spiral.imperial.ac.uk€¦ · web viewcan activity monitors predict outcomes in patients with...
TRANSCRIPT
Can Activity Monitors Predict Outcomes in Patients with Heart
Failure? A Systematic Review
Authors:
Matthew K.H. Tan1, Joanna K.L. Wong1, Kishan Bakrania2,3, Yusuf Abdullahi1, Leanne Harling2,3,4,
Roberto Casula2,3,4, Alex V. Rowlands2,3,4, Thanos Athanasiou1, Omar A. Jarral1
1Department of Surgery and Cancer, Imperial College London, London, United Kingdom.
2Diabetes Research Centre, University of Leicester, Leicester General Hospital, Leicester, UK
3NIHR Leicester Biomedical Research Centre, UK
4Alliance for Research in Exercise, Nutrition and Activity (ARENA), Sansom Institute for Health
Research, Division of Health Sciences, University of South Australia, Adelaide, Australia
Corresponding Author: Omar A. Jarral,
Department of Surgery and Cancer,
Imperial College London,
London, W2 1NY.
Email: [email protected];
Tel: +447855773118
ABSTRACT
Background: Actigraphy is increasingly incorporated into clinical practice to monitor intervention
effectiveness and patient health in congestive heart failure (CHF). We explored the prognostic impact
of actigraphy-quantified physical activity (AQPA) on CHF outcomes.
Methods: PubMed and Medline databases were systematically searched for cross-sectional studies,
cohort studies or randomised controlled trials from January 2007 to December 2017. We included
studies that used validated actigraphs to predict outcomes in adult HF patients. Study selection and
data extraction were performed by two independent reviewers.
Results: A total of 17 studies (15 cohort, 1 cross-sectional, 1 randomised controlled trial) were
included, reporting on 2,759 CHF patients (22-89 years, 27.7% female). Overall, AQPA showed a
strong inverse relationship with mortality and predictive utility when combined with established risk
scores, and prognostic roles in morbidity, predicting cognitive function, New York Heart Association
functional class and intercurrent events (e.g. hospitalisation), but weak relationships with health-
related quality of life scores. Studies lacked consensus regarding device choice, time points and
thresholds of PA measurement, which rendered quantitative comparisons between studies difficult.
Funding: No specific funding was provided for this review
Conclusions: AQPA has a strong prognostic role in CHF. Multiple sampling time points would allow
calculation of AQPA changes for incorporation into risk models. Consensus is needed regarding
device choice and AQPA thresholds, while data management strategies are required to fully utilise
generated data. Big data and machine learning strategies will potentially yield better predictive value
of AQPA in CHF patients.
Registration: Nil
Keywords: actigraphy, physical activity, congestive heart failure, prognostic impact, quality of life
Word count: 250 words
1. INTRODUCTION
Cardiovascular disease (CVD) is the most common cause of death in Europe despite the
substantial decrease in its mortality rates over the last decade (1). In 2012, it was the most common
cause of death in women and second most in men in the United Kingdom (2). National Health Service
(NHS) England spent £6.4b in the financial year 2012/3, representing a huge economic burden (2).
Congestive heart failure (CHF) holds the greatest burden with 30-40% mortality at 1-year(3, 4) and
makes up 1-2% of the annual healthcare budget in Western countries (5, 6). There has been a surge in
uptake of cardiac rehabilitation programmes in the UK which focuses on improving physical activity
(PA) as PA has been shown to be a predictor of premature CVD-associated morbidity and mortality
(7), with risk reduction attributed to, amongst other effects, increased cardiac output through left
ventricular dilatation and hypertrophy and decreased arterial stiffness through oxidative stress
reduction and release of vasodilatory mediators (8).
The role of PA is two-fold: not only does a lack of PA act as a risk factor of CVD, but it also
acts as a quantitative parameter of effectiveness of the interventions designed to improve PA in CVD
patients. There exists a substantial amount of prospective clinical evidence for both sedentary
behaviour and moderate-to-vigorous PA (MVPA) with regards to CVD events in non-CVD patients.
Several reviews have found that sedentary behaviour consistently increases the risk of both non-fatal
and fatal CVD events in the general adult population (9-12). Conversely, PA level has been shown to
be inversely related to CVD risk in non-CVD populations by multiple prospective studies (13-15),
including the Whitehall II study (16). Indeed, MVPA has been used as the basis for PA guidelines by
the American College of Sports Medicine/American Heart Association for adults (17), with other PA
guidelines developed to improve cardiovascular health in the general population (18).
These findings hold true for CHF patients. In a population of black adult men, a dose-
response relationship is seen between increasing MVPA levels and protection from heart failure
hospitalisations (19). Evidence from male participants belonging to a Swedish population-based
cohort suggested that both extremities of PA compared to moderate levels could increase the risk of
HF in men (20). Recognising such evidence, PA has been incorporated into management plans as
formal cardiac rehabilitation or exercise training following CVD events. Strategies used to quantify
PA involve subjective (or indirect) measurements, using self-reported techniques which are in the
form of questionnaires and interviews. These techniques are widely used due to their cost-
effectiveness, low patient burden and practicality. These have been reported to be less reliable, as
shown by Healy et al. who report a wide range of test-retest correlation coefficients (21), Fjeldsoe et
al. observing variation in PA recall depending on participants’ activity levels (22), and Tomaz et al.
showing overestimation of vigorous PA (23). The subjective nature of these strategies predispose
participants to recall and response biases which lead to the over- or underestimation of PA data (24),
and results in weak correlation to CVD outcomes (25, 26).
An alternative strategy is therefore needed to objectively and representatively quantify PA.
Technological advancement has given rise to actigraphs – a small device worn on various parts of the
body (most commonly the hip or wrist) – that monitors human activity (27). It generally consists of
three main components: 1) a piezoelectric or capacitive accelerometer, a device that measures
acceleration, 2) on-board memory to store recorded values, and 3) a wired or wireless interface from
which data can be downloaded (28).
Actigraphs measure acceleration along multiple axes which allow for the quantification of
step and activity count, and for posture to be detected (29, 30). Taken together, these data are more
meaningful than subjective and objective PA measurements using CPET, as it allows for the
situational contextualisation of PA. There is therefore potential for actigraphs to be incorporated in
clinical practice, as their multifunctionality means that they can more accurately and consequentially
predict CVD outcomes and monitor the effectiveness of cardiac rehabilitation programmes.
Physicians can provide appropriate advice to patients, and patients can take responsibility of their own
health monitoring. The utility of actigraphs is increasingly recognised, demonstrated by top industry
players’ interest in the actigraph device market.
Whilst there is a great amount of literature on the predictive role of subjectively measured PA
in CVD events (31-33), there is limited literature on the predictive role of objectively measured PA,
with only a review looking at the clinical utility of actigraphs in CHF (34). Recognising the
significance of actigraphs in the future of cardiovascular medicine, we aim to review the predictive
value of actigraphs in CHF patients, as expenditure on this group of CVD patients poses the highest
economic burden. Therefore, this systematic review aims to provide a comprehensive overview of all
studies in the past 10 years measuring the prognostic value of actigraphy-quantified physical activity
(AQPA) mortality, morbidity and health-related quality of life (HRQoL) outcomes of CHF patients,
provide insight into measurement techniques and protocols of actigraphy studies, and create a
framework to provide recommendations for future work.
