Gastrointestinal dysfunctions as a risk factor for sleep disorders in children with idiopathic autism spectrum disorder: A retrospective cohort study

Abstract

Sleep disorders often co-occur with autism spectrum disorder. They further exacerbate autism spectrum disorder symptoms and interfere with children’s and parental quality of life. This study examines whether gastrointestinal dysfunctions increase the odds of having sleep disorders in 610 children with idiopathic autism spectrum disorder, aged 2–18 years, from the Autism Genetic Resource Exchange research program. The adjusted odds ratio for sleep disorder among those with gastrointestinal dysfunctions compared to those without was 1.74 (95% confidence interval: 1.22–2.48). In addition, the odds of having multiple sleep disorder symptoms among children with gastrointestinal dysfunctions, adjusted for age, gender, behavioral problems, bed wetting, current and past supplements, and current and past medications for autism spectrum disorder symptoms were 1.75 (95% confidence interval: 1.10–2.79) compared to children without gastrointestinal dysfunctions. Early detection and treatment of gastrointestinal dysfunctions in autism spectrum disorder may be means to reduce prevalence and severity of sleep problems and improve quality of life and developmental outcomes in this population.

Keywords

autism gastrointestinal dysfunction idiopathic autism spectrum disorder multiplex families severity sleep disorder

Sleep disorders are widespread problems in children with neurodevelopmental disorders. Overall, sleep problems are more profound between the ages 2 and 6 years and decline thereafter childhood (Kothare and Kotagal, 2011). However, sleep problems tend to persevere into adulthood in individuals with neurodevelopmental disorders and autism spectrum disorder (ASD) in particular (Kothare and Kotagal, 2011; Sivertsen et al., 2012). Frequent night arousal, difficulty initiating sleep, and sleep maintenance disorder were listed among the most prevalent sleep disorders associated with ASD, with prevalence ranging from 40% to 88% (Cortesi et al., 2010; Ming and Walters, 2009; Richdale, 1999). Studies using animal models, as well as those with samples drawn from the general population, found that sleep abnormalities were linked to an array of abnormal behaviors such as inattention, hyperactivity, and restricted/repetitive and self-injurious behaviors (Barnard and Nolan, 2008; Goldman et al., 2009; Johnson et al., 2009).

Poor sleep in children with ASD further exacerbates ASD symptoms. Studies have shown that children with ASD and sleep disorders were more hyperactive and more compulsive and ritualistic in their behavior than children with ASD without sleep disorders (Cortesi et al., 2010; Goldman et al., 2009; Kothare and Kotagal, 2011). Furthermore, sleep disorders in children with ASD impact the quality of parents’ sleep increasing parental burden and family stress (Doo and Wing, 2006; Meltzer, 2008). There are several medical factors that are causally implicated in sleep disorders in ASD. These include neurological problems, such as abnormalities in neurotransmitter systems (e.g. melatonin, GABA, and serotonin); medical disorders that disrupt sleep continuity (e.g. epilepsy, gastrointestinal (GI) dysfunctions, and obstructive sleep apnea); and the medications used to treat these and other co-occurring conditions. Some of these conditions both contribute to insomnia and are aggravated by insomnia (Goldman et al., 2009; Kothare and Kotagal, 2011; Liu et al., 2006; Oblak et al., 2010).

Among the factors implicated in sleep disorders, GI dysfunctions are of particular interest. The role of GI health in ASD and its impact on neuronal functioning and other physiological systems is a growing area of research. How GI dysfunctions are related to ASD is not yet clear. Many patients with ASD have a history of GI disorder symptoms, feeding issues, and unique intestinal bacterial populations, which have been shown to relate to variable symptom severity (Macfabe, 2013).

Serotonin and melatonin are derived from tryptophan. Melatonin is a major hormone involved in sleep initiation and circadian rhythm and is often prescribed to children with ASD who have sleep problems. About 90% of serotonin is produced in enterochromaffin cells of the gut (Donnerer and Lembeck, 2006). Enterochromaffin cells are also immunoreactive. High rates of immune disorders, low levels of tryptophan and its metabolites melatonin and serotonin, coupled with the fact that children with ASD have high prevalence of GI dysfunctions warrants a closer study of the role GI dysfunctions may play in comorbidities, and sleep disorders in particular present in ASD population.

A limited number of studies have addressed the role of GI dysfunctions in conditions that co-occur with ASD (Brown and Mehl-Madrona, 2011; Maenner et al., 2012; Mannion et al., 2013; Ming et al., 2008; Wang et al., 2011). Several studies reported significant associations between GI dysfunctions and sleep disorders (Maenner et al., 2012; Mannion et al., 2013; Ming et al., 2008). The studies differed in respect to sample size, selection criteria, definition of sleep disorders, source of data (medical records vs parental reports), and specific types of sleep problems assessed. Many studies did not differentiate between primary (idiopathic) and secondary ASD diagnoses. Yet, when assessing risk factors for sleep disorders, it is important to distinguish whether the ASD diagnosis is primary, or secondary to another neurological disorder, as the etiology and the associated pathophysiology may differ between idiopathic ASD and ASD secondary to neurological disorders such as cerebral palsy, tuberous sclerosis, or fragile X syndrome. Finally, none of these studies assessed the association of GI dysfunctions with the severity of sleep disorder in children with ASD.

To bridge the gap in current knowledge, our study used an ASD population with uniformly applied and verified ASD diagnoses based on gold standard diagnostic instruments. We determined whether GI dysfunctions are associated with higher odds of having sleep disorder symptoms, as well as the severity of sleep problems (defined as the number of sleep disorder symptoms), in a cohort of children with idiopathic ASD. We hypothesized that children with idiopathic ASD with GI dysfunctions will have higher odds of reporting at least one sleep disorder symptom compared to children with idiopathic ASD without this risk factor. In addition, those with GI dysfunctions will have higher odds of reporting multiple sleep disorder symptoms compared to those without GI dysfunctions. Our study is the largest study to date reporting the association between sleep disorders, individual sleep disorder symptoms, and sleep severity with GI dysfunctions in children with idiopathic ASD.

