Pica, Autism, and Other Disabilities

The SEED methodology has been previously described.²⁴  Briefly, SEED data were collected in 2 phases (2007–2012 and 2013–2016) from 6 US study sites (located in CA, CO, GA, MD, NC, and PA). At each site, children with ASD and other DDs were recruited from multiple special education and clinical sources that provide services to children with disabilities. Children recruited from each source were those with select special education or International Classification of Disease codes indicative of ASD or other DDs often seen as precursors or cooccurring diagnoses in children eventually diagnosed with ASD. POP group children were recruited on the basis of randomly selected birth records; they were sampled from the same birth years and study sites but were not otherwise matched to the case groups. This study was approved by Institutional Review Boards at the Centers for Disease Control and Prevention and each site.

Children eligible for SEED enrollment were those who lived in the respective site’s study area both at birth and at enrollment and lived with a caregiver since at least 6 months of age who could provide legal consent and communicate in English (all sites) or Spanish (2 sites). The child was required to live with the biological mother in the second phase of SEED data collection but not the first; nonetheless, the respondent for 98% of children in the first phase was the biological mother. Therefore, in this report, we refer to the respondent as the mother. Children were 30 to 68 months of age (median = 55.4 months) at the time of their SEED developmental assessment.

This analysis included children from all 3 study groups (ASD, DD, and POP) who were not missing data on pica. Although children were initially identified as eligible for 1 of the 3 study groups, final study classification was determined by standardized research developmental assessment results.²⁵  After enrollment, mothers of all children were administered the Social Communication Questionnaire (SCQ)²⁶  to screen for ASD symptoms in their child. Children with scores ≥11 were considered potential ASD cases regardless of their initial classification. Additionally, children who had a previous ASD diagnosis or special education classification were considered potential ASD cases regardless of their SCQ scores. All children were administered an in-person general developmental assessment, the Mullen Scales of Early Learning.²⁷  Children in the potential ASD group were additionally administered the Autism Diagnostic Observation Schedule,²⁸  and their caregivers were administered the Autism Diagnostic Interview-Revised.²⁹  Children who met the study criteria for ASD on the basis of their Autism Diagnostic Observation Schedule and Autism Diagnostic Interview-Revised assessments were classified as ASD cases²⁵ ; children who had been classified as potential ASD cases on the basis of symptoms or past diagnoses but did not meet the study ASD case criteria were classified as DD with ASD characteristics. Children selected for the DD group who had an SCQ score <11 and no previous ASD educational classification or diagnosis were classified as DD without ASD characteristics. A small number of children recruited to the POP group screened positive on the SCQ and met study criteria for ASD; they received a final ASD study classification.

We further divided children with ASD, DD with ASD characteristics, and DD without ASD characteristics according to whether they had cooccurring ID. ID was defined as a Mullen Scales of Early Learning developmental score (which is considered comparable to IQ) <70.²⁷  Thus, we examined 6 subgroups of children: (1) ASD with ID, (2) ASD without ID, (3) DD with ID and ASD characteristics, (4) DD with ID but without ASD characteristics, (5) DD with ASD characteristics but without ID, and (6) DD with neither ID nor ASD characteristics.

Mothers completed structured telephone interviews and self-administered forms. Data collected included sociodemographic characteristics and child health conditions and behaviors. The Child Behavior Checklist,³⁰  a standardized self-administered form, asked respondents to indicate whether their child had exhibited problem behaviors within the previous 2 months. Response options for each item were “not true,” “somewhat or sometimes true,” and “very true or often true.” Pica was ascertained via the item, “child eats or drinks things that are not food–not including sweets.” If this item was answered affirmatively, the respondent was asked to describe the consumed substance. We considered a child to have pica if the mother responded, “somewhat or sometimes true” or “very true or often true.” We reviewed free text responses on item(s) the child had eaten to assess the accuracy of responses to the pica question. Although free text data were collected in both phases of SEED, only data from the second phase of SEED had been processed and made available for this assessment; nonetheless, we have no reason to believe responses differed between the 2 study phases. Our review indicated the Child Behavior Checklist pica question was accurately capturing pica; >95% of respondents specified their child was consuming 1 or more nonfood items, only 2% incorrectly specified a food item, and a few responses were too vague to classify. Examples of nonfood items reported are presented in Supplemental Table 3.

