Some 0.8%–1.5% of all infants in developed countries are born with very low birth weight (VLBW), defined as a birth weight below 1500 g (1, 2). Developments in perinatal and neonatal care in recent decades have allowed an increasing number of VLBW infants to survive, and these individuals now constitute a substantial portion of the adult population. Although most such infants survive without major disabilities, VLBW children, adolescents, and young adults are at increased risk for psychopathology, with both internalizing and externalizing symptoms (3–9). Studies of their mental health outcomes have produced inconsistent results, however (10, 11). One reason for this may lie in the heterogeneity of VLBW infants. Some have suffered from intrauterine growth retardation in addition to being born preterm. Indeed, there is evidence that VLBW individuals who are born small for gestational age differ in later mental health outcomes, such as depression (12).

Studies also suggest that VLBW children are at increased risk for attention deficit hyperactivity disorder (ADHD) or related behavioral symptoms (3, 5–9, 13, 14). Data on whether these behavioral symptoms persist into adulthood are scanty, however. Young adults with VLBW have reported more difficulties in their ability to concentrate than have comparison subjects born at term (11). Recently, Hack et al. (4) found that more parents reported inattention in VLBW young adults than in term comparison subjects but found no group differences in self-reported behavioral symptoms of ADHD.

Our first major aim in this study was to investigate self-reported behavioral symptoms of ADHD in a well-characterized cohort of young adults born with VLBW or born at term. Because the effects of VLBW on behavioral symptoms of ADHD may vary according to intrauterine growth pattern (4, 15), we further examined whether the effects of VLBW differ in VLBW adults whose birth weight was small for gestational age (SGA) as compared with those whose birth weight was appropriate for gestational age (AGA). Moreover, whereas ADHD and its behavioral symptoms are often comorbid with risk-taking behaviors, data demonstrate (4, 10, 14, 16, 17), with few exceptions (11), that VLBW adolescents and young adults exhibit fewer delinquent and other risk-taking behaviors. Thus, our second major objective in this study was to examine whether VLBW is associated with risk-taking behaviors. Our secondary aim was to examine whether additional pregnancy-related factors, such as preeclampsia and maternal smoking during pregnancy, predict similar behavioral symptoms of ADHD and risk-taking behaviors.

Participants

The original cohort comprised 335 VLBW infants born between 1978 and 1985 in the province of Uusimaa, Finland, and treated at Helsinki University Central Hospital (18). In 2004, we traced 95.1% of the original cohort and invited those 255 residing in the greater Helsinki area to attend a clinical study aimed at a comprehensive assessment of their somatic and psychological health (12, 18–21). Of those invited, 166 (65.1%; 95 women and 71 men) participated. Their gestational age at birth ranged from 24 to 36 weeks and their birth weights from 600 to 1500 g. According to Finnish birth weight standards (22), 55 (33.1%) had a relative birth weight lower than two standard deviations below the mean and were classified as SGA, whereas the remaining 111 (66.9%) were classified as AGA. In the original cohort, mortality after hospital discharge up to 2004 was 1.8%.

Term comparison subjects (gestational age ≥37 weeks, birth weight at or above two standard deviations below the mean) were recruited in the same hospitals; each was the next eligible singleton infant of the same gender born after each corresponding VLBW birth. The recruitment procedure has been described in detail elsewhere (18). The comparison subjects’ gestational age at birth ranged from 37 to 43 weeks and their birth weight from 2560 to 4930 g. In 2004, their mortality after hospital discharge was 1.0%. Overall, 96.8% of these subjects were successfully traced, and of the 314 individuals invited to participate in the study, 172 (54.8%; 103 women and 69 men) did so.

No differences were observed between those who agreed to participate and those who declined in neonatal variables (birth weight, birth weight standard deviation score, gestational age, maternal preeclampsia, bronchopulmonary dysplasia, days of mechanical ventilation, oxygen treatment, and days at discharge from the neonatal intensive care unit). Participants in the VLBW group had a lower incidence of cerebral palsy at age 15 months (N=10 compared with N=17, p=0.005) (18). In the VLBW group only, smoking during pregnancy was more common among nonparticipants’ mothers (31.8% compared with 19.9%, p=0.04).

