Timeline Follow-back Captures the Timing and Frequency of Prenatal Cannabis Exposure, Not Its Dose

Question: How do studies using the Timeline Follow-back method characterize the dose, timing, and frequency of prenatal cannabis exposure, and how do these exposure metrics relate to infant neurodevelopmental outcomes?

Answer

Across the retrieved papers, Timeline Follow-back (TLFB) functions in prenatal research as a calendar of when and how often cannabis was used, and almost never as a measure of how much THC reached the fetus. The one retrieved study that applies TLFB directly to an infant neurodevelopmental outcome operationalizes exposure three ways — presence, gestational timing, and frequency within exposed infants — and reports only small volumetric and microstructural differences in neonatal brain [[doi:10.64898/2026.03.02.26347436. Where frequency is analyzed in a pregnancy cohort, the contrast is coarse and categorical (greater than weekly versus less than monthly), and only the high-frequency stratum is associated with lower birthweight [[doi:10.64898/2026.03.01.26347369. Quantitative THC dose is measurable by TLFB — via quantity, frequency and potency converted to standard THC units [[doi:10.1101/2025.05.21.25328059, or as milligrams of THC per day anchored to creatinine-normalized urine THC-COOH [[doi:10.1101/2025.03.14.25323995] — but that work has been done in non-pregnant adult and adolescent samples, so the dose axis is effectively absent from the prenatal-infant literature. The collective position supported here is that timing is the best-evidenced exposure dimension, frequency is a weak and mostly dichotomous modifier, and dose in milligrams is an unexamined variable in infant outcomes.

Frequency bands, not milligrams, are the exposure axis that predicts infant growth

The strongest frequency evidence comes from a prospective cohort that stratified prenatal cannabis exposure by use greater than weekly versus less than monthly: high-frequency exposure was associated with lower birthweight (Beta = −0.13) while low-frequency exposure was not [[doi:10.64898/2026.03.01.26347369. That same cohort found the largest birthweight reduction in cotinine-positive exposed pregnancies (Beta = −0.20) and the highest small-for-gestational-age risk in the cannabis-plus-nicotine group, yet effects persisted in cotinine-negative pregnancies [[doi:10.64898/2026.03.01.26347369. This design treats frequency as an ordinal proxy for dose and cannot distinguish a true dose-response from a threshold, and it does not quantify THC content at all [[doi:10.64898/2026.03.01.26347369.

Gestational timing dissociates white-matter microstructure from regional volume

In the largest retrieved TLFB-based study — 1,782 mother-infant dyads, 221 exposed, with natural-sleep MRI in the first month of life — first-trimester exposure was primarily associated with white-matter microstructural alterations in cerebellar and association tracts, whereas exposure sustained throughout pregnancy was associated with volumetric differences across cerebellar and subcortical regions [[doi:10.64898/2026.03.02.26347436. The volumetric signature itself was bidirectional, with smaller amygdala, thalamic and cerebellar vermis volumes and larger caudate, hippocampal and cerebellar cortex volumes [[doi:10.64898/2026.03.02.26347436. The authors characterize the effect sizes as small but statistically significant [[doi:10.64898/2026.03.02.26347436, which matters for how the timing claim should be weighted.

Timing-specificity is corroborated by tissue-level and preclinical evidence that TLFB cannot supply

Human fetal brain multi-omics found minimal molecular effects in female first-trimester tissue but pronounced system-level disruption in male second-trimester tissue, with disrupted pathways carrying signatures of autism spectrum disorder, schizophrenia-related pathology, and cortical connectivity disorders [[doi:10.64898/2026.07.15.738779. Rodent work that manipulated the gestational window directly showed that postnatal versus prenatal THC exposure produced inverse changes in microglial phagocytosis and juvenile social behavior, which the authors read as evidence that gestational timing, not merely total exposure, determines outcome direction [[doi:10.1101/2025.02.17.638714. Longitudinal MRI in mice exposed prenatally to THC found ventriculomegaly in embryos followed by decelerated brain growth sustained into adulthood, with females more affected [[doi:10.1101/2024.11.02.621669]. These studies support the trimester-specific interpretation of the TLFB findings while fixing dose by design rather than measuring it.

Quantified THC dose exists in the TLFB literature, but only outside pregnancy

The Enhanced Cannabis Timeline Followback yields quantity, frequency and potency converted into weekly standard THC units, and those units discriminated cannabis use disorder with areas under the curve above 0.70, supporting thresholds of 8.26 units per week in adults and 6.04 in adolescents [[doi:10.1101/2025.05.21.25328059. A calendar-based TLFB summarized as milligrams of THC per day was associated with creatinine-normalized urine THC-COOH among exclusive flower users, with effect sizes explored by gender [[doi:10.1101/2025.03.14.25323995. That biomarker correspondence was established in people with HIV, not in pregnancy, and the authors note the long-standing difficulty of quantifying cannabis consumption given product variety and variability in consumption behavior [[doi:10.1101/2025.03.14.25323995. Converting self-reported flower mass into THC mass also inherits potency error: independent testing of 74 Colorado samples found labeled THC values higher than observed values in flower and edible products, with flower falling outside the legally allowable 15% band [[doi:10.1101/2025.03.14.25323943.

