Preterm Birth and Adult Intelligence Outcomes

The Eves et al. (2021) individual participant data (IPD) meta-analysis published in JAMA Pediatrics answers a question that single cohorts have not been able to answer cleanly: do the cognitive deficits associated with very preterm (VPT, <32 weeks) or very low birth weight (VLBW) birth persist into full adult cognitive maturity, and which neonatal and social risk factors carry the bulk of the variance? Pooling 8 longitudinal birth cohorts from the RECAP-Preterm and APIC consortia (total N = 2,135 adults; mean age 24.6 ± 4.3 years), the analysis confirms a substantial and durable IQ disadvantage—on average 12 IQ points or 0.78 standard deviations below term-born peers—and identifies five risk factors whose effects partition cleanly within the VPT/VLBW group.

What earlier meta-analyses had established

The clinical case that preterm birth is associated with cognitive deficits has been settled since at least Bhutta, Cleves, Casey, Cradock, and Anand’s (2002) JAMA meta-analysis of school-age outcomes. That synthesis covered 17 case-control studies of children born preterm assessed at ages 5-16 and documented deficits in IQ (typically 0.5-1.0 SD below term-born controls), as well as elevated rates of attention, behavioral, and academic problems. The Bhutta paper made preterm-related cognitive impairment a clinically recognized condition rather than an intermittent finding.

The follow-up question—whether neonatal-care advances since the 1990s had attenuated those deficits—was addressed by Twilhaar et al. (2018) in JAMA Pediatrics. Their meta-analysis of children born extremely (<28 weeks) or very preterm since 1990 found that, despite substantial improvements in survival and short-term morbidity, the cognitive deficit at school age has remained essentially stable at approximately 0.86 SD below term-born controls. Better neonatal care kept more very-preterm infants alive without proportionally reducing the cognitive sequelae among survivors.

The remaining open question—whether the school-age deficits decay, persist, or amplify into adulthood—is what Eves et al. (2021) settle.

The IPD meta-analysis design

Individual participant data meta-analysis is the gold-standard form of evidence synthesis when feasible: rather than combining published effect sizes (which are constrained by what original authors chose to report), it harmonizes raw participant-level data across cohorts and re-analyzes the pooled dataset under a common analytical protocol. Eves et al. drew from the RECAP-Preterm and APIC research networks—international consortia of preterm birth cohorts established in the 1980s and 1990s—and identified 8 cohorts with available adult IQ data (born 1978-1995, assessed in adulthood between approximately 2000 and 2019).

The harmonized sample comprised 1,068 VPT (<32 weeks) or VLBW (<1,500 g) participants and 1,067 term-born controls, totaling 2,135 adults. Mean age at outcome assessment was 24.6 ± 4.3 years, well into the period of stable adult cognitive functioning. Full-scale IQ scores from age-appropriate Wechsler and other validated batteries were converted to z-scores within each cohort to permit pooled analysis. Analyses adjusted for cohort, sex, and—where available—socioeconomic and educational covariates.

The headline IQ gap

Pooled across cohorts, adults born VPT or with VLBW scored 0.78 standard deviations (approximately 12 IQ points) below term-born adults at mean age 24.6 years. The magnitude is essentially unchanged from the school-age gaps documented by Bhutta et al. (2002) and Twilhaar et al. (2018), which means: the cognitive disadvantage does not narrow with development. Children with neonatal-related cognitive impairments do not, on the population scale, catch up to term-born peers as they mature into adulthood.

The aggregate-meta complement to the IPD analysis (drawing on 13 published studies) produced an effect size in the same range, indicating that the IPD findings are not artifacts of which cohorts contributed individual-level data.

