Slower Reaction Times and Impaired Learning in Young Adults With Birth Weight <1500 g
OBJECTIVE: Children with very low birth weight (VLBW; <1500 g) perform worse on cognitive tests than do children who are born at term. Whether this difference persists into adulthood has been little studied. We assessed core neurocognitive abilities (processing speed, working memory, attention, and learning capacity) in young adults with VLBW and in term-born control subjects.
METHODS: In conjunction with the Helsinki Study of Very Low Birth Weight Adults, 147 VLBW and 171 control subjects who were aged 18 to 27 years and did not have neurosensory impairments performed a computerized test battery (CogState Ltd, Melbourne, Australia). T tests and linear regression models were used. Cohen's d was used to express effect size (ES).
RESULTS: VLBW adults had slower reaction times than did control subjects on all 5 tasks: simple reaction time (mean difference: 4.0% [95% confidence interval (CI): 1.1%–7.0%]; ES: 0.30), choice reaction time (mean difference: 3.2% [95% CI: 0.3%–6.2%]; ES: 0.24), working memory (mean difference: 8.4% [95% CI: 3.7%–13.4%]; ES: 0.40), divided attention (mean difference: 7.2% [95% CI: 2.7%–11.9%]; ES: 0.36), and associated learning reaction time (mean difference: 6.4% [95% CI: 1.3%–11.9%]; ES: 0.28). In addition, VLBW adults showed impaired learning abilities on the associated learning task (percentage of correct responses: 85.7 vs 80.2; P < .001; ES: 0.64). The results were little affected by adjustment for confounders.
CONCLUSIONS: Nonimpaired VLBW individuals exhibited slower psychomotor speed and lower accuracy on the associated learning task. These results indicate that very preterm birth, even when obvious neurosensory deficits are absent, may have long-term consequences on core neurocognitive abilities.
Approximately 0.8% to 1.5% of infants in high-income countries are born at very low birth weight (VLBW; birth weight <1500 g),1,2 amounting to, for example, 62000 annual births in the United States and 46000 in the European Union.3 The development of neonatal intensive care has lead to remarkable improvements in the survival of VLBW infants, from ∼40% in the 1960s to almost 90% today4,5; therefore, their long-term prognosis is of concern for health professionals, the VLBW individuals themselves, and society as a whole.
Preterm VLBW infants are deprived of the last weeks or months of intrauterine life, a period of rapid development of the fetal brain.6 Previous reports on their long-term cognitive abilities have similarly indicated lower mean IQ and reduced scholastic performance among children who were born preterm.7 Some studies reported that there might be some catchup of cognitive abilities during childhood.8 Little is reported, however, about the adult cognitive outcomes of prematurity,9–12 because the first survivors of modern neonatal intensive care have not entered adulthood until recently.
Our aim was to study core neurocognitive abilities in ostensibly unimpaired young adults with VLBW and control subjects who were born at term. We used a computerized test battery that focused on visual attentional and memory domains including psychomotor processing speed, working memory, attention, and associated learning. Our secondary aim was to study the association between test performance and perinatal risk factors that are associated with severe prematurity.
A flowchart in Fig 1 depicts the Helsinki Study of Very Low Birth Weight Adults, described in detail previously.13–15 Briefly, it consists of 166 VLBW young adults and 172 term control subjects who were born between 1978 and 1985 in a geographically defined region in the province of Uusimaa, Finland. During 2004 and 2005, at the ages of 18 to 27 years, the study participants underwent a comprehensive clinical assessment of their physical and psychological health.
Nonparticipants did not differ from study participants with regard to neonatal variables (birth weight, birth weight SD score, gestational age, presence of maternal preeclampsia, bronchopulmonary dysplasia, days of mechanical ventilation, oxygen treatment, or days at discharge from the NICU), although in the VLBW group, participants had a lower incidence of cerebral palsy as evaluated at 15 months of age (6.0% vs 19.1%; P[r] = .005).13 In addition, in the VLBW group, maternal smoking during pregnancy was more common among nonparticipants (31.8% vs 19.9%; P = .04).
Of the 338 study participants who attended the clinic, 5 VLBW adults did not complete the cognitive assessment (3 with severe visual impairment, 1 with moderate mental retardation, and 1 due to equipment-related technical problems). Because we aimed to compare cognitive abilities explicitly among unimpaired VLBW adults with their term-born peers, we excluded from the analyses an additional 15 participants with neurosensory impairments (1 with deafness and 14 with cerebral palsy, 1 of whom also had developmental disability). Thus, we included 318 participants in the analyses (Fig 1), and their demographic and clinical characteristics are presented in Table 1. Of the VLBW individuals, 49 (33.3%) were born small for gestational age (SGA; birth weight SD score less than −2 SD) and 98 (66.7%) appropriate for gestational age (AGA; birth weight SD score within ±2 SD), according to Finnish growth charts.16 The study protocol was approved by the local ethics committee, and all participants signed an informed consent.
