Cardiac Health Assessment using a wearable device before and after TAVI

Introduction

As transcatheter aortic valve implantation (TAVI) for severe aortic stenosis is increasingly used in the elderly population, including a high percentage of patients with substantial co-morbidity, improvement in quality of life is just as important as extending life expectancy.^{1, 2} Not all TAVI patients benefit from improved physical activity as assessed by a 6-minute walking test (6MWT) or quality of life (QoL) questionnaire.^{3, 4} However, these tests could be influenced by other factors and comorbidities like peripheral vascular disease for the 6MWT or depression for the QoL questionnaire. Another concern with such tools is that they merely provide a snapshot of a patient’s life, and might change under different circumstances. Consequently, an unbiased and longer-term tool to anticipate the benefit from TAVI would allow physicians and patients to personalize treatment and expectations.

In recent years digital health has begun to transform the medical world.⁵ Smart phones and health watches, in particular, have found their way into the clinic.⁶ These devices can detect atrial fibrillation⁷ and predict the wearer’s 5-year risk of dying.⁸ The wearable device used in this study, the Philips Health Watch,⁹ continuously measures physical parameters like heart rate, number of steps, and amount of physical activity. Combining parameters from the health watch might provide a more physiologic and comprehensive assessment of functional status before and after TAVI. After intervention for aortic stenosis, patient symptoms often improve, but do they objectively become more active as measured by a wearable tracker?

In this study, we evaluated the change in parameters collected by the Philips Health Watch in patients before and after TAVI in comparison to standard clinical and research tests (6MWT, quality of life questionnaire). We hypothesized that after a TAVI procedure, physiologic parameters such as step count, total activity time, and daily total energy expenditure would increase, whereas respiration rate and heart rate would decrease.

METHODS

This prospective, exploratory study sought to identify parameters from the Philips Health Watch (DL8791, Philips, Stamford, CT, USA) that changed after successful TAVI. The study was performed in compliance with the Declaration of Helsinki and local regulations. All subjects gave written informed consent as approved by an independent medical ethics committee (MEC-U approval ID:W16.141).

Analysis of the health watch data

The Philips Health Watch is a wrist-worn, photoplethysmography based, heart rate and activity monitor (Figure 1). Once per minute it measures parameters like heart rate, respiration rate, step count, and total energy expenditure (number of calories needed to carry out physiologic functions like breathing and physical activity, but excluding the energy required for digesting food) as described previously by Hendrikx et al.⁹ Parameters are measured at a 1-Hz sampling rate and stored on the device as 1-minute average values. Data can be extracted via Bluetooth by means of an iPod, using a proprietary iOS application from Philips, and sent to a Philips Research server for use in analyses.

• Download figure
• Open in new tab

Figure 1. Philips Health Watch.

The Philips Health Watch is a wrist-worn, photoplethysmography based, heart rate and activity monitor. It measures parameters like heart rate, respiratory rate, step count, total energy expenditure, and activity time. Measurements use a 1-Hz sampling rate and are stored on the device as 1-minute average values. Data can be extracted in the hospital via Bluetooth using the iPad application Watch Control and sent over Wi-Fi to the Philips Research server for analysis.

A full report including primary data from the watch and derived parameters consists of a summary averaged over one day (Table 3), distributions of heart rate and respiration rate (Figure 2), and log plot of the heart rate and total energy expenditure (Figure 3). Total energy expenditure is divided into subcategories of the Metabolic Equivalent Task (MET) scale. As the elderly TAVI population of this cohort seemed rather inactive, a subdivision of the MET scale was designed: basal activity corresponded to a MET of 1.5 to 2, light activity from 2 to 3, moderate activity from 3 to 6, and high activity from 6 upwards (we used standard thresholds for the last two categories).¹¹

• Download figure
• Open in new tab

Figure 2. Distribution of heart rate and respiratory rate.

This page of the output report from the health watch depicts density plots of heart rate and respiration rate during the day and during sleep. It also shows the frequency of resting heart rate and the distribution of activity levels for heart rate observations > 100 beats per minute.

• Download figure
• Open in new tab

Figure 3. Log plot of heart rate vs. total energy expenditure.

Another page of the report plots each heart rate and corresponding total energy expenditure on a log plot, divided into subcategories of the Metabolic Equivalent Task (MET) scale. The fitted line relates to the energy efficiency of the cardiovascular system, as detailed in the text.

Each red dot in Figure 3 represents a particular measurement: the 1-minute average heart rate and corresponding energy expenditure level. The fitted line quantifies the cardiac energy expenditure slope (CEES): as heart rate rises, more energy is needed to maintain the resulting hemodynamic state. When the slope is less steep, more energy is needed to maintain a heart rate of, for example, 60/minute. Conversely, when the slope is steeper, less energy is needed to maintain the same hemodynamic state. Potentially the steepness of the slope (CEES) serves as an indicator for the energy efficiency of the cardiovascular system.