What predicts variation within the VPT/VLBW group

The most actionable part of the analysis is the partitioning of risk within the VPT/VLBW sample. Five factors emerged as reliably associated with adult IQ; their standardized regression coefficients (within-VPT/VLBW models) are:

• Lower gestational age: β = 0.11 SD per week earlier (95% CI: 0.07-0.14). A baby born at 26 weeks scores, on average, 0.66 SD lower in adulthood than one born at 32 weeks, all else equal.
• Lower birth weight z-score: β = 0.21 SD per 1.0 SD lower (95% CI: 0.14-0.28). The within-gestational-age component of small size carries its own independent effect.
• Bronchopulmonary dysplasia (BPD): β = −0.16 SD (95% CI: −0.30 to −0.02). Chronic lung disease of prematurity, even when survived, predicts a meaningful adult IQ decrement.
• Intraventricular hemorrhage (IVH, any grade): β = −0.19 SD (95% CI: −0.33 to −0.05). The magnitude here likely understates the effect for severe IVH, which the analysis combines with milder grades.
• Lower maternal educational level: β = 0.26 SD per education-category step (95% CI: 0.17-0.35). The single largest standardized effect among the five identified factors.

The maternal-education finding is the least intuitive and the most policy-relevant. It indicates that the social environment surrounding the VPT/VLBW infant is at least as predictive of adult cognitive outcomes as the neonatal medical complications traditionally treated as the primary risk factors. This is consistent with the broader nurturing-care literature, but its specificity to preterm cohorts has been less clearly established before.

Comparison with the broader early-adversity cognitive literature

The 0.78 SD adult IQ gap reported by Eves et al. (2021) is, by the standards of cognitive epidemiology, a large effect. It exceeds the typical adult cognitive consequences of moderate prenatal alcohol exposure (≈0.3-0.5 SD), childhood lead exposure at currently regulated levels (≈0.1-0.2 SD), or moderate household poverty without other adversities (≈0.3-0.5 SD). It is comparable in magnitude to the effects of severe early-childhood institutional deprivation, severe iodine deficiency, or fetal alcohol syndrome (rather than fetal alcohol spectrum disorder). In other words, the cognitive consequence of being born very preterm or with very low birth weight, in the absence of any other identifiable adversity, places the affected individual in the same general magnitude range as those who experienced major early-life environmental insults.

What distinguishes the preterm/VLBW case from those comparators is that the proximate cause—premature departure from gestation, with attendant immature organ development—is medical rather than environmental, and most of the modifiable risk factors operate either before birth (preventing preterm delivery) or in the immediate neonatal period (NICU care quality). The maternal-education finding adds a postnatal social dimension that places the preterm/VLBW case partly within the broader nurturing-care framework documented in middle-income-country cohorts and that suggests the social environment is not merely a separate confound but an active modifier of the medical-origin cognitive trajectory.

What this changes about cognitive prognosis after preterm birth

The clinical translation of these findings depends on three of their features. First, the persistence of the gap into adulthood means that early-childhood cognitive trajectories are predictive of adult outcomes; the school-age IQ measurement is not an under-estimate of adult risk. Second, the biological and social factors are independently predictive, which means that targeting either category alone leaves substantial preventable variance unaddressed. Third, the maternal-education effect is the largest single factor, suggesting that social-determinants interventions—home visiting programs, structured early-learning environments, parental education support—may yield larger marginal returns than further refinements to neonatal medical care.

For families and clinicians, the operational implication is that VPT/VLBW status should be treated as a long-term cognitive-risk factor that warrants developmental monitoring through at least the early school years, and ideally into adolescence, with parallel attention to the social environment of the home as a modifiable amplifier of either resilience or risk.

What the analysis cannot establish

Three limitations qualify the interpretation. First, the cohorts contributing to the IPD pool were born between 1978 and 1995, a period that predates several major neonatal-care advances (surfactant therapy became standard in the early 1990s; less-invasive ventilation strategies emerged in the 2000s). Whether cohorts born after 2000 will show the same adult-IQ trajectory is not yet established, though Twilhaar et al.’s (2018) finding that school-age deficits have not narrowed across the 1990s suggests the persistence is robust.

