Systematic review investigating the relationship between autism spectrum disorder and metabolic dysfunction

Abstract

The objective of this systematic review is to examine metabolic dysfunction, specifically metabolic syndrome and its components, as well as type 2 diabetes mellitus (T2DM) as it relates to individuals with a diagnosis of Autism Spectrum Disorder (ASD). We searched PubMed, Embase, Cochrane, PsychInfo, and Scopus from January 1, 1998 to October 12, 2018 for English, peer-reviewed, original articles containing adult and pediatric populations with any form of ASD and metabolic dysfunction, including T2DM, hyperglycemia, hypertension, dyslipidemia, or central obesity. Exclusion criteria included studies without ASD-specific results, basic science research, review papers, case studies, and medication clinical trials. Eight studies were included in this review, with a total of 70,503 participants with ASD and 2,281,891 in comparison groups. Within ASD populations, higher prevalence for metabolic syndrome components hyperglycemia, hypertension, and dyslipidemia were observed, as well as increased incidence and prevalence of T2DM. However, heterogeneity of study definitions and measurements should be noted. While there is evidence of increased prevalence of T2DM, hyperglycemia, hypertension, and dyslipidemia for those with ASD, the relationship is poorly understood. There is also lack of research investigating central obesity and risk of metabolic syndrome as a diagnosis. More research addressing these gaps is warranted to evaluate the risk of metabolic dysfunction in populations with ASD.

1. Introduction

In the past twenty years, autism spectrum disorder (ASD) has been a prominent topic of discussion in the realm of psychiatry. Widespread diagnosis of ASD began in the last few decades, and many are being diagnosed not only in childhood, but also in adulthood. Globally, 1–2 % of adults and children are estimated to have ASD (CDC, 2016).

Comorbidities associated with ASD, such as obesity, hypertension and dyslipidemia, can decrease life expectancy (Hwang et al., 2019; Smith DaWalt et al., 2019). Metabolic syndrome and T2DM are two disease states of metabolic dysfunction that are widely researched because they are chronic and costly diseases affecting many people globally and can lead to more serious health conditions such as myocardial infarction, heart disease, or other adverse events (O’Neill & O’Driscoll, 2015; Rao Kondapally Seshasai et al., 2011). In this review, we define metabolic dysfunction as encompassing type 2 diabetes (T2DM) and the following components of metabolic syndrome (Alberti et al., 2009): hyperglycemia (fasting blood glucose ≥100 mg/dL), central obesity, dyslipidemia (triglycerides ≥150 mg/dL (1.7 mmol/L) or high density lipoprotein (HDL) <40 mg/dL (1.0 mmol/L) in males; <50 mg/dL (1.3 mmol/L) in females) and hypertension (systolic ≥130 and/or diastolic ≥85 mm Hg), and type 2 diabetes (T2DM).

In 2009, a statement released jointly by a combination of health organizations stated that there is no obligatory component for metabolic syndrome, but individual components should be considered for risk prediction, and ethnic variations in measurements should be considered (Alberti et al., 2009). Thus, this review will focus not only on a diagnosis of metabolic syndrome using criteria from this joint statement, but individual components as well. While metabolic syndrome includes hyperglycemia as a component, it is also important to include T2DM separately as an indicator of metabolic dysfunction because of potentially differing diagnostic criteria and pathogenesis that can lead to insulin resistance (Olokoba et al., 2012).

Few studies have explored metabolic dysfunction in individuals with a diagnosis of ASD. Interventions focusing on these indicators of metabolic dysfunction may help significantly decrease comorbidity in individuals with ASD and prevent serious adverse health events from occurring (O’Neill & O’Driscoll, 2015). Thus, the purpose of this study was to summarize current research investigating if populations with ASD have an increased risk and prevalence of metabolic dysfunction, specifically metabolic syndrome and its components, such as hyperglycemia, hypertension, dyslipidemia, and central obesity, in addition to T2DM.

2. Methods

2.1. Search strategy and selection criteria for systematic review

Searches were conducted on October 12, 2018 within five databases: Pubmed, Scopus, Embase, Cochrane, and PsychInfo for articles published in the last two decades (1998–2018). Search strings followed a basic structure of [autis* OR asperger* OR pervasive developmental disorder OR PDD OR ASD] and [metabolic OR obesity OR diabetes OR hyperglycemia OR hypertension OR hyperlipidemia OR dyslipidemia]. When applicable, other search terms were used to filter by date and study subject. Specific search strings for each database may be found in a supplemental file. Endnote X8 was used as the citation manager for all steps of the review, including importation of search results, merging, deduplication, screening and sorting, and full text storage.

Papers were included if they met all of the following inclusion criteria:

1. Published between January 1, 1998 and October 12, 2018

2. Human population studied, including child or adult participants with any type of ASD (such as Asperger’s or pervasive developmental disorder) using a comparison group without a diagnosis of ASD.

