{
  "study_id": "NCT04312269",
  "agent": "baseline",
  "run": 2,
  "p_primary_hypothesis_supported": 0.12,
  "effect_direction": "favours intervention",
  "effect_size_90pct_interval": {"low": -0.45, "high": 0.75, "metric": "Cohen's d for between-group difference in WMFT change (time-based), baseline to 6 weeks, positive = more improvement with TMR"},
  "moderator_prediction": "none detectable; if anything SWS-only TMR roughly equals all-phase TMR, since lab TMR benefits come from NREM2/SWS and REM cueing adds nothing, but arms of ~17 cannot distinguish them",
  "claim_ids_used": [],
  "rationale": "Both arms receive identical MyoCI training, so TMR is an add-on with a small expected lab effect (g~0.29), likely smaller for generalized arm function in chronic stroke (non-memory, transfer outcome). Four arms with ~70 total leaves ~17 per arm; detecting d=0.3 needs ~175 per arm, so power is roughly 10-15%. WMFT time scores are noisy and skewed in moderate-severe patients. Small pilot-scale effects may appear numerically but significance is unlikely. Pooling TMR arms raises power only modestly. Some chance of outcome switching to FMA-UE or subgroup reporting, but resolution requires WMFT at 6 weeks."
}