Antecedents of screening positive for attention deficit hyperactivity disorder in 10-year old children born extremely preterm

Abstract

Background

Attention deficit hyperactivity disorder (ADHD) incidence is higher among children born very preterm than among children who are mature at birth.

Methods

We studied 583 ten-year old children born before 28 weeks of gestation whose IQ was above 84 and had a parent-completed Child Symptom Inventory-4 (CSI-4), which allowed classifying the child as having or not having symptoms of ADHD. For 422 we also had a teacher report, and for 583 we also had a parent report of whether or not a physician made an ADHD diagnosis.

Results

The risk profile of screening positive for ADHD based on a parent’s report (P) differed from the risk profile based on the teacher’s (T) report, while the risk profile according to a physician (MD), and according to any 2 observers (2) closely resembled the parent-report profile. Among the statistically significant risk factors were young maternal age (P, MD, 2), maternal obesity (P, MD, 2), maternal smoking (P, MD, 2), magnesium given at delivery for seizure prophylaxis (P, 2), recovery of Mycoplasma sp from the placenta (T, 2), low gestational age (P, 2), low birth weight (T, MD), singleton (P, MD, 2), male (P,T, MD, 2), mechanical ventilation on postnatal day 7 (MD), receipt of a sedative (P, 2), retinopathy of prematurity (P), necrotizing enterocolitis (MD), antibiotic receipt (MD, 2), and ventriculomegaly on brain scan (P, 2).

Conclusion

The multiplicity of risk factors identified can be subsumed as components of four broad themes, low socioeconomic state, immaturity/vulnerability, inflammation, and epigenetic phenomena.

Introduction

The lower the gestational age, the higher the rate of attention deficit (hyperactive) disorder (ADHD) [1]. Nevertheless, we are not aware of any study of the epidemiology of ADHD among individuals born extremely prematurely. The ELGAN (Extremely Low Gestational Age Newborn) Study of children born before 28 weeks of gestation provided an opportunity to identify risk factors for ADHD in this high-risk group. When this cohort was evaluated at 10 years of age, parents and teachers completed a Child Symptom Inventory-4 (CSI-4), which provided information about symptoms of ADHD based on DSM-5 criteria. Parents also reported whether a physician had made a diagnosis of ADHD, and whether medication had been prescribed.

Here we explore relationships between prospectively collected information about potential antecedents and a parent’s, a teacher’s, and a physician’s perception of whether or not the child had ADHD symptoms and/or an ADHD diagnosis. We also explore to what extent the antecedent risk factors for ADHD behaviors differed among three sets of observers, (i.e., parents, teachers, and physicians) and combinations of these observers.

Methods

Participants

The enrollment and consent procedures for all aspects of this multi-center prospective, observational study of the risk of structural and functional neurologic disorders in extremely preterm infants [2] were approved by the institutional review boards of all participating institutions. At age 10-years, 889 children had a neurocognitive assessment [3], while a parent completed questionnaires regarding the child’s medical and neurological status and behavior, including the Child Symptom Inventory-4 Parent Checklist (CSI-4)[4]. Parents also completed a questionnaire that asked the following two questions. Since age 2 years, have you been told by a doctor or other health professional that your child has or has had attention deficit or hyperactivity (ADD or ADHD)? Does your child currently take or has your child taken since age 2 years any daily medications for attention deficit or hyperactivity (ADD or ADHD)?

Of the children assessed, 583 had a DAS IQ ≥ 85 and a parent report about ADHD. They constitute the sample for this report. Of these children, 422 also had a teacher report. Additional details about data collected for this report can be found in the Supplementary Material.

Data analysis

We evaluated the null hypothesis that ELGANs who screen positive for a diagnosis of ADHD have the same demographic, pregnancy and perinatal characteristics as their peers who screen negative. We limited our attention to the entity of ANY ADHD, and not individual subtypes (i.e., inattentive, hyperactive-impulsive, and combined forms) for three reasons. First, we did not have information from the physician about diagnostic details, only that s/he did or did not make a diagnosis of ADHD. Second, the relatively small numbers of children in each subgroup (inattentive, hyperactive-impulsive, and combined), appreciably limited power and did not allow robust analyses. Third, we are supported by a literature emphasizing a unitary concept of ADHD [5], and factor analyses that have a heavy “common factor” loading [6, 7].