Perinatal characteristics of the study participants, as well as pregnancy-related characteristics of their mothers, were derived from hospital records. Maternal smoking history during pregnancy was obtained before delivery by self-report (yes/no), and preeclampsia was defined according to current criteria.17 Adult characteristics of the study participants were gathered in conjunction with the clinic study either by self-report (handedness, school history, current health, and parental educational attainment [calculated as the highest education achieved by either parent]) or from the clinical examination (head circumference).
We used a computerized cognitive test battery (CogState Ltd, Melbourne, Australia), designed to be sensitive and rapid and have good reliability with minimal practice effects (www.cogstate.com).18–21 This test battery targets major cognitive domains such as psychomotor function, attention, working memory, learning, and aspects of executive functions. The battery in this study consisted of 5 tasks that are based on playing cards presented on a computer screen, to be culture-neutral and independent of language as well as socioeconomic background of the subject. The tasks have proved sensitive in detecting mild cognitive impairment,22 as well as in detecting cognitive changes in athletes with concussion,23 people who are exposed to fatigue and alcohol,24 and patients with HIV infection.20 As compared with traditional paper-and-pencil neuropsychological tests, a computerized test has the advantage of being faster and more instructor-independent,25 in addition to causing fewer data entry errors and allowing for registration of reaction times. In this study, we used CogHealth 3.0.5, which consists of the following tasks:
Simple reaction time (SRT): In this psychomotor speed task, a playing card is shown face down on the computer screen, and the subject is instructed to press the “yes” key as quickly as possible whenever the card turns face up. This procedure is repeated randomly 35 times. The SRT task is repeated again as the final task of the test battery.
Choice reaction time (CRT): In this task that measures simple visual attention and psychomotor speed, a face-down card is shown on the screen. When the card flips over, the subject is asked to determine as quickly as possible whether the face-up card is red. If the card is red, then the subject should press the “yes” key, and if it is not red, then the subject should press the “no” key. This procedure is repeated 30 times.
One back working memory (WM): The third task of the test measures mainly working memory but also psychomotor speed and visual attention. One card at a time is shown on the screen, and the subject is asked to press as quickly as possible the “yes” key or the “no” key according to whether the card exactly matches the 1 presented just before. The procedure is repeated 30 times.
Divided attention (DA): This is a vigilance task that assessed divided attention, which is a somewhat more demanding dimension of visual attention, as well as psychomotor processing. Five cards are shown next to each other on the screen, with 1 horizontal white line above and another below the cards. The cards start moving randomly up and down, and the subject is asked to press the “yes” key as soon as a moving card touches a white line. Thus, the subject needs to monitor all 5 cards and both lines simultaneously. The trial is repeated 30 times.
Associated learning (AL): Measuring visual learning and declarative memory, this represents the most difficult task of the test. Five pairs of cards are shown in the top half of the screen, and additionally 1 random pair is shown below these 5 pairs, in the bottom half of the screen. When the random pair in the bottom half of the screen turns face up (a procedure repeated 50 times during the task), the subject has to decide whether these face-up cards match any of the pairs above by pressing the “yes” key or the “no” key. After each match, the matching pair turns face down, and from then on the subject has to recall the paired cards to produce correct responses. When a correct response is given, the pair of cards turns face up for a short while, enabling learning during the task.
The participants arrived at the clinic in the morning, and the cognitive testing was performed after participants had breakfast at the clinic. Before testing, all participants had at least 1 practice round, during which they were verbally instructed on each task by a research nurse who was blinded to birth status. Practice effects are seen from the first to second administration but rarely thereafter, so at least 1 practice test is recommended.21 During testing, the participants were alone at the computer, wearing headphones to achieve optimal concentration and to hear the error beep resulting from inaccurate responses. Preceding every task during the test, written instructions appeared on the computer screen and familiarization trials with both auditory and visual feedback on false/accurate responses were provided. Thereafter, the trials to be recorded for the task began and the text instructions disappeared. Total test time was ∼15 to 20 minutes. After completion of the test, the participants received a report on their own performance.