3. English, peer reviewed, original research, with clinical or translational focus (observational or longitudinal studies only)

4. T2DM or metabolic syndrome associated components including hypertension, hyperglycemia, dyslipidemia, or central obesity as study outcome or measurement

Exclusion criteria included animal studies, basic science research, review papers, case studies or case series less than ten participants, medication clinical trials, studies focusing only on body mass index (BMI) determined obesity for adults and children, studies focusing on only type 1 diabetes (T1D), and genetic studies that included populations with disorders other than ASD specifically. These study types were excluded because our study focused on a specific definition of metabolic dysfunction expressed phenotypically and at a population level. Medication trials were excluded because populations were not specific for ASD. Investigation of child obesity by BMI was not evaluated in this study even though there is no consensus in defining metabolic syndrome in pediatric populations. There are many recently conducted meta analyses summarizing research on BMI, which have found strong associations between overweight or obesity defined by BMI and ASD in both adults and children (Kahathuduwa et al., 2019; Zheng et al., 2017). Our operational definition of metabolic dysfunction in adults focuses on central obesity because the accepted definition of metabolic syndrome is defined as such (Alberti et al., 2009) and does not consider BMI determined obesity, and we want to be consistent in children. We also emphasize central obesity because recent research indicates it is a stronger predictor for cardiovascular disease and related sequela compared to BMI (Lee et al., 2008; Welborn & Dhaliwal, 2007). Because repetition of recently conducted reviews on BMI is not the focus of our paper, which is more so on metabolic syndrome and T2DM, we have not included BMI analyses in our review. Central obesity may be measured using waist circumference or waist-to-hip ratio (Alberti et al., 2009).

All available results were independently screened (by first authors AYC and BMB). Both reviewers initially screened by title, abstract and keywords independently. Full texts were retrieved in the second stage of review and assessed independently by the two reviewers. Papers deemed to meet eligibility criteria were compared and a final agreement was reached through discussion, with input from other study authors. Both reviewers independently extracted data on included papers, which was then consolidated and interpreted by the first author AYC.

2.2. Data analysis

Data included study type, time of data collection, location of study, study size, age, sex, race, definition or type of ASD, comparison group criteria, metabolic dysfunction related findings of interest and method of categorization, and limitations. Excel sheets were used to chart extracted data. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were used as a guide and checklist for creation of the manuscript. Study authors were emailed questions if data were not clear and clarifications were obtained, documented in detail in the results section.

3. Results

As seen in the PRISMA diagram in Fig. 1, 6,089 papers were found in the initial search across all databases. After deduplication, title/abstract/keyword screening, and full text screening, eight papers were found eligible for inclusion in the final review.

Fig. 1.

PRISMA diagram showing each step of the search process. Papers from other sources consist of papers that were found in the reference list of papers meeting the inclusion criteria found in the original search process.

3.1. Study characteristics

Of the eight papers, seven were observational (Croen et al., 2015; Dziobek et al., 2007; Flygare Wallén et al., 2018; Kim et al., 2010; Moses et al., 2014; Shedlock et al., 2016; Tolchard & Stuhlmiller, 2018), and one was a prospective cohort (Chen et al., 2016). Studies were conducted in the United States, Korea, Taiwan, Sweden, Israel, and Australia. Studies either analyzed only a pediatric population (Kim et al., 2010; Shedlock et al., 2016), only an adult population (Croen et al., 2015; Dziobek et al., 2007; Moses et al., 2014; Tolchard & Stuhlmiller, 2018), both a pediatric and adult population (Chen et al., 2016), or a mixed-age population containing both children and adults together without distinguishing results by age (Flygare Wallén et al., 2018). For studies with a pediatric population, study populations contained between 29–48,762 subjects. Age ranges were 2–18 years old, and 80 %–100 % male. Adult studies contained 22–1507 subjects, with most tending towards less than 100 subjects. These studies had an age range between 18 to over 70, with most studies tending towards having a larger proportion of younger adults. Most study populations were predominately male (43 %–80 %). Almost all comparison groups were matched by age and sex. Most studies did not specifically identify race or ethnicity for study participants except for Croen et al. (2015), which had a 65.6 % White, non-Hispanic, 11 % Asian, 4 % Hispanic, and 7.6 % Black ASD population and Tolchard and Stuhlmiller (2018), which had a 67 % Aboriginal or Torres Strait Islander population. Dziobek et al. (2007) and Moses et al. (2014) explicitly state within their methods the exclusion of participants that use lipid lowering or glucose lowering medications. Study characteristics and demographic information can be found in Table 1.

Table 1.

Paper Characteristics.