Methods

This study is a secondary analysis of data from the Autism Genetic Resource Exchange (AGRE) project. The AGRE project is a registry-based retrospective cohort study. We analyzed data from 731 children with ASD for whom phenotypic data were collected (Autism Genetic Resource Exchange Data Source, 2013). AGRE granted permission to download the phenotypic data. AGRE collected these data from patient records, parental reports, and on-site exams. AGRE is the largest DNA repository and family registry, housing a database of genotypic and phenotypic information made available to autism researchers worldwide. It includes over 1700 well-characterized families with children with ASD. During AGRE recruitment, families identified the diagnosis they received from their physician or autism specialist. Families with one or more children diagnosed under the Diagnostic and Statistical Manual of Mental Disorders (4th ed., text rev.; DSM-IV-TR with an ASD, including autism, pervasive developmental disorder not otherwise specified, or Asperger’s syndrome) were accepted into the program, as long as they had not participated in any other autism gene bank. The ASD diagnosis was confirmed by AGRE’s physicians who were trained by a research-certified Autism Diagnostic Observation Schedule (ADOS)/Autism Diagnostic Interview-Revised (ADI-R) trainer and who underwent reliability reviews throughout the data collection process. We abstracted data for our study from the AGRE electronic database and determined the risk of developing non-febrile seizures in children with idiopathic ASD. We limited our study to children with well-documented idiopathic ASD to exclude other factors that may influence risk for sleep disorders (such as found in children with ASD secondary to other neurological conditions).

Sample

We drew our study sample from the subset of AGRE participants with ASD who had registered with AGRE at any point in time from inception of the program to the date when AGRE suspended further recruitment (1999–2009) and for whom phenotypic data were collected. This population includes children diagnosed with ASD from families with one or more affected children for whom medical histories, environmental exposure histories, and physical/neurological exam data were available. AGRE also collected data that may flag possible non-idiopathic autism using the following indicators: known neurogenic disorders, abnormal neurological exams, abnormal imaging, molecular or medical tests, significant dysmorphology, fragile X (full mutation), pre- or peri-natal injuries, prematurity less than 35 weeks, or small nuclear ribonucleoprotein polypeptide N (SNRPN) duplications. Subjects flagged by AGRE as having possible non-idiopathic autism were excluded from this study.

Inclusion and exclusion criteria

Children were included in the study sample if they were of age 2 to 18 years and diagnosed with an ASD. A minimum age of 2 years was chosen to ensure that diagnostic criteria involving language function have been evaluated accurately.

Children were excluded who had diagnoses of tuberous sclerosis, Landau–Kleffner Syndrome, Rett syndrome, cerebral palsy, Lennox–Gastaut syndrome, West syndrome, and continuous spike-and-wave syndrome (CSWS) as these diagnoses are indicative of secondary ASD. Individuals with any history of status epilepticus (Zeng et al., 2007), neonatal seizures, infantile spasms, and neurofibromatosis-1 (NF-1) were also excluded because such history may also indicate severe brain abnormalities and secondary (non-idiopathic) autism.

Measures

AGRE verified the ASD diagnosis received by participants from the physician or autism specialist using the ADOS and the ADI-R, which are both based on the DSM-IV diagnostic criteria for ASD (De Bildt et al., 2004; Gotham et al., 2008; Lord, 1991; Lord et al., 1994, 1997). In our sample, everyone had a diagnosis of an ASD prior to enrollment with AGRE and had met the following criteria during AGRE assessment: (1) minimum scores required for autism on ADI-R and/or ADOS or (2) minimum required scores for autism spectrum on ADOS.

ADOS

The ADOS is a semi-structured assessment of communication, social interaction, repetitive behavior, and play characteristics for individuals suspected of having autism or other pervasive developmental disorders. The ADOS consists of four modules, each of which is appropriate for children and adults of differing developmental and language levels, ranging from nonverbal to verbally fluent (Lord, 1991).

ADI-R

The ADI-R is a clinical diagnostic instrument for assessing autism in children and adults. The ADI-R provides a diagnostic algorithm for autism as described in both the International Classification of Diseases-10th Edition (ICD-10) and DSM-IV. The instrument focuses on behavior in three main areas: qualities of reciprocal social interaction; communication and language; and restricted and repetitive, stereotyped interests and behaviors. The ADI-R is appropriate for children and adults with mental ages from about 18 months and above (Lord et al., 1994).

Medical history data

AGRE pediatric neurologists conducted an in-home structured medical history interview to collect retrospective health information on each study participant. In our study, the sleep disorder outcome variable was coded as “yes” if the AGRE records indicated presence of at least one of the six sleep disorder symptoms (difficulty falling asleep, interrupted sleep, night terrors, excessive movements, undergoing a sleep study, or having a formal sleep disorder diagnosis). We also created a measure of the sleep disorder severity based on the number of sleep disorder symptoms present.

In the AGRE database, two variables contained data on GI dysfunctions. The first is a binary variable coded as “yes” if any of the GI symptoms were present. The second is a GI symptom variable consisting of the following categories: gastroesophageal (GE) reflux, peptic ulcer disease, irritable bowel syndrome, inflammatory bowel disease, chronic diarrhea, constipation, other, unknown, and multiple.

We also included variables to assess for potential confounding effects such as febrile and non-febrile seizures, current and past medications and supplements, behavioral problems, and nightly bed wetting. The nightly bed wetting variable was defined based on AGRE variable indicating failure to achieve toilet training for bladder at night or computed as failure to achieve such training by the age of 60 months.