Because including children with a low IQ [8] might lead us to identify the antecedents of low IQ as well as the antecedents of ADHD, we limited our analyses to children whose IQ was 85 or above. We began with univariable analyses (Supplementary Material Tables), and then selected likely candidates for multivariable analyses. Because postnatal phenomena can be influenced by antepartum phenomena, we created multinomial logistic regression models in which risk factors are ordered in a temporal pattern, so that the earliest occurring predictors/covariates of any ADHD at age 10 years are entered first and are not displaced by later occurring covariates [9]. For these time-oriented logistic regression risk models, we categorized sets of antecedents/covariates by the time they occur, or are identified (e.g., antenatal, early postnatal, and later postnatal). We used a step down procedure seeking a parsimonious solution without interaction terms. These models allowed us to calculate odds ratios and 95% confidence intervals. Results are displayed as forest plots for any ADHD according to a parent, teacher, physician, any two, or by receipt of medication.

Results

We limited these analyses to the 583 children for whom we had a parent-completed CSI-4 and whose DAS verbal and non-verbal scores were both ≥ 85. The CSI-4 does not make a diagnosis of ADHD. Rather, it screens children. Parent reports screened18% (104/583) of children as positive for ADHD (ADHD-SP), while teachers screened 14% (60/422) as ADHD-SP.

The supplement provides tables and text describing what we found on univariable analyses. Based on these findings we constructed time-oriented risk models that evaluate risk factors in the order they occur or are first identified (Table 1). What we call the first epoch consists of maternal demographic and pregnancy characteristics, while the second epoch consists of perinatal exposures and characteristics first identified at the time of birth, and the third epoch includes exposures and characteristics from the first postnatal month.

Table.

Time-oriented risk models of any ADHD according to the reporter. These are odds ratios and 95% confidence intervals comparing children identified by parents, teachers, physicians, any two of these reporters, and by receipt of medication for ADHD to children not identified as having any ADHD according to any of these criteria. In time-oriented risk models, risk factors are organized in temporally ordered epochs. Predictors/covariates of the outcome (e.g., report of ADHD) in the earliest epoch are entered first and are NOT displaced by covariates in later epochs. Each epoch here is based on including the variables from the previous epochs. The final models with all variables included at one time are in Supplement Table K.

		Parent	Teacher	Physician	Any 2	ADHD Rx
Epoch 1
Maternal age	< 21	2.3 (1.1, 4.7)		2.3 (1.3, 4.3)	2.7 (1.4, 5.3)	3.3 (1.8, 6.2)
Mother’s BMI	≥ 30	2.4 (1.4, 4.3)		1.7 (1.1, 2.8)	2.8 (1.6, 4.8)	1.7 (1.00, 3.0)
Smoked during preg	Yes	2.2 (1.1, 3.3)		2.0 (1.2, 3.5)	2.1 (1.1, 3.9)
Mycoplasma	Yes		2.5 (1.1, 5.9)		2.2 (1.01, 4.7)
Epoch 2
Magnesium for sz	prophylaxis	1.7 (1.1, 2.5)			1.5 (1.02, 2.2)
Sex	Male	1.9 (1.2, 3.2)	3.3 (1.8, 6.1)	1.6 (1.05, 2.5)	2.0 (1.2, 3.3)	1.8 (1.1, 2.9)
Singleton	Yes	2.5 (1.4, 4.6)		1.8 (1.1, 3.1)	2.6 (1.3, 5.0)
Gestational age, wks	23–24	2.1 (1.2, 3.6)			2.2 (1.2, 4.0)	1.8 (1.03, 3.2)
Birthweight, grams	≤ 750		2.4 (1.3, 4.4)	2.1 (1.3, 3.3)
Epoch 3
Any sedative d 1–28	Yes	3.0 (1.8, 5.1)			1.9 (1.1, 3.4)
Ventriculomegaly**	Yes	3.2 (1.5, 6.7)			2.7 (1.2, 6.0)
ROP: pre-threshold‡‡	Yes	2.0 (1.02, 4.0)
Mech ventilation Day 7	Yes			2.1 (1.2, 3.6)		2.3 (1.3, 4.3)
Antibiotic, week 2–4	Yes			3.5 (1.6, 7.6)	3.6 (1.3, 9.5)	2.9 (1.2, 6.6)
NEC	IIIb			3.3 (1.1, 9.6)
Echolucent lesion**	Yes				3.5 (1.1, 11)
ROP: plus disease	Yes				2.6 (1.2, 5.7)
Epoch 4
M-CHAT positive†	Yes					2.2 (1.1, 4.4)
Model N		547	422	560	546	560

^**Alone or with other ultrasound scan abnormalities

^‡‡Satisfied ET-ROP criteria for ablative surgery (pre-threshold disease)

^†Among those with GMFCS < 1 and no vision or hearing abnormality

The risk factor patterns for ADHD-SP based on parent report and physician diagnosed ADHD have similarities, while the pattern for teacher-identified ADHD-SP is unlike that of either parent or physician. Consequently, the pattern for any two reporters closely approximates that of parent and physician. As expected, the pattern associated with prescribed medication most closely resembles that of physician-identified ADHD.