Study	Country	Study Type	Sample Size (ASD/Control, N)	Sex (ASD/Control, % male)	Age (ASD/Control, mean or range years)	ASD measured by	Metabolic Dysfunction Components	Metabolic Dysfunction components measured by
Chen et al. (2016)	Taiwan	Prospective Cohort	6122	80 %	79 % 10–17 21 % 18–29	(ICD-9-CM) 299 between Jan 2002-Dec 2009	T2DM, HTN, dyslipidemia	(ICD-9-CM) 250.x0 and 250.x2, x = 0–9 recorded twice in medical record
			24488	sex matched	age matched
Croen et al. (2015)	United States	Observational	1507	73.1 %	52.4 % 18–24 38.1 % 25–48 9.5 % >49	(ICD-9-CM) 299.0, 299.8, 299.9 recorded twice in medical record	T2DM, HTN, dyslipidemia	Diabetes registry, ICD-9-CM, medication reports
			15070	sex matched	age matched
Dziobek et al. (2007)	United States	Observational	22	77.2 %	40.8 ± 10.8 44.6 ± 14.8	DSM-IV and ADI-R	T-chol, LDL, HDL, TG, FBG	Laboratory results§
			22	sex matched	44.6 ± 14.8
Flygare Wallén et al. (2018)	Sweden	Observational	13921	65.6 %	57 % (0–18) 43 % >18	ICD-10 (F84.0), (F84.5), (F84.1) (F84.9), (F84.3), (F84.8)	DMa and HTN	(ICD-10-CM) E10-E14, I10-I15
			1996140	49.2 %	21.4 % (0–18) 78.6 % >18
Kim et al. (2010)	Korea	Observational	29	100 %	10.1 ± 1.3	DSM-III and DSM-IV	FBG, T-chol, HDL, LDL, TG	Laboratory results
			29	sex matched	age matched
Moses et al. (2014)	Israel	Observational	80	70 %	31.1 ± 8.25	DSM-IV	FBG, T-chol, LDL, HDL, TG	Laboratory results§
			828	55 %	35.6 ± 8.59
Shedlock et al. (2016)	United States	Observational	48762	80 %	2–18	(ICD-9-CM) 299.90, 299.91, 299.80, 299.81, 299.00, 299.01	T2DM, HTN, hyperlipidemia	(ICD-9-CM) for T2DM and hyperlipidemia, HTN recorded twice in medical record
			243810	sex matched	age matched
Tolchard et al. (2018)	Australia	Observational	60	“more male”	0–100 “younger than control”	Not specified	DMb, HTN, FBG and “blood sugar level”	Not specified how DM and HTN were determined, laboratory results
			1504	43.1 %	N/A

ASD = autism spectrum disorder, HTN = hypertension, T-chol = total cholesterol, T2DM = type 2 diabetes mellitus, DM = diabetes mellitus, LDL = low density lipoprotein, HDL = high density lipoprotein, TG = triglycerides, FBG = fasting blood glucose, ICD = International Classification of Diseases, DSM = Diagnostic and Statistical Manual of Mental Disorders.

^§Studies explicitly stating exclusion of participants on relevant medications that would affect laboratory results (glucose, blood pressure, and lipid lowering medications).

^aAuthors of this paper have been contacted and believe that a majority of adult participants with DM have T2DM, however, the proportion is unknown for pediatric participants.

^bAuthors of this paper have been contacted and have confirmed that a large majority (98.7 %) of participants had T2DM.

3.2. Metabolic dysfunction findings

Studies found a relationship between ASD and metabolic dysfunction. There was higher incidence and higher odds for those with ASD of having T2DM/hyperglycemia (six out of eight studies), hypertension (four out of six studies), and/or dyslipidemia (four out of five studies) compared to the comparison group for both pediatric and adult populations, though results are heterogeneous for metabolic syndrome components. No studies evaluated the relationship between ASD and metabolic syndrome as a diagnosis or central obesity. Other factors investigated included age, sex, medical comorbidities, and use of antipsychotic medication; however, neither race nor ethnicity were investigated.

3.2.1. T2DM or hyperglycemia

Six out of eight studies investigating the outcome of T2DM or fasting blood glucose in ASD and comparison groups found significant differences that indicated higher fasting blood glucose and a higher prevalence, incidence, or odds of T2DM for the group with ASD. Each of these studies’ findings are summarized in Table 2.

Table 2.

Summary of T2DM/Hyperglycemia Findings by Study.