Statistical analyses

All analyses were performed using SAS version 9.4. Frequency distributions and prevalence rates were calculated for each of the characteristics of interest. Bivariate analyses were performed to measure the association of covariates with presence or absence of sleep disorder symptoms. A prevalence odds ratio (OR) was calculated to test for a significant association. Categorical variables were assessed using Rao–Scott adjusted χ² test for association. Group means were assessed using cluster-adjusted Wald F tests for linear regression models.

Two models were created: Model 1, binary logistic regression for presence of at least one sleep disorder symptom; Model 2, multi-category logistic regression for the severity of sleep disorder, defined as the number of sleep disorder symptoms present (single, multiple, and none). To measure the association between presence of sleep disorder symptoms with GI dysfunctions, and severity of sleep disorder with GI dysfunctions, logistic regression analyses were used, treating each family as a cluster. This approach allowed full statistical control for the main effects of measured and unmeasured familial confounders.

The AGRE data include a majority of families with more than one child with ASD, so not all observations were independent. The non-independence of observations was accounted for in all analyses by adjusting variance estimations using SAS PROC SURVEY procedures (PROC SURVEYFREQ and PROC SURVEYLOGISTIC). The SAS PROC SURVEY methods were developed for complex survey data in which the assumptions of random population sampling and independence of observations are not met (Chen and Gorrell, 2008). The procedures account for the random effects of within-family covariates and, subsequently, adjust the variance, significance levels, and confidence intervals (CIs) for family clusters. All statistical tests were two-tailed with statistical significance attained when the two-tailed p-value was less than 0.05, unless otherwise stated.

Results

Sample characteristics

After selection criteria were applied, the total sample consisted of 610 children with ASD aged from 2 to 18 years with a mean age of 8.34 (standard error (SE) of mean = 0.16) years, from 324 families. Our sample included 78.0% boys and 22.0% girls (3.54:1 M:F ratio). This ratio is consistent with the known 3:1 to 4:1 male–female ratio for the full spectrum of ASD (Gillberg, 1990; Volkmar et al., 1993). In our sample, 582 (95.4%) children were from families with more than one child affected with ASD (multiplex families) and 28 (4.6%) were from families with only one affected child (simplex families).

GI dysfunctions

The total prevalence of GI dysfunctions in our sample was 42.5% (251/591). Table 1 shows relationships between individual sleep disorder symptoms and presence or absence of GI dysfunctions in our sample. There were significantly more sleep disorder symptoms among children with GI problems (64.1%; 150/234) than for those without GI problems (50.8%; 156/307). Among the types of GI problems reported, 2.0% were GE reflux (12/610), 0.3% irritable bowel syndrome (2/610), 10.8% chronic diarrhea (66/610), 14.1% constipation (86/610), 5.4% other GI dysfunction (33/610), and 8.5% multiple GI problems (52/610). Since the specific GI symptoms were not balanced across the categories, with the irritable bowel syndrome category having only two observations, and other AGRE categories such as “other GI dysfunctions” and “multiple GI problems” not containing details on what specific GI conditions the categories included, for our further analyses, we chose the binary GI dysfunction variable, which indicated presence or absence of any of the above-described GI symptoms.

Table 1.

Individual sleep symptom variables by GI dysfunctions (n = 610).

GI affected status
Sleep variable	Total	Affected # (%)	Unaffected # (%)	Prevalence odds ratio (95% CI)
Sleep disorder symptoms^a,*	541
Yes		150 (64.1)	156 (50.8)	1.73 (1.21–2.47)
No		84 (35.9)	151 (49.2)	1.73 (1.21–2.47)
Difficulty falling asleep	525
Yes		101 (44.1)	108 (36.5)	1.37 (0.95–1.99)
No		128 (55.9)	188 (63.5)	1.37 (0.95–1.99)
Interrupted sleep*	525
Yes		99 (43.2)	86 (29.1)	1.86 (1.30–2.67)
No		130 (56.8)	210 (70.1)	1.86 (1.30–2.67)
Night terrors	524
Yes		24 (10.5)	26 (8.8)	1.22 (0.72–2.08)
No		204 (89.5)	270 (91.2)	1.22 (0.72–2.08)
Excessive movements	514
Yes		17 (7.6)	19 (6.6)	1.16 (0.59–2.29)
No		208 (92.4)	270 (93.4)	1.16 (0.59–2.29)
Sleep study	230
Yes		2 (1.8)	5 (4.1)	0.43 (0.08–2.32)
No		107 (98.2)	116 (95.9)	0.43 (0.08–2.32)
Sleep disorder diagnosis	522
Yes		9 (4.0)	13 (4.4)	0.90 (0.41–1.99)
No		217 (96.0)	283 (95.6)	0.90 (0.41–1.99)

GI: gastrointestinal; CI: confidence interval.

Primary outcome variable.

Sleep disorder symptom variable with a significant association with GI dysfunctions.

Sleep disorders

In our sample, 56.3% reported at least one sleep disorder symptom (306/544). Among those who reported sleep problems, 39.6% reported difficulty falling asleep (209/528), 35.0% reported interrupted sleep (185/528), 9.5% reported night terrors (50/527), 7.0% reported excessive movements (36/517), 4.2% had a sleep disorder diagnosis (22/525), and 3.0% had undergone a sleep study (7/231). Of the six individual sleep disorder symptoms, only interrupted sleep was significantly associated with the presence of GI dysfunctions (OR: 1.86; 95% CI: 1.30–2.67) (Table 1). The denominator for reported prevalence varies according to the number of participants who responded to specific question. To make the best use of all information available, for our subsequent analyses, we chose a variable denoting presence of at least one sleep disorder symptom as a primary outcome. A bivariate analysis of relationships between sleep disorder and presence or absence of GI dysfunctions showed that those with GI dysfunctions had 73% greater odds of having had at least one sleep disorder symptom compared to those without GI dysfunctions (odds ratio (OR): 1.73; 95% CI: 1.21–2.47) (Table 1).