The test results were transmitted as encrypted data files to the CogState server for automatic data extraction and then combined with clinical data. Initially, the data were visually inspected and cleaned with identification of outliers and inconsistencies. The WM task results for 2 participants were excluded because their total errors on that task were >40, suggesting that they did not understand the task requirements. Similarly, the AL task result for 1 participant was excluded because it was very slow: mean reaction time exceeded 4500 milliseconds. SRT was calculated as the mean of the 2 identical SRT tasks. Four participants, however, had discontinued the second SRT task, and thus their SRT was based on the first SRT task only.
For all tasks, reaction times were recorded in milliseconds, and because of slightly positively skewed distributions, mean values of logarithm-transformed (base 10) reaction times were used as outcome variables. The accuracies (percentage of correct responses) for 4 of the tasks (SRT, CRT, WM, and DA) were negatively skewed with a ceiling effect, such that the majority of participants performed near the highest possible accuracy for the test. Because the reproducibility and utility of these accuracy variables with ceiling effects are equivocal (the task seemed too easy for the majority, making group differences unlikely), we decided not to use them as outcome variables in the analyses.19,23 An exception is the accuracy for the AL task, which in a previous and in this study had an acceptable distribution and lacked a ceiling effect and thus was used as an untransformed continuous outcome variable.19
T tests were used for group comparisons on continuous variables, and χ2 tests were used for comparisons on categorical variables. When adjusting for confounding factors (gender, age, parental educational attainment [serving as an index of socioeconomic status], and adult head circumference), linear regression models were fitted. The independent variables in the models were chosen mainly on theoretical grounds because of the associations of gender,26 age,27 parental educational attainment,8,28 and adult head size29 with cognitive abilities. When no group × gender interactions were found, the results are reported with men and women pooled together. Cohen's d effect size (ES) measures were calculated in order to demonstrate the relative magnitude of the group differences and to illustrate their clinical meaningfulness.30 Background variables as predictors of test performance were examined using Pearson's or Spearman's correlation coefficients and multiple linear regression with adjustment for gender and birth status. All analyses were performed with SPSS 16.0 for Windows (SPSS, Inc, Chicago, IL). For avoiding type I error because of multiple testing, the α level was set to .01.
Cognitive Performance Among VLBW and Term-born Young Adults
The VLBW group had consistently longer reaction times than the term-born group on all tasks (unadjusted P < .04; Table 2, Fig 2). The VLBW group also performed worse with regard to the accuracy of test performance on the AL task (P < .001; Table 2). Within the VLBW group, individuals who were born SGA and AGA did not significantly differ from each other (unadjusted P > .30).
When adjusting for gender and age at examination in model 1 and additionally for parental educational attainment in model 2, the results remained essentially the same (Table 3). After additional adjustment for adult head circumference in model 3, the results for the WM and AL accuracy tasks remained unchanged, whereas the others were slightly attenuated (SRT, CRT, DA, AL). Again, within the VLBW group, SGA and AGA did not significantly differ from each other. No interaction effects between gender × group were found (P > .13), indicating that the direction of the difference between VLBW and control groups were similar in men and women. The ESs (Cohen's d) for the group differences ranged between 0.24 and 0.63, which are regarded as small-sized to medium-sized effects (Table 2).30
Background Variables as Predictors of Test Performance
Table 1 shows the background characteristics of the participants. Men had shorter reaction times on some tasks (on the SRT, WM, DA, and AL tasks in the control group [r = 0.21–0.36, P ≤ .006] and on the WM task in the VLBW group [r = 0.24, P = .003]) and higher error rates on some (on the SRT and WM in the control group; P ≤ .003 for trend), whereas the AL accuracy did not differ between genders. Older age was related to longer reaction times on the AL task, and maternal smoking during pregnancy was associated with shorter reaction times on the CRT and WM tasks. One school grade higher corresponded to 1.7 percentage points higher AL accuracy (P = .01, adjusted for gender and birth status). With regard to current smoking, handedness, parental educational attainment, firstborn order, and maternal preeclampsia during pregnancy, no association with test performance was found. Among the VLBW group, few neonatal factors were associated with test performance: only a longer duration of mechanical ventilation was related to longer reaction time on the DA task (r = 0.27, P = .001), whereas bronchopulmonary dysplasia, retinopathy of prematurity, and blood culture–verified sepsis were unrelated to test performance.
Our findings show that even in the absence of neurosensory impairments, VLBW birth constitutes a risk factor for slower psychomotor processing speed and impaired visual learning abilities in young adulthood. We found few associations with other perinatal risk factors. For example, there was no evidence of an effect of intrauterine growth retardation (for which SGA- status served as a proxy) on these cognitive outcomes over and above the effect of preterm birth.
