Prospective (Not Retrospective) Studies of Childhood Maltreatment and Adult Chronic Disease

What the evidence says

The strongest signal in this claim set is that how maltreatment exposure is measured changes what turns up. Newbury (2017) followed a birth cohort with informant-based exposure reports collected prospectively at ages 5, 7, 10, and 12, then checked psychiatric status at 18. Prospective reports of maltreatment increased the odds of depression (OR 2.37 unadjusted, OR 1.81 after adjusting for the corresponding retrospective report) and self-injury (OR 2.42 unadjusted, OR 1.70 adjusted) in early adulthood. But across all five psychiatric outcomes examined, prospective reports produced smaller effect sizes than retrospective self-reports of the same exposures, and several prospective associations shrank toward the null once retrospective reports were controlled for in the same model. Eldeeb (2026) reinforces the pattern from a different angle: in a cohort followed from ages 18 months–6 years to ages 36–46, prospective parent reports of physical abuse were uncorrelated with adult depression (r = 0.04–0.07), while adult self-report of the same abuse, measured concurrently with the depression outcome, was the strongest correlate (r = 0.26; a one-unit increase in this measure raised depression scores by 4.41 points). A latent construct pooling all reporter types across the 40+-year follow-up still increased depression scores (0.22 points per unit), but adult self-reports loaded far more heavily onto that construct (β = 0.75–0.78) than prospective parent reports did (β = 0.26–0.29).

Biological-marker studies (Lang, 2019) point in the same associational direction for cellular aging and DNA methylation but with thinner sample sizes and lower confidence. Exposure to two or more types of violence associated with shortened telomere length in children (n = 39), and prenatal maternal stress from a natural disaster associated with differential methylation at 957 gene loci in offspring (n = 36), predominantly immune-related. A reduction in harsh parenting, tracked longitudinally in a larger sample (n = 399), associated with reduced epigenetic cellular aging. Pain exposure in the neonatal intensive care unit associated with a trend toward greater telomere erosion in very preterm infants, though this did not reach conventional statistical thresholds across the full sample.

At the molecular level, Radtke (2015) found childhood maltreatment positively correlated with methylation at cg17860381 in the glucocorticoid receptor gene promoter, and this methylation combined with maltreatment history associated with increased overall psychopathological symptoms. For borderline personality disorder symptoms specifically, maltreatment (β = 0.5, p ≤ .01) and cg17860381 methylation (β = 0.5, p ≤ .01) each contributed independently, with no interaction term reaching threshold — an additive rather than multiplicative relationship.

For clinical outcomes, Trotta (2015) followed first-presentation psychosis patients for one year after a retrospective childhood-adversity assessment. Seventy-one percent of the 285 patients reported at least one adversity type (physical abuse, sexual abuse, parental separation, parental death, disrupted family arrangements, or being taken into care). Parental separation specifically increased the odds of longer psychiatric hospitalization (≥49 days) by roughly two-fold (p = .012) and increased the odds of medication noncompliance by roughly two-fold. No adversity measure showed a measured effect on illness course or remission over the same year.

Across the underlying literature, prospective designs remain the minority: of 37 studies in the review underlying the Lang claims, only 9 gathered exposure data during childhood itself, and only 7 of those were longitudinal. Most of the field's evidence on maltreatment and adult outcomes still relies on retrospective adult recall, which the claims above show produces different — generally larger — effect estimates than prospective measurement of the same exposures.

Where the evidence splits

Newbury and Eldeeb disagree on whether prospectively measured childhood maltreatment predicts adult depression at all. Newbury's prospective informant reports (ages 5–12, outcome at 18) increased the odds of depression (OR 1.81, adjusted). Eldeeb's prospective parent reports (ages 18 months–6 years, outcome at 36–46) were uncorrelated with adult depression (r = 0.04–0.07). The claims point to design differences that could account for this: Newbury measured general maltreatment with a shorter follow-up to early adulthood, while Eldeeb measured physical abuse specifically with a 40+ year follow-up and modeled prospective and retrospective reports jointly through a latent construct in which retrospective self-report dominated. Neither source explains the gap directly against the other, so this split is not resolved by the claims themselves — only the design contrasts they separately report.

Thresholds and dose-response

Two claims establish a specific exposure count at which an effect is reported. Exposure to two or more types of violence in childhood associates with shortened telomere length in children (Lang, 2019, n = 39). Separately, two or more prospectively reported forms of maltreatment lose their independent association with depression, anxiety, self-injury, and alcohol/cannabis dependence once the corresponding retrospective report is added to the model (OR falls to 1.44, 95% CI 0.65–3.20, for depression) — the same "two or more" threshold that shows an effect in the unadjusted prospective analysis does not hold up once retrospective report is controlled for (Newbury, 2017).

Open empirical questions

• None of the sources here measure adult physical chronic disease endpoints (cardiovascular disease, diabetes, metabolic syndrome) — the biomarker findings (telomere length, methylation) are not linked in these claims to a downstream disease diagnosis.
• Why prospective parent-reported physical abuse shows no measured correlation with adult depression (Eldeeb) while prospective informant-reported general maltreatment increases depression odds (Newbury) is not explained by either source.
• Whether the telomere and methylation changes reported in children persist into, or predict, adult outcomes is not measured in the current sources — all telomere/methylation claims are child- or infant-stage measurements.
• Whether the psychosis-outcome associations (hospitalization, noncompliance) found in a first-presentation South London sample generalize to other populations or hold beyond one year of follow-up is not measured here.