Study	Population	Metabolic Dysfunction Component	Main Analysis (ASD vs control)	Covariates used in Analysis	Secondary analyses
					Age	Sex	Atypical Antipsychotics	Other
Chen et al. (2016)	Pediatric and Adult	T2DM	Higher incidence* (2.63 vs 0.63 per 1,000 personyears)	Demographic data, atypical antipsychotics, medical comorbidities	Adolescent: 2.71 (1.64–4.48)	Both sexes with ASD have higher risk of T2DM compared to controls Comorbid obesity and dyslipidemia increase T2DM risk in males with ASD; HTN and dyslipidemia in females with ASD	Both short-term and longterm use increases T2DM risk Short term use: 1.97 (1.20–3.23) Long term use: 1.64 (1.02–2.63)	Comorbidities: HTN, obesity, dyslipidemia increase T2DM risk
			Shorter onset* (3.42 ± 2.11 vs 5.35 ± 2.31 years)	Hazard Ratio: 3.25 (95 % CI 2.23–4.75)	Young adult: 5.31 (2.85–9.90)	Comorbid obesity and dyslipidemia increase T2DM risk in males with ASD; HTN and dyslipidemia in females with ASD
Croen et al. (2015)	Adult	T2DM	Higher rate* (7.6 % vs 4.3 %)	Sex, age, race/ethnicity		Both sexes with ASD have increased odds of T2DM compared to control Females with ASD have increased odds of T2DM compared to males with ASD Male OR: 1.74 (1.21–2.50) Female OR: 3.74 (2.23–6.27)
			Higher odds* 2.18
Flygare Wallén et al. (2018)	Mixed	DM	Higher odds* Male: 1.705 (1.504–1.933) Female: 1.596 (1.329–1.916)	Age Sex		Both sexes with ASD have increased odds of T2DM compared to control		Comobidities: in ASD group, 3.6 % had all obesity, HTN, and DM, 5.9 % had obesity and HTN, 4.8 % had DM and HTN, and 3.2 % had obesity and DM
Shedlock et al. (2016)	Pediatric	T2DM	Higher odds* 2.68 (2.41–2.99)	No analysis				Medication Use: Those with ASD and T2DM have greater relative risk of being prescribed medication to treat conditions.
Moses et al. (2014)	Adult	FBG§	Lower* (77.8 ± 11.39 vs 90.1 ± 9.52)	Age, sex, BMI, with Bonferroni correction
Kim et al. (2010)	Pediatric	FBG	No significant difference (result N/A)	No analysis
Dziobek et al. (2007)	Adult	DM	No significant difference (79 ± 8.9 vs 79 ± 9.3)	No analysis
		DM	Higher odds* 3.34 (1.79–11.42)	No analysis
Tolchard et al. (2018)	Adult	FBG	Higher (6.34 ± 3.0 vs 6.14 ± 4.1)	No analysis
		“blood sugar level”	Higher (7.56 ± 3.2 vs 6.89 ± 3.8)

Means and standard deviations represented as X ± SD, odds ratios represented as OR (95 % Confidence interval) unless otherwise stated as 99 %.

All measurements of FBG are measured in units of mg/dL.

ASD = autism spectrum disorder, HTN = hypertension, T2DM = type 2 diabetes mellitus, DM = diabetes mellitus, FBG = fasting blood glucose, OR = odds ratio, CI = Confidence Interval, BMI = body mass index.

^§Exclusion of glucose lowering drugs explicitly stated.

^*Indicates significant finding (95 % CI or p ≥ 0.05).

Shedlock et al. (2016) found increased odds for T2DM in children. Similarly, Flygare Wallén et al. (2018) looked at diabetes mellitus (DM) and reported increased odds for a mixed age population with ASD as well. These relationships still persisted when accounting for possible confounding factors such as atypical antipsychotic use or BMI. Chen et al. (2016) found that a cohort of adolescents and young adults with ASD had increased incidence of T2DM at follow-up when adjusting for demographic variables, atypical antipsychotics use, and medical comorbidities. Finally, the duration of onset of T2DM was shorter for adults with ASD compared to the comparison group (Chen et al., 2016).

Tolchard and Stuhlmiller (2018) observed that adults with ASD exhibited higher fasting blood glucose levels and had higher odds for T2DM. In contrast, no significant difference in fasting blood glucose was found for children with ASD compared to the comparison group by Kim et al. (2010), and fasting blood glucose was found to be lower for adults with ASD compared to adults without ASD after controlling for age, sex, and BMI (Moses et al., 2014). Furthermore, Moses et al. (2014) state that treatment with antipsychotic medication does not alter fasting blood glucose significantly; however, it is unknown if this was analyzed with the inclusion of participants with other forms of intellectual disability. Dziobek et al. (2007) also did not find a significant difference between those with Asperger’s disorder (a condition now included within the definition of ASD) and the comparison group. Because their primary focus was on dyslipidemia, Dziobek et al. (2007) did not pursue further analyses of fasting blood glucose adjusting for covariates.