In bivariate analyses, we assessed other possible risk factors for sleep disorders in ASD, such as seizures (febrile and non-febrile), behavioral problems, gestational age, past and current medications and supplements, enuresis, age, and gender. Only behavioral problems, nightly bed wetting, and current supplements showed significant relationships with presence or absence of sleep disorder (OR: 2.31; 95% CI: 1.59–3.37, OR: 1.81; 95% CI: 1.26–2.58, and OR: 1.58; 95% CI: 1.08–2.32, respectively) (Table 2). Measures of intellectual disability and motor deficit were not available (Maenner et al., 2012); therefore, these factors were not assessed in our models. The measure of cognitive functioning collected by AGRE was the Raven Color Picture Test. However, since a large portion of our sample had missing observations for this variable, the Raven test score was not included in our final analyses.

Table 2.

Sleep disorder symptoms (yes/no) by covariates (n = 610).

Sleep disorder symptoms
Covariate	Total	Affected # (%); mean (SE)	Unaffected # (%); mean (SE)	Prevalence odds ratio (95% CI)
GI problems^a	541
Yes		150 (49.0)	84 (35.7)	1.73 (1.21–2.47)
No		156 (51.0)	151 (64.3)	1.73 (1.21–2.47)
Age (years) at AGRE medical assessment**	544	8.48 (0.21)	8.30 (0.24)	1.02 (0.97–1.07)
Gestational age at birth (weeks)	542	39.17 (0.12)	39.17 (0.13)	1.00 (0.91–1.11)
Gender**	544
Male		241 (78.8)	185 (77.7)	1.06 (0.71–1.60)
Female		65 (21.2)	53 (22.3)	Reference
Race_b	544
White		248 (81.0)	177 (74.4)	1.47 (0.95–2.28)
Other race, multiple race, or unknown		58 (19.0)	61 (25.6)	Reference
Ethnicity	524
Hispanic or Latino		59 (20.0)	50 (21.8)	0.90 (0.56–1.42)
Not Hispanic or Latino		236 (80.0)	179 (78.2)	Reference
Febrile seizures	535
Yes		11 (3.7)	6 (2.6)	1.45 (0.54–3.95)
No		289 (96.3)	229 (97.4)	Reference
Non-febrile seizures	544
Yes		25 (8.2)	20 (8.4)	0.97 (0.50–1.87)
No		281 (91.8)	218 (91.6)	Reference
Behavioral problem symptoms*	536
Yes		203 (67.4)	111 (47.2)	2.31 (1.59–3.37)
No		98 (32.6)	124 (52.8)	Reference
Nightly bed wetting*	544
Yes		149 (48.7)	82 (34.5)	1.81 (1.26–2.58)
No		157 (51.3)	156 (65.5)	Reference
Past ASD medications	544
Yes		137 (44.8)	87 (36.6)	1.41 (0.97–2.05)
No		169 (55.2)	151 (63.4)	Reference
Current ASD medications	544
Yes		125 (40.8)	81 (34.0)	1.34 (0.91–1.98)
No		181 (59.2)	157 (66.0)	Reference
Current seizure medications	544
Yes		25 (8.2)	12 (5.0)	1.68 (0.86–3.25)
No		281 (91.8)	226 (95.0)	Reference
Past supplement	544
Yes		110 (35.9)	65 (27.3)	1.49 (1.00–2.23)
No		196 (64.1)	173 (72.7)	Reference
Current supplements*	544
Yes		126 (41.2)	73 (30.7)	1.58 (1.08–2.32)
No		180 (58.8)	165 (69.3)	Reference

SE: standard error; GI: gastrointestinal; CI: confidence interval; AGRE: Autism Genetic Resource Exchange; ASD: autism spectrum disorder.

Primary outcome variable.

Race and Ethnicity categories were collapsed to avoid empty cells. Race categories “American Indian/Alaskan Native” (n = 2), “Asian” (n = 26), “Black or African American” (n = 8), “More than One Race” (n = 46), “Native Hawaiian or other Pacific Islander” (n = 6), “Unknown” (n = 36) were combined under “Other race, multiple race, or unknown.” For our Ethnicity variable, the category “Unknown” (n = 22) was set to missing.

Significant covariate.

Variables included into full model.

In Model 1, to assess confounding effects of each covariate described in Table 2, we added each one to the main effects model, and if addition of the covariate changed the point estimate of the main effect by 10% or more, it was considered a confounder to be adjusted for in the final model (Rothman et al., 2008). None of the variables tested changed the crude OR by more than 10%. A goodness-of-fit test was calculated comparing the model fit with the data. The model with age and gender was compared to the main effects model. Then, the most complex model that included age, gender, behavioral problems, bed wetting, and current supplements was compared to the main effects model with age and gender added. The saturated model had a significant chi-square deviance test indicating a better fit than the simpler model (Table 4). Although age and gender did not meet the 10% change-in-effect criterion, nor did the extended model with age and gender show an improvement in fit compared to the main effects model (Table 4), these two variables were included into the final model as valuable demographic information. The odds of having at least one sleep disorder symptom among children with GI dysfunctions, adjusted for age, gender, behavioral problems, bed wetting, and current use of supplements were 1.68 (95% CI: 1.16–2.44) compared to children without GI dysfunctions.

In addition to the main model, the model was tested measuring the association of the severity of sleep disorder, defined as the number of sleep disorder symptoms present, with covariates of interest (Table 3).

Table 3.

Sleep severity variable by covariates (n = 610).