We compared distributions of sociodemographic factors according to whether the child had pica and by main study group (ASD versus POP and DD versus POP) using χ² tests. We assessed the prevalence of children who engaged in pica by main study group and for each of the aforementioned subgroups. We calculated unadjusted prevalence ratios (PRs) and adjusted prevalence ratios (aPRs) and 95% confidence intervals, in which the prevalence of pica for each ASD and DD case group and subgroup was compared with the prevalence for the POP group. aPRs were calculated by using modified Poisson regression³¹ ; in addition to ASD or DD cases status, models included child sex, child and maternal age at enrollment, maternal race or ethnicity, maternal education, and household income.

We conducted supplemental analyses to examine (1) whether associations observed in our main analysis varied within various demographic strata (same factors as included in adjustment models) and (2) whether associations were similar using a more restrictive pica definition (by which only “very true or often true” counted as pica). For stratified analyses, we calculated stratum-specific PRs and used the Mantel-Haenszel method³²  to estimate pooled aPRs.

Overall, 4739 children were included in this analysis: 1426 with ASD, 1735 with other DDs, and 1578 POPs (Fig 1).

In all 3 study groups, most mothers were >35 years of age, were non-Hispanic white (NHW), had a bachelor’s degree or higher, and had a household income >$50 000 (Table 1). Nonetheless, we observed study group differences; mothers of children with ASD and DD were significantly less likely to be NHW, have a bachelor’s degree, or have a high household income than mothers of POP group children. There were markedly more males in the ASD and DD groups than the POP group. Mothers of children with pica were significantly younger and less likely to be NHW, have a bachelor’s degree, or have a high household income than mothers of children without pica. Children with pica were more likely to be male. More than 70% of children in all study and pica groups were 48 to 68 months of age with slight differences observed between study groups.

Pica was reported in 23.2%, 8.4%, and 3.5% of children in the ASD, DD, and POP groups, respectively (Fig 2, Table 2). Within the ASD group, pica was reported in 28.1% of children with ID and 14.0% of children without ID. Within the DD group, pica was reported in 26.3% of children with both ID and some ASD characteristics, 12.0% of children with ASD characteristics but without ID, 9.7% of children with ID but without ASD characteristics, and 3.2% of children with neither ID nor ASD characteristics.

Unadjusted PRs indicated associations between pica and all ASD groups and subgroups and between pica and DD overall and 3 of the 4 DD subgroups. There was no association between pica and DD with neither ID nor ASD characteristics (Table 2). Despite some variability due to small sample sizes, the findings from stratified analyses were generally comparable to the unadjusted analyses (Supplemental Tables 4 and 5).

Multivariable aPRs were also similar to the unadjusted PRs (Table 2). Pica remained strongly associated with ASD overall, ASD with ID, ASD without ID, and DD with both ID and some ASD characteristics (aPRs 4.4–8.0). Pica was moderately associated with DD overall, DD with ASD characteristics without ID, and DD with ID but not ASD characteristics (aPRs 1.9–2.5). Pica was not associated with DD with neither ID nor ASD characteristics.

The differences between ASD and DD case groups and subgroups and the POP group were even more pronounced when we used a more restrictive pica classification, (ie limited to a maternal response of “very true or often true” to the pica question) (Supplemental Table 6).

Our findings suggest that pica is more common in preschool-aged children with ASD, whether or not ID was present, and with other types of DD in which the child had some ASD characteristics, ID, or both. In comparison with the POP group, pica was strongly associated with ASD and with other DDs characterized by both ASD symptoms and ID and moderately associated with DD with ASD characteristics only or ID only. Notably, pica was not elevated in children with DDs that did not include ASD characteristics or ID. Thus, our findings suggest that not all children with DDs are at increased risk for pica. Using a more conservative definition of pica did not change the relative findings.

Our pica prevalence estimate in children with ASD is higher than the prevalence reported by Emond et al²²  in their cohort study of children with ASD (12.3% and 12.5%), and markedly higher than the prevalence reported by Neumeyer et al²³  in their study of children in an autism treatment network (3.0% and 2.4%). Likewise, although our estimates fall within the range of estimates from studies presented by Matson et al¹  in their review of pica in children or adults with ASD and/or ID (4% to 26%) it is notably higher than several estimates from that review. Our findings expand on these reports in several ways. Whereas many previous studies were limited by nonstandardized classification of ASD or ID, we collected objective developmental data as part of the SEED protocol to systematically classify ASD and ID. Most past studies examined a limited subset of pica, such as pica that came to medical attention and/or required treatment; in contrast, with SEED we examined pica behavior ascertained from a standardized developmental questionnaire³⁰  regardless of whether the child had developed adverse consequences from their pica. Another strength of this study in comparison with others is the inclusion of a general population comparison group, which was actively recruited from the same study areas as the ASD and DD case groups. Finally, SEED’s large sample size and detailed developmental data collection allowed us to obtain stable estimates of pica among children in various disability subgroups and thus hone in more clearly than past studies on types of disability that were and were not associated with pica.