Factors that may moderate or mediate the relationship between ASD and T2DM include age, atypical antipsychotic use, medical comorbidities, and sex. Atypical antipsychotic use was found to be associated with a higher risk of developing T2DM for both short-term and long-term users (Chen et al., 2016). However, Moses et al. (2014) did not see a change in fasting blood glucose associated with antipsychotics. Medical comorbidities were also found to increase risk of T2DM as well, increasing the likelihood of having a diagnosis of hypertension, dyslipidemia, and/or obesity (Chen et al., 2016). When looking at age effects, Chen et al. (2016) found that young adults with ASD between the ages of 18–29 exhibited higher risk for T2DM compared to adolescents with ASD. For sex, results were heterogeneous. Chen et al. (2016), Croen et al. (2015), and Flygare Wallén et al. (2018) found that both males and females with ASD were associated with a higher risk of T2DM. Croen et al. (2015) found that females with ASD have statistically significantly higher rates compared to males, but Chen et al. (2016) did not find evidence that risk was independently associated with sex. However, they found that obesity and dyslipidemia predicted subsequent T2DM in males, whereas hypertension and dyslipidemia predicted subsequent T2DM in females, and stated this could be an indicator of differing pathophysiology due to an unknown sex effect (Chen et al., 2016).

3.2.2. Hypertension

Four out of six studies found associations between ASD and hypertension, summarized in Table 3. Croen et al. (2015) reported significantly higher rates of hypertension in adults who had ASD compared to adults without ASD when adjusting for age, sex, and race/ethnicity. Similarly, other groups also found that children and adults with ASD had higher odds of hypertension compared to comparison groups (Flygare Wallén et al., 2018; Shedlock et al., 2016; Tolchard & Stuhlmiller, 2018). Flygare Wallén et al. (2018) adjusted for age and sex.

Table 3.

Summary of Hypertension Findings by Study.

Study	Population	Metabolic Dysfunction Component	Main Analysis (ASD vs control)	Covariates used in Analysis	Secondary analyses
					Age	Sex	Atypical Antipsychotics	Other
Croen et al. (2015)	Adult	HTN	Higher rate* (25.6 % vs 15.6 %)	Sex, age, race/ethnicity		Both sexes have increased odds of HTN compared to control
						Females with ASD have similar odds of HTN compared to males with ASD Male OR: 2.19 (1.75–2.76) Female OR: 2.19 (1.56–3.08)
Flygare Wallén et al. (2018)	Mixed	HTN	Higher odds, males only* Male: 1.175 (1.050–1.315) Female: 0.880 (0.756–1.024)	Age Sex		Only males with ASD are have increased risk of HTN compared to controls		Comobidities: in ASD group, 3.6 % had all obesity, HTN, and DM, 5.9 % had obesity and HTN, 4.8 % had DM and HTN, and 3.2 % had obesity and DM
Shedlock et al. (2016)	Pediatric	HTN	Higher odds* 2.04 (1.84–2.27)	No covariates				Medication Use: Those with ASD and HTN have greater relative risk of being prescribed medication to treat conditions.
Tolchard et al. (2018)	Adult	HTN	Higher odds* 6.27 (1.99–19.18)	No covariates
Chen et al. (2016)	Pediatric and Adult	HTN	No significant difference (1 % vs 1 %)	No covariates
Dziobek et al. (2007)	Adult	Systolic and diastolic blood pressure	No significant difference (Systolic 124 ± 14 vs 119 ± 16, diastolic 76 ± 10 vs 74 ± 10)	No covariates

Means and standard deviations represented as X ± SD, odds ratios represented as OR (95 % Confidence interval) unless otherwise stated as 99 %. ASD = autism spectrum disorder, HTN = hypertension, DM = diabetes mellitus, OR = odds ratio, CI = Confidence Interval.

^*Indicates significant finding (95 % CI orp ≥ 0.05).

In contrast, Chen et al. (2016) found similar rates of hypertension (1 %) in both the ASD and control group, and Dziobek et al. (2007) observed no significant difference between those who had Asperger’s and those without when looking at systolic and diastolic blood pressure. However, these analyses did not test for covariables or control for any other factors.

Sex is a well-known moderator for hypertension; however, results were heterogeneous. Croen et al. (2015) found that females with ASD had higher rates than males with ASD even though both exhibited risk, but this contrasts directly with Flygare Wallén et al. (2018), who found that only males with ASD had increased odds of hypertension. No analyses looked at antipsychotic use in the context of hypertension in those with ASD.

