New multisensory telemonitoring device accurately identifies risk of worsening HFNews - May 3, 2017
Paris, France | The HeartLogic multi-sensor algorithm as an automatic predictor of heart failure events: results from the MultiSENSE Trial.
Presented at ESC Heart Failure 2017 by Roy Stuart GARDNER (Linithgow, United Kingdom)
The typical journey of a HF patient is characterised by large fluctuations in function and quality of life. Over the course of disease progression, better periods are alternated with hospitalisation for acute decompensation episodes, and over time a chronic, gradual decline is seen. It is notably difficult to predict episodes. Efforts have been made to telemonitor patients, but the achieved benefit has been modest so far. More accurate and daily measurements may be needed.
The MultiSENSE HF diagnostic solution HeartLogic has been developed. It captures heart sounds and data on respiration, thoracic impedance, heart rate and activity. Multiple changes from these sensors are aggregated and given a weight based on an individual’s daily risk for worsening HF to create a composite HeartLogic index. This index is updated daily and an alert is issued when the index crosses a user-definable threshold. Data have recently been published on use of the HeartLogic, and the primary endpoints were met, with sensitivity of 70% (95% CI: 55.4-82.1%) and unexplained alert rate of 1.47 (95%CI: 1.32-1.55). False positive rate was 1.56 (95%CI: 1.41-1.77).
This is a posthoc analysis of MultiSENSE study date, in which sensor data and clinical events from 900 patients with NYHA class II, III or IV within the last 6 months, and an implanted CRT-D system were followed up for 1 year. HF events encompassed hospital admission with primary cause of HF, or unscheduled administration of IV medications or ultrafiltration. 192 HF events, of which 145 were usable for HeartLogic occurred, in 106 and 102 patients, respectively. Event rates were calculated for patients when IN or OUT of HeartLogic Alert State. Event Rate Ratio (ERR) was calculated as the event rate IN Alert State divided by OUT of Alert State. The effect on ERR of each programmable threshold from 10 to 40 was assessed. Similarly, ERR of baseline NT-proBNP measurement calculated using above or below threshold (range 500-1500 pg/mL) was assessed.
- At nominal HeartLogic threshold = 16, the HF event rate was ten times higher in the IN Alert State (0.80 events/patient-year: PY) as compared to the OUT of Alert State (0.08 events/PY).
- ERRs ranged between 8.4 and 12.6 across all thresholds.
- Event rates as a function of baseline NT-proBNP was not as informative; ratio of high vs. low NT-proBNP at different cutoffs did not predict event rates as well as did ERR.
- In multivariate analyses, a HeartLogic threshold of 16 had an adjusted ERR of 5.91 (95%CI: 3.63- 9.62, P<0.0001).
This analysis shows that HeartLogic predicts the risk of an HF event independent of baseline variables. The Event Rate was ten times higher when HeartLogic was IN Alert vs. OUT of Alert State. HeartLogic was a better prognosticator than a baseline NT-proBNP.
A drawback of the comparison with analysis of NT-proBNP is that NT-proBNP was only measured at baseline, while the HeartLogic Index was measured daily. Plus, physicians were not blinded to NT-proBNP, which may have modified therapy, thereby possibly limiting its predictive accuracy.
Further studies are needed to establish if HeartLogic can improve patient outcomes.
Discussant Martin COWIE (London, GB) congratulated Gardner for extracting so much from these data. The challenge in telemonitoring studies has always been to get sense out of all the noise. The challenge lies in that that demands more than just risk stratification; you need to know what leads to meaningful clinical improvement.
He noted that the false positive rate seemed low in MultiSENSE, and he called these results one step ahead in the field of telemonitoring. A caveat lies in whether improvements in clinical outcomes justify the cost and logistical investments. Important to realise is also that not alerts are immediately actionable; a physician still needs to think about what causes the alert in this patient.
It was also noted that this system might be useful in acute heart failure, as decongestion often develops in the week before the acute event.
Our coverage of ESC HF 2017 is based on the information provided during the congress.