Study design
This is a population-based follow-up study of two low birth weight groups and a control group, all born between 1986 and 1988. One group was born preterm with very low birth weight (VLBW, birth weight: ≤ 1500 g) and the other was born small for gestational age (SGA) at term (birth weight < 10th percentile adjusted for gestational age (GA), sex and parity). The controls were born at term (birth weight ≥ 10th percentile, adjusted for GA, sex and parity).
During the enrolment period, all VLBW participants were admitted to the referral neonatal intensive care unit for the counties of North and South Trøndelag at The University Hospital in Trondheim, Norway. In the same period, the SGA and control participants were enrolled as part of a multicenter study comprising participants from Uppsala, Sweden, and Trondheim and Bergen, Norway [
28]. A 10% random sample of women was selected for follow-up during pregnancy. At birth, all children of mothers in the random sample and all children born SGA in the nonrandom sample were included for follow-up. Of these, only SGA and control participants recruited in Trondheim were included in the present study. Data were collected between autumn 2006 and autumn 2008.
Study population
The preterm VLBW Group
During 1986–88, 99 VLBW children were admitted to the neonatal intensive care unit, of whom 23 died and one with a congenital syndrome was excluded. At follow-up, additionally two were excluded because of severe cerebral palsy (CP) and/or mental retardation. Of the 73 eligible, 14 could not be traced. Hence, 59 were invited to participate, of whom 43 (17 men, 26 women) consented to participate in this part of the study (73% of 59 invited and 59% of 73 eligible). Eleven (26%) were born SGA (VLBW–SGA). The reference standards for classification of SGA (below 10th percentile) were specific for each sex and gestational week based on data from the Norwegian Medical Birth Registry [
29]. Group characteristics were mean (SD) birth weight: 1236 g (205), gestational age: 29.0 (2.5) weeks, assessment age: 19.5 (0.6) years and parental socioeconomic status (SES,
n = 41): 3.3 (1.4).There were no significant differences compared with controls in assessment age or parental SES. Two VLBW participants had CP (both bilateral spastic subtype, one of these with four limbs affected).
The term SGA Group
Among eligible women in the Trondheim part of the multicenter study, 104 of the 1200 (9%) gave birth to an SGA child at term. One newborn with a congenital syndrome was excluded. Of 103 eligible, 17 had moved or could not be traced. Hence, 86 were invited to participate, of whom 55 (24 men, 31 women) consented to participate in this part of the study (64% of 86 invited and 53% of 103 eligible participants). Group characteristics were mean (SD) birth weight: 2911 g (240), gestational age: 39.6 (1.2) weeks, assessment age: 19.8 (0.7) years and mean parental SES (n = 47): 3.5 (1.2). There were no significant differences compared with the control group in gestational age, assessment age or parental SES. One participant had CP (spastic bilateral subtype).
The control group
This group comprised 120 children recruited from the random sample. Two with a congenital syndrome were excluded. Of 118 eligible participants, 16 had moved or could not be traced. Hence, 74 of 102 invited subjects (31 men and 43 women) participated in this part of the study (73% of 102 invited and 63% of 118 eligible). Group characteristics were mean (SD) birth weight: 3716 g (473), gestational age: 39.8 (1.2) weeks, mean age at assessment: 19.7 (0.5) years and parental SES (n = 67): 3.6 (1.0).
Nonparticipants
There were no statistically significant differences in birth weight, gestational age and head circumference at birth between participants and those who did not consent to participate (data not shown). In the VLBW group, the proportion of men was higher among those who declined to participate (12/16 (75%) compared with those who participated (17/43 (40%), p = 0.015).
Outcome measures
Self-reported mental health was obtained by the Achenbach System of Empirically Based Assessment (ASEBA), Adult Self-Report (ASR, age range from 18–59) [
30], which comprises 123 problem items rated as “not true” (0), “somewhat or sometimes true” (1), or “very true or often true” (2). Co-occurring problems are clustered into the following Syndrome Profiles according to the manual: Anxious/Depressed, Withdrawn, Somatic Complaints (comprising the composite scale for Internalizing Problems), Aggressive Behavior, Rule-Breaking Behavior and Intrusive Behavior (comprising the Externalizing Problems composite scale), Thought Problems and Attention Problems. The Total Problems score is the sum of all problem items. Higher scores indicate more problems.