2. METHODS
2.1 Search Strategy
This systematic review was performed in accordance with the Preferred Reporting Items for
Systematic Reviews and Meta-Analyses (PRISMA) guidelines (35).
PubMed and Medline databases were searched between January 2007 and December 2017 to
capture studies performed in the last 10 years to reflect contemporary device usage using a search
algorithm detailed in the online supplement S1. References of selected papers were manually searched
to ensure maximum coverage.
Two reviewers (M. T., J. W.) independently performed the literature search, removed
duplicates by title review, and reviewed the abstracts and full texts to determine the inclusion of
studies according to the pre-determined criteria. The final list of included studies from both reviewers
were then merged, and conflicts between reviewers were discussed in person with a third reviewer (O.
J.) until agreement was reached.
2.2 Inclusion and Exclusion Criteria
An initial literature review was performed to determine the scope of available literature and
pre-determine inclusion and exclusion criteria. Both non-randomised and randomised controlled trials
(RCTs) in English reporting the prognostic impact of AQPA on morbidity (defined as functional
impairment or intercurrent events), mortality or health-related quality of life (HRQoL) as measured by
validated instruments (e.g. Kansas City Cardiomyopathy Questionnaire, Minnesota Living with Heart
Failure (MLHF) score) in adult CHF were considered in this review. Both wearable or implantable
accelerometers were included. Case studies or case series, and studies reporting on children (<18-
years-old) and on devices which only measured other physiological parameters (e.g.
electrocardiography) were excluded.
2.3 Outcomes of Interest and Data Extraction
Data from included studies were independently extracted by two reviewers (M.T., J.W.). The
two datasets were combined and any conflict between reviewers was discussed in person with a third
reviewer (O.J.) until agreement was reached.
The following data were extracted: author, publication year, study period (defined as period in
which patients were recruited), study design and research question, number of patients and their
characteristics, device used, PA measurement duration and time points of measurement, PA analysis
method, primary outcome measured and main findings relating primary outcomes to AQPA. Authors
of the included studies were contacted if the data required was not found in the published literature.
2.4 Risk of Bias Assessment
Risk of bias assessment was performed using the Newcastle-Ottawa scale for non-randomised
studies. Assessment of RCTs was performed using the Cochrane Risk of Bias tool. These scores were
then converted to the Agency for Healthcare Research and Quality (AHRQ) standards, with studies
rated as “poor”, “fair” or “good quality”.
3. RESULTS
3.1 Selected Studies
The literature search identified 406 studies, of which 98 were duplicates. From the 308
studies reviewed, 17 were included (36-52) (Fig. 1). Data from these studies are summarised in the
online supplementary Table S1.
3.2 Study Objectives, Designs and Population
All studies observed the prognostic impact of objectively measured PA in CHF. 13 were
prospective cohort studies (36-43, 45-47, 50, 52), two were retrospective cohort studies (48, 49) and
one was cross-sectional (44). The remaining study was an ancillary study using patients from a
randomised controlled trial (RCT) (51). These studies recruited a total of 2,759 patients but only
2,270 (82.3%) were included in the analyses due to technical difficulties precluding complete datasets
from being collected from the excluded patients.
3.2.1 Risk of Bias Assessment
Studies included in this review were generally of good quality. According to the AHRQ
standards, two studies were rated “poor” (11.8%), one study was rated “fair” (5.9%) and 15 studies
were rated “good” (82.4%).
3.3 Devices Used
A total of 12 different devices were used and are detailed in Table 1(readers are invited to
review the reliability of most of these devices in a previous review by Van Remoortel et al. (53)). Six
were triaxial, one was biaxial, four were uniaxial, and one was omnidirectional. 10 devices were worn
on either the wrist (n=2), or waist (n=8), with one worn on both. An implantable cardioversion device
(ICD) or cardiac resynchronisation therapy defibrillator (CRT-D) that monitored PA was used in one
study (41).
3.4 PA Quantification Methods
Four main methods of quantifying PA were used: activity counts stratified by intensity, vector
magnitudes and frequencies, walking parameters, and metabolic equivalent of task (MET). Three
studies by Alosco et al. used both walking parameters and activity counts to quantify PA (37-39). One
study did not report its activity analysis method (40) and another study only considered movement-
related activity duration as a percentage of 24-hours (52). Actigraphy data was stratified according to
activity levels which had different definitions across studies.
3.4.1 Activity Counts Stratified by Intensity
Non-specific uniaxial ActiGraph activity counts were used by eight studies to quantify PA
and stratify activity by intensity (37-39, 46, 48, 49). All studies considered <100 counts/minute to be
sedentary behaviour but varied in defining light, moderate and vigorous intensity PA.
Light intensity PA (LIPA) was defined by Alosco et al. (37-39) as 100 to 760 counts/minute
based on Matthew’s calibration of accelerometer output for adults (54). However, a higher upper
threshold was used in other studies, with activity considered to be LIPA up to 1,951 (46) or 2,019
counts/minute (48, 49) based on the National Health and Nutrition Examination survey (NHANES)
methodology by Troiano et al. (55).
Further discrepancies arose in defining moderate and vigorous intensity PA. Loprinzi’s group
combined the two activity levels, considering >2,020 counts/minute to be MVPA.(48, 49) Alosco et
al.(37-39) observed two moderate intensity PA levels based on calibration studies, namely Matthew’s
moderate intensity between 760 to 5,724 counts/minute (54) and Freedson’s moderate intensity
between 1,952 to 5,724 counts/minute (56). These studies defined vigorous intensity PA to be >5,724
counts/minute.
Melin et al. (46) was the sole study which determined intra-individual variability through
quantifying the kurtosis and skewness of actigraphy data.
3.4.2 Vector Measurements
Three studies considered vector magnitudes and frequency per minute (41, 47, 51). If a pre-
defined threshold was reached, this was considered an active minute. Conraads et al. was the only
study that used implanted devices, which measured the number of minutes the patient is active per day
(41). A patient was considered “active” if pre-determined thresholds of both number and magnitude of
accelerometer deflections were reached. Waring et al. used a triaxial accelerometer, calculating the
sum of vector magnitude units (VMUs) from three planes of movement per minute (47). This study
defined the threshold for higher-intensity PA to be ≥3,000 VMUs and dichotomised their patients into
more active (≥60 minutes of higher-intensity activity) and less active (<60 minutes) groups. Finally,
Snipelisky et al. determined activity using “arbitrary accelerometer units” which were determined by
cumulative triaxial vector measurements, averaging values obtained in 15-minute epochs to give
average daily accelerometer units. All epochs with >50 units were also averaged to give hours active
per day (51).
