Accuracy of Seattle Heart Failure Model and HeartMate II Risk Score in Non-Inotrope-Dependent Advanced Heart Failure Patients: Insights From the ROADMAP Study (Risk Assessment and Comparative Effectiveness of Left Ventricular Assist Device and Medical Management in Ambulatory Heart Failure Patients)

David E Lanfear; Wayne C Levy; Josef Stehlik; Jerry D Estep; Joseph G Rogers; Keyur B Shah; Andrew J Boyle; Joyce Chuang; David J Farrar; Randall C Starling

doi:10.1161/CIRCHEARTFAILURE.116.003745

Accuracy of Seattle Heart Failure Model and HeartMate II Risk Score in Non-Inotrope-Dependent Advanced Heart Failure Patients: Insights From the ROADMAP Study (Risk Assessment and Comparative Effectiveness of Left Ventricular Assist Device and Medical Management in Ambulatory Heart Failure Patients)

Circ Heart Fail. 2017 May;10(5):e003745. doi: 10.1161/CIRCHEARTFAILURE.116.003745.

Authors

Affiliations

¹ From Henry Ford Hospital, Detroit, MI (D.E.L.); University of Washington, Seattle, WA (W.C.L.); University of Utah, Salt Lake City (J.S.); Houston Methodist Hospital, TX (J.D.E.); Duke University, Durham, NC (J.G.R.); Virginia Commonwealth University, Richmond (K.B.S.); Thomas Jefferson University, University Hospital, Philadelphia, PA (A.J.B.); St. Jude Medical, Pleasanton, CA (J.C., D.J.F.); and Cleveland Clinic, OH (R.C.S.). dlanfea1@hfhs.org.
² From Henry Ford Hospital, Detroit, MI (D.E.L.); University of Washington, Seattle, WA (W.C.L.); University of Utah, Salt Lake City (J.S.); Houston Methodist Hospital, TX (J.D.E.); Duke University, Durham, NC (J.G.R.); Virginia Commonwealth University, Richmond (K.B.S.); Thomas Jefferson University, University Hospital, Philadelphia, PA (A.J.B.); St. Jude Medical, Pleasanton, CA (J.C., D.J.F.); and Cleveland Clinic, OH (R.C.S.).

PMID: 28465311
DOI: 10.1161/CIRCHEARTFAILURE.116.003745

Abstract

Background: Timing of left ventricular assist device (LVAD) implantation in advanced heart failure patients not on inotropes is unclear. Relevant prediction models exist (SHFM [Seattle Heart Failure Model] and HMRS [HeartMate II Risk Score]), but use in this group is not established.

Methods and results: ROADMAP (Risk Assessment and Comparative Effectiveness of Left Ventricular Assist Device and Medical Management in Ambulatory Heart Failure Patients) is a prospective, multicenter, nonrandomized study of 200 advanced heart failure patients not on inotropes who met indications for LVAD implantation, comparing the effectiveness of HeartMate II support versus optimal medical management. We compared SHFM-predicted versus observed survival (overall survival and LVAD-free survival) in the optimal medical management arm (n=103) and HMRS-predicted versus observed survival in all LVAD patients (n=111) using Cox modeling, receiver-operator characteristic (ROC) curves, and calibration plots. In the optimal medical management cohort, the SHFM was a significant predictor of survival (hazard ratio=2.98; P<0.001; ROC area under the curve=0.71; P<0.001) but not LVAD-free survival (hazard ratio=1.41; P=0.097; ROC area under the curve=0.56; P=0.314). SHFM showed adequate calibration for survival but overestimated LVAD-free survival. In the LVAD cohort, the HMRS had marginal discrimination at 3 (Cox P=0.23; ROC area under the curve=0.71; P=0.026) and 12 months (Cox P=0.036; ROC area under the curve=0.62; P=0.122), but calibration was poor, underestimating survival across time and risk subgroups.

Conclusions: In non-inotrope-dependent advanced heart failure patients receiving optimal medical management, the SHFM was predictive of overall survival but underestimated the risk of clinical worsening and LVAD implantation. Among LVAD patients, the HMRS had marginal discrimination and underestimated survival post-LVAD implantation.

Clinical trial registration: URL: http://www.clinicaltrials.gov. Unique identifier: NCT01452802.

Keywords: area under the curve; death; heart failure; survival analysis.

Publication types

Multicenter Study
Observational Study

MeSH terms

Aged
Disease Management*
Female
Follow-Up Studies
Heart Failure / mortality
Heart Failure / therapy*
Heart Ventricles
Heart-Assist Devices*
Humans
Male
Middle Aged
Outpatients*
Prospective Studies
ROC Curve
Risk Assessment / methods*
Risk Factors
Survival Rate / trends
Treatment Outcome
United States / epidemiology

Associated data

ClinicalTrials.gov/NCT01452802

Grants and funding

R01 HL103871/HL/NHLBI NIH HHS/United States