The HeartLogic Multi-Sensor Algorithm Significantly Augments the Prognosis of a Baseline NT-proBNP Assessment for Heart Failure Events

Background: Guidelines carry a class IA recommendation for the use of natriuretic peptides (BNP or NT-proBNP) to establish prognosis or disease severity in chronic Heart Failure (HF). However, these reflect a snapshot assessment at the time of blood draw, which lose relevance over time with changes in the patient's condition. We recently reported on a multi-sensor HF Index and alert algorithm (HeartLogic) using hardware already present in an implantable defibrillator that detected HF events with 70% sensitivity and 34 day advance warning, and identified patients with 10-fold increased risk of worsening HF. In this analysis, our objective was to evaluate whether HeartLogic augments the prognosis of a single baseline NT-proBNP assessment for HF events over the course of a year. Methods: The MultiSENSE trial enrolled patients in North America, Europe, and Asia implanted with CRT-Ds that enabled multi-sensor data collection for up to one year. Patient demographics, clinical history, and baseline measurements including NT-proBNP were obtained at enrollment. HF events were defined as either HF admissions or unscheduled visits with augmented intravenous HF treatment, and were independently adjudicated. The HeartLogic algorithm continuously measured sensor data including heart sounds, respiration rate and tidal volume, thoracic impedance, heart rate, and activity; sensor changes from the patient's own baseline were aggregated and weighted on the basis of an individual daily risk to calculate the daily HeartLogic HF index and compared against a user-configurable alert threshold. Clinicians were blinded to the sensor data and HeartLogic index and alert. HF event rates (expressed as events/pt-yr) were calculated for patients with baseline NT-proBNP above and below 1000 pg/mL and periods with HeartLogic in or out of alert relative to the nominal threshold of 16. Results: The study enrolled 900 patients for up to one year (72.7% male; age 66.6 ± 10.5 years; NYHA Class II/III: 67.2%/26.8%; LVEF 30.0 ± 11.4%) resulting in a total of 192 HF events (average event rate: 0.23 events/pt-yr). The HF event rates are shown in figure (right), with the proportion of patient follow-up for each risk stratification listed below the figure. About half of follow-up time (53%) was within the lowest risk group (0.02 events/pt-yr): out of alert and NT-proBNP <1000 pg/mL. An active HeartLogic alert in patients with low NT-proBNP resulted in a 23.5x increased risk of a HF event (0.47 events/pt-yr). In patients with elevated NTproBNP an active HeartLogic alert was associated with a 50.0x increased risk of an HF event (1.00 events/pt-yr) relative to the lowest risk group. Conclusion: HeartLogic alerts significantly augment the ability of baseline NT-proBNP to identify periods with an elevated risk of HF event over up to a year. B-type natriuretic peptides are the current gold standard marker of adverse heart failure risk, but the prognostic ability decreases over time. Dynamic assessment using HeartLogic alerts in conjunction with intermittent/sparse NT-proBNP could automatically identify periods of time in which patients are at significantly increased risk of worsening HF and help better triage resources to this vulnerable patient population.