3.4.3 Walking Parameters
Daily step counts are one of the simplest PA measurements and were used in six studies (36-
39, 43, 44). All used a daily step count averaged over the measurement period to categorise patients
into different PA levels based on predefined step counts.
Walking speed was used in a study by Jehn et al. (45) to distinguish New York Heart
Association (NYHA) functional class. This was achieved by quantifying the total amount of time
spent in each activity level: “passive” (not defined), “active” (not defined), “walking” (0 to
80m/minute) and “fast walking” (83 to 115m/minute).
3.4.4 Metabolic Equivalent of Task (MET)
Two studies converted sensor data to MET values (42, 50) which expresses the energy cost of
PA. Studies using METs mostly utilised proprietary software, and therefore unique algorithms, related
to the specific accelerometer used. Howell et al. (42) used the Actical with the following algorithm:
METs = 2.384 + (0.0007341 x activity counts) (57). Miyahara et al. determined METs every 10
seconds before stratifying patients into high (>8.4 MET-hours) or low activity (<8.4 MET-hours)
groups (50).
3.5 Prognostic Impact of AQPA
AQPA shows significant but variable prognostic value in CHF depending on the parameter
considered. However, most CHF patients generally exhibit low levels of PA and the studies
demonstrate an inverse relationship between PA and CHF outcomes (Fig. 2).
The studies have been classified by their observed primary outcomes into three categories:
mortality, morbidity and HRQoL.
3.5.1 Mortality
Five studies determined the prognostic value of AQPA in mortality (41-43, 46, 49). Three
studies looked specifically at mortality (43, 46, 49) while two (41, 42) examined it in combination
with intercurrent events (e.g. hospitalisation).
In a Japanese CHF population, Izawa et al. performed a multivariate analysis including brain
natriuretic peptide, peak oxygen uptake, minute ventilation/carbon dioxide production (VE/VCO2)
slope and number of steps taken per day into the model (43). This analysis revealed that number of
steps, dichotomised into ≤4,889.4 and >4,899.4 steps/day, was the only independent variable that had
a strong significant predictive ability for mortality. Patients with ≤4,889.4 steps/day had a mortality of
88%, 25 percentage points higher than that of the comparison group. Similarly, in a retrospective
analysis of a Western CHF population, Loprinzi reported that an increase in AQPA by 60 minutes per
day reduced the risk of all-cause mortality by 35% (49). Conraads et al. observed a similar
improvement, showing a 5% relative risk reduction in mortality or CHF-related hospitalisation with
each 10-minute increase in PA (41). Finally, Melin et al. observed the predictive value of PA
variability in CHF as determined by actigraphy (46). Skewness, indicating high PA recorded in one
period and not others, was shown to be a significant predictor of mortality.
Melin et al. also added skewness to the Heart Failure Survival Score (46). The score, being
one of the most validated prediction models in CHF (58, 59), was independently associated with death
incidence in the studied population with a c-index (area under the Receiver Operating Characteristic
curve) of 0.71 (46). This was improved to 0.74 on addition of peak 3-hour skewness, showing that
AQPA had an incremental predictive value in this CHF population. Similarly, Conraads et al. added
AQPA to the Candesartan in Heart Failure Assessment of Reduction in Mortality and Morbidity
programme risk score (60). This improved its predictive ability of death or CHF-related
hospitalisation from a c-index of 0.61 to 0.65 (41).
3.5.2 Morbidity
Five studies focused on cognitive function in CHF patients (36-40). Four studies investigated
AQPA’s relationship to hospitalisations and other intercurrent events in CHF (41, 42, 47, 50) and two
determined the ability of AQPA in predicting NYHA functional class (45, 51).
Alosco et al. also found AQPA was predictive of cognitive function independent of
anatomical or physiological changes; daily step count was associated with attention and executive
function (β=0.31, p=0.03) and language ability (β=0.35, p=0.01). This was the only study to show
prognostic value on memory, specifically episodic memory (β=0.27, p=0.049) (36). Fulcher et al. also
showed an independent predictive value of daily step count on global cognitive function (β=0.28,
p<0.01), attention and executive function (β=0.29, p<0.01) and processing speed (β=0.22, p<0.05).
However, this study showed no predictive value for memory tests (β =0.12, p=0.057) (40). Alosco et
al. further reported that a lower step count (β=0.17, p=0.057) and less time spent in Matthew’s
moderate activity (β=0.19, p<0.05) at baseline predicted attention and executive function up to 12
months later (37). In another study by the same group, decreases in either light intensity PA and
moderate-to-vigorous PA over 12 weeks also predicted decline in attention and executive function
(38).
AQPA significantly predicted risk of intercurrent events following CHF diagnosis. In these
follow-up studies, Conraads et al. showed 3% relative risk reduction for CHF-related hospitalisation
with each 10-minute increase in PA (41). This was consistent with findings from a Japanese
population, with Miyahara et al. showing a 6-fold higher CHF-related hospitalisation rate in low
activity patients when compared to those who were physically active (50). This study also showed
total physical activity to be the only significant predictor of rehospitalisation on multivariate analysis
(Odds Ratio=0.65, p=0.03) and LIPA to best predict CHF rehospitalisation (OR=0.60, p=0.03).
However, while Waring et al. showed that patients with lower PA had a higher odds ratio of
readmission within 30 days, ranging from 4.8 to 16.9 over the four weeks of data collection, this study
showed higher-intensity PA (defined as VMUs ≥3,000) to be the predictive parameter for CHF
rehospitalisation (47), and not LIPA as seen in Miyahara et al. (50). Finally, Howell et al. considered
multiple non-elective intercurrent events including mortality, hospitalisations, emergency department
visits, intercurrent illness and outpatient procedures (42). Inclusion of the magnitude of the most
active 6-minute block of the day in multivariate analysis was enough to significantly predict these
events (Hazard Ratio=2.73; 95% Confidence Interval=1.10-6.77; p=0.03), although its predictive
value for individual events was not specified.
AQPA was also able to predict NYHA functional class. Lower average daily accelerometer
units and hours active per day was shown by Snipelisky et al. to be associated with higher NYHA
class (51). Jehn et al. observed time spent at two walking speeds: “walking” (0 to 80m/minute) and
“fast walking” (83 to 115m/minute) (45). “Total walking”, “walking” and “fast walking” duration
could differentiate between NYHA class I and III, II and III (both p=0.001) but not I and II.
Importantly, monitoring for a duration of only four days for “fast walking” was enough for a
specificity and sensitivity of 74% in discriminating moderate CHF (NYHA class III).