Number of sleep disorder symptoms
Covariate	Total	None # (%); mean (SE)^a	Single # (%); mean (SE); prevalence odds ratio (95% CI)	Multiple # (%); mean (SE); prevalence odds ratio (95% CI)
GI problems^b	512
Yes		81 (35.8)	77 (49.0)80 (51.0)1.73 (1.21–2.47)	67 (51.9)62 (48.1)1.93 (1.26–2.98)
No		145 (64.2)	77 (49.0)80 (51.0)1.73 (1.21–2.47)	67 (51.9)62 (48.1)1.93 (1.26–2.98)
Age (years) at AGRE medical assessment	515	8.32 (0.25)	8.65 (0.28)1.03 (0.97–1.09)	8.31 (0.30)1.00 (0.94–1.06)
Gestational age (weeks)	510	39.21 (0.13)	39.18 (0.16)0.99 (0.88–1.11)	39.19 (0.20)0.99 (0.86–1.15)
Gender	515
Male		177 (77.3)	123 (78.3)34 (21.7)1.06 (0.64–1.78)	104 (80.6)25 (19.4)1.22 (0.73–2.05)
Female		52 (22.7)	123 (78.3)34 (21.7)1.06 (0.64–1.78)	104 (80.6)25 (19.4)1.22 (0.73–2.05)
Race	515
White		170 (74.2)	123 (78.3)34 (21.7)1.25 (0.77–2.04)	107 (82.9)22 (17.1)1.69 (0.95–3.01)
Other race, multiple race, or unknown		59 (25.8)	123 (78.3)34 (21.7)1.25 (0.77–2.04)	107 (82.9)22 (17.1)1.69 (0.95–3.01)
Ethnicity	496
Hispanic or Latino		50 (22.7)	32 (21.3)118 (78.7)0.92 (0.54–1.59)	23 (18.3)103 (81.7)0.76 (0.43–1.35)
Not Hispanic or Latino		170 (77.3)	32 (21.3)118 (78.7)0.92 (0.54–1.59)	23 (18.3)103 (81.7)0.76 (0.43–1.35)
Febrile seizures	509
Yes		6 (2.7)	3 (1.9)151 (98.01)0.73 (0.18–2.93)	7 (5.4)122 (94.6)2.10 (0.70–6.33)
No		220 (97.3)	3 (1.9)151 (98.01)0.73 (0.18–2.93)	7 (5.4)122 (94.6)2.10 (0.70–6.33)
Non-febrile seizures	515
Yes		18 (7.9)	13 (8.3)144 (91.7)1.06 (0.45–2.39)	12 (9.3)117 (90.7)1.20 (0.55–2.62)
No		211 (92.1)	13 (8.3)144 (91.7)1.06 (0.45–2.39)	12 (9.3)117 (90.7)1.20 (0.55–2.62)
Behavioral problem symptoms*	512
Yes		103 (45.4)	96 (61.5)60 (38.5)1.93 (1.25–2.97)	97 (75.2)32 (24.8)3.65 (2.21–6.02)
No		124 (54.6)	96 (61.5)60 (38.5)1.93 (1.25–2.97)	97 (75.2)32 (24.8)3.65 (2.21–6.02)
Nightly bed wetting*	515
Yes		80 (34.9)	65 (41.4)92 (58.6)1.32 (0.85–2.03)	78 (60.5)51 (39.5)2.85 (1.83–4.43)
No		149 (65.1)	65 (41.4)92 (58.6)1.32 (0.85–2.03)	78 (60.5)51 (39.5)2.85 (1.83–4.43)
Past ASD medications*	515
Yes		83 (36.2)	67 (42.7)90 (57.3)1.31 (0.84–2.05)	63 (48.8)66 (51.2)1.68 (1.06–2.66)
No		146 (63.8)	67 (42.7)90 (57.3)1.31 (0.84–2.05)	63 (48.8)66 (51.2)1.68 (1.06–2.66)
Current ASD medications*	515
Yes		77 (33.6)	58 (36.9)99 (63.1)1.16 (0.73–1.83)	60 (46.5)69 (53.5)1.72 (1.06–2.78)
No		152 (66.4)	58 (36.9)99 (63.1)1.16 (0.73–1.83)	60 (46.5)69 (53.5)1.72 (1.06–2.78)
Current seizure medications	515
Yes		12 (5.2)	10 (6.4)147 (93.6)1.23 (0.54–2.80)	14 (10.9)115 (89.1)2.20 (0.95–5.13)
No		217 (94.8)	10 (6.4)147 (93.6)1.23 (0.54–2.80)	14 (10.9)115 (89.1)2.20 (0.95–5.13)
Past supplements*	515
Yes		63 (27.5)	51 (32.5)106 (67.5)1.27 (0.80–2.01)	58 (45.0)71 (55.0)2.15 (1.31–3.53)
No		166 (72.5)	51 (32.5)106 (67.5)1.27 (0.80–2.01)	58 (45.0)71 (55.0)2.15 (1.31–3.53)
Current supplements*	515
Yes		71 (31.0)	66 (42.0)91 (58.0)1.61 (1.04–2.50)	57 (44.2)72 (55.8)1.76 (1.11–2.81)
No		158 (69.0)	66 (42.0)91 (58.0)1.61 (1.04–2.50)	57 (44.2)72 (55.8)1.76 (1.11–2.81)

SE: standard error; GI: gastrointestinal; CI: confidence interval; AGRE: Autism Genetic Resource Exchange; ASD: autism spectrum disorder.

Reference category for outcome variable.

Primary outcome variable.

Significant covariate.

In bivariate analyses, the variables significantly associated with severity of sleep disorder included the following: behavioral problems, nightly bed wetting, past and current ASD medications, and past and current supplements. The odds of having single or multiple sleep disorder symptoms among children with behavioral problems increased two-fold and three-fold, respectively, compared to children without behavioral problems. Current supplements were also associated with increased odds of having single or multiple disorder symptoms (Table 3). The odds of having multiple sleep disorder symptoms increased three-fold among children with nightly bed wetting compared to those who did not have nightly bed wetting problem. Those who reported taking ASD medications, currently or in the past, as well as taking supplements in the past, also had increased odds of having multiple sleep disorder symptoms compared to children not taking the respective medications (Table 3). The covariates, which showed significant associations with the severity of sleep disorder, were tested for potential confounding effects on the relationship between GI dysfunctions and severity of sleep disorder. In Model 2, none of the variables tested resulted in a significant change to the point estimate. Results of the goodness-of-fit test, however, indicated a better fit of the saturated model, with all covariates in question, compared to the main effect model or compared to main effect with age and gender model. For reasons similar to those expressed for Model 1, age and gender were included in the final model (Table 4). Subsequently, the odds of having multiple sleep disorder symptoms among children with GI dysfunctions, adjusted for age, gender, behavioral problems, bed wetting, current and past supplements, and current and past ASD medications were 1.75 (95% CI: 1.10–2.79) compared to children without GI dysfunctions.