Nonetheless, we were not able to explore associations between pica and specific DDs. The SEED DD group, identified from diagnostic and special education codes, is heterogeneous, including children with many different (and sometimes multiple) disabilities. We were unable to assess specific non-ASD DD diagnoses separately because of sample size limitations and complexities in categorizing the various diagnostic and education disability codes into meaningful subgroups. However, our subdivisions of the DD group according to the presence of ASD characteristics objectively measured on a standardized instrument and on whether the child’s developmental test scores were in the ID range allowed us to evaluate important subgroups of children with disability defined by rigorous standardized assessments.

Given the young age of the participants in this study sample, it is possible that pica prevalence in this study will not reflect pica prevalence at older ages. Nonetheless, in stratified analyses, we assessed whether the associations varied within the narrow range of ages included in our study and found that all associations observed in the main analysis were observed among the oldest stratum of children, those 48 to 68 months of age. Although we cannot assess whether affected children will continue to manifest pica as they age, our data nonetheless indicate it is a serious health concern in preschool-aged children. Notably, all children in this study were much older than 18 months, the age at which ingestion of foreign objects is deemed inappropriate.³³ 

Although we were unable to examine underlying reasons for the high pica prevalence we observed in children with ASD, ASD characteristics, and ID, Mayes and Zickgraf³⁴  found that pica in children with autism was one component of a larger pattern of atypical eating behaviors, such as limited food preferences and hypersensitivity to food textures. Furthermore, sensory processing difficulties are commonly reported in children with ASD and may result in both atypical eating and pica behavior.³⁵  Additionally, some children with ASD and/or ID may not understand the difference between edible and inedible objects.^33,35  Treatments for pica that have empirical support include applied behavior analysis¹  and functional analysis, which helps identify triggers for pica and determine if pica is secondary to sensory-seeking behavior, automatic reinforcement, or social factors such as seeking attention from caregivers.^7,33  Parent prevention measures include closely monitoring children, keeping items out of reach, using childproof locks, finding activities that occupy children’s attention, and informing other caregivers of concerns.³³ 

Despite our study’s strengths, we also note several limitations. SEED developmental data are cross-sectional, limiting causal inference. Although we defined pica on the basis of a standardized developmental assessment questionnaire,³⁰  pica was a single behavioral item on this instrument; while our internal assessment of available text responses about nonfood items was reassuring about the validity of the pica responses, we lacked data on clinical diagnoses and sequelae of pica and the typical quantity of nonfood items ingested. We also lacked data on some established biological risk factors for pica such as nutritional deficiencies and lead toxicity. It is possible that ascertainment of pica varied for cases and controls because caregivers might more closely monitor children with ASD and ID than children without these conditions. However, the impact of differential monitoring on pica prevalence estimates could have been in either direction; if children with ASD or ID were more closely monitored, caretakers had not only increased opportunity to observe pica, but also increased opportunity to prevent pica. Moreover, the difference between the level of observation for case and control children was likely not large, given that all children were young and thus likely to have been monitored fairly closely. Because SEED sought to enroll a diverse population of children recruited from a wide range of education and clinical sources, numerous families targeted for potential recruitment could not be located or contacted; however, a previous study in one SEED site indicated that a large proportion of these families were likely no longer eligible for SEED because they moved out of the study area; additionally, that study documented that although maternal age and education were associated with differential participation between ASD cases and POP controls, several other demographic and health factors were not associated with participation, and findings from most SEED analyses were likely robust.³⁶  We adjusted for both maternal age and education in this analysis. This study takes place in 6 different population-based US sites; however, results may not be generalizable throughout the United States.

We found that preschool-aged children with ASD, ASD characteristics, and ID more commonly engage in pica and therefore could benefit from close monitoring, strategies to restrict access to dangerous items and possibly behavioral therapy.^1,5,7,37  These findings can help promote awareness about pica among pediatricians and the disability community. Given the current paucity of research, more studies of the prevalence, correlates, and clinical outcomes of pica in children with ASD and other DDs are needed.

We thank the SEED Data Coordinating Center team at the Clinical and Translational Sciences Institute of Michigan State University for their support throughout this study.

aPR

adjusted prevalence ratio

ASD

autism spectrum disorder

DD

developmental disability

ID

intellectual disability

NHW

non-Hispanic white

POP

general population-based control

PR

prevalence ratio

SCQ

Social Communication Questionnaire

SEED

Study to Explore Early Development