Wearable-Tracked Walking Behavior Hints at Subclinical HF

Walking patterns captured by a fitness tracker may detect signs of subclinical heart failure (HF), according to a new analysis. Researchers—many of them employees of the device manufacturer—say that this behavioral information, which goes beyond step count, might one day be used by clinicians in conjunction with in-office testing to identify and monitor HF even before symptoms develop.

The data, published online recently in the Journal of Cardiac Failure, come from the Project Baseline Health Study, led by Verily Life Sciences in partnership with Duke University School of Medicine, Stanford Medicine, and the American Heart Association.

Sooyoon Shin, PhD (Verily Life Sciences, South San Francisco, CA), said that while research on wearables’ role in advancing cardiometabolic health has increased exponentially in recent years, their study is unique in that it looked at subclinical heart failure and did so in a real-life setting with a diverse population, and with an eye toward patient-reported outcomes.

The 18 measures tracked by the accelerometer-equipped Verily smartwatch spanned different concepts related to walking behavior, including exercise capacity, daily quantities of activity, and endurance, among others.

“Other consumer wearables, they all have the same sensors—accelerometer, PPG [photoplethysmography]—as our device. However, our device has a little more granularity,” in that it generates step counts every 10 seconds, she said. “That’s why we could generate more-detailed measures.”

It will be important to test, going forward, how their own calculations compare with what could be obtained by less-granular fitness trackers, Shin added.

She said more study also is needed before the tool tested here could be available for consumers to use. “To be honest,” Shin said, “I would say there will be multiple things that need to happen before we really incorporate this into the clinical flow.”

Understandably, “not all clinicians really welcome this [type of data], because it may not be reliable, [but] we tried our best,” she commented. By validating the various measures within the current dataset, she added, at a minimum, “we know that it’s reliable and accurate in this population.”

Shin stressed that it will be necessary to pinpoint what kinds of differences in walking behavior are meaningful and actionable. “I wouldn’t say that we are quite there with this first set of evidence from [our] research, but hopefully this serves as the first step to go there,” she said.

As reported by TCTMD, investigators and companies around the world are exploring numerous roles for wearable devices, from atrial fibrillation screening and management to ECG-based detection of subclinical LV dysfunction. Still to be worked out is how clinicians will absorb and apply the deluge of data headed their way, the best way to ensure patients will have equal access to novel tech’s benefits, and whether wearables save money or raise healthcare costs.

‘Subtle but Holistic Signs’

Shin and colleagues analyzed 1,265 participants in the Project Baseline Health Study who were without heart failure risk factors (stage 0), at risk for heart failure (stage A), or considered pre-heart failure (stage B) or had an adaptive remodeling phenotype known as RF-/ECHO+ (no risk factors but mild echocardiographic change at their first in-clinic assessment). Mean age was 47.3 years, and women made up 57.6% of the cohort. In terms of race/ethnicity, 67.6% were white, 8.5% were Black/African American, 8.5% were Asian, and 15% were “other.”

All participants wore a Verily watch that used an accelerometer to detect daily values for 18 walking measures related to step count, periods of activity, and intensity, among other metrics.

With multivariable adjustment, each standard deviation decrease in 17 of the measures was linked to significantly increased risks of both stage B heart failure (ORs ranging from 1.18 to 2.10) and stage A heart failure (ORs ranging from 1.07-1.45), as well as decreased the risk of RF-/ECHO+ (ORs ranging from 0.80 to 0.93).

The strongest predictor of heart failure was peak 30-minute pace, with adjusted odds ratios of 2.10 for stage B (95% CI 1.74-2.53) and 1.43 for stage A (95% CI 1.23-1.66).

Decreases in 13 of the 18 measures, meanwhile, were significantly linked to higher likelihood of stage B versus stage A heart failure.

The wearable-device measurements correlated most strongly with in-clinic testing (eg, 6-minute walk test and exercise capacity), echocardiographic testing, and participant-reported outcomes (ie, the EQ-5D-5L and World Health Organization Disability Assessment Schedule questionnaires) for stage B heart failure, less strongly for stage A heart failure, and the least for stage 0.

“These correlations suggest that there could be subtle but holistic signs (spanning from physical capacity to self-perceived health status and quality of life) that clinicians could monitor before the onset of fully symptomatic HF,” the authors conclude. “To our knowledge, this is the first [study] investigating associations between sensor-based in real life physical activity measures and presymptomatic HF.”

They specify that “at-risk and early-stage HF [are] a clinical setting primed for interventions.”

What their approach does not replace, said Shin, is the clinician’s judgment, nor does it preclude the need for diagnosis. Rather, “in the future it can be used as supplementary information in the context of the patient,” she explained. So, if the patient has known risk factors or abnormalities on echo, monitoring their walking behavior could help encourage more-timely intervention or spur an in-office visit for further testing.

Walking behavior, of course, could simply be marker for frailty, lack of fitness, or a variety of diseases, not just a harbinger of heart failure. But Shin said that in their study there appear to be clear links between their measurements and subclinical HF.

“It is true that our sample had other comorbidities,” including cardiometabolic conditions. “However, when we generated the model, we tried to do careful controlling of all the known covariates” to account for things like sex, age, and body mass index, said Shin. “The significance was still there, so we can say that it’s at least associated—we cannot really talk about causality yet, but there’s a significant association with preclinical heart failure stages.”

Spencer Carter, MD (University of Utah, Salt Lake City), and Jennifer T. Thibodeau, MD (University of Texas Southwestern Medical Center, Dallas), writing in an editorial, agree a decrease in activity could be a “warning sign” of impending HF.

“Because symptoms of HF can develop gradually, it can sometimes be hard for a patient to know that they aren’t doing as well as they had been,” they observe. “The information from the activity tracker could be monitored over time to identify new limitations in activity. This could prompt the medical team to evaluate heart function and change treatment in order to help a patient to feel better and live longer.”

Already it makes sense for patients to ask their physicians if a fitness tracker could be useful in measuring activity levels that could inform the care they receive, Carter and Thibodeau suggest.