Reviewed aggregate research analysis
Case-control analysis of T1-weighted MRI-derived SynthSeg volumes in 265 public ds000030 participants. Inferential hierarchy, null results, and limitations remain visible—not buried.
Secondary evidence
—
The association is literature-concordant and technically robust within this dataset. It remains secondary and may reflect treatment exposure, illness duration, chronicity, or other unmeasured factors—not disease-specific anatomy.
The cited large-consortium literature cannot establish participant-level independence from this public cohort, so this is not labeled an independent replication.
Triggered after the secondary result; these checks cannot promote it to confirmatory evidence.
Prespecified primary family
Swipe horizontally to view all effect estimates and corrected p-values.
| Hypothesis | Contrast / outcome | Adjusted effect | 95% CI | Raw p | Holm p | Conclusion |
|---|
Effects are adjusted percentage differences from log-volume models. Every primary model includes all 265 participants; case and control counts remain visible in each row.
Cohort audit
Swipe horizontally to view acquisition-cell counts.
| Group | N | Age, mean ± SD | Reported F / M | Scanner 35343 / 35426 | No ghost / ghost |
|---|
Controls are concentrated on scanner 35343, whereas clinical groups are more evenly split. Scanner and ghost label were therefore mandatory covariates; adjustment cannot guarantee elimination of residual confounding.
Measurement checks
The absent QC association reduces concern about gross diagnosis-differential segmentation quality. It does not establish manual-reference accuracy.
Power context
Approximate unadjusted Cohen's d needed for 80% power at the conservative five-test α reference. Covariate adjustment can change actual power.
Statistical contract
Mandatory cautions
Reproducibility
Literature anchors