ADHD symptoms and diagnosis in adult preterms: systematic review, IPD meta-analysis, and register-linkage study

Background This study examined differences in ADHD symptoms and diagnosis between preterm and term-born adults (≥18 years), and tested if ADHD is related to gestational age, birth weight, multiple births, or neonatal complications in preterm borns. Methods (1) A systematic review compared ADHD symptom self-reports and diagnosis between preterm and term-born adults published in PubMed, Web of Science, and PROQUEST until April 2021; (2) a one-stage Individual Participant Data(IPD) meta-analysis (n = 1385 preterm, n = 1633 term; born 1978–1995) examined differences in self-reported ADHD symptoms[age 18–36 years]; and (3) a population-based register-linkage study of all live births in Finland (01/01/1987–31/12/1998; n = 37538 preterm, n = 691,616 term) examined ADHD diagnosis risk in adulthood (≥18 years) until 31/12/2016. Results Systematic review results were conflicting. In the IPD meta-analysis, ADHD symptoms levels were similar across groups (mean z-score difference 0.00;95% confidence interval [95% CI] −0.07, 0.07). Whereas in the register-linkage study, adults born preterm had a higher relative risk (RR) for ADHD diagnosis compared to term controls (RR = 1.26, 95% CI 1.12, 1.41, p < 0.001). Among preterms, as gestation length (RR = 0.93, 95% CI 0.89, 0.97, p < 0.001) and SD birth weight z-score (RR = 0.88, 95% CI 0.80, 0.97, p < 0.001) increased, ADHD risk decreased. Conclusions While preterm adults may not report higher levels of ADHD symptoms, their risk of ADHD diagnosis in adulthood is higher. Impact Preterm-born adults do not self-report higher levels of ADHD symptoms, yet are more likely to receive an ADHD diagnosis in adulthood compared to term-borns. Previous evidence has consisted of limited sample sizes of adults and used different methods with inconsistent findings. This study assessed adult self-reported symptoms across 8 harmonized cohorts and contrasted the findings with diagnosed ADHD in a population-based register-linkage study. Preterm-born adults may not self-report increased ADHD symptoms. However, they have a higher risk of ADHD diagnosis, warranting preventive strategies and interventions to reduce the presentation of more severe ADHD symptomatology in adulthood.

Previous evidence has consisted of limited sample sizes of adults and used different methods with inconsistent findings. This study assessed adult self-reported symptoms across 8 harmonized cohorts and contrasted the findings with diagnosed ADHD in a population-based register-linkage study.
• Preterm-born adults may not self-report increased ADHD symptoms. However, they have a higher risk of ADHD diagnosis, warranting preventive strategies and interventions to reduce the presentation of more severe ADHD symptomatology in adulthood.

BACKGROUND
Individuals born preterm (<37 + 0 wks + days gestational age) have an increased risk for attention-deficit hyperactivity disorder (ADHD) and elevated symptom levels in childhood. 1 This increased risk not only characterizes tshose born very preterm (VP; <32 + 0 wks + days) or with very low birth weight (VLBW; <1500 g), but the risk increases linearly with each declining week of gestation. 1 Although well established in childhood, there is conflicting evidence as to whether the risk for ADHD persists in preterm adults. The conflicting information may reflect the varying quality of evidence, diverse methods, and small sample sizes of the cohort studies with adult follow-ups, 2,3 whereas the populationbased register-linkage studies have examined ADHD diagnosis in samples including both children and adults. 1,4 Studying ADHD in preterm borns beyond childhood is important as the prevalence rate of ADHD diagnosis in the general population declines with age. 5,6 Moreover, not all children with ADHD diagnosis in childhood continue to meet the diagnostic criteria in adulthood and the symptom profiles may change with inattention showing the highest and impulsivityhyperactivity the lowest continuity. 7 Also, the predominately male presentation of ADHD in childhood may no longer be present in adulthood. 6 We conducted a systematic review of existing studies examining associations between preterm birth and ADHD symptoms and diagnosis in adulthood, with evidence quality evaluation. We performed an Individual Participant Data (IPD) meta-analysis to determine if preterm and term-born adults differed in selfreported ADHD symptoms. In a population-based register-linkage study, we investigated the risk of receiving an ADHD diagnosis in preterm compared to term-born adults. We examined the effects of gestational age, birth weight, multiple birth status, bronchopulmonary dysplasia (BPD), and intraventricular hemorrhage (IVH) on later ADHD symptoms and diagnosis for preterm-born adults.

