{
 "seed": 20260924,
 "amendment": "PRE-REGISTRATION AMENDMENT 1 (26 September 2026, before any exposure-outcome model was\nfitted; only the table's row structure had been inspected).\n\nThe pre-registration stated that MetaCardis Supplementary Table 12 \"carries 1,882 rows\nover 1,087 unique IDs (343 once, 693 twice, 51 three times)\" and therefore REQUIRED a\nparticipant-level random intercept for repeated measures. That reading is WRONG.\n\nVerified here: the duplicated rows are EXACT duplicates, not repeat visits. Across 357\nmultiply-listed participants the within-ID standard deviation of the endpoint is exactly\n0.0, and the species abundances are identical within ID in 400 of 400 checked. The\nduplication is structural: MetaCardis lists a participant once per overlapping cohort\ngroup, and the groups nest exactly (IHD372 = CIHD158 + ACS112 + HF102 = 158+112+102).\n\nConsequence: there are no repeated measures to model. The primary analysis is therefore\nordinary least squares on ONE ROW PER PARTICIPANT, which is also what the discovery\ncohort used - so the replication now matches the discovery specification more closely\nthan the pre-registration anticipated, rather than less. The mixed model is dropped as\ninapplicable, not as a modelling preference.",
 "n_participants_total": 994,
 "n_bacteroides_features": 30,
 "n_species_features": 729,
 "primary_ols": {
  "n": 544,
  "beta": -0.03206112862606294,
  "se": 0.020273759044514218,
  "p": 0.1143741601736055,
  "r2": 0.08440250826129225,
  "std_beta": -0.06825356919221076,
  "ci": [
   -0.0717976963533108,
   0.007675439101184928
  ]
 },
 "sensitivity_no_lvef": {
  "n": 972,
  "beta": -0.03059278695046239,
  "se": 0.015337115813931828,
  "p": 0.046357940825009464,
  "r2": 0.09443311588963288,
  "std_beta": -0.06363009976682955,
  "ci": [
   -0.06065353394576877,
   -0.0005320399551560079
  ]
 },
 "secondary_hf_stratum": {
  "n": 71,
  "beta": -0.06449651473738051,
  "se": 0.050529893802663664,
  "p": 0.20649833792913522,
  "ci": [
   -0.16353510659060128,
   0.03454207711584027
  ]
 },
 "permutation_p": 0.12658734126587343,
 "bootstrap_ci": [
  -0.07040850830407622,
  0.005136201642569122
 ],
 "discovery_beta": 0.4201,
 "discovery_beta_in_ci": false,
 "verdict": "not_replicated",
 "secondary_hf_stratum_4cov": {
  "n": 83,
  "beta": -0.07485690235841347,
  "se": 0.04379556334567283,
  "p": 0.09148811905373967,
  "ci": [
   -0.16069620651593222,
   0.010982401799105285
  ]
 },
 "hf_stratum_note": "HF102 rows selected BEFORE de-duplication; the earlier run lost the HF label because de-duplication keeps the first table row, which for most HF participants is their parent IHD372 row. Effect size only, never a hypothesis test."
}