The association of low birth weight and preterm birth with cognitive impairments in childhood was recognized long ago.31 In 1936, Brander32 reported reduced IQ among 7- to 15-year-old children with birth weight <2500 g. Since the early years of prematurity research, a substantial number of follow-up studies on children born preterm have been published.7 In contrast, studies that extend into adulthood are remarkably scarce.9,10,33 In this study, our main aim was to examine adult outcomes of severe prematurity. The study outcomes—processing speed and learning capacity—represent core cognitive abilities that are likely to be required for successful performance, for instance, at school, at work, and in everyday life.
In the study by Hack et al,9 20-year-old VLBW young adults had lower IQ and academic achievement scores than did control subjects, even after exclusion of individuals with neurosensory impairments or subnormal IQ. Adults who were born preterm, even those without impairments, attain on average a lower educational level than their term-born counterparts.34 These differences may have enduring ramifications for the society. For example, low IQ and low educational attainment in young adult life are important predictors of adverse health outcomes, including all-cause mortality.35
Our findings also parallel previous observations of impaired visual memory, visual processing, and learning in 14-year-old VLBW adolescents.28 Furthermore, the findings are in agreement with those by Rose and Feldman,36 who noted slower processing speed and poorer performance on 4 computerized memory tasks in 11-year-old VLBW children than in control subjects. Nosarti et al12 reported impairments in aspects of executive functioning skills, including processing speed, in healthy 22-year-olds who were born very preterm (<33 weeks of gestation). No association between perinatal variables and executive functions were found. Imaging studies provide support for a relationship between impaired executive functions and white matter abnormalities in very preterm preschool children and VLBW adolescents.37,38
The strengths of the study lie in the large and well-characterized study sample; the good return rate; and the use of a sensitive, validated, and computerized test. Furthermore, the extensive background data available allowed exploration of the effects of perinatal factors and current health on cognitive abilities.
The study is limited by the absence of full-scale IQ of the participants, which excludes the possibility to control for general intelligence. Lacking cognitive data of the parents, we are unable to explore a possible common genetic background for prematurity and cognitive functioning. Today's neonatal intensive care differs from that experienced by our cohort; therefore, the results may not be directly applicable to preterm individuals who were born more recently. Moreover, as discussed by Collie et al,18 individuals who were familiar with video games may have been advantaged when performing the test, although this would introduce bias only if the proportion of these individuals differed between the groups. All tasks were visual. Ceiling effects precluded use of accuracies of tasks other than AL as outcome variables. Finally, we cannot exclude participation bias, although participants and nonparticipants were similar on most perinatal variables.
As a guide to the clinical significance of the reported cognitive impairment, the magnitude of the group difference was equivalent to a change in test performance caused by a blood alcohol concentration of 0.02% to 0.03%.24 The ESs were moderate, although we emphasize that we studied high-functioning VLBW adults who were born in a high-income country. Nevertheless, even moderate effects may be of major clinical importance on a population level if the outcome measured is critical for a person's life.39 Given the impact of cognitive abilities on, for instance, occupational success and mortality, the outcomes measured here can indeed be regarded as critical.39 It is interesting that in a study that examined the association between IQ and mortality while controlling for confounding factors, Deary and Geoff40 found that information processing efficiency at 56 years of age (simple and 4-choice reaction times) was an even better predictor of survival until age 70 than was IQ. Thus, it is plausible that impairments in reaction times, which may reflect fundamental functions of the brain such as information processing speed, are of importance.
Being born at VLBW, even when neurosensory impairments are absent, is associated with subtle but significant neuropsychologic deficits that last into young adulthood. The results highlight the need for keeping a low threshold for providing support, such as special educational services, to this particular risk group.
This study was funded by grants from the Academy of Finland; the Biomedicum Helsinki Foundation; the Finnish Concordia Foundation; Finska Läkaresällskapet; the Finnish Medical Society Duodecim; the Finnish Foundation for Pediatric Research; the Finnish Special Governmental Subsidy for Health Sciences; the Finnish National Graduate School of Clinical Investigation; the Jalmari and Rauha Ahokas Foundation; the Juho Vainio Foundation; the Novo Nordisk Foundation; the Päivikki and Sakari Sohlberg Foundation; the Pediatric Graduate School and the Clinical Graduate School in Paediatrics and Obstetrics/Gynaecology, University of Helsinki; the Perklén Foundation; the Research Foundation for the Orion Corporation; the Signe and Ane Gyllenberg Foundation; the Sigrid Juselius Foundatiom; the Waldemar von Frenckell Foundation; Vasa Nation at Helsinki University; the Wilhelm and Else Stockmann Foundation; and the Yrjö Jahnsson Foundation.