The report and concept of CEES was proposed and made available as data derived from the raw data from the health watch by (the author from) Philips Research Eindhoven, and used in clinical data analysis at Catharina Hospital in Eindhoven.

RESULTS

A total of 100 subjects were enrolled into the study. Their demographics and medications (before and after TAVI) are displayed in Table 1 and Table 2, respectively. After TAVI, 11 patients died and 14 patients were lost to follow-up. Data extraction for 3 patients failed before TAVI. Complete watch data were thus obtained in 97 (pre TAVI) versus 75 (post TAVI) patients. The population consisted of more men (57%) than women. Demographic characteristics were representative of a clinical TAVI population with a median age of 81.0 years, NYHA class II or higher in 92%, hypertension in 66%, and dyslipidemia in two-thirds of patients. All patients fulfilled guideline criteria for severe aortic stenosis. Operative characteristics can be found in the supplemental material (Procedural data).

DISCUSSION

To our knowledge, this is the first study in which extensive one-week physiologic data before and after TAVI was assessed using a sophisticated wearable sensor, the Philips Health Watch. Watch parameters like activity time and step count did not increase after TAVI for the overall group, in contrast to significant improvements in 6MWT (53 meters, or 18% over baseline) and physical limitation score from the questionnaire (7 points, or 13% over baseline). The increase in 6MWT mirrors results obtained in a randomized trial: 254 meters at baseline, 288 meters at 30 days, and 297 meters 1 year after TAVI. ¹²

One explanation for our findings is that a relatively older population truly does not increase daily activity after TAVI, simply because they do not have to or do not want to (i.e. lack of necessity or motivation). In such cases, step count and heart rate will not and do not have to increase. However, 6MWT increased significantly after 1 year in a cohort randomized to TAVI compared to no change in those randomized to medical therapy¹³, arguing against a Hawthorne effect. A second explanation for our findings is that quality of life questionnaires and 6MWT⁴ produce an abundance of motivation in the hospital setting that does not correspond to daily life. In this case the findings on standard tests (6MWT, QoL questionnaires) would appear improved, whereas parameters on the health watch would remain unchanged. Thus both tests may be correct given their circumstances. A third explanation is that the health watch parameters contain bias or imprecision that limits their paired comparison, even in a reasonably sized cohort such as our own. This seems to be an interesting topic for further research.

Novel metrics like the Cardiovascular Energy Expenditure Slope (CEES) may be valuable parameters in the future or in other settings for evaluating the energy efficiency of the heart. The steepness of the slope objectively quantifies the relationship between heart rate and total energy expenditure. A less steep slope corresponds with a lower CEES value. In such case more energy is needed to maintain heart rate than when the slope is steeper (and CEES value higher). There was a significant increase in CEES for the good responders cohort and this was also accompanied by a significant increase in moderate activity time. There was no significant improvement in CEES or moderate daily activity time in patients that had no or moderate (< 40 meters) improvement in the 6MWT after TAVI. This novel metric could be used in future research as a tool to identify patient improvement after TAVI intervention, independent of subjective variables. More research on this metric is warranted.

How to best identify patients whose symptoms benefit from treatment with TAVI remains an important and unanswered clinical question. A post-hoc analysis of a randomized trial of TAVI compared to a surgical valve procedure demonstrated that 36% of patients had no change in 6MWT after 30 days and 12 months, and 23-28% demonstrated no improvement on their QoL questionnaire scores (albeit using a different tool than in our cohort). ¹² When considering an intervention, both procedural risks and economic costs should be balanced against the potential improvement in quality of life. Since the patients that commonly qualify for TAVI treatment are relatively older and more frail, an increase in physical performance can be equally or even more important than extending life expectancy. With increasing costs in health care, the benefit of an intervention should be clear and personalized¹⁴, and for TAVI this cost-benefit ratio has been disputed¹⁵. Unfortunately, the overall findings from this study cannot identify patients using the health watch who would be expected to have an above average response.

The impact of health watches and other sensor technologies on cardiologic care warrants more research. The Apple Heart study⁷ found an irregular heart rhythm in only 0.52% of over 400,000 people followed for 8 months, of whom just 21% completed further testing and of these a 34% minority were ultimately diagnosed with atrial fibrillation – a paltry yield of 153 people, or <<0.1% of the total. Findings such as these show the inevitable tradeoffs between mass testing versus pre-test probability of an actionable diagnosis.

CONCLUSIONS

This is one of few studies before and after TAVI with extensive, one-week functional assessment with a sophisticated wearable sensor, the Philips Health Watch. Data from the health watch did not register an increase in activity time, total step count, or other parameters after TAVI, whereas traditional 6MWT and QoL assessment did improve. Watch-based parameters such as these might be less appropriate for measurement of treatment effect in the TAVI population. However our findings relating to the good responder subpopulation suggest that using data such as the CEES parameter derived from the data from a wearable device, might become useful to objectively identify patient improvement after TAVI intervention. This seems to be interesting for further research.

SUPPLEMENTAL MATERIAL