3.2.3. Dyslipidemia

Four out of five studies found an association between ASD and altered lipid profiles, summarized in Table 4. Dziobek et al. (2007) found elevated levels of total cholesterol, low-density lipoprotein (LDL), and triglycerides (TG) in adults with Asperger’s compared to the comparison group. They controlled for physical activity, which resulted in high-density lipoprotein (HDL) and triglycerides no longer being significantly different between those with Asperger’s and the comparison group; however, total cholesterol and LDL remained significantly different (Dziobek et al., 2007). They also investigated whether social anxiety, and obsessive-compulsive behavior could account for these findings, but group differences for total cholesterol and LDL still remained significant between Asperger’s group and controls (Dziobek et al., 2007). Finally, they found that diet, antipsychotic medication, and smoking were not confounding influences (Dziobek et al., 2007). Kim et al. (2010) found the opposite, male children with ASD did not exhibit higher fasting plasma total cholesterol or LDL compared to the comparison group, but did exhibit higher TG, LDL/HDL ratio, and lower HDL levels compared to the comparison group. They also found an interaction between autism and triglyceride levels that predicted HDL levels. In a model, this interaction along with LDL/HDL ratio, triglycerides, and autism explained 53 % variation of HDL (Kim et al., 2010). Authors interpret the interaction as there being a bigger difference in plasma HDL between those with ASD and the comparison group when TG levels are low compared to at higher levels. In other words, when TG is low, those with ASD have even lower HDL than the comparison group than when TG is high, indicating dyslipidemia (Kim et al., 2010). It is unknown if Kim et al. excluded or controlled for use of lipid-lowering medication, however. Other studies had similar observations, finding a significantly higher rate of dyslipidemia in adults with ASD (Chen et al., 2016; Croen et al., 2015), and higher odds of hyperlipidemia in children with ASD compared to respective comparison groups (Shedlock et al., 2016).

Table 4.

Summary of Dyslipidemia Findings by Study.

Study	Population	Metabolic Dysfunction Component	Main Analysis (ASD vs control)	Covariates	Secondary Analyses
					Age	Sex	Atypical Antipsychotics	Other
Croen et al. (2015)	Adult	Dyslipidemia	Higher rate* (22.8 % vs 15.1 %)	Sex, age,race/ethnicity		Both sexes with ASD have increased odds of dyslipidemia compared to control
			Higher odds* 2.12 (1.74–2.60)			Females with ASD have increased odds of dyslipidemia compared to males with ASD Male OR: 1.89 (1.49–2.40) Female OR: 2.87 (1.97–4.19)
Dziobek et al. (2007)	Adult	T-chol§	Higher* (195.0 ± 31.0 vs 171.8 ± 24.8)	Social anxiety, obsessive-compulsive behavior, physical activity, diet, antipsychotic medication, and smoking				Physical Activity: group differences remained significant for TC, LDL, but not for HDL or TG
		HDL	Lower* (45.5 ± 10.9 vs 53.6 ± 13.4)
		LDL	Higher* (124.3 ± 27.9 vs 101.9 ± 22.8)
		TG	Higher* (126.2 ± 66.8 vs 81.5 ± 38.2)
Kim et al. (2010)	Pediatric	T-chol	Lower (170.5 ± 32.3 vs 179.6 ± 27.3)	BMI, age, socioeconomic status				HDL related: Interaction between TG and autism, plus variables LDL/HDL ratio significantly predict HDL. When TG is low, bigger difference in HDL in ASD vs control compared to when TG is high
		HDL	Lower* (48.8 ± 11.9 vs 60.5 ± 10.9
		LDL	Lower (101.2 ± 29.1 vs 101.2 ± 29.1)
		TG	Higher* 102.4 ± 52.4 vs 70.6 ± 36.3
Moses et al. (2014)	Adult	T-chol§	Lower (168.3 ± 32.78 vs 185.4 ± 40.49)	Age, sex, BMI, with Bonferroni correction After controlling, no significance in for all lipid profile measurements			Medication type has no effect on lipid profile
		HDL	Higher (50.3 ± 13.89 vs 48.1 ± 13.19)
		LDL	Lower (97.9 ± 24.96 vs 112.5 ± 32.92)
		TG	Lower (98.0 ± 48.92 vs 129.4 ± 72.29)
Shedlock et al. (2016)	Pediatric	Hyperlipidemia	Higher odds* 2.01 (1.90–2.13)	No covariates				Those with ASD and hyperlipidemia have greater relative risk of being prescribed medication to treat conditions.
Chen et al. (2016)	Pediatric and Adult	Dyslipidemia	Higher rate* (2.0 % vs. 1.6 %)	No covariates

Means and standard deviations represented as X ± SD, odds ratios represented as OR (95 % Confidence interval) unless otherwise stated as 99 %.

All measurements of T-chol, LDL, HDL, TG are measured in units of mg/dL.

ASD = autism spectrum disorder, T-chol = total cholesterol, LDL = low density lipoprotein, HDL = high density lipoprotein, TG = triglycerides, OR = odds ratio, CI = Confidence Interval BMI = body mass index.

^§Exclusion of lipid lowering drugs explicitly stated.

^*Indicates significant finding (95 % CI orp ≥ 0.05).

One study with contrasting results indicated lower levels of total cholesterol in adults with ASD compared to the comparison group, but differences no longer exist after covariates were analyzed (Moses et al., 2014). The authors claim that this result, in addition to analysis on medication use, indicates no evidence of increased total cholesterol after antipsychotic use, though formal results are not shown, and it is unknown if this analysis includes participants with other intellectual disabilities.