3.5.3 HRQoL
Five studies predicted HRQoL outcomes based on AQPA (39, 44, 48, 51, 52). Alosco et al.
showed inactivity to be predictive of worse HRQoL as measured by the short form-12 (SF-12) health
survey. Patients in the inactive group could be distinguished from limited physical activity and
physically active groups based on the SF-12 physical composite score (39). Edwards and Loprinzi
observed a similar relationship using the Centres for Disease Control HRQoL measure (48), with
sedentary behaviour weakly but significantly associated with worse HRQoL (β=0.004, p=0.03) even
in a bivariate model with MVPA. Izawa et al. (44) showed a strong correlation between SF-36 mental
health components and daily steps or energy expenditure on PA. Snipelisky et al. showed greater
average daily accelerometer units and hours active per day to be predictive of improved HRQoL
scores using the Kansas City Cardiomyopathy Questionnaire and MLHF (51). Only van den Berg-
Emons et al. showed no significant relationship between MLHF scores and HRQoL (52).
4. DISCUSSION
4.1 PA Measurement Methods
While videotaping and counting steps or other activity measures is arguably the “gold
standard” for quantifying PA, this is clearly only feasible in a controlled laboratory setting and not
applicable in quantifying patients’ day-to-day activities. As mentioned previously, current studies
commonly use self-reported assessments, ranging from questionnaires to patient diaries. These
methods face subjective errors which can be overcome by objective quantification of PA using
actigraphy. However, these devices too have limitations (e.g. inability to distinguish between different
sedentary activities). A range of monitors and data analysis methods exist, making comparison
between studies difficult. Furthermore, one weakness of AQPA is the inability to analyse data based
on situational context which may be resolved using pattern recognition and machine learning. This
review will make recommendations to ease future efforts in the field.
4.1.1 Time Points of PA Measurements
Most studies took PA measurements from a single baseline period, characterising patient
activity levels at the time of study inclusion (36, 39, 40, 43-49, 61) or following intervention (41).
Only two studies by Alosco et al. quantified PA at two distinct periods, observing patient PA at
baseline and 12 months (37) and at baseline and 12 weeks in a different study (38). This may be due to
the relatively high cost of actigraphy, limiting AQPA sampling. We believe that multiple
measurements of PA are valuable as AQPA trends potentially contain prognostic value; further
obtaining repeat measurements is realistic as AQPA is becoming more feasible. For example, large-
scale objective measurement of PA is now increasingly incorporated in large surveys (e.g. UK
Biobank (62), US NHANES (63)).
Measurements of short time periods appear feasible in allowing discrimination of patients at
risk of poor cardiovascular outcomes, with most studies included in this review using a time period of
seven days (36-40, 43, 46, 48, 49) or including patients if they had at least three valid days of wear
(36-40, 45, 46).
It must be appreciated that PA in CHF patients is not a static but a dynamic parameter that
changes with time and health status. While it is promising that cross-sectional AQPA can predict risk
of negative CHF outcomes, this merely reinforces the progression of patients with poorer health and
functional status towards poorer outcomes. PA decline over 12 weeks, independent of absolute PA
quantity, was shown by Alosco et al. (38) to predict poorer cognitive function in CHF patients.
Quantifying the rate of decline might prove to be more useful in determining and predicting disease
severity and outcomes.
4.1.2 Defining Levels of PA
Thresholds which define PA levels differ between populations, most noticeably between
paediatric and adult populations. Thresholds vary because of differences in the calibration samples
and/or protocols. It should be noted that no single regression equation is able to accurately predict
energy expenditure on PA, nor is any set of cut-off points able to accurately quantify time spent in
different intensity categories across a wide range of activities (64). Multiple thresholds were used by
studies despite only including adult patients.
Notably, thresholds used by the included studies are taken from calibration studies in non-
CHF populations in laboratory environments (54). To allow for quantitative analysis of the prognostic
impact of AQPA in the future, these definitions must be standardised in future studies, with specific
thresholds tailored to different diseases. The recent move to accelerometers that store high-resolution
raw acceleration signals in non-proprietary units provides opportunities to improve the
characterisation of PA and device comparability (63).
Presently, all studies included have utilised AQPA thresholds to categorise patients, followed
by the observation of the prognostic impact of these categories on CVD outcomes. The categorisation
of patients based only on AQPA thresholds indicates the potentially missed opportunity of data
optimisation, as the wealth of data generated could contain more nuanced relationships between PA
and CVD outcome. For example, AQPA be incorporated as a continuous statistic in survival analysis,
providing a more precise predictive value in future prognostic algorithms.
4.2 Prognostic Impact of AQPA in CVD
This systematic review has shown that AQPA has an independent prognostic value for
mortality, morbidity and HRQoL outcomes in CHF patients, extending the studies discussed above.
This similar conclusion is unsurprising, with the unique difference being the objective PA
measurement method, facilitating more accurate and precise measures of PA.
4.2.1 Mortality
AQPA has been shown to have independent predictive value for mortality in CHF. In line with
the physiological benefits of PA (8) and consistent with previous studies using subjective PA
measures, greater levels of AQPA show better prognosis in CHF mortality risk. Additionally, high
variability in AQPA (i.e. intervening epochs of sedentary behaviour between LIPA or MVPA periods)
was also predictive of greater mortality (46). This suggests that quantity is not the only characteristic
of AQPA that has prognostic value in CHF, and it behoves future studies to observe nuances in AQPA
data when considering its prognostic value in CHF and CVD.
The addition of AQPA to existing risk scores is a potential application of PA data, providing it
adds value to and is not seen as burdensome compared to existing risk assessments. While only two
studies investigated and observed enhanced predictive abilities of risk scores (41, 46) this could pave
the way for a convenient and non-invasive method of quantifying risk in the future.
4.2.2 Morbidity
As with mortality, AQPA has independent predictive value for morbidity, particularly towards
cognitive function, intercurrent events, and functional classification.
While cognitive impairment is a common morbidity in CHF patients, its aetiology has not
been well-elucidated, but is thought to be related to structural alterations (65-68) and cerebral
perfusion (61, 68). As discussed in the introduction, PA causes physiological changes in
cardiovascular parameters, including vascular dilatation, which could contribute to increased cerebral
perfusion. AQPA provides an easy, non-invasive method to provide objective values to predict such
changes and intervene earlier in patients at risk of such impairment.
This too holds true for non-fatal intercurrent events and functional classification. AQPA
provides an objective method to predict the risk of individual patients. From a clinical perspective, this
may allow additional resources to be allocated to high-risk patients, with aims to reduce functional
impairment and unplanned visits to hospital.
4.2.3 HRQoL
Across the studies, there exists a positive correlation between PA and HRQoL, with sedentary
behaviour resulting in poorer HRQoL overall. It was not possible to make quantitative comparisons
between studies due to the variety of instruments used, including the SF-12, SF-36, MLHF and the
CDC HRQoL measure. However, it should be noted that these HRQoL instruments are based on
patient reporting, thus incurring a psychological element in their reports. It is uncertain to what extent
their reporting is attributed to the actual physiological limitation of CHF and the psychological
perception of their disease (69). Though AQPA can quantify the decrease in PA, the magnitudes
attributable to disease or patient perception have not been elucidated. The relationship between AQPA
and physical function should be examined further.
5. SUGGESTIONS FOR FUTURE RESEARCH
Although actigraphy has been used in research for over 20 years, its clinical application is still
relatively novel due to limitations discussed above. Based on these, we suggest improvements to this
promising field of research (Fig. 3).