Second, the IPD meta-analysis treats VPT (<32 weeks) and VLBW (<1,500 g) as a combined category. These conditions overlap heavily but are not identical; some VLBW infants are not preterm (small-for-gestational-age at term), and the analyses that pool them assume the cognitive consequences are similar across the two routes to neonatal vulnerability. Subgroup decomposition is feasible with the IPD data but is not the focus of the 2021 paper.

Third, the maternal-education finding, while statistically robust, may absorb confounding with other socioeconomic variables (household income, neighborhood resources, parental engagement) that the harmonization procedure could not consistently separate across cohorts. The maternal-education effect is best read as a proxy for the broader social environment rather than as a specific causal claim about formal education per se.

The trajectory the field is on

The progression Bhutta (2002) → Twilhaar (2018) → Eves (2021) represents one of the cleaner cumulative-evidence stories in pediatric cognitive epidemiology. Each successor study answered a more specific version of the question its predecessor opened: from “are there cognitive deficits at school age?” (yes, ~1 SD), to “have those deficits been reduced by neonatal-care advances?” (no, they have remained ~0.9 SD), to “do they persist into adulthood?” (yes, ~0.8 SD with identifiable modifiable predictors).

The next research-design step suggested by Eves et al. (2021) is intervention rather than further descriptive cohort work. The risk factors they identify—maternal education, BPD, IVH, gestational age, birth weight—differ in modifiability. Maternal education and the broader social environment are most amenable to public-health intervention; BPD is amenable to neonatal pulmonary-care optimization; IVH is in part preventable by careful neonatal hemodynamic management. Whether targeted intervention on any of these factors would attenuate the adult IQ gap is the experimental question that the descriptive epidemiology now positions the field to ask.

Frequently asked questions

How much lower is adult IQ after very preterm or very low birth weight birth?

The Eves et al. (2021) IPD meta-analysis of 8 cohorts (N = 2,135) found adults born very preterm (<32 weeks) or with very low birth weight (<1,500 g) scored on average 0.78 standard deviations—approximately 12 IQ points—below term-born peers at mean age 24.6 years. The aggregate-meta complement using 13 published studies produced an effect size in the same range.

Do preterm-related cognitive deficits diminish with age?

No. The 0.78 SD adult gap reported by Eves et al. (2021) is essentially unchanged from the school-age gaps documented by Bhutta et al. (2002, ~0.5–1.0 SD) and Twilhaar et al. (2018, ~0.86 SD). Children with neonatal-related cognitive impairments do not, on the population scale, catch up to term-born peers as they mature into adulthood.

What factors predict adult cognitive outcomes within preterm survivors?

Five factors emerged with reliable associations. Lower gestational age (β = 0.11 SD/week earlier) and lower birth weight z-score (β = 0.21 SD/SD lower) predict worse outcomes; bronchopulmonary dysplasia (β = −0.16 SD) and intraventricular hemorrhage (β = −0.19 SD) carry independent decrements; lower maternal educational level (β = 0.26 SD/category) is the largest single standardized effect.

Why is maternal education such a strong predictor?

Maternal education is best read as a proxy for the broader social environment surrounding the preterm infant—home learning resources, parental engagement, household income, neighborhood resources—rather than as a specific causal claim about formal schooling. Its effect indicates that the social environment is at least as predictive of adult cognitive outcomes as the neonatal medical complications traditionally treated as the primary risk factors.

Has neonatal care reduced these cognitive deficits over time?

Twilhaar et al. (2018) found that despite substantial improvements in survival and short-term morbidity since 1990, the cognitive deficit at school age has remained essentially stable. Better neonatal care has kept more very preterm infants alive without proportionally reducing the cognitive sequelae among survivors.

What does this mean for clinical follow-up?

Very preterm or very low birth weight status warrants long-term cognitive-risk monitoring through at least the early school years and ideally into adolescence. Because biological and social factors are independently predictive, follow-up should attend to both neonatal medical history and the home social environment as a modifiable amplifier of either resilience or risk.