Only sex was observed to have a differential effect for dyslipidemia. Croen et al. (2015) found that females with ASD had higher odds of having dyslipidemia compared to males with ASD, though both sexes had increased odds compared to the comparison group. Antipsychotic use and age were not investigated in the context of dyslipidemia in those with ASD.

3.2.4. Central obesity

No studies investigated the relationship between ASD and waist circumference, waist-to-hip ratio, or any other central obesity measurements.

4. Discussion

This review found preliminary evidence for increased prevalence of metabolic dysfunction associated with ASD, specifically for hyperglycemia, hypertension, and dyslipidemia, in pediatric and adult populations, though results were heterogeneous, indicating need for standardization of study measures and further investigation. The review found evidence for increased prevalence of T2DM associated with ASD as well. Because none of the included papers explicitly discussed metabolic syndrome itself–only individual components–it is unknown whether ASD is associated with increased prevalence of concurrent instances of metabolic dysfunction that define metabolic syndrome. No research was found investigating central obesity in these populations. Potential moderating factors included age, sex, atypical antipsychotic usage, and comorbid medical illnesses, however, studies varied on inclusion of these cofactors within analysis. Furthermore, there is lack of research investigating the effects of race and ethnicity on these relationships.

The evidence for a relationship between ASD and T2DM and hyperglycemia is relatively concordant, especially when looking at a diagnosis of T2DM. This relationship was observed for both pediatric and adult populations, warranting a need for earlier detection and management in clinic to prevent some of the more severe symptoms that arise and allow for better care coordination. While there was some heterogeneous evidence when analyzing random blood glucose levels, using such measurements may not be as accurate an indicator of hyperglycemia or T2DM because it may not be true fasting blood glucose or there may be participant variability based on time of last meal. Future studies investigating this relationship should aim to use strict criteria in defining T2DM, such as using HbA1c or glucose tolerance tests rather than self-report (WHO, 2006). Studies also varied in using covariates, which may be an important consideration in future analyses because of the multifactorial characteristics of T2DM.

Currently, there are many specific factors that need further investigation, such as age, race, sex, atypical antipsychotic medication, and other medical comorbidities. Some of these have already been investigated in the context of moderation in non-ASD populations (Freyberg et al., 2017; Kautzky-Willer et al., 2016). For example, it is known that there are sex differences in risk for T2DM in the general population (Meisinger et al., 2002). The studies in this review had differing results, finding that either females were at higher risk compared to males, both sexes were at similar high risk, or sexes had different risk factors that predicted subsequent development of T2DM. Thus, further investigating the effect of sex in the pathogenesis of T2DM in ASD populations may be of interest.

It may be informative to additionally compare prevalence and onset data using study populations more similar in age, sex, ethnicity, race, medication usage, or pathology so that mechanisms for similarities or differences can be narrowed down. Stricter age, sex, ethnicity and race matching may determine whether these associations between ASD and metabolic dysfunction are independent or not. Because rates of T2DM are increasing in general, there may be other confounding factors at play that can be considered when observing prevalence, which could be informative to studies evaluating mechanisms (Thibault et al., 2016).

These recommendations may be extended to research into dyslipidemia and hypertension, for which only a few studies have investigated. Because studies used variable measures to define dyslipidemia or hypertension, using stricter definitions reduces variability based on single visit metrics, especially when looking at in-office blood pressure, which may not be as accurate an indicator of hypertension (Pioli et al., 2018). Using 24-hr blood pressure monitoring in both adults and children may be more informative than isolated measurements to define a disease state (Flynn et al., 2014; Hermida et al., 2015). These markers may be harder to evaluate in pediatric patients due to compliance, hence the lack of studies done so far; however, research on younger adult populations could potentially identify preliminary markers for earlier intervention. Other challenges may include sensory sensitivity, language impairment, and hyperactivity or impulsivity that are experienced by many individuals with ASD as well. Investigation of dyslipidemia and hypertension must also include incorporation of important factors such as age, sex, race, and use of antipsychotic medication.

Atypical antipsychotic medications are an important factor to study because they are a useful short-term indication to treat irritability for those with ASD (Posey et al., 2008). Because people with ASD are frequently prescribed atypical antipsychotics, this may be one mechanism underlying metabolic dysfunction in those with ASD. Metabolic dysfunction has been known to affect many other populations that are prescribed antipsychotic medications (Freyberg et al., 2017; Pramyothin & Khaodhiar, 2015). However, there is inconclusive research found by this review investigating atypical antipsychotic use influencing or modifying risk for metabolic dysfunction in populations with ASD, especially when looking at biomarkers such as cholesterol, triglycerides, or blood pressure. Because there is evidence that second generation antipsychotics may lead to metabolic alterations or may have some cardiovascular toxicity (Ijaz et al., 2018; Rojo et al., 2015), further research is warranted to confirm whether this mechanism is found in those with ASD taking these medications. Metabolic monitoring occurs in adults and children prescribed antipsychotics (Tso et al., 2017), recommending that prescribers screen patients at baseline, 12 weeks post-initiation of medication, and annually thereafter (Consensus Development Conference on Antipsychotic Drugs and Obesity and Diabetes, 2004), but there may be gaps in offering appropriate interventions. In addition, rates of pre-initiation and post-initiation monitoring are often low globally (approximately 50 % or less) (Mitchell et al., 2012), even after guidelines have been put in place, especially in children (<20 %) (Connolly et al., 2015; Coughlin et al., 2018). Some evidence indicates that rates of metabolic monitoring following antipsychotic medication initiation can be optimized in children and adults with ASD through the use of an integrated care model (Ruiz et al., 2016). Currently there is no standard preventative strategy implemented to lower the risk of metabolic dysfunction and associated comorbidities observed in people with ASD taking atypical antipsychotic medications; in the future, closer monitoring is needed to deter further metabolic dysfunction in this population.