5.1 AQPA Measurement
Current threshold definitions are based on healthy populations in laboratory-based validation
studies. PA limitations vary with pathology and these quantitative definitions may not be
representative when applied to patient populations. These levels may also lack sensitivity to stratify
patients who are all physically limited and may explain why most studies found low activity levels in
their cohorts. Furthermore, dysfunction in physical activity occurs not only in CHF but in almost all
CVDs, including CAD, arrhythmic, valvular, aortic and peripheral vascular disease. Therefore, AQPA
metrics and level thresholds need to be standardised, not for CVD as a whole but for individual
diseases.
5.2 Study Design
Studies included in this review analysed small to moderate sized patient cohorts (range 36-
836) and further studies in this field should ideally determine the relationships of AQPA to CHF and
CVD outcomes in larger patient populations. Additionally, as most studies used a single timepoint for
AQPA measurement, further studies would benefit from multiple measurement timepoints to
determine how changes in AQPA affect disease severity and outcomes. AQPA studies to investigate
the effect of any CVD intervention could be performed to relate AQPA improvements with
intervention, given that interventions including CABG, valve replacement/repair or angioplasty are
known to reduce mortality and morbidity. AQPA has also been shown to be related to cardiovascular
biomarkers such as serum apolipoproteins in elderly patients (70) and may hold such a relationship in
all CVD patients.
Future studies should strive to use wireless technologies which permit continuous data
transfer, allowing for real-time analysis instead of post-hoc determination of risk. Having shown that
both quantitative and qualitative descriptors of AQPA have predictive ability for mortality, morbidity
and HRQoL, these studies should guide the next step in determining the temporal relationship between
AQPA and CHF outcomes. Conraads et al. for example showed that early AQPA, defined as <30 days
post-intervention, had significant impact on death or CHF hospitalisation (41). It would be interesting
to observe if such a chronological correlation holds for other time epochs.
Studies should also consider devising methods to engage patients in the use of actigraphy data.
Remote monitoring, without the input of nurses or doctors, has already been shown to have significant
benefit in patients’ clinical outcomes (71). Given the tangible nature of the information provided by
actigraphs, the information could empower them to engage in desirable activities solely due to
awareness their personal PA levels or through methods such as gamification (72).
5.3 Data Management
Greater inclusion of actigraphy in clinical practice represents diagnostic, prognostic and
management potential. However, with this comes logistical and analytical challenges – large quantities
of data are generated with data being continually recorded over the measurement period. In larger
epidemiological studies, this can run into the region of terabytes even if recorded for a week. Current
strategies of big data management and analytics in CVD have been reviewed here (73, 74).
Application of “smart” algorithms could be valuable for predicting morbidity in CHF patients.
Machine learning is increasingly applied to various industries, including healthcare. Although still
controversial due to data protection issues, machine learning has been shown to be an accurate means
of identifying at-risk patients in both inpatient (e.g. acute kidney injury (75)) and outpatient (e.g.
breath sound telemonitoring predicting chronic obstructive pulmonary disease exacerbation (76)) non-
CVD contexts. Current morbidity and mortality risk scores are static and rely on fixed equations.
Though extensively validated, these scores are decades old and may no longer represent the population
today. “Smart” algorithms on the other hand can evolve constantly with each patient added to the
database, allowing dynamic characterisations of patients. AQPA shows longitudinal predictive value
for cognitive function and mortality and might be incorporated into these algorithms to identify at-risk
patients prior to decline. Earlier identification would allow prompt intervention and prevention of
impairment.
6. CONCLUSION
AQPA is increasingly feasible with advancements and integration of technology into clinical
practice. Early studies show clear predictive ability of these objectively measured PA parameters on
mortality, morbidity and HRQoL, with poorer prognosis associated with lower free-living PA. The
results are consistent with current established knowledge and mirror findings from self-reported PA
measures but has an advantage in providing quantifiable values. This objective data may therefore
play a role in predicting morbidity and mortality outcomes in CHF patients and may also provide
more accurate risk stratification through complementing pre-existing risk scores. There needs to be
increased technology adoption in CVDs besides HF. Increased use of actigraphy, potentially with
inclusion of commercial devices, will result in greater data generation, making big data strategies
relevant to realise its full prognostic potential.
7. ACKNOWLEDGEMENTS AND FUNDING
This project was jointly funded by the National Institute for Health Research (NIHR)
Biomedical Research Centre (based at Imperial College London and Imperial College Healthcare
NHS Trust) and the NIHR Cardiovascular Biomedical Research Unit (based at the Royal Brompton
and Harefield NHS Foundation Trust). A.R. and K.B. are with the NIHR Biomedical Research Centre
(based at University Hospitals of Leicester and Loughborough University), the NIHR Collaboration
for Leadership in Applied Health Research and Care – East Midlands and the Leicester Clinical Trials
Unit. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or
the Department of Health.
8. AUTHOR CONTRIBUTION
M.T. and J.W. contributed to study selection, data extraction, data analysis and preparation of the
manuscript.
O.J. contributed to study selection, and preparation of the manuscript.
9. CONFLICT OF INTERESTS
None declared.
10. ETHICS APPROVAL
No formal ethics approval was required.
REFERENCES
1. Townsend N, Nichols M, Scarborough P, Rayner M. Cardiovascular disease in Europe 2015: epidemiological update. European heart journal. 2015;36(40):2673-4.2. Bhatnagar P, Wickramasinghe K, Williams J, Rayner M, Townsend N. The epidemiology of cardiovascular disease in the UK 2014. Heart. 2015;101(15):1182-9.3. The survival of patients with heart failure with preserved or reduced left ventricular ejection fraction: an individual patient data meta-analysis. European heart journal. 2012;33(14):1750-7.4. Centre NCG. CHRONIC HEART FAILURE. National clinical guideline for diagnosis and management in primary and secondary care. 2010.5. Braunschweig F, Cowie MR, Auricchio A. What are the costs of heart failure? Europace : European pacing, arrhythmias, and cardiac electrophysiology : journal of the working groups on cardiac pacing, arrhythmias, and cardiac cellular electrophysiology of the European Society of Cardiology. 2011;13 Suppl 2:ii13-7.6. Liao L, Allen LA, Whellan DJ. Economic burden of heart failure in the elderly. PharmacoEconomics. 2008;26(6):447-62.7. Dalal HM, Doherty P, Taylor RS. Cardiac rehabilitation. BMJ : British Medical Journal. 2015;351.8. Lavie CJ, Arena R, Swift DL, Johannsen NM, Sui X, Lee DC, et al. Exercise and the cardiovascular system: clinical science and cardiovascular outcomes. Circ Res. 2015;117(2):207-19.9. Biswas A, Oh PI, Faulkner GE, Bajaj RR, Silver MA, Mitchell MS, et al. Sedentary time and its association with risk for disease incidence, mortality, and hospitalization in adults: a systematic review and meta-analysis. Ann Intern Med. 2015;162(2):123-32.10. Ford ES, Caspersen CJ. Sedentary behaviour and cardiovascular disease: a review of prospective studies. Int J Epidemiol. 2012;41(5):1338-53.11. Grontved A, Hu FB. Television viewing and risk of type 2 diabetes, cardiovascular disease, and all-cause mortality: a meta-analysis. JAMA. 2011;305(23):2448-55.12. Wilmot EG, Edwardson CL, Achana FA, Davies MJ, Gorely T, Gray LJ, et al. Sedentary time in adults and the association with diabetes, cardiovascular disease and death: systematic review and meta-analysis. Diabetologia. 2012;55(11):2895-905.13. Lee IM, Rexrode KM, Cook NR, Manson JE, Buring JE. Physical activity and coronary heart disease in women: is "no pain, no gain" passe? JAMA. 2001;285(11):1447-54.
