Childhood adversity and vagal regulation: A systematic review and meta-analysis

The first of the three RSA meta-analyses queued in raw/papers/Polyvagal/ to reach the wiki, and the one that asks the sharpest question of the vagal-psychophysiology literature: does one of the field’s most robust developmental risk factors — childhood adversity — actually move vagal tone? Two separate three-level meta-analyses, 115 and 75 articles, ~27,000 participants pooled. The answer, at the population level, is no.

What was tested

Vagal regulation is read here through vagally-mediated heart-rate variability (vmHRV) — RSA, RMSSD, HF-HRV — split into two constructs both sides of the vagal literature agree on:

• Baseline vagal activity — resting parasympathetic tone; higher is generally taken to reflect homeostatic capacity and regulatory resource.
• Vagal reactivity — the change in vmHRV from rest to a challenge (cognitive task, emotion-eliciting film, stress test); vagal withdrawal is the canonical adaptive response, so its reduction is read as poor flexible engagement.

The theoretical warrant for expecting childhood adversity (CA) to lower both is drawn from Polyvagal Theory (Porges 2003, 2007) and the Neurovisceral Integration Model (Thayer & Lane 2000; Thayer is an author) — adversity is held to disrupt the cortico-limbic circuits that control the heart via the vagus, leaving a downstream vmHRV signature. The meta-analysis is a test of that downstream prediction, not of PVT’s brainstem anatomy (see is-polyvagal-theory-valid).

The central result: two nulls

association	k (studies)	effect sizes	N	pooled r	95% CI	p
CA — baseline vagal activity	90	250	18,716	−.015	[−.033, .003]	.109
CA — vagal reactivity	55	190	8,741	−.017	[−.039, .005]	.127

Neither pooled association differs from zero, and both survive a sensitivity analysis excluding prenatal/adulthood-adversity contamination. No publication-bias signal on either (Egger’s test and Trim-and-Fill both clean). This is a large, well-powered, pre-registered null on a hypothesis the developmental-psychophysiology literature has long treated as established — the paper is explicit that it “was unexpected,” given the structural and functional brain alterations documented in adversity-exposed individuals.

Where a small effect does emerge

Significant heterogeneity meant the null could hide subgroup effects, and it does — but the surviving effects are small and specific:

• Psychopathology is the strongest moderator of baseline tone. CA tracks lower resting vmHRV in samples where some participants carry a psychiatric diagnosis (r = −.106) but not in non-clinical samples (r = −.005). The adversity→low-vagal-tone link is a feature of clinical populations, not the general population — consistent with a stress-sensitization reading (alterations appear where adversity plus psychopathology co-occur) rather than adversity acting alone. See interoceptive-psychopathology.
• Direct/severe adversity matters, indirect does not. Maltreatment (r = −.058) and direct adversities (r = −.037) move baseline tone; poverty, neighbourhood violence and other indirect adversities (r ≈ −.001) do not. Deprivation/mixed exposures moved it more than threat. In the joint model, psychopathology and proximity retain unique effects.
• Reactivity effects are age- and method-dependent. CA tracks lower vagal reactivity in older samples and when adversity occurred less recently — and, tellingly, only when reactivity is operationalized as task mean levels of vmHRV (r = −.079), not as difference or residualized-change scores (both null). The authors flag this last point against themselves: a task-mean-level association with no baseline association is hard to interpret as genuine blunted withdrawal, and a post-hoc analysis found the mean-level effect may be an artefact of higher baseline vmHRV in adversity-exposed subsamples.

Why it matters here

Two threads.

For the RSA/vagal literature. This is a first-hand, well-powered demonstration that RSA’s association with a canonical real-world risk factor is essentially null at the population level and small even where present. Whatever RSA indexes, its predictive traffic with childhood adversity is far thinner than the framing literature (including the Neurovisceral Integration and polyvagal traditions the paper invokes) assumes — an instance of the recurring lesson that vagal-tone measures carry less individual-difference signal than the theories built on them require. It attaches to is-polyvagal-theory-valid on the evidential-robustness axis, not the mechanism axis: it does not test PVT’s dorsal/ventral anatomy, and it is authored from inside the mainstream vmHRV tradition, so it is a null within the paradigm, not a critique of it.

For the wiki’s childhood-adversity page. early-adverse-life-events currently carries a “blunted physiology, amplified report” narrative — adversity lowering autonomic/HPA reactivity while raising subjective distress (Gooding et al. 2016; Carroll et al. 2017). This meta-analysis qualifies that story for the vagal channel specifically: the blunting is not a general population fact but a clinical-sample, direct-adversity phenomenon, and even then small. It does not contradict the interoceptive-accuracy findings on that page (RSA is an efferent autonomic readout, not an interoceptive-accuracy score — the distinction the respiratory-sinus-arrhythmia page insists on), so it sharpens rather than overturns.

Reliability of the finding, and its brakes

Strengths: pre-registered, two independent meta-analyses, three-level model allowing multiple effect sizes per study, unpublished data solicited to counter the file-drawer, rigorous moderator testing in a combined model. Brakes, mostly the authors’: it covers one ANS aspect only (parasympathetic vmHRV), omits vagal recovery, controls none of smoking/alcohol/fitness/daytime, uses only linear HRV indices, and — the sharpest self-criticism — pools reactivity as group comparisons/correlations that cannot distinguish vagal withdrawal from vagal augmentation, so a real but heterogeneous reactivity effect (some exposed individuals over-withdrawing, others over-augmenting) would average to the null they report. They recommend a pattern-based reactivity approach for future work. Respiration was frequently unmeasured, leaving RSA’s respiratory confound live throughout.

Data and analysis scripts open on OSF (osf.io/cuynb); open access under CC BY.