Background Variables

Information on maternal, perinatal, and neonatal characteristics were drawn from hospital records. Maternal smoking during pregnancy was based on a self-report obtained antenatally (yes/no), and maternal preeclampsia was defined according to current criteria (23). Information about participants’ adult characteristics, including data on health, medical history, current medication, and school history, as well as parental educational attainment (highest level of education achieved by either parent), was obtained by self-report. The Beck Depression Inventory (24) was administered, and body mass index (BMI) was measured during a clinical examination. The maternal, neonatal, and adult characteristics of the VLBW and comparison groups are summarized in Table 1. Each participant gave informed consent. The study protocol was approved by the Ethics Committee for Children and Adolescents’ Diseases and Psychiatry at Helsinki University Central Hospital.

Main Outcome Measures

During the clinic visit, participants completed a psychological survey, with questionnaires that included questions pertaining to behavioral symptoms of ADHD and to substance use. Two VLBW participants who did not complete the questionnaire and another two who had attended training school because of developmental delay were excluded, leaving 162 VLBW participants (93 women and 69 men) and 172 term comparison subjects eligible for the study (age range=18–27 years, mean=22.4, SD=2.2). The survey included the Adult Problem Questionnaire (APQ) (25), which measures behavioral symptoms of ADHD. The APQ is a 43-item self-report with Likert scale ratings ranging from 0, “Not at all present,” to 3, “Very much.” Before analysis, items with missing values (N=11, <0.001%) were replaced by the participant’s individual mean score. In this study, the APQ had good internal consistency; the general coefficient of reliability was 0.96, calculated by the procedure of Tarkkonen and Vehkalahti (26).

To identify the latent variable structure of the APQ, all items were subjected to exploratory factor analysis. Each latent underlying factor derived comprises a group of highly intercorrelated variables representing an entity; for instance, APQ items measuring executive dysfunctioning are expected to form a factor. We applied the maximum likelihood estimation method with varimax rotation. At the initial step, items with communalities below 0.3 (N=8) were removed (item numbers 26–28, 34, 38–41). The final factor solution (shown in the table in the data supplement that accompanies the online edition of this article) yielded six factors that explained 53.2% of the total variance. Factor scores—representing sums of items weighted by their loading on the factor—for each subscale were computed and served as outcome variables in all analyses. Reliability coefficients (26) for factor scores are also listed in the supplemental table.

Risk-taking behavior was operationalized in terms of substance use. Each participant completed a questionnaire inquiring about alcohol use (quantity and frequency of use and frequency of becoming drunk) and use of tobacco (quantity) and of recreational drugs (the questionnaire asked, “Have you ever tried marijuana, hash, or other narcotics, or for instance sniffed glue?” with possible responses ranging from “never” to “≥20 times”). Before analysis, outcomes were dichotomized.

Statistical Analysis

With a sample size exceeding 145, a group difference of 0.38 standard deviations is detectable with 90% power at an alpha level of 0.05. Group comparisons of continuous outcome variables (behavioral symptoms of ADHD) were carried out by univariate analysis of variance and comparisons of categorical outcome variables (risk-taking behavior) by chi-square tests. Analyses of covariance and logistic regression analyses were used when group differences were tested in the presence of covariates, which included gender, age at testing, BMI, participation in a modified curriculum at school, current parental educational attainment, and maternal smoking during pregnancy. We adjusted for age, gender, socioeconomic status, and maternal smoking during pregnancy, because previous research has identified these as risk factors for ADHD (27–30). BMI was adjusted for because of the associations between obesity, negative emotions, and excess use of alcohol, aspects embedded in the APQ (31). Finally, we controlled for participation in a modified school curriculum as a proxy for cognitive performance, since ADHD-like symptoms may be secondary to cognitive problems.