METHODS
The systematic review was performed in line with the preferred reporting items for Systematic Review and Meta-Analyses of Individual Participant Data (PRISMA IPD). 8 The IPD meta-analysis and population-based registerlinkage study were a part of the Research on European Children and Adults born Preterm (RECAP) consortium (https://recap-preterm.eu/) and Adults Born Preterm International Collaboration (APIC) (https://www.apic-preterm. org/members/). The analysis plan was submitted and approved by the RECAP consortium prior to the study (04/05/2019).

Systematic search
We conducted systematic searches in PubMed, Web of Science, and PROQUEST from Inception to October 2019 and an updated search in April 2021 ( Fig. 1 and Supplemental Table S1). Duplicate studies were removed, with remaining studies independently assessed in duplicate (RR, AL) first by title and abstract, followed by full-text assessment (Fig. 1  studies were identified via the references of included studies and from relevant systematic reviews and meta-analyses.

Eligibility criteria
Cohort, case-control, register-linkage studies, and meta-analyses assessing ADHD self-reported symptoms or diagnosis in adulthood (≥18 years) with data on gestational age, were included in the systematic review. Eligible symptom measures included a validated scale providing an overall ADHD symptom score, inattention, hyperactivity/impulsivity, attention problem scores, or clinical cutoffs. Eligible ADHD diagnosis included data from medical records, medical registers, or a structured diagnostic interview. Attention scales derived from neurocognitive tests were beyond the scope of this review. Due to profound changes in medicine during recent decades, we excluded studies with participants born before 1970. There were no language restrictions. Studies were excluded based on criteria presented in Supplemental Table S2. Self-reported symptom studies, which met inclusion criteria for the systematic review and were a part of the RECAP project and APIC consortia were eligible to participate in the IPD meta-analysis. Cohorts with adult follow-up were invited to provide a data dictionary, which was checked for ADHD symptom scales. Studies that declined to provide a data dictionary, did not have an adult ADHD measure, or were not a part of the RECAP/ APIC consortia were excluded.
In the population-based register-linkage study, the participants were identified from the Finnish Medical Birth Register. We included all live births in Finland from 01/01/1987 to 31/12/1998.

Outcome measures
For the IPD, all included cohort studies utilized at least one ADHD selfreported symptom scale in adulthood (Supplemental Table S3). Primary outcomes in the IPD were total ADHD symptom z-score and probable clinical ADHD based on clinical cutoffs. Secondary outcomes were inattention, hyperactivity/impulsivity, and attention problems z-scores.
For the Finnish population-based register-linkage study, the primary outcome was primary or subsidiary ADHD diagnosis derived from the Care Register for Health Care with follow-up data at age 18 years or older available until 31/12/2016. This register carries all diagnoses registered during inpatient and outpatient visits in public specialized medical care classified using the International Statistical Classification of Diseases and Related Health Problems 10th (ICD-10) Revision (code: F90 for ADHD). The Care Register for Health Care is a validated research tool. 9

Quality of evidence assessment
Two reviewers (RR, AL) independently assessed the evidence quality of studies included in the systematic review using the Newcastle-Ottawa-Scale (NOS). 12 The assessment criteria for each NOS domain (Selection, Comparability, Outcome) were set out at the onset of the study (Supplemental Table S4). Assessment disagreements between the two reviewers (RR, AL) were resolved based on consensus with a third reviewer (ML-P, KH).