14. Paffenbarger RS, Jr., Hyde RT, Wing AL, Lee IM, Jung DL, Kampert JB. The association of changes in physical-activity level and other lifestyle characteristics with mortality among men. N Engl J Med. 1993;328(8):538-45.15. Rockhill B, Willett WC, Manson JE, Leitzmann MF, Stampfer MJ, Hunter DJ, et al. Physical activity and mortality: a prospective study among women. Am J Public Health. 2001;91(4):578-83.16. Batty GD, Shipley MJ, Marmot M, Smith GD. Physical activity and cause-specific mortality in men: further evidence from the Whitehall study. Eur J Epidemiol. 2001;17(9):863-9.17. Haskell WL, Lee IM, Pate RR, Powell KE, Blair SN, Franklin BA, et al. Physical activity and public health: updated recommendation for adults from the American College of Sports Medicine and the American Heart Association. Circulation. 2007;116(9):1081-93.18. Global Recommendations on Physical Activity for Health. WHO Guidelines Approved by the Guidelines Review Committee. Geneva2010.19. Koo P, Gjelsvik A, Choudhary G, Wu WC, Wang W, McCool FD, et al. Prospective Association of Physical Activity and Heart Failure Hospitalizations Among Black Adults With Normal Ejection Fraction: The Jackson Heart Study. Journal of the American Heart Association. 2017;6(9).20. Rahman I, Bellavia A, Wolk A, Orsini N. Physical Activity and Heart Failure Risk in a Prospective Study of Men. JACC Heart failure. 2015;3(9):681-7.21. Healy GN, Clark BK, Winkler EA, Gardiner PA, Brown WJ, Matthews CE. Measurement of adults' sedentary time in population-based studies. Am J Prev Med. 2011;41(2):216-27.22. Fjeldsoe BS, Winkler EA, Marshall AL, Eakin EG, Reeves MM. Active adults recall their physical activity differently to less active adults: test-retest reliability and validity of a physical activity survey. Health Promot J Austr. 2013;24(1):26-31.23. Tomaz SA, Lambert EV, Karpul D, Kolbe-Alexander TL. Cardiovascular fitness is associated with bias between self-reported and objectively measured physical activity. Eur J Sport Sci. 2016;16(1):149-57.24. Prince SA, Adamo KB, Hamel ME, Hardt J, Connor Gorber S, Tremblay M. A comparison of direct versus self-report measures for assessing physical activity in adults: a systematic review. Int J Behav Nutr Phys Act. 2008;5:56.25. Kowalski K, Rhodes R, Naylor PJ, Tuokko H, MacDonald S. Direct and indirect measurement of physical activity in older adults: a systematic review of the literature. Int J Behav Nutr Phys Act. 2012;9:148.26. Alessa HB, Chomistek AK, Hankinson SE, Barnett JB, Rood J, Matthews CE, et al. Objective Measures of Physical Activity and Cardiometabolic and Endocrine Biomarkers. Med Sci Sports Exerc. 2017;49(9):1817-25.
27. Yang C-C, Hsu Y-L. A Review of Accelerometry-Based Wearable Motion Detectors for Physical Activity Monitoring. Sensors (Basel, Switzerland). 2010;10(8):7772-88.28. Chen KY, Janz KF, Zhu W, Brychta RJ. RE-DEFINING THE ROLES OF SENSORS IN OBJECTIVE PHYSICAL ACTIVITY MONITORING. Medicine and science in sports and exercise. 2012;44(1 Suppl 1):S13-S23.29. Yang CC, Hsu YL. A review of accelerometry-based wearable motion detectors for physical activity monitoring. Sensors (Basel). 2010;10(8):7772-88.30. Luinge HJ, Veltink PH. Measuring orientation of human body segments using miniature gyroscopes and accelerometers. Med Biol Eng Comput. 2005;43(2):273-82.31. Agha G, Loucks EB, Tinker LF, Waring ME, Michaud DS, Foraker RE, et al. Healthy Lifestyle and Decreasing Risk of Heart Failure in Women: The Women's Health Initiative Observational Study. Journal of the American College of Cardiology. 2014;64(17):1777-85.32. Kraigher-Krainer E, Lyass A, Massaro Joseph M, Lee Douglas S, Ho Jennifer E, Levy D, et al. Association of physical activity and heart failure with preserved vs. reduced ejection fraction in the elderly: the Framingham Heart Study. European Journal of Heart Failure. 2014;15(7):742-6.33. Djoussé L, Driver JA, Gaziano J. Relation between modifiable lifestyle factors and lifetime risk of heart failure. JAMA. 2009;302(4):394-400.34. Alharbi M, Straiton N, Gallagher R. Harnessing the Potential of Wearable Activity Trackers for Heart Failure Self-Care. Current Heart Failure Reports. 2017;14(1):23-9.35. Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gotzsche PC, Ioannidis JP, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. Ann Intern Med. 2009;151(4):W65-94.36. Alosco ML, Brickman AM, Spitznagel MB, Sweet LH, Josephson R, Griffith EY, et al. Daily Physical Activity Is Associated with Subcortical Brain Volume and Cognition in Heart Failure. J Int Neuropsychol Soc. 2015;21(10):851-60.37. Alosco ML, Spitznagel MB, Cohen R, Raz N, Sweet LH, Josephson R, et al. Decreased physical activity predicts cognitive dysfunction and reduced cerebral blood flow in heart failure. J Neurol Sci. 2014;339(1-2):169-75.38. Alosco ML, Spitznagel MB, Cohen R, Sweet LH, Hayes SM, Josephson R, et al. Decreases in daily physical activity predict acute decline in attention and executive function in heart failure. J Card Fail. 2015;21(4):339-46.39. Alosco ML, Spitznagel MB, Miller L, Raz N, Cohen R, Sweet LH, et al. Depression is associated with reduced physical activity in persons with heart failure. Health Psychol. 2012;31(6):754-62.