Because symptoms of neurosensory impairment and depression may overlap with symptoms of ADHD, we then reran these analyses after exclusion of individuals with neurosensory impairment (13 with cerebral palsy, one who was deaf, two who were blind, and two with developmental delay mild enough to allow participation in a standard curriculum at school and who were therefore initially included in analyses) and those with severe depression (four with scores ≥30 on the Beck Depression Inventory).

Statistical analyses were performed with SPSS for Windows, version 13.0 (SPSS, Chicago). All significance tests were based on a two-sided p value <0.05.

Initial Analyses

Before proceeding to analyses targeting the major study objectives, we tested associations between behavioral symptoms of ADHD, substance use, and pregnancy-related factors, such as maternal smoking and preeclampsia. Maternal smoking during pregnancy was associated with an increase in ADHD-related symptoms (total sumscore, p=0.005, executive dysfunctioning, p=0.003). Maternal preeclampsia was associated with higher scores on the emotional instability subscale (p=0.05) but not with any other ADHD-related symptoms, nor with substance use, and it did not serve as a covariate in the analyses.

Behavioral Symptoms of ADHD

Table 2 shows the distribution of behavioral symptoms of ADHD between VLBW individuals and term comparison subjects. VLBW individuals had lower total symptom scores than comparison subjects. Group comparisons of the factor analysis-derived subscale scores (pertaining to various behavioral dimensions of ADHD) revealed that VLBW individuals scored significantly lower than comparison subjects on the alcohol use subscale only (adjusted p=0.006), which contributed to their lower total score. No differences emerged between the VLBW and comparison groups on other subscales.

We then divided the VLBW group into SGA and AGA subgroups (Table 3, Figure 1). The AGA subgroup had lower total symptom scores than both the SGA subgroup and the comparison group. On the executive dysfunctioning and emotional instability subscales, the SGA subgroup scored higher than the AGA subgroup and the comparison group, whereas the corresponding scores for the comparison group fell between the scores of the other two groups. While the SGA and AGA subgroups did not differ on the alcohol use subscale, they both scored lower than the comparison group (adjusted p=0.01 and p=0.04, respectively). No between-group differences were observed on other subscales. These associations were not moderated by gender, with the exception of the inattention/restlessness subscale (Table 3). These data are adjusted for covariates (as explained in the Method section), and the adjusted and unadjusted results were similar.

We reran the analyses in Table 3 after excluding those 21 individuals with neurosensory impairment or severe depression. Group differences with regard to total sumscore were attenuated (adjusted mean differences between the AGA and SGA subgroups: –5.51, 95% CI=–11.94 to 0.93; between the comparison group and the AGA subgroup: 4.59, 95% CI=–0.04 to 9.22; and between the comparison group and the SGA subgroup: –0.92, 95% CI=–6.96 to 5.12). The SGA subgroup still scored higher than the AGA subgroup on the executive dysfunctioning subscale (adjusted p=0.019), higher than the AGA subgroup and the comparison group on the emotional instability subscale (adjusted p=0.002 and p=0.02, respectively), and lower than the comparison group on the alcohol use subscale (adjusted p=0.04). Finally, we reran the analyses in Table 3 with additional adjustment for use of antidepressants, which was more common in the SGA subgroup (12). Changes to the results were marginal (data not shown).

Risk-Taking Behavior in Terms of Substance Use

Table 4 shows the association of VLBW with substance use. Fewer VLBW individuals (in both the AGA and SGA subgroups) than comparison subjects reported using alcohol or getting drunk monthly. Likewise, having tried recreational drugs was less common for VLBW individuals. Fewer VLBW individuals (in both subgroups) reported having smoked tobacco, and the proportion of daily smokers was smaller (although significant in the SGA subgroup only). Within the VLBW group, the SGA and AGA subgroups did not significantly differ in substance use. These associations were not modified by gender (the interaction term for gender by study group was nonsignificant). The results listed in Table 4 remained similar after exclusion of individuals with neurosensory impairment and severe depression, except for the proportion of daily smokers, which became nonsignificant.