Table 4.

Model comparison.

	−2Log L	#Parameters	Chi-square deviance	df	p-value
Model 1, sleep disorder symptoms
Main effect only	731.003	1	–	–	–
Main effect + age and gender	730.426	3	0.58	2	0.7483
Main effects + age and gender + all significant covariates^a,*	686.214	6	44.21	3	< 0.001
Model 2, sleep disorder symptom severity
Main effect only	1085.371	2	–	–	–
Main effect + age and gender*	1083.536	6	1.84	4	0.7652
Main effects + age and gender + all significant covariates^b,*	1016.628	18	66.91	12	< 0.001

Variables that showed a significant association with presence of sleep disorder symptoms in bivariate analyses (see Table 2).

Variables that showed a significant association with the sleep severity in bivariate analyses (see Table 3).

Final model.

To determine the best models for the data, the goodness-of-fit model comparison test was used in which each of the two main effect models were compared to two extended models that included the following: (1) age and gender and (2) age, gender, and significant covariates from a bivariate test described in Tables 2 and 3. Based on this test, the most complex models showed a better fit than the simpler modes, thus providing assurance that the chosen models are adequate (Table 4).

Based on these results, after adjusting for age, gender, behavioral problems, bed wetting, and medication and supplement use, children with idiopathic ASD who had GI dysfunctions had a 68% increase in odds of having at least one sleep disorder symptom compared to children without GI dysfunctions. In addition, children with idiopathic ASD and GI dysfunctions had a 75% increase in odds of having multiple sleep disorder symptoms compared to children without GI dysfunctions.

Discussion

Sleep disorders among children with ASD have profound impact on child’s health and development. They contribute to the severity of the ASD symptoms and have a negative impact not only with children’s but also the parents’ quality of life. GI dysfunctions were found to be important risk factors for a number of medical and psychiatric conditions that co-occur with ASD, including sleep disorders. GI dysfunctions are potentially treatable and, if the association proves causal, treatment could reduce the risk of developing sleep disorders in ASD. Our study is the largest study to date estimating the effects of GI dysfunctions on sleep disorders and severity of sleep disorders in children with idiopathic ASD. In both the crude model and the model adjusted for age and gender where we controlled statistically for familial clustering, the odds of having sleep disorders was significantly higher among children with ASD who had GI dysfunctions (OR_(crude) 1.73; 95% CI: 1.21–2.47 and OR_(adjusted) 1.68; 95% CI: 1.16–2.44, respectively). In addition, children with GI dysfunctions had nearly twice the odds of having multiple sleep disorder symptoms (OR_(adjusted) 1.75; 95% CI: 1.10–2.79) compared to children without GI dysfunctions. Of the individual sleep disorder symptoms, only interrupted sleep was found to be significantly associated with GI dysfunctions (OR_(crude) 1.86; 95% CI: 1.30–2.67). The association of GI dysfunctions with interrupted sleep in our ASD sample is consistent with the results from the general population that linked certain types of GI dysfunction with increased night awakening and pain (Johnson, 2005). Clinical diagnosis of sleep disorder and undergoing a sleep study, individually or combined, were not found to be significantly associated with the presence or absence of GI dysfunctions possibly due to lack of statistical power to detect the effect (Table 1). We found that behavioral problem symptoms, nightly bed wetting, and current supplements were significantly associated with higher odds of having sleep disorders in our bivariate analyses (Table 2). These variables, as well as the current use of medications for ASD behaviors, were also associated with the severity of sleep disorder (Table 3). The observed relationship between the presence or absence of sleep disorders and the use of supplements may be due to the fact that supplements such as melatonin are often prescribed for children with ASD and sleep issues (Malow et al., 2016). The list of supplements contained in AGRE data was very broad and included dietary supplements, vitamins, minerals, and melatonin. Isolating just the supplements specific to sleep was not feasible. The relationship between the use of medications given for ASD behavior and the severity of sleep disorder also needs to be treated with caution. Some medications for ASD symptoms have either insomnia or sedation as a side effect. Therefore, it is important to take side effects into consideration when examining relationship between sleep and use of medication/supplements in ASD.

Several studies reported significant associations between GI dysfunctions and sleep disorders (Kang et al., 2014; Maenner et al., 2012; Mannion et al., 2013; Ming et al., 2008). A number of studies linked the presence of GE reflux with the increased prevalence of sleep problems, including sleep apnea, in general, and ASD populations (Johnson and Malow, 2008; Jung et al., 2010). However, low prevalence of GE reflux in our sample (2%) does not allow drawing definitive conclusions about the role of this specific GI disorder in the observed relationship between sleep problems and GI dysfunctions.

The inconsistency in the specific types of sleep problems studied, differences in measures of associations reported across the studies, and the lack of information on whether presence of GI dysfunctions impacts the severity of sleep disturbances in children with ASD, warranted a more narrow study of the relationship between sleep disorders and GI dysfunctions. In addition, previous studies often included both children with idiopathic ASD and those with ASD secondary to such neurological disorders as cerebral palsy, Landau–Kleffner, tuberous sclerosis, or fragile X syndrome, all of which present an independent risk for sleep disorders, thus potentially distorting the reported relationships between sleep disorders and GI dysfunctions. For instance, a study by Maenner et al. (2012) based on Autism and Developmental Disabilities Monitoring (ADDM) network data included both primary and secondary ASD. And while the sample used included only children who had medically documented GI problems, the reported odds of having sleep disorders for those with GI dysfunctions were higher than the results found in our study (OR: 3.1; 95% CI: 1.5–6.4) (Maenner et al., 2012). Such inconsistencies across the studies in reported association between sleep disorders and GI dysfunctions emphasize greater need for making distinction between idiopathic and non-idiopathic ASD when studying comorbidities in this population, as the risk factors for some of the co-occurring conditions may differ in primary versus secondary ASD. A study by Kang et al. (2014) used a sample with idiopathic ASD only, and while they reported higher rates of sleep disorders among those with any symptoms of GI disorder, the study contained no information on individual types of sleep problems and it did not address the relationship between GI dysfunctions and severity of sleep disorder.