The risk profile associated with parent-reported ADHD-SP has 9 variables, 3 from the first epoch (i.e., low maternal age at delivery, pre-pregnancy maternal obesity, and smoking during pregnancy), 3 from the second epoch (magnesium for seizure prophylaxis, male sex and singleton) and 3 from the third epoch (receipt of a sedative, ventriculomegaly, and pre-threshold retinopathy of prematurity). The risk profile associated with teacher-reported ADHD-SP has only 3 variables (recovery of Mycoplasma sp from the placenta parenchyma, male sex, and low birth weight). The risk profile associated with physician diagnosed ADHD shares 3 variables with the parent-reported ADHD-SP profile (low maternal age and mother’s pre-pregnancy obesity, singleton) and has 3 third epoch variables that are not part of the parent-reported or teacher-reported profiles (mechanical ventilation on day 7, antibiotic receipt during weeks 2-4, and “surgical” necrotizing enterocolitis).

Discussion

The major findings of this study are that the risk profiles differ by who identified the ADHD-SP, and that both antenatal and postnatal risk factors are probably important.

The organizations that provide guidance about making a diagnosis of ADHD recommend observation in two settings [10], or gathering information from both a classroom teacher and from parents [11]. In the ELGAN Study cohort, the kappa for parent and teacher agreement about ADHD-SP (regardless of subtype) was 0.23, which indicates only a “fair level” of agreement [12]. This might reflect nothing more than the different settings in which teacher and parent see the child. It is also possible that the teacher completing the report has seen the child only when the child is medicated, whereas the parent and the physician has seen the child when not medicated. Then again, parents and teachers might really differ in the behaviors they consider acceptable. We explored the possibility that each behavioral set has a relatively unique risk profile.

We now review the individual risk factors identified, documenting which of these risk factors have been reported by others, and which are new. Then we offer a synthesis of all our findings.

Socio-economic information

Three of the maternal (antenatal) risk factors for parent-identified ADHD-SP (young age at the time of delivery, pre-pregnancy obesity, pregnancy smoking) convey information about socio-economic status, a well-known correlate of ADHD [13], as well as inflammation [14], and epigenetic phenomena phenomena [15, 16].

Magnesium for seizure

Even though antenatal magnesium exposure can be neuroprotective [17], here receipt of magnesium for seizure prophylaxis is likely a surrogate for preeclampsia and fetal growth restriction, risk factors for ADHD among children born at term [18].

Male

Both parents and teachers identified ADHD-SP more commonly among boys than among girls. Indeed, in one study of children born preterm, fully 84% of the children with ADHD were boys [1]. One group of authors postulated a “female protective effect against attention-deficit/hyperactivity disorder” [19].

Singleton

Among children born at term, however, singletons had higher rates than twins of ADHD symptoms in three studies [20–22], lower rates in one study [23], and no difference in one study [24]. We did not find any other assessment of the relationship between twinning and ADHD among children born very preterm.

Gestational age and birth weight

Attention problems and diagnoses of ADHD are more common in preterm and low birthweight children than in their term-born peers [1, 25–27]. Even among children not selected by gestational age or birthweight, low birthweight has been identified as an ADHD antecedent [28, 29].

Any sedative

Because very preterm newborns in the intensive care nursery are exposed to stressful and painful procedures during neonatal intensive care, they are often given analgesics and sedatives. Both fentanyl [30] and midazolam [31] exposure have been associated with brain imaging abnormalities and poorer outcomes. What remains unclear is the magnitude of the contribution from confounding by indication [32].

For example, recipients of these drugs are more likely than others to have remained on the ventilator for extended periods of time and to have had bowel disease requiring surgery. Both of these exposures were associated in the ELGAN Study with prolonged inflammation [33, 34] and developmental adversities [35, 36].

