs1143634, another intronic IL1B SNP located 3,128 base pairs downstream from rs1143629 but not in linkage disequilibrium with it. Genotype at this particular site is associated with pain intensity (George et al., 2014; Oliveira et al., 2014). Yılmaz et al. (2010) found that rs1143634 genotype was significantly associated with the risk of migraine, again highlighting the broader implications of variation in IL1B for pain.

EDNRA remains another intriguing candidate gene. Our previous work identified an association between rs5333 and child self-report of pain (Kleiber, Schutte, et al., 2007). Analysis of the full data set in the present study further supported this possible association (p = .02). Results also identified relationships between other SNPs in EDNRA and all phenotypes of interest, supporting further examination of the possible role of EDNRA in pain and anxiety.

Genes involved in peripheral nociception at the site of topical analgesic action, such as those in the endothelin pathway, are candidates for influencing pain perception and effectiveness of topical anesthetics. Endothelin 1 (ET-1) codes for a potent vasoconstrictive peptide that induces pain through a direct effect on sodium-channel polarization in sensory neurons (Gokin et al., 2001; Zhou, Davar, & Strichartz, 2002). The pain-signaling action of endothelin is mediated through two receptors, EDNRA and EDNRB (Khodorova et al., 2003; Pomonis, Rogers, Peters, Ghilardi, & Mantyh, 2001; Zhou et al., 2002), which may play a role in migraine susceptibility (Lemos et al., 2011; Miao, Wang, & Fang, 2012; Tikka-Kleemola et al., 2009) and cancer-related pain (Viet et al., 2011). Although migraine-related pain differs qualitatively from acute procedural pain, further examination of the role of EDNRA and EDNRB in pain phenotypes is warranted.

One unique aspect of this data set was the measurement of multiple behavioral phenotypes. Genes important in the stress response via the HPA axis (BDNF, CRHR1, and SERPINA6), the catecholamine pathway (COMT), and cortisol response (FKBP5) may play a role in anxiety responses to painful medical procedures. Of particular interest was rs6265 (BDNF), which may be related to high trait anxiety, although other studies have been inconclusive (Arias et al., 2012; Colzato, Van der Does, Kouwenhoven, Elzinga, & Hommel, 2011; Montag, Basten, Stelzel, Fiebach, & Reuter, 2010). Other analyses (Bredemeier, Beevers, & McGeary, 2014) identified possible interactions of variants in BDNF with the 5-HTTLPR region in SCL6A4; replication of these findings should be pursued in other populations.

While whole-genome and whole-exome studies are becoming more prevalent, the present analysis highlights the continued importance of candidate-gene studies for identifying pathways and mechanisms of interest. Genomic studies of behavioral phenotypes in children are rare; leveraging existing data is vital for advancing future work in this area. One anticipated outcome of such research is identification of at-risk individuals and development of individualized interventions. Candidate-gene studies can help identify those at high risk and highlight pathways of interest for development of pharmacologic therapies and other interventions.

Genotype–phenotype studies of behavioral phenotypes are also complicated by the difficulty in defining and analyzing such phenotypes. This difficulty highlights the importance of careful and consistent definition of behavioral phenotypes and phenotype harmonization in future studies (National Institute on Aging, 2012), which will facilitate sample combination and generate more statistical power. Another alternative is the use of intervening phenotypes or endophenotypes. For example, pain sensitivity may be more readily measured and may demonstrate stronger genotype–phenotype relationships than pain.

Finally, while the effect sizes we identified in the present analysis are small, they do point to biological factors related to behaviors of interest. Results of exploratory analyses not only improve our understanding of the biological basis of children’s procedural pain but also help identify preliminary targets for intervention.

Limitations

Findings reported here highlight limitations inherent in incorporating genotype–phenotype analyses within the context of intervention trials. The need to control for study-intervention effect presented challenges for analysis and application of findings in future studies and clinical settings. In addition, other variables that could affect the phenotype of interest, including the reason for IV insertion, presence of chronic health conditions, and presence of comorbidities, were not collected as part of the primary study. Studies of genomic variants also face the challenge of population stratification; our study incorporated TDT analysis to limit these concerns. Sample size was determined by the primary outcomes of interest, not the genomic analyses, and was unalterable. The limited sample size resulted in preliminary findings to consider including in future, larger studies. Additional replication and functional studies of these and related genomic variants will help advance science and facilitate translation into clinical practice. As noted, the vast majority of our sample self-identified as White/non-Hispanic; however, this subgroup did not differ substantially from the overall study sample.

Nursing Implications

While this study was exploratory, findings from this and similar analyses could impact clinical care and future nursing-driven research. Results of genomic analyses of pain and distress phenotypes in children could be useful in developing predictive screens for pain response, effectiveness of topical analgesics, and anxiety during painful medical procedures, which would benefit clinical practitioners, children, and their families. Integrating biological markers into clinical screening tools would further advance translation of genomic findings into practice, aligning with the national All of Us Research Program (formerly the Precision Medicine Initiative; https://www.nih.gov/allofus-research-program) to ensure implementation of appropriate interventions. Future studies defining additional common variants of modest impact or rare high-impact variants that affect behavior (e.g., monoamine oxidase A [MAOA]; Ficks & Waldman, 2014; Holz et al., 2016) will provide additional opportunities to identify risk profiles and may suggest alternative pain- and anxiety-management strategies. The role of environmental triggers requires better definition to fully identify causes of and approaches to prevention and treatment of distress. Robust and fully specified models, incorporating biological, psychological, and social interactions leading to pain, anxiety, and distress, can be incorporated into parsimonious clinical screens used to guide intervention selection, implementation, and evaluation.