Joint associations of recovery capital and polygenic risk with electronic health record observed remission and relapse in substance use disorder: Longitudinal evidence from the All of Us research program
Janaka V. Kosgolla, Douglas C. SmithAbstract
Background and aims
Recovery capital and genetic liability may be related to substance use disorder (SUD) remission through psychosocial and biologic pathways, but their joint association with remission course remains poorly understood. This question is clinically important because people entering remission differ in recovery resources and inherited liability. We examined associations of baseline recovery capital index (RCI) and a multitrait polygenic risk score (PRS) with 12‐month electronic health record (EHR) observed remission and described the treatment‐related pathway component captured by EHR recorded service contact.
Design
National cohort study using All of Us electronic health record, survey and genomic data. Weighted relapse hazards were estimated with inverse probability of treatment and censoring weights, followed by bootstrap g‐computation to estimate 12‐month remission summaries.
Setting
United States health care systems participating in All of Us.
Participants
Adults with non‐nicotine SUD evidence and baseline RCI and PRS data. Among 7257 eligible participants, 2413 contributed 5124 constructed remission episodes and 7287‐episode month observations before final restrictions; bridge models used 7068 observations from 5123 episodes.
Measurements
Exposures were baseline RCI, a 0 to 1 mean across 21 retained survey items and an addiction PRS standardized, residualized on 10 ancestry principal components and normalized to a 0 to 1 scale. Outcomes were EHR observed relapse and 12‐month remission summaries from remission qualifiers, active use diagnoses and toxicology.
Findings
Comparing the 90th with the 10th percentile of RCI was associated with higher 12‐month remission probability by 7.9 percentage points [95% confidence interval (CI) = 1.5–14.8] at low PRS and 10.6 points (95% CI = 3.9–17.8) at median PRS. At high PRS, the contrast was 6.4 points (95% CI = −2.6 to 16.1). The pattern was non‐monotonic, with less favorable remission at lower to mid RCI combined with mid to higher PRS and more favorable remission at high RCI with low PRS. PRS added modest information beyond RCI; severity adjusted point estimates preserved positive RCI contrasts.
Conclusions
Recovery capital and genetic liability appear to be jointly associated with 12‐month electronic health record observed substance use disorder remission projections in a nonlinear pattern. Higher recovery capital appears to be associated with more favorable projections across most polygenic risk score levels. Contrasts are modest and sensitive to censoring assumptions.