The studies in this research area are primarily focused on adults, despite the fact that a majority of research in ASD is focused on pediatric populations (Matson & LoVullo, 2009). Significant research on the long-term effects of metabolic dysfunction may be limited due to the small quantity of literature on adult ASD populations in general. Looking at metabolic dysfunction in adults with ASD may be more common because the diagnostic criteria for metabolic syndrome are relatively more established compared to in children. Research in children may also be limited because metabolic symptoms may not present until adulthood. However, a later onset of disease may increase the potential for prevention strategies to be implemented earlier in children with ASD who are at risk. Early intervention, such as monitoring diet, sleep, and physical activity may help prevent the development of T2DM commonly seen in the adult ASD population. Some therapeutics such as metformin may be indicated, as there is evidence that it can prevent or minimize weight gain, especially with medications like antipsychotics, however potential risks associated with medication use must be evaluated, such as irritability, nausea, and gastrointestinal upset (Anagnostou et al., 2016; de Silva et al., 2016). With our current research base however, it is harder to make any conclusion on risk experienced by children compared to adults for metabolic dysfunction.

It is established that individuals with ASD are at higher risk for obesity, and BMI should continue to be monitored for optimal care (Zheng et al., 2017). However, the relationship between ASD and central obesity in adults is unknown. For the purpose of our review, we excluded measures of obesity that were not focused on central obesity, which is more specific and strongly associated with risk for many chronic diseases in addition to being a component of metabolic syndrome (Lee et al., 2008). Obesity in children with ASD has also been studied and summarized (Kahathuduwa et al., 2019), indicating that children with ASD have an increased risk of having a higher BMI that is considered overweight or obese. There is no research investigating central obesity in children with ASD, and no consensus yet for measuring central obesity in regard to metabolic syndrome in pediatric populations. Thus, the lack of data on central obesity measures in both adult and pediatric populations must be considered in future research, especially when considering the ease of collecting related metrics such as waist circumference (Appel et al., 2004).

Limitations to the conclusions within this review include varied determinants of ASD diagnosis, as well as varied measurement and determinants of metabolic dysfunction components or T2DM. Thus, it is hard to compare the effect sizes of each study, and difficult to determine whether contrasting results are due to methodology or true heterogeneity. Evidence for mechanisms and covariate factors was not a focus in any paper for a metabolic component, nor did all analyses check for confounding or interaction, however, this may have been due to some studies having limited statistical power from a small sample size. There is lack of prospective analysis, and observational analyses are unable to give information that is helpful for evaluating the relationship past an increase in risk. There is also no research on central obesity, which is especially important data to obtain as collection can be incorporated into routine medical check-up. Association between ASD and some measurements (such as blood pressure or fasting blood glucose) may be blunted or reversed due to the patient controlling it through medication. Metabolic syndrome as a diagnosis is hard to assess due to the lack of consensus for a single definition. Race and ethnicity are known risk factors for metabolic dysfunction; however, no papers investigated either as a risk factor (Gurka et al., 2014).

5. Conclusion

Current research has demonstrated that those with ASD have increased rates of metabolic dysfunction, including T2DM, hyperglycemia, hypertension, and dyslipidemia. More research, especially on central obesity and metabolic syndrome as a diagnosis, is warranted to investigate metabolic dysfunction in people with ASD in order to guide prevention, monitoring, diagnosis, and treatment. Future work may further investigate factors such as age, sex, race/ethnicity, and antipsychotic medication in their analyses. Other biomarkers, measurements, mechanisms, and factors that may influence these relationships may also be a target of study to gain a more comprehensive understanding and build upon the research that has been conducted thus far. With the growing prevalence of metabolic dysfunction and the large burden these illnesses place on individuals and society, these findings highlight the importance of addressing risk in people with ASD who already face many comorbidities to lessen the burden. This will benefit the immediate goal of identifying these metabolic comorbidities in this specific population and the long-term goal of preventing progression of these comorbidities in the hopes of providing optimal patient care.