We owe our sincere gratitude to all participants of the study; to research nurses Paula Nyholm, Anne Kaski, Hilkka Puttonen, and Marita Suni; and to Sigrid Rosten for data management.
- Accepted August 7, 2009.
- Address correspondence to Sonja Strang-Karlsson, MD, Children's Hospital, Biomedicum 2, Tukholmankatu 8C, Helsinki University Central Hospital, PO Box 705, FI-00029, Helsinki, Finland. E-mail:
Financial Disclosure: Financial Disclosure: Dr Darby is a shareholder and an employee of CogState Ltd; the other authors have no financial relationships relevant to this article to disclose.
What's Known on This Subject:
Children born with VLBW gain on average lower scores on cognitive tests. Whether such deficits are compensated for in adulthood has been little studied.
What This Study Adds:
Unimpaired young adults born severely preterm have on average longer reaction times and impaired learning abilities compared with term-born control subjects. This underlines a need for keeping a low threshold for support for children and adolescents born preterm.
- ↵National Research and Development Centre for Welfare and Health: Parturients, deliveries and births 2007. Available at: www.stakes.fi/EN/tilastot/statisticsbytopic/reproduction/parturients.htm. Accessed March 8, 2009
- ↵March of Dimes. PeriStats. Available at: www.marchofdimes.com/peristats. Accessed March 8, 2009
- ↵Euro-Peristat. European perinatal health report. Available at: www.europeristat.com/publications/european-perinatal-health-report.shtml. Accessed March 8, 2009
- ↵Lee KS, Kim BI, Khoshnood B, et al. Outcome of very low birth weight infants in industrialized countries: 1947–1987. Am J Epidemiol.1995;141 (12):1188– 1193
- ↵Allin M, Walshe M, Fern A, et al. Cognitive maturation in preterm and term born adolescents. J Neurol Neurosurg Psychiatry.2008;79 (4):381– 386
- ↵Kajantie E, Hovi P, Räikkönen K, et al. Young adults with very low birth weight: leaving the parental home and sexual relationships—Helsinki Study of Very Low Birth Weight Adults. Pediatrics.2008;122 (1). Available at: www.pediatrics.org/cgi/content/full/122/1/e62
- ↵Cysique LA, Maruff P, Darby D, Brew BJ. The assessment of cognitive function in advanced HIV-1 infection and AIDS dementia complex using a new computerised cognitive test battery. Arch Clin Neuropsychol.2006;21 (2):185– 194
- ↵Darby D, Maruff P, Collie A, McStephen M. Mild cognitive impairment can be detected by multiple assessments in a single day. Neurology.2002;59 (7):1042– 1046
- ↵Moriarity J, Collie A, Olson D, et al. A prospective controlled study of cognitive function during an amateur boxing tournament. Neurology.2004;62 (9):1497– 1502
- ↵Collie A, Maruff P. Computerized neuropsychological testing. Br J Sports Med.2003;37 (1):2– 3
- ↵Machin S, Pekkarinen T. Assessment: global sex differences in test score variability. Science.2008;322 (5906):1331– 1332
- ↵Thalheimer W, Cook S. How to calculate effect sizes from published research articles: a simplified methodology; August 2002. Available at: http://work-learning.com/effect_sizes.htm. Accessed November 21, 2008
- ↵Ylppö A. Brain and neural defects caused by delivery in preterm infants [in Finnish]. Duodecim.1920;36 :171– 181
- ↵Brander T. Studies on development of intelligence in preterm children. A contribution to the knowledge about the origin of particularly lighter degrees of exogenously conditioned subnormal intelligence [doctoral thesis; in German]. Helsinki, Finland: Akademische Abhandlung aus der Universitäts-Kinderklinik in Helsingfors; 1936
- ↵Ericson A, Källén B. Very low birthweight boys at the age of 19. Arch Dis Child Fetal Neonatal Ed.1998;78 (3):F171– F174
- ↵Roberts BW, Kuncel NR, Shiner R, Caspi A, Goldberg LR. The power of personality: the comparative validity of personality traits, socioeconomic status, and cognitive ability for predicting important life outcomes. Perspect Psychol Sci.2007;2 (4):313– 345
- ↵Deary IJ, Geoff D. Reaction time explains IQ's association with death. Intelligence.2005;16 (1):64– 69
- Copyright © 2010 by the American Academy of Pediatrics