Data sources
IPD and data dictionaries were provided by each of the participating cohorts via the RECAP platform data nodes. The adult data were cleaned and harmonized by the University of Helsinki (RR and reviewed by PG). Perinatal data for the IPD were harmonized by the University of Warwick. The register data were cleaned and harmonized by the Finnish Institute for Health and Welfare. The data underlying this article cannot be shared publicly due to the personal and sensitive information of the participants. If requested, access to this data is subject to data sharing agreements.

Statistical analysis
The IPD meta-analysis was conducted using a one-stage meta-analytic approach where IPD from all cohort studies were analyzed simultaneously. We applied linear mixed-effects models with the restricted maximum likelihood estimation method for the covariance parameters to examine mean differences with 95% confidence intervals (95% CI) in ADHD symptoms between preterm and term-born adults. We applied generalized linear mixed-effects models for binomial data to estimate the odds ratio (OR) with 95% CI of scoring above the clinical cutoff on any of the ADHD scales in preterm compared with term-born adults. In mixed-effects models, random intercepts were specified to account for the clustering of participants within cohorts.
ADHD scores were normalized with rank normalization and standardized to the mean of 0 and SD of 1 within each of the studies. As all but two (VICS, NZ VLBW) of the included cohorts had two different ADHD measures, the mean of available standardized ADHD scores was taken.
In the register study, we applied binomial logistic regression to estimate the relative risk (RR) with 95% CI of being registered with ADHD diagnosis in public specialized medical care in adulthood between preterm compared with term-born adults.
In both study designs, we tested whether the associations varied according to the degree of prematurity and SGA status. Then we tested whether gestational age, birth weight z-score, multiple births, BPD, and IVH were associated with ADHD symptoms and diagnosis among adults born preterm. The effect sizes were estimated in adjusted models. Finally, in sensitivity analyses, we tested whether excluding participants with NSI affected the associations and in the register study, we also excluded participants who died or emigrated at 18 years of age or older.

Systematic review
Eleven individual studies and one meta-analysis met our inclusion criteria, representing 9 clinical cohorts examining 12 adult selfreported outcomes and two diagnostic interviews (Table 1). Overall, included studies were mostly moderate to high methodological quality (Table 1 and Supplemental Table S4) with scores ranging from 3 to 8 out of 9 possible points. Regarding Selection, five cohorts focused on VP/VLBW, while three cohorts focused on EP/ ELBW participants. None of the studies received all possible points, particularly due to the lack of adjustment for familial confounding either via parental ADHD or sibling comparisons. However, eight of the cohorts adjusted their analysis for at least three of the following factors, including participant age, sex, maternal age at birth, maternal/parental education, and family socioeconomic status. Regarding the Outcome, only two studies 3,13 had more than 50% attrition rate. Neither of the two cohorts (NTNU LBW Life, McMaster) which utilized diagnostic interviews (MINI), observed differences between VLBW/ELBW preterms and normal birth weight (NBW) term controls. 14,15 Of the 11 studies, 2,3,13,16-21 with adult  Meta-analyses on adult ADHD symptoms (Note: The meta-analyses include some of the empirical studies listed above. Meta-analyses that did not meta-analyze adult data are not included, nor studies without any self-reported ADHD symptoms). p Normal birth weight. q Achenbach system of empirically based assessment-adult self report. r Achenbach system of empirically based assessment-young adult self report.   self-reported questionnaires, only three cohorts (BLS, EPICure, NTNU LBW Life) found differences between preterms and term controls ( Table 1). As all but two of the studies 3,21 were previously metaanalyzed, 22 and no difference was observed, an aggregate data meta-analysis was not duplicated here.
All IPD data sets were checked for data integrity (RR, PG). Any questions were resolved with representatives of specific cohorts. Ultimately, no major issues were identified with the data sets.
Characteristics of the cohorts included in the IPD meta-analysis are in Table 2. Raw cohort-specific mean values of ADHD total and subscale scores for preterm and term-born controls are shown in Supplemental Table S3 and the number of participants within each cohort and across the different cohorts contributing to the IPD meta-analysis according to the degree of preterm birth are in Supplemental Table S5. Of the covariates, lower parental education, and female sex were associated with higher total ADHD symptoms z-score and probability to score above the ADHD clinical cutoff (Supplemental Table S6).
Preterm birth was not associated with total ADHD symptoms zscore, symptoms above the clinical cutoff (Fig. 2), hyperactivityimpulsivity, inattention, or attention problems z-scores (Supplemental Fig. 1). We individually compared EP, VP, MLP, SGA preterms, and preterms not born SGA groups with term controls (Supplemental Table S7). There were no differences, except adults born EP had higher odds of scoring above the ADHD clinical cutoff. However, EP and term controls showed no differences when we excluded participants with NSI (n = 148) (p = 0.18; data not shown).
In preterm only analyses, gestational age, birth weight z-score, multiple birth, and BPD were not associated with ADHD symptoms in adulthood (Table 3). ADHD symptoms z-scores were higher for preterm borns with IVH than with no IVH (Table 3). This difference remained when we excluded participants with NSI (p = 0.008; data not shown).