Ventriculomegaly

To our knowledge, no report other than ours identifies ventriculomegaly as an antecedent of attention problems. We emphasize that our sample for these analyses consists only of children whose IQ was ≥ 85, thus eliminating intellectual disability or low cognitive abilities (sometimes associated with ventriculomegaly) as the primary reason for these limitations.

Retinopathy of Prematurity (ROP)

Prethreshold ROP had been the intensity of disease that prompted surgical intervention when these data were collected. Since then, “plus disease,” defined as abnormal dilation and tortuosity of the blood vessels, is now the primary indication for laser treatment [37]. Consequently, both of these ROP variables probably provide similar risk information.

We are not aware of any report that identified ROP as an antecedent of ADHD. The most plausible explanations for their co-occurrence in any child invoke exposure to a common antecedent, such as relative hyperoxia [38], or a common vulnerability associated with especially low gestational age [39].

Mechanical ventilation on day 7

The mechanical ventilation variable likely conveys information that is similar to the variable for receipt of sedative described above. We are not aware of any report of mechanical ventilation or bronchopulmonary dysplasia (a correlate of mechanical ventilation) as risk factors for ADHD.

Antibiotic treatment during weeks 2-4

The antibiotic variable conveys information about bacteremia, as well as low gestational age [40]. Thus, this variable can convey information not only about inflammation, but also immaturity/vulnerability (above and beyond that conveyed by low gestational age and related variables).

Necrotizing enterocolitis (NEC)

The NEC variable might also carry information about inflammation [34], although it might also convey information about anesthetic-induced developmental neurotoxicity [41, 42]. We are aware of only one report of NEC placing the very preterm newborn at increased of an impairment of selective attention [43].

An echolucent image, the ultrasound image of periventricular leukomalacia

We found only one report of an association between periventricular leukomalacia and an impairment of selective attention in children born extremely preterm [43].

Synthesis

In an effort to synthesize what might appear to be disparate isolated findings, we offer the following four broad themes.

The first, low socioeconomic status and perhaps genetic contributions, too, are conveyed by three maternal characteristics (i.e., young age at the time of delivery, pre-pregnancy obesity, pregnancy smoking) [13].

The second theme is endogenous vulnerability/immaturity [26]. The most obvious of the variables conveying such information are low gestational age and low birthweight. On the other hand, each of the postnatal variables also has the potential to carry this information because each occurs most commonly in the least mature. This applies to receipt of a sedative, ventriculomegaly, ROP, mechanical ventilation on day 7, receipt of an antibiotic, and NEC.

The third major theme is inflammation [44]. Among the variables that are likely to carry information about inflammation are maternal obesity [14], maternal smoking during pregnancy [45, 46], ventriculomegaly [47], ROP [48], mechanical ventilation on day 7 [33], NEC [34], and of course, recovery of Mycoplasma sp from the placenta parenchyma, and receipt of an antibiotic.

The fourth major theme is epigenetic phenomena [49]. The variables likely to convey this information include maternal obesity [50], maternal smoking during pregnancy [51], magnesium for seizure prophylaxis (as a surrogate for preeclampsia) [52], and ROP (via its association with fetal growth restriction) [53]. Epigenetic processes have been invoked to explain the relationships between impaired development and social status [54], endogenous immaturity [55], and inflammation [56].

Strengths and limitations

Our study has several strengths. First, it is comprised of a large number of infants. Second, infants were selected according gestational age criteria and not birth weight, thereby minimizing confounding associated with fetal growth restriction 16). Third, all antecedent data were collected years before ascertainment of ADHD-SP. Fourth, attrition was modest.

Our study also has several limitations. First, we were not able to obtain teacher assessments of 252 children. Although these children do not appear to have had neonatal characteristics that differ from children whose teacher provided a report, their absence reduces the power and has the potential to be a source of bias. Second, the CSI-4, which we used to identify children who screened positive for ADHD based on the parent’s or teacher’s report, has not been adequately compared to other ADHD assessment instruments [57]. Third, our reliance on assessments of behaviors is not in keeping with the claims of others that a neurobiologic perspective should be preferred [58]. Fourth, we did not have any information about family history of ADHD or about alleles of selected genes and so could not assess familial contributions or gene-environment interactions.

Conclusion

The risk profiles of 10-year old children born before the 28^th week of gestation identified by parents appear to differ from the risk profile of their peers identified as ADHD by teachers. This raises the possibility that the parents and teachers identified different phenotypes, each with its own risk profile. The multiplicity of risk factors identified can be subsumed as components of four broad themes, low socioeconomic state, immaturity/vulnerability, inflammation, and epigenetic phenomena.