40. Fulcher KK, Alosco ML, Miller L, Spitznagel MB, Cohen R, Raz N, et al. Greater physical activity is associated with better cognitive function in heart failure. Health Psychol. 2014;33(11):1337-43.41. Conraads VM, Spruit MA, Braunschweig F, Cowie MR, Tavazzi L, Borggrefe M, et al. Physical activity measured with implanted devices predicts patient outcome in chronic heart failure. Circ Heart Fail. 2014;7(2):279-87.42. Howell J, Strong BM, Weisenberg J, Kakade A, Gao Q, Cuddihy P, et al. Maximum daily 6 minutes of activity: an index of functional capacity derived from actigraphy and its application to older adults with heart failure. J Am Geriatr Soc. 2010;58(5):931-6.43. Izawa KP, Watanabe S, Oka K, Hiraki K, Morio Y, Kasahara Y, et al. Usefulness of step counts to predict mortality in Japanese patients with heart failure. Am J Cardiol. 2013;111(12):1767-71.44. Izawa KP, Watanabe S, Oka K, Hiraki K, Morio Y, Kasahara Y, et al. Association between mental health and physical activity in patients with chronic heart failure. Disabil Rehabil. 2014;36(3):250-4.45. Jehn M, Schmidt-Trucksass A, Schuster T, Weis M, Hanssen H, Halle M, et al. Daily walking performance as an independent predictor of advanced heart failure: Prediction of exercise capacity in chronic heart failure. Am Heart J. 2009;157(2):292-8.46. Melin M, Hagerman I, Gonon A, Gustafsson T, Rullman E. Variability in Physical Activity Assessed with Accelerometer Is an Independent Predictor of Mortality in CHF Patients. PLoS One. 2016;11(4):e0153036.47. Waring T, Gross K, Soucier R, ZuWallack R. Measured Physical Activity and 30-Day Rehospitalization in Heart Failure Patients. J Cardiopulm Rehabil Prev. 2017;37(2):124-9.48. Edwards MK, Loprinzi PD. Sedentary behavior & health-related quality of life among congestive heart failure patients. Int J Cardiol. 2016;220:520-3.49. Loprinzi PD. The effects of free-living physical activity on mortality after congestive heart failure diagnosis. Int J Cardiol. 2016;203:598-9.50. Miyahara S, Fujimoto N, Dohi K, Sugiura E, Moriwaki K, Omori T, et al. Post-discharge Light-Intensity Physical Activity Predicts Rehospitalization of Older Japanese Patients With Heart Failure. Journal of cardiopulmonary rehabilitation and prevention. 2017.51. Snipelisky D, Kelly J, Levine JA, Koepp GA, Anstrom KJ, McNulty SE, et al. Accelerometer-Measured Daily Activity in Heart Failure With Preserved Ejection Fraction. Clinical Correlates and Association With Standard Heart Failure Severity Indices. 2017;10(6):e003878.52. van den Berg-Emons RJ, Bussmann JB, Balk AH, Stam HJ. Factors associated with the level of movement-related everyday activity and quality of life in people with chronic heart failure. Phys Ther. 2005;85(12):1340-8.53. Van Remoortel H, Giavedoni S, Raste Y, Burtin C, Louvaris Z, Gimeno-Santos E, et al. Validity of activity monitors in health and chronic disease: a
systematic review. International Journal of Behavioral Nutrition and Physical Activity. 2012;9(1):84.54. Matthew CE. Calibration of accelerometer output for adults. Med Sci Sports Exerc. 2005;37(11 Suppl):S512-22.55. Troiano RP, Berrigan D, Dodd KW, Masse LC, Tilert T, McDowell M. Physical activity in the United States measured by accelerometer. Med Sci Sports Exerc. 2008;40(1):181-8.56. Freedson PS, Melanson E, Sirard J. Calibration of the Computer Science and Applications, Inc. accelerometer. Med Sci Sports Exerc. 1998;30(5):777-81.57. Heil DP. Predicting activity energy expenditure using the Actical activity monitor. Res Q Exerc Sport. 2006;77(1):64-80.58. Aaronson KD, Schwartz JS, Chen TM, Wong KL, Goin JE, Mancini DM. Development and prospective validation of a clinical index to predict survival in ambulatory patients referred for cardiac transplant evaluation. Circulation. 1997;95(12):2660-7.59. Lund LH, Aaronson KD, Mancini DM. Validation of peak exercise oxygen consumption and the Heart Failure Survival Score for serial risk stratification in advanced heart failure. Am J Cardiol. 2005;95(6):734-41.60. Pocock SJ, Wang D, Pfeffer MA, Yusuf S, McMurray JJ, Swedberg KB, et al. Predictors of mortality and morbidity in patients with chronic heart failure. European heart journal. 2006;27(1):65-75.61. Alosco ML, Spitznagel MB, Raz N, Cohen R, Sweet LH, Garcia S, et al. The interactive effects of cerebral perfusion and depression on cognitive function in older adults with heart failure. Psychosom Med. 2013;75(7):632-9.62. Doherty A, Jackson D, Hammerla N, Plotz T, Olivier P, Granat MH, et al. Large Scale Population Assessment of Physical Activity Using Wrist Worn Accelerometers: The UK Biobank Study. PLoS One. 2017;12(2):e0169649.63. Troiano RP, McClain JJ, Brychta RJ, Chen KY. Evolution of accelerometer methods for physical activity research. Br J Sports Med. 2014;48(13):1019-23.64. Crouter SE, Clowers KG, Bassett DR, Jr. A novel method for using accelerometer data to predict energy expenditure. J Appl Physiol (1985). 2006;100(4):1324-31.65. Pan A, Kumar R, Macey PM, Fonarow GC, Harper RM, Woo MA. Visual assessment of brain magnetic resonance imaging detects injury to cognitive regulatory sites in patients with heart failure. J Card Fail. 2013;19(2):94-100.66. Vogels RL, van der Flier WM, van Harten B, Gouw AA, Scheltens P, Schroeder-Tanka JM, et al. Brain magnetic resonance imaging abnormalities in patients with heart failure. Eur J Heart Fail. 2007;9(10):1003-9.