Population-based register-linkage study: ADHD diagnosis
The register study characteristics are described in Table 2. Supplemental Table S5 provides the number of participants according to the degree of preterm birth. Of the entire population of 729,154 individuals, 4683 (0.64%; 308 of 37,538 [0.82%] preterm, 4375 of 691,616 [0.63%] term) were registered with an ADHD diagnosis while attending specialty care at age 18 years or older. Of the covariates, lower maternal education, male sex, and older age were associated with higher RR for ADHD diagnosis in adulthood (Supplemental Table S6).
Compared to term-born controls, the RR of ADHD diagnosis in adulthood was higher for preterms born SGA and not SGA and for those born VP (Supplemental Table S7).
In the preterm-only analyses, the RR for ADHD diagnosis decreased according to each increasing week of gestation and SD in birth weight z-score (Table 3). Multiple births, BPD, or IVH were not associated with the risk of ADHD diagnosis in adulthood ( Table 3).
The associations remained when we excluded those individuals who died (n = 25) or emigrated (n = 90) after 18 years of age and those with NSI (n = 10157) (p-values < 0.01; data not shown). After excluding individuals with NSI, two associations became apparent: RR of ADHD diagnosis was higher for EP than term-born controls (RR = 2.06, 95% CI 1.08, 3.94, p = 0.03) and for preterms with any BPD compared with those with no BPD (RR = 2.52, 95% CI 1.26, 5.05, p = 0.01).