Our VLBW young adults born SGA reported more executive dysfunction and emotional instability than did their AGA peers and term comparison subjects, suggesting poorer adult outcomes for those born both too early and too small. VLBW individuals whose birth size was AGA, in contrast, fared well. VLBW young adults as a group did not report more ADHD-related symptoms; they did, however, report less substance use. Among determinants of fetal growth, maternal smoking during pregnancy was associated with increased ADHD-related behavioral symptoms.

Studies show that VLBW individuals are at risk for ADHD symptoms in childhood and adolescence (3, 5–9, 13–15). While “common” ADHD is usually strongly associated with elevated rates of substance use (27), research also indicates, somewhat counterintuitively, decreased risk-taking behavior among adolescents and young adults born preterm (4, 10, 14, 16, 17, 32). Indeed, it has been proposed that the ADHD of preterm children is more “pure” (13), characterized by a more even gender distribution and by less hyperactivity in relation to inattention, and that it is less frequently accompanied by comorbid disorders (7–9, 13). This proposed distinct phenotype may reflect different biological mechanisms underlying the disorder. In our study, the combination of increased executive dysfunction, emotional instability, and less alcohol use among VLBW-SGA adults thus fits the theory of a distinct phenotype. Our other findings from the substance use survey, showing substantially less substance experimentation and use in the VLBW group, further support this idea of fewer comorbid disorders.

In disagreement with reports pointing to an ADHD with more inattention in relation to hyperactivity among former VLBW infants (4, 7, 9), we found no increase specifically in symptoms of inattention. It is thus conceivable that either the inattention/restlessness subscale fails to identify inattention satisfactorily or that self-rating questionnaires in general are an unsuitable method for exploring this particular trait.

While the relationship between VLBW and ADHD symptoms in childhood is established, this relationship in adulthood is less clear. ADHD usually declines with age, and the risk of symptom persistence is greatest among children with the hyperactive delinquent type of ADHD (28). In agreement with our results, the few findings published on ADHD symptoms in young adulthood have failed to demonstrate any excess of self-reported ADHD symptoms in VLBW individuals relative to comparison subjects (4, 15). However, in contrast to self-reports, the parents of these VLBW individuals rated their young adults as having significantly more ADHD symptoms than did parents of comparison subjects. In a study of extremely-low-birth-weight (<1000 g) adolescents, study subjects scored higher on ADHD symptoms according to parent report but not according to self-report (14). The reason for this discrepancy between parent reports and self-reports is unclear. It may be related to parental overprotection after a preterm birth or to denial (14) or recalibration (14) among the young VLBW adults. In any case, in the absence of studies with objective criterion variables, we cannot know whose ratings may be more biased, the parents’ or the VLBW young adults’. Recently, Dalziel et al. (32) reported that a group of adults born at a mean gestational age of 34.1 weeks and a mean birth weight of 1946 g scored nonsignificantly lower than term comparison subjects on the Brown Attention Deficit Disorder scale, a result in line with ours. Previous observations have shown that extremely-low-birth-weight young adults and comparison subjects rate their health-related quality of life equally high, despite more disability in the extremely-low-birth-weight group (3, 33). Possibly, this phenomenon affected the ratings on the APQ in our study as well, with VLBW individuals in general and VLBW-AGA in particular reporting low ADHD-related behavior relative to term comparison subjects.

The underlying mechanisms for VLBW-SGA individuals reporting excessive ADHD-related behaviors may be biological as well as psychosocial (6). Biological programming of the developing brain due to a nonoptimal fetal environment or prematurity-associated illness in the postnatal period may interfere with neuronal organization and thus modify the later psychological phenotype. In studies of low-birth-weight children (<2500 g) (30), even of those within the term-born range (34), small body size at birth predicts behavioral symptoms of ADHD. Our findings suggest that intrauterine growth retardation, for which SGA serves as a proxy, is a more important predictor for later ADHD-related behavioral traits than is prematurity per se. Furthermore, we cannot exclude the possibility that ADHD, being a highly heritable disorder (27), and prematurity have a common genetic origin. After birth, the psychosocial environment and parenthood also affect the development of the VLBW individual: being born preterm—especially if SGA—is associated with less favorable parent-child and family interactions at age 4 months (35). A VLBW birth initially gives rise to maternal psychological distress, with high-risk infants causing them the most distress (36).

