Use Kuo 2021 clinical PhenoAge calibration

README.md

@@ -9,7 +9,7 @@ tags:
 
 # phenoage
 
-Clinical Phenotypic Age combines chronological age with nine blood biomarkers. The coefficients and mortality-to-age equation follow the [2019 Liu et al. correction](https://doi.org/10.1371/journal.pmed.1002760). The printed equation approaches a lower bound of 141.50 years as modeled mortality approaches zero. Supply raw C-reactive protein in mg/dL. pyaging floors it at 0.01 mg/dL before taking its natural logarithm. Use pyaging >=0.6.1 for the corrected calibration, which runs in the package class.
+Clinical Phenotypic Age combines chronological age with nine blood biomarkers. The coefficients and age conversion follow [Kuo et al. (2021), section 4.3](https://doi.org/10.1111/acel.13376). The leading 1 inside the logarithm in the 2019 erratum, used in pyaging 0.6.1, compressed scores near 141.50 years. In an unweighted check of 5,645 NHANES 2009–2010 adults, 98.2% of 0.6.1 scores fell within 0.1 years of 141.50. The Kuo equation restored the age scale while preserving mortality ranking. Supply raw C-reactive protein in mg/dL. pyaging floors it at 0.01 mg/dL before taking its natural logarithm. The Kuo calibration requires pyaging >=0.6.2.
 
 Model weights retain the original authors' terms; the pyaging software license does not relicense them.
 

config.json

@@ -10,7 +10,7 @@
  "last_author": "Steve Horvath",
  "model_type": "penalized hazards regression with Gompertz calibration",
  "n_features": 10,
- "notes": "Clinical Phenotypic Age combines chronological age with nine blood biomarkers. The coefficients and mortality-to-age equation follow the [2019 Liu et al. correction](https://doi.org/10.1371/journal.pmed.1002760). The printed equation approaches a lower bound of 141.50 years as modeled mortality approaches zero. Supply raw C-reactive protein in mg/dL. pyaging floors it at 0.01 mg/dL before taking its natural logarithm. Use pyaging >=0.6.1 for the corrected calibration, which runs in the package class.",
+ "notes": "Clinical Phenotypic Age combines chronological age with nine blood biomarkers. The coefficients and age conversion follow [Kuo et al. (2021), section 4.3](https://doi.org/10.1111/acel.13376). The leading 1 inside the logarithm in the 2019 erratum, used in pyaging 0.6.1, compressed scores near 141.50 years. In an unweighted check of 5,645 NHANES 2009\u20132010 adults, 98.2% of 0.6.1 scores fell within 0.1 years of 141.50. The Kuo equation restored the age scale while preserving mortality ranking. Supply raw C-reactive protein in mg/dL. pyaging floors it at 0.01 mg/dL before taking its natural logarithm. The Kuo calibration requires pyaging >=0.6.2.",
  "platform": [
   "clinical laboratory assays"
  ],
@@ -31,6 +31,6 @@
  "unit": [
   "years"
  ],
- "version": "0.6.1",
+ "version": "0.6.2",
  "year": 2018
 }

phenoage.provenance.json

@@ -1,11 +1,11 @@
 {
-  "source": "https://doi.org/10.1371/journal.pmed.1002760",
+  "source": "https://doi.org/10.1111/acel.13376",
   "accessed": "2026-11-03",
-  "locator": "Page 1, corrected log hazard and Phenotypic Age equations",
+  "locator": "Section 4.3, log hazard and Phenotypic Age equations",
   "gompertz_gamma": 0.0076927,
   "gompertz_numerator": 1.51714,
   "time_horizon_months": 120,
-  "formula_runtime_minimum": "0.6.1",
-  "artifact_sha256": "73aeaac3f00926d9358d7f20dc71b6eafaf3912e15abb1508a6db715ea4e1d62",
+  "formula_runtime_minimum": "0.6.2",
+  "artifact_sha256": "86b0b7418082b7ebf1a9ca6054a827b26c5c3442448991c2d895f9790665e41d",
   "validation": "Notebook executed in full. Postprocessing checked against the printed mortality equations and fixed 80-digit Decimal reference values."
 }

phenoage.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:73aeaac3f00926d9358d7f20dc71b6eafaf3912e15abb1508a6db715ea4e1d62
-size 5037
+oid sha256:86b0b7418082b7ebf1a9ca6054a827b26c5c3442448991c2d895f9790665e41d
+size 5229