DISCUSSION
This systematic review highlighted the conflicting evidence of previous studies that varied in methodological quality from moderate to high. In the IPD meta-analysis the level of selfreport ADHD symptoms in adulthood was similar in preterm and term controls. In contrast, the population-based register-linkage study revealed that adults born preterm, EP, VP, and preterms born SGA and preterms not born SGA had a higher RR for being registered with ADHD diagnosis in public specialized medical care in adulthood compared to term controls. Moreover, among preterms in the register-linkage study, the RR of ADHD diagnosis in adulthood decreased according to each weekly increase in gestation and SD increase in birth weight z-score. The associations in the IPD meta-analysis or the register-linkage study were not explained by participant's sex, age, and parental/maternal education. Moreover, the associations changed only a little when we excluded participants with NSI.
The conflicting information of the previous studies identified by the systematic review reflects methodological differences between the studies and limitations that relate to small sample sizes and different control group recruitment approaches. Several studies also lack standardized adjustments for covariates, gestational age group assessments, and adult-specific follow-ups. We were able to address some of the limitations of the individual cohort studies in the IPD meta-analysis and the registerlinkage study.
The discrepant IPD meta-analysis and the register-linkage study findings may relate to differences in the severity of ADHD symptoms detected by the meta-analysis of self-reports and register-linkage study of diagnosis. ADHD self-reports may have captured milder, sub-clinical symptoms, whereas the registerlinkage study on ADHD diagnosis may have captured the most severe end of the ADHD spectrum, attending public specialized medical care. Only 0.64% of the individuals in our study had been registered with an ADHD diagnosis in adulthood. This is slightly lower than the estimated adult prevalence of 2.5-4.2% in highincome countries. 6 Hence, the specialized medical care diagnosis may miss a broader spectrum of ADHD cases. Furthermore, we cannot preclude informant bias, as self-reports are subject to response-bias. Reported simultaneously, adult self-reports and parent reports of the same participants frequently vary. 2 In order to receive an ADHD diagnosis in adulthood, information is  Fig. 2 ADHD symptoms and diagnosis by Cohort. Mean differences between preterm (<37 weeks) and term-born (37 + 0-41 + 6 wks + days) adults (≥18 years) in self-reported attention-deficit hyperactivity disorder (ADHD) symptoms z-scores and in the odds to score above versus below the ADHD clinical cutoffs in the Individual Participant Data meta-analysis of cohort studies, and in the relative risk to receive ADHD diagnosis in adulthood in the population-based register-linkage study. Error bars refer to 95% confidence intervals.
gathered from multiple informants, most importantly from the individual themself, but also from partners, parents, siblings, or school records. Then a physician makes the diagnosis according to the specific diagnostic criteria for ADHD. For the IPD meta-analysis, we cannot rule out selective data attrition in the cohorts, as adults with more severe ADHD symptoms may have been more likely to drop out, eliminating the possible detection of associations. In the register-linkage study, data attrition was minimal. While this precludes bias of selective drop-out, we cannot rule out that preterms may have been more likely to be registered with ADHD diagnosis in adulthood, if they have visited health care more frequently for other preterm-related conditions. Although the IPD meta-analysis (Cohen d = 0.10, OR = 1.34) and the register-linkage studies (RR = 1.20) provided ample statistical power (80% at alpha = 0.05) to detect small effect sizes, the statistical power in the register-linkage study was superior. This is particularly true for comparisons of ADHD symptoms above the clinical cutoff in the IPD meta-analysis and ADHD diagnosis in the register-linkage study. The statistical power is, however, smaller in comparisons of different gestational age groups and in the comparisons limited to preterm borns only. Moreover, the discrepancies may also be attributable to the composition of the total IPD sample which comprised preterm and term-born individuals from seven different countries, while the register-linkage study comprised only Finns. It is also important to note, that due to differences in the study designs, the samples included in the IPD meta-analysis comprised different numbers of different gestational age groups. For instance, 36.7% of the total IPD EP participants came from the UK, 20.7% from Australia, 17.9% from New Zealand, 13.7% from Finland, 5.6% from Norway, and 5% from Germany., whereas over 70% of MLP and term-born gestational age groups came from Finland. That preterm and term-born adults self-reported equal levels of ADHD symptoms may also suggest that ADHD symptoms do not hinder the daily functioning of adults born preterm. While ADHD in the general population is associated with risk-taking behaviors and criminality, [31][32][33] preterm-born individuals report less externalizing behaviors, 22 less smoking and substance abuse, 19,34 fewer contacts with police, 34 have less often experienced sexual intercourse 35,36 by young adulthood, and they also less often have substance use disorders and criminal convictions. 4 The lack of differences may also reflect preterm personality: in comparison to term borns preterm borns report being more conscientious, shyer, and less impulsive, less excitement seeking, and less open to new experiences. 37,38 The lack of differences between preterm and term adults in selfreported symptomatology in the IPD meta-analysis are consistent with a previous meta-analysis, 22 which however included fewer cohorts with adult follow-up than included here. The register-linkage study findings indicating higher RR for ADHD diagnosis for preterms than term controls are consistent with another Finnish register study, which examined ADHD specialty care diagnosis between ages 3 and 19, 1 and with a Swedish register study assessing specialty care diagnoses at ages 5-19 y. 4 Our findings are also in agreement with Norwegian, and one Swedish, register studies assessing psychostimulant medication use at 18-38 y, 39 30 y, 40 and at 26-33 y. 41 Highlighting further differences between the results of the IPD meta-analysis and register-linkage study, women in the IPD metaanalysis of cohorts reported higher levels ADHD symptoms, while in the register-linkage study men had a higher RR for ADHD diagnosis in adulthood. The higher ratio of men to women, may suggest that women may be more likely to be underdiagnosed. 6 This may be due to women with ADHD, at least in childhood, presenting less disruptive ADHD-like behaviors, 42 and in parent and teacher reports girls with ADHD show lower levels of hyperactivity, impulsivity, inattention and externalizing symptoms and higher levels of internalizing symptoms. 32 However, it has also been suggested that differences between men and women in ADHD may level off as the individuals' age, and some studies have reported that in adulthood women report higher ADHD symptomatology than men. 6,43 The only agreed finding between the cohort IPD meta-analysis and the register-linkage study was that participants with lower Table 3. Associations between gestational age, birth weight z-score, multiple births, and neonatal complications with self-reported attention-deficit/ hyperactivity disorder (ADHD) symptoms in adulthood in the individual participant data meta-analysis of cohort studies and with ADHD diagnosis in adulthood in the population-based register-linkage study among those born preterm.