The Adult Autism Spectrum Quotient (AQ) (self-report) [
31] was used to assess traits within the autistic spectrum. The AQ consists of 50 items, scored 0 or 1, grouped into five different domains: social skills, attention switching, attention to detail, communication and imagination. The maximum total score is 50; higher scores indicate more autistic traits. A clinical cutoff value is set at a total AQ score ≥ 32 [
31]. A Norwegian pilot version was used with permission from the original author. We (the first author and a colleague) have since slightly revised and translated the AQ according to standard procedure, and this version is available at
http://www.autismresearchcentre.com. This questionnaire presumes normal intelligence. Analyses were performed only on those participants who had also been tested cognitively. As we wanted to explore results for our whole sample, scores are given both including and excluding those with cognitive disability (defined as IQ < 2 SD of mean IQ score) in the control group.
Health-related QoL was evaluated by the Short Form 36 (SF-36). This questionnaire comprises 36 statements summarized in eight transformed subscales, ranging from 1–100 [
32]: mental health (5 items), vitality (4 items), social functioning (2 items) and role limitations due to emotional problems (3 items), general health (5 items), physical functioning (10 items), bodily pain (2 items) and role limitations due to physical problems (4 items). Higher scores indicate favorable health outcomes. Psychometric properties have been evaluated in a Norwegian representative sample and Norwegian normative data are available [
33].
Self-esteem was evaluated through the Self-Perception Profile for Adolescents-Revised (SPPA-R) [
34]. SPPA-R consists of 35 statements rated from 1–4, divided into seven domain-specific subscales: close friends, school competence, social acceptance, athletic competence, physical appearance, romantic appeal and global self-worth. Higher scores indicate positive self-perceptions. The psychometric properties of SPPA-R have been evaluated in a Norwegian national representative sample for ages 13–19 years, and Norwegian norms are available [
34]. Analyses were conducted with and without individuals with CP.
Information on relations to friends and family was obtained by ASR Adaptive scores. Questions on friends included number of close friends, frequency of contact with friends, getting along with friends, visits from friends, rated 0–3. A total score for relation to friends was calculated with a maximum score of 12, higher scores indicate better adaptive functioning. Information on family relations included the self-perceived relationship with parents and/or siblings compared with others, and was rated worse (0), average (1) or better (2). In addition, participants were asked study-specific questions about current housing and daily occupation.
The Wechsler Adult Intelligent Scale (WAIS III) [
35] was administered by a trained neuropsychologist to measure the participants’ full-scale intelligence quotient (IQ) based on age-appropriate US norms. We obtained data on IQ for 41 VLBW, 53 SGA and 73 control participants. Both low birth weight groups had significantly lower IQ scores than controls; results are presented in the Additional file
1: Table S1.
Parents’ socioeconomic status (SES) was recorded according to Hollingshead’s two-factor index of social position rated from 1 (lowest) to 5 (highest) based on the parent’s education and occupation [
36]. The SES score was calculated at 14 years of age and supplemented with data at 20 years. Information on SES was available for 41 VLBW, 47 SGA and 67 control participants.
Statistics
The groups were compared using one-way ANOVA and Scheffe’s post hoc test for normally distributed data and Kruskal–Wallis test and Mann–Whitney test for data not normally distributed. Comparisons of proportions were made by Pearson’s chi-squared test or the exact unconditional z-pooled test. Normality of the ASR variable was checked by visual inspection of Q–Q plots. Although ASR Total Problems score was not normally distributed, the square root transformed score was approximately normally distributed. This score was consequently used in linear regression analysis to compare the ASR Total Problems score in the low birth weight groups with the control group as reference. Potential confounders (assessment age, SES (parental) and sex) and IQ were each adjusted for separately.
Missing items were handled according to the manual for ASR and SF-36. In ASR, one subject in the VLBW group had a missing item on “friends visit” and ASR problem scores were not calculated for one subject in the VLBW and one in the control group because of missing items. On SF-36, one subject in the control group had too many missing items to be included. Data were missing on AQ for three subjects in the control group. Missing items were handled by simple mean imputation on AQ and SPPA-R. On AQ, subscales were only calculated if fewer than five items were missing in total [
37] and if ≥ half of items in each subscale were completed. On the SPPA-R, subscales were only calculated if ≥ 80% was completed.
Cognitive disability was defined as IQ < 2 SD of the mean IQ score in the control group (i.e. IQ < 77.14). Cognitive disability was found in nine of 41 (22%) VLBW and four of 53 (8%) SGA participants, but in none of the control group.
We used SPSS version 17.0 (PASW statistics) (SPSS Inc., Chicago, IL) for data analysis. A two-sided p-value < 0.05 was considered significant and 95% confidence intervals are given when appropriate. Due to the large number of comparisons p-values between 0.01 and 0.05 should be interpreted with caution.