Biomarker-anchored exposure dissociates brain connectivity from neonatal behavior

Urine COOH-THC measured at 8–14 weeks' gestation in 42 Black mother-infant dyads was significantly associated with altered resting-state connectivity in hippocampus, insula and caudate at approximately one month postpartum, but those connectivity patterns were not correlated with NICU Network Neurobehavioral Scale indices [[doi:osf:3e9md_v1. This is a dose-adjacent metric — a metabolite concentration rather than a self-reported quantity — and it shows that a biological exposure gradient can register in brain circuitry without a parallel behavioral signal at that age [[doi:osf:3e9md_v1. It also shows that early-pregnancy exposure alone, without later-trimester data, is sufficient to detect connectivity differences [[doi:osf:3e9md_v1.

Binary exposure flags retain associations with child phenotype after genetic and familial adjustment

In ABCD, prenatal cannabis exposure ascertained by retrospective maternal report accounted for significant variance in autism spectrum traits and total Social Responsiveness Scale score above polygenic risk, and remained positively associated with total SRS and several traits after adding propensity-score adjustment for familial confounders, though not with social communication specifically [[doi:osf:wpng5_v1. This is the longest-horizon outcome in the retrieved set, and it comes at the cost of the timing and frequency resolution that TLFB provides — the exposure is retrospective and not dose-resolved [[doi:osf:wpng5_v1.

Conflicts and gaps

The retrieved evidence does not contain a single study that relates a TLFB-derived THC dose to an infant neurodevelopmental outcome; the TLFB-in-infant study models presence, timing and frequency only [[doi:10.64898/2026.03.02.26347436], and the dose-quantified TLFB work is in non-pregnant populations [[doi:10.1101/2025.05.21.25328059] [[doi:10.1101/2025.03.14.25323995]. Frequency is operationalized incompatibly across studies — greater-than-weekly versus less-than-monthly strata in one cohort [[doi:10.64898/2026.03.01.26347369], frequency-within-exposed-infants in another [[doi:10.64898/2026.03.02.26347436], and no frequency metric at all in the biomarker and ABCD studies [[doi:osf:3e9md_v1] [[doi:osf:wpng5_v1]. Whether neurodevelopmental effects are cannabis-specific or nicotine-driven is unresolved: the birthweight association was strongest in cotinine-positive pregnancies while still present without cotinine [[doi:10.64898/2026.03.01.26347369], and no retrieved infant neuroimaging study reports cotinine-adjusted estimates [[doi:10.64898/2026.03.02.26347436] [[doi:osf:3e9md_v1]. Sex-specificity is asserted at the tissue and rodent level [[doi:10.64898/2026.07.15.738779] [[doi:10.1101/2024.11.02.621669] but the retrieved TLFB cohort abstract does not report sex-stratified infant findings [[doi:10.64898/2026.03.02.26347436]. Finally, the direction of dose-response is untested in humans: preclinical studies that vary dose show that low, clinically relevant exposures (10 mg/kg/day CBD) alter offspring behavior [[doi:10.1101/2025.04.30.651495] and that CBD dose modulates THC-induced behavioral effects in a dose-dependent manner [[doi:10.64898/2026.03.05.709920], which implies a non-monotonic or interaction structure that categorical frequency bands cannot capture [[doi:10.64898/2026.03.01.26347369].

What would change the answer

Applying the Enhanced Cannabis Timeline Followback — quantity, frequency and potency converted to milligrams THC per day and to standard THC units [[doi:10.1101/2025.05.21.25328059] — within an existing infant neuroimaging cohort such as HBCD would convert the current presence/timing/frequency model [[doi:10.64898/2026.03.02.26347436] into a dose-response analysis, ideally with distributed-lag or gestational-week-weighted exposure terms rather than trimester bins. Such a study must pair self-report with repeated trimester-specific urine THC-COOH [[doi:osf:3e9md_v1] and cotinine [[doi:10.64898/2026.03.01.26347369], and should use measurement-error or latent-class models to estimate exposure misclassification rather than assuming TLFB quantity is unbiased. Because potency conversion is the weakest link, parallel independent assay of the products participants actually use — the approach that exposed label inaccuracy in Colorado flower and edibles [[doi:10.1101/2025.03.14.25323943] — is required before milligrams THC per day can be treated as a real dose. Sibling-comparison and paternal-exposure negative-control designs would address the confounding that propensity-score adjustment only partially absorbs [[doi:osf:wpng5_v1], and dose-graded preclinical designs using clinically relevant low exposures [[doi:10.1101/2025.04.30.651495] with explicit gestational-window manipulation [[doi:10.1101/2025.02.17.638714] would specify the shape — threshold, monotonic, or non-monotonic — that human frequency bands are currently too coarse to distinguish.