Predictors
Individual participant data meta-analysis (n = 1385) Estimate refers to mean difference in ADHD symptoms between categorical predictors and standard deviation unit change per each unit change in the predictor. All associations are adjusted for participant's sex, age (in adulthood follow-up in the meta-analysis of cohorts and age at death, emigration or end of follow-up [31/12/2016] in the register-linkage study), and parental education (of either parent in the meta-analysis of cohorts and of maternal in the registerlinkage study). OR odds ratio, RR relative risk, 95% CI 95% confidence interval.
parental/maternal education had higher ADHD symptoms and RR for diagnosis. A strength of our systematic review and IPD meta-analysis includes a focused age group, allowing for a better understanding of the symptomatology of adults, instead of summarizing the symptoms or diagnosis across the entire lifespan as done in other studies. 44 This is important, as generally, the rates of ADHD tend to decline with age. 5 Additionally, the limited range of birth years included allows contextualization of the medical practices or interventions, which may have played a role in the outcomes observed in this generation. Furthermore, performing an IPD meta-analysis allows for the same covariates and confounders to be accounted for across all studies, typically a weakness in standard aggregate data meta-analyses. Moreover, in both the IPD meta-analysis and the register-linkage study, we were able to uniquely parcel out gestational age and birth weight categories and study neonatal complications in relation to the risk of ADHD symptoms and diagnosis in those born preterms, for which individual cohort studies lack the power to do.
A common weakness, our results do not represent low-or middle-income countries, as the IPD meta-analysis included only high-income country populations. 44 Hence, the generalizability of the findings is limited. Additionally, although we had IPD data, we were not able to account for additional potential confounders, and residual confounding cannot be excluded. We were not able to obtain IPD from all cohorts with self-reported ADHD symptom scales in adulthood with preterm or VLBW participants. It is unlikely that the inclusion of these three studies would have changed the results, since these studies have previously reported no differences. 13,16,21 Finally, while the Care Register for Health Care is a validated research tool, including ADHD diagnosis in childhood, 9 comparable data on adulthood ADHD diagnosis is lacking.
Considering the substantial economic burdens associated with preterm birth 45 and with ADHD, 46 understanding the intersections of the two is important. Pinpointing that preterms or specific preterm and birth weight groups are at a slightly increased risk of receiving ADHD diagnosis in adulthood than others may pave the way for targeted interventions to reduce the burden. On the other hand, the lack of differences in self-reported ADHD symptoms between preterm and term-born adults delivers a positive message for the preterm-born individuals themselves, their families, and healthcare professionals.