The growing number of findings of the relationship between sleep problems and GI dysfunctions in children with ASD lead to questions about the possible mechanism that may underlie both conditions so frequent in this population. Some studies suggest that autonomic arousal is implicated in both GI function and sleep disorders in children with ASD and that further research into common neural mechanisms that regulate these processes is warranted (Mazurek et al., 2013; Mazurek and Petroski, 2015). In addition, low levels of tryptophan and its metabolites, serotonin and melatonin, found in children with ASD merits further research into the role of metabolic status in sleep disorders and GI dysfunctions, as well as a number of other comorbidities found in this population.

Strengths and limitations

Limitations

Although AGRE participants were recruited nationwide, the sample is not representative of the US population as most of the families were recruited in regions where Autism Speaks has significant activity. Since the majority of families had more than one child with ASD, generalizability to children with ASD to families with only one child with ASD is limited. In addition, some of the data are based on parental report rather than direct assessment. In general, the use of secondary data poses certain limitations. For our study, one limitation was that some of the variables needed were unavailable. For example, prior studies identified, along with age and gender, moderate-to-severe intellectual disability and motor deficits among the risk factors for sleep disorders in the ASD population (Maenner et al., 2012; Mannion et al., 2013). Since there were no specific measures for these conditions in the AGRE database, we could not include these factors in our models. In addition, the GI symptoms variable had poorly defined categories in which symptoms were lumped together into “multiple” and “other” categories without providing any additional information regarding which specific symptoms they contained, rendering this variable unusable for our analyses. Apart from these instances, the variables needed to address our aims were well represented in the AGRE database. Finally, medical histories were obtained retrospectively.

Strengths

Unlike previous studies, our study used a well-defined ASD population while excluding those with secondary ASD diagnoses to eliminate potential confounding factors associated with the primary neurological disorders. This approach, while producing more conservative estimates, resulted in greater precision. AGRE conducted systematic re-rating of the ADOS and ADI-R assessments for each audit while using electronic scoring and validation through the Internet System for Assessing Autistic Children (ISAAC), which resulted in data that is both valid and highly reliable. In addition, AGRE being one of the largest data repositories containing genetic and phenotypic data on ASD allowed a large sample for our analyses, making our study the largest study to date that assessed the effect GI dysfunctions have on the risk for and severity of sleep disorders in children with ASD.

The use of SAS methods for survey data allowed us to cluster families while computing between-family ORs of having sleep disorders and adjust for the effects of non-independent observations. Treating each family as a cluster provided full statistical control of not only the main effects of measured family-level confounding factors but also unmeasured factors such as the aggregation of phenotypical characteristics in multiplex families.

Finally, although previous studies reported a positive association between GI dysfunctions and sleep disorders in children with ASD, our study produced additional evidence for the effect GI dysfunctions have on the severity of sleep disorders in this population. Our study also describes the association between GI dysfunctions and individual sleep disorder symptoms—data that had not been consistently reported in previous studies.

Conclusion

Our study found that GI dysfunctions increase the odds of reporting sleep disorders in children with idiopathic ASD, independent of age and gender. This study provides key evidence for clinicians as well as parents of children with ASD that suggests higher risk of sleep disorders among children with ASD who have GI dysfunctions. Sleep disorders further exacerbate autism symptomatology. Therefore, early intervention would be expected to result in significant improvement in long-term developmental outcomes in children with ASD.

More research is needed to determine whether or not there is a causal relationship between GI dysfunctions and the onset or severity of sleep disorders. If there is a causal link between GI health and sleep disorders, early screening, monitoring, and treatment of children with GI dysfunctions could be effective in reducing the presence and severity of sleep disorders in ASD. Such research has the potential to identify risk early in life, possibly ameliorate the risk, and improve quality of life not only for the child but for the parents as well.

Footnotes

Acknowledgements

The authors gratefully acknowledge the resources provided by the Autism Genetic Resource Exchange (AGRE) Consortium* and the participating AGRE families. We thank Dr Thomas E Conturo for his contributions to this study.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Autism Genetic Resource Exchange is a program of Autism Speaks and is supported, in part, by grant 1U24MH081810 from the National Institute of Mental Health to Clara M. Lajonchere (PI). This work was supported by NIH grant R21 MH105822.

References

Autism Genetic Resource Exchange Data Source (2013) Available at: https://research.agre.org/agredownload/datacenter.cfm?do-pheno (accessed 8 August 2013).

Barnard

Nolan

(2008) When clocks go bad: neurobehavioural consequences of disrupted circadian timing. PLoS Genetics 4(5): e1000040.

Brown

Mehl-Madrona

(2011) Autoimmune and gastrointestinal dysfunctions: does a subset of children with autism reveal a broader connection? Expert Review of Gastroenterology & Hepatology 5(4): 465–477.

Chen

Gorrell

(2008) An introduction to SAS survey analysis PROCs. In: NESUG proceedings. Available at: http://www.lexjansen.com/nesug/nesug08/sa/sa06.pdf

Cortesi

Giannotti

Ivanenko

et al . (2010) Sleep in children with autistic spectrum disorder. Sleep Medicine 11(7): 659–664.