Our results parallel prior findings on the psychological profile among VLBW young adults from this same cohort. We recently demonstrated (12) that their depression was modified by their intrauterine growth pattern, as reflected in SGA status. The VLBW group reported less depression, although as in the present study, this was confined to those born AGA; those born SGA presented with more depression (12). With regard to their temperamental characteristics, VLBW young adults (both AGA and SGA) reported less fun-seeking behavior than did term comparison subjects (21), a trait fitting well their lower rate of substance experimentation reported here. As for personality, the VLBW group displayed more conscientiousness and less impulsiveness, excitement seeking, hostility, and openness to experience (19). They also leave the parental home and start cohabiting with an intimate partner at a later age (20). We can conclude from these studies that VLBW adults are, on average, more dutiful, more cautious, and less prone to risk-taking behaviors such as substance use, whereas their psychopathology seems to vary according to intrauterine growth patterns: the higher degrees of depression and ADHD symptoms are confined to those born SGA. Obviously, replication of these findings is necessary to confirm these conclusions.

The strengths of this study include a well-characterized cohort with comprehensive perinatal data, which allows us to explore the impact of early life events on adult health. In particular, separating the SGA and AGA subgroups in the analyses broadens the perspectives of how VLBW may affect adult health, since it has been suggested that preterm individuals suffering from intrauterine growth retardation are at higher risk for decreased cognitive capacity in childhood than their AGA peers (37). Valuable information may be overlooked in failing to separate SGA and AGA subgroups. It should be borne in mind, however, that definitions of SGA are accompanied by uncertainty about whether the infants actually suffer from intrauterine growth retardation. As we used a conservative definition of SGA (lower than two standard deviations below the mean), most if not all of our SGA infants are likely to have suffered from intrauterine growth retardation, whereas the AGA group may include some intrauterine growth retardation individuals who were less growth retarded.

Our results must be considered in the light of some limitations. First, we aimed to assess ADHD-related behaviors, not make a clinical diagnosis. We therefore used a questionnaire, and questionnaires have proven to be important and helpful tools in the assessment of ADHD trends in population studies (38, 39). Fortunately, the APQ we used in this study was originally designed to assess ADHD-related behaviors in adulthood specifically (25). Since many ADHD questionnaires, as well as the DSM-IV diagnostic criteria, were originally designed for children, new and different questionnaires are needed for assessing specifically adult ADHD. Second, we lack retrospective data on childhood ADHD, data needed according to the DSM-IV criteria for a valid diagnosis in adulthood. The validity of retrospective diagnoses has been questioned, however, because of susceptibility to recall bias (40). Third, as in similar study settings, a participation bias toward healthier participants cannot be ruled out, although our participants and nonparticipants did not differ with regard to most of the perinatal variables listed in Table 1. Fourth, since the participants were young adults themselves, we did not approach their parents and are therefore unable to adjust for possible genetic susceptibility indicated by any existing parental ADHD symptoms. Finally, the ADHD-related symptoms were assessed by self-report. When diagnosing ADHD in children, parent and teacher reports are considered more important than self-reports, whereas for adults, clinicians must rely to a greater extent on the patients’ subjective opinions of their own symptoms. Adults are rarely monitored by a parent or an employer as consistently as children are by their parents or teachers, making observer reports less accurate and self-reports of greater value (25, 38). Still, ADHD remains a clinical diagnosis (38, 39), and studies of VLBW adults utilizing a more extensive assessment battery of ADHD remain to be done.

In sum, in contrast to previous findings in VLBW children, we found no evidence of increased behavioral symptoms of ADHD in VLBW adults. However, when we contrasted AGA and SGA subgroups with each other and with term comparison subjects, we found that the SGA group showed higher scores for executive dysfunction and emotional instability. Although these findings are reassuring for a large proportion of VLBW children and adults and their families, they also call for studies elucidating the pathophysiological and neuropsychological mechanisms underlying such differences.