De Bildt

Sytema

Ketelaars

et al . (2004) Interrelationship between Autism Diagnostic Observation Schedule-Generic (ADOS-G), Autism Diagnostic Interview-Revised (ADI-R), and the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR) classification in children and adolescents with mental retardation. Journal of Autism and Developmental Disorders 34(2): 129–137.

Donnerer

Lembeck

(2006) The Chemical Languages of the Nervous System: History of Scientists and Substances. Basel; New York: Karger.

Doo

Wing

(2006) Sleep problems of children with pervasive developmental disorders: correlation with parental stress. Developmental Medicine and Child Neurology 48(8): 650–655.

Gillberg

(1990) Autism and pervasive developmental disorders. Journal of Child Psychology and Psychiatry 31(1): 99–119.

10.

Goldman

Surdyka

Cuevas

et al . (2009) Defining the sleep phenotype in children with autism. Developmental Neuropsychology 34(5): 560–573.

11.

Gotham

Risi

Dawson

et al . (2008) A replication of the Autism Diagnostic Observation Schedule (ADOS) revised algorithms. Journal of the American Academy of Child and Adolescent Psychiatry 47(6): 642–651.

12.

Johnson

(2005) Gastroesophageal reflux disease and sleep disorders: a wake-up call for physicians and their patients. Reviews in Gastroenterological Disorders 5(Suppl. 2): S3–S11.

13.

Johnson

Malow

(2008) Sleep in children with autism spectrum disorders. Current Treatment Options in Neurology 10(5): 350–359.

14.

Johnson

Giannotti

Cortesi

(2009) Sleep patterns in autism spectrum disorders. Child and Adolescent Psychiatric Clinics of North America 18(4): 917–928.

15.

Jung

Choung

Talley

(2010) Gastroesophageal reflux disease and sleep disorders: evidence for a causal link and therapeutic implications. Journal of Neurogastroenterology and Motility 16(1): 22–29.

16.

Kang

Wagner

Ming

(2014) Gastrointestinal dysfunction in children with autism spectrum disorders. Autism Research 7(4): 501–506.

17.

Kothare

Kotagal

(2011) Sleep in Childhood Neurological Disorders. New York: Demos Medical Publishing.

18.

Liu

Hubbard

Fabes

et al . (2006) Sleep disturbances and correlates of children with autism spectrum disorders. Child Psychiatry & Human Development 37(2): 179–191.

19.

Lord

(1991) Methods and measures of behavior in the diagnosis of autism and related disorders. The Psychiatric Clinics of North America 14(1): 69–80.

20.

Lord

Pickles

McLennan

et al . (1997) Diagnosing autism: analyses of data from the Autism Diagnostic Interview. Journal of Autism and Developmental Disorders 27(5): 501–517.

21.

Lord

Rutter

Le Couteur

(1994) Autism Diagnostic Interview-Revised: a revised version of a diagnostic interview for caregivers of individuals with possible pervasive developmental disorders. Journal of Autism and Developmental Disorders 24(5): 659–685.

22.

Macfabe

(2013) Autism: metabolism, mitochondria, and the microbiome. Global Advances in Health and Medicine 2(6): 52–66.

23.

Maenner

Arneson

Levy

et al . (2012) Brief report: association between behavioral features and gastrointestinal problems among children with autism spectrum disorder. Journal of Autism and Developmental Disorders 42(7): 1520–1525.

24.

Malow

Katz

Reynolds

et al . (2016) Sleep difficulties and medications in children with autism spectrum disorders: a registry study. Pediatrics 137(Suppl. 2): S98–S104.

25.

Mannion

Leader

Healy

(2013) An investigation of comorbid psychological disorders, sleep problems, gastrointestinal symptoms and epilepsy in children and adolescents with Autism Spectrum Disorder. Research in Autism Spectrum Disorders 7(1): 35–42.

26.

Mazurek

Petroski

(2015) Sleep problems in children with autism spectrum disorder: examining the contributions of sensory over-responsivity and anxiety. Sleep Medicine 16(2): 270–279.

27.

Mazurek

Vasa

Kalb

et al . (2013) Anxiety, sensory over-responsivity, and gastrointestinal problems in children with autism spectrum disorders. Journal of Abnormal Child Psychology 41(1): 165–176.

28.

Meltzer

(2008) Brief report: sleep in parents of children with autism spectrum disorders. Journal of Pediatric Psychology 33(4): 380–386.

29.

Ming

Walters

(2009) Autism spectrum disorders, attention deficit/hyperactivity disorder, and sleep disorders. Current Opinion in Pulmonary Medicine 15(6): 578–584.

30.

Ming

Brimacombe

Chaaban

et al . (2008) Autism spectrum disorders: concurrent clinical disorders. Journal of Child Neurology 23(1): 6–13.

31.

Oblak

Gibbs

Blatt

(2010) Decreased GABA(B) receptors in the cingulate cortex and fusiform gyrus in autism. Journal of Neurochemistry 114(5): 1414–1423.

32.

Richdale

(1999) Sleep problems in autism: prevalence, cause, and intervention. Developmental Medicine & Child Neurology 41(1): 60–66.

33.

Rothman

Greenland

Lash

(2008) Modern Epidemiology. Philadelphia, PA: Lippincott Williams & Wilkins.

34.

Sivertsen

Posserud

Gillberg

et al . (2012) Sleep problems in children with autism spectrum problems: a longitudinal population-based study. Autism 16(2): 139–150.

35.

Volkmar

Szatmari

Sparrow

(1993) Sex differences in pervasive developmental disorders. Journal of Autism and Developmental Disorders 23(4): 579–591.

36.

Wang

Tancredi

Thomas

(2011) The prevalence of gastrointestinal problems in children across the United States with autism spectrum disorders from families with multiple affected members. Journal of Developmental and Behavioral Pediatrics 32(5): 351–360.

37.

Zeng

Rensing

et al . (2007) Kainate seizures cause acute dendritic injury and actin depolymerization in vivo. The Journal of Neuroscience 27(43): 11604–11613.