67. Woo MA, Kumar R, Macey PM, Fonarow GC, Harper RM. Brain injury in autonomic, emotional, and cognitive regulatory areas in patients with heart failure. J Card Fail. 2009;15(3):214-23.68. Kumar R, Yadav SK, Palomares JA, Park B, Joshi SH, Ogren JA, et al. Reduced regional brain cortical thickness in patients with heart failure. PLoS One. 2015;10(5):e0126595.69. MacMahon KA, Lip GH. Psychological factors in heart failure: A review of the literature. Archives of Internal Medicine. 2002;162(5):509-16.70. Ahmed K, Rask P, Hurtig-Wennlof A. Serum apolipoproteins, apoB/apoA-I ratio and objectively measured physical activity in elderly. Scand Cardiovasc J. 2011;45(2):105-11.71. Noah B, Keller MS, Mosadeghi S, Stein L, Johl S, Delshad S, et al. Impact of remote patient monitoring on clinical outcomes: an updated meta-analysis of randomized controlled trials. npj Digital Medicine. 2018;1(1):20172.72. Jaarsma T, Klompstra L, Ben Gal T, Boyne J, Vellone E, Bäck M, et al. Increasing exercise capacity and quality of life of patients with heart failure through Wii gaming: the rationale, design and methodology of the HF-Wii study; a multicentre randomized controlled trial. European Journal of Heart Failure. 2015;17(7):743-8.73. Hemingway H, Asselbergs FW, Danesh J, Dobson R, Maniadakis N, Maggioni A, et al. Big data from electronic health records for early and late translational cardiovascular research: challenges and potential. European heart journal. 2017:ehx487-ehx.74. Alexander C, Wang L. Big Data Analytics in Heart Attack Prediction. J Nurs Care. 2017;6(393):2167-1168.1000393.75. Kate RJ, Perez RM, Mazumdar D, Pasupathy KS, Nilakantan V. Prediction and detection models for acute kidney injury in hospitalized older adults. BMC Med Inform Decis Mak. 2016;16:39.76. Fernandez-Granero MA, Sanchez-Morillo D, Leon-Jimenez A. Computerised Analysis of Telemonitored Respiratory Sounds for Predicting Acute Exacerbations of COPD. Sensors (Basel). 2015;15(10):26978-96.
Figure Legend
Figure 1: Study selection process.
Figure 2: Outcomes of cardiovascular disease predicted by activity monitors.
Figure 3: Recommendations for future research.
Figure 2
Table 1: Summary of devices used.
Device Manufacturer Accelerometer Type No. of Axes
Sensor Placement (*: wear-sites in included studies)
Data Transmission
Weight (g)
Dimensions (mm)
Included Studies
Kenz Lifecorder uniaxial accelerometer
Suzuken Co. Ltd., Nagoya, Japan
Piezoelectric 1 Waist* USB 60 72.5 x 41.5 x 27.5
(43, 44)
ActiGraph 7164 accelerometer
Actigraph, Pensacola, FL, USA
Bimorph piezoelectric cantilever beam
1 Wrist, waist*, thigh, ankle USB 43 51 x 14 x 15 (48, 49)
ActiGraph GT1M Actigraph, Pensacola, FL, USA
Capacitive: ADXL320 (Analog Devices, Norwood, MA)
2 Wrist, waist*, thigh, ankle USB 27 38 x 37 x 18 (36-40)
ActiGraph GT3X Actigraph, Pensacola, FL, USA
Capacitive: ADXL335 (Analog Devices, Norwood MA)
3 Wrist, waist*, thigh, ankle USB 27 38 x 37 x 18 (46)
ActiGraph GT3X+ Actigraph, Pensacola, FL, USA
Capacitive 3 Wrist*, waist*, thigh, ankle USB 19 46 x 33 x 15 (47)
ActiGraph wGT3X-BT
Actigraph, Pensacola, FL, USA
Capacitive 3 Wrist*, waist*, thigh, ankle USB 19 46 x 33 x 15 (47)
Actical actigraph device
Minimitter Inc., Respironics, Bend, OR, USA
Bimorph lead zirconate titanate piezoelectric: muRata Piezotite® (Kyoto, Japan) PKGS-LD-R series
Omni Ankle, waist, wrist* 9-pin RS-232 serial port
17 28 x 27 x 10 (42)
ICD/CRT-D with the OptiVol feature
Medtronic Inc, Minneapolis, MN, USA
NA 1 Implanted Wireless NA NA (41)
Kinetic activity monitor (specific model not named)
Kersh Health, Plano, TX, USA
Capacitive (KXUD9-2050, Kionix, Ithaca, NY, USA)
3 Waist* USB NA NA (51)
HJA-350IT Omron Healthcare Co. Ltd., Kyoto, Japan
NA 3 Waist* NA 60 74 x 46 x 34 (50)
Activity monitor (specific device not
Temec Instruments, the Netherlands
Capacitive 1 Waist* (note: placement of accelerometers in study was on
NA 500 NA (52)
named) sternum and thigh)Accelerometer (specific device not named)
Aipermon® GmbH, Germany
NA 3 NA NA NA NA (45)
Supplemental Material
S1: Terms used in the search strategy.
(angina OR “coronary artery disease” OR CAD OR “heart failure” OR HF OR “ischaemic heart
disease” OR IHD OR “pulmonary hypertension” OR arrhythmia OR “atrial fibrillation” OR
“coronary artery bypass graft” OR CABG OR “valve repair” OR “valve replacement” OR “valve
stenosis” OR “valve regurgitation” OR “aortic aneurysm” OR “aortic dissection” OR “aortic
rupture” OR “carotid artery stenosis” OR “peripheral artery disease” OR “intermittent claudication”)
AND (“single axis” OR triaxial OR accelerometry OR accelerometer OR “activity monitor” OR
geneactiv OR actigraph OR actiwatch OR activinsights OR hexoskin OR “lumo back” OR “shine
misfit” OR “stayhealthy RT3” OR RT3)
Table S1: Summary of findings of included studies.
Author, year of publication, study period and design,
study quality
Research question and patient no. and characteristics
Device used and wear-site
Duration of measurement Primary outcomesMeasurement time points
Main findings related to activity monitorsActivity analysis method
Alosco et al. 2015 (36)
Study period not reported
Prospective cohort study
Good quality
Examine the benefits of physical activity on the brain in CHF patients and related cognitive implications
92 patients recruited, 50 patients included in analysis 68.2±9.32 years old 62.0% male
GT1M accelerometer (Actigraph, Pensacola, FL)
Right waist
7 days Subcortical brain volume: whole-brain standard 3D T1-weighted images
Cognitive function: attention/executive function, episodic memory, language function
Baseline Average 4,348.49±2092.08 steps/day (24.0% sedentary, 40.0% limited physical activity, 36.0% physically active)
Greater daily steps/day predicted increased subcortical, thalamus and ventral diencephalon volume
Daily step count was positively correlated with cognitive function in all domains:- Attention/executive function (β=0.31, p=0.03)- Episodic memory (β=0.27, p=0.049)- Language (β=0.35, p=0.01)
Daily step count: 0-2,499 steps: sedentary 2,500-4,999 steps: limited
physical activity 5,000-12,000 steps: physically
active
Alosco et al. 2014 (37)
Study period not reported
Prospective cohort study
Good quality
Determine whether baseline levels of physical activity predict cognitive function and cerebral perfusion at 12 months in CHF patients
145 patients recruited, 65 patients included in analysis 69.8±10.1 years old 72.3% male
GT1M accelerometer (Actigraph, Pensacola, FL)
Right waist
7 days Cognitive function: MMSE and Trail Making Test A and B
Cerebral blood flow (CBF): transcranial Doppler US measured cerebral blood flow velocity
Baseline and 12-months Incomplete actigraphy data due to mechanical issues and/or invalid wear resulted in 48 patients excluded from analysis
Reduced baseline daily step count (p=0.04) and less time spent in Matthews' moderate activity at baseline (p=0.049) was a significant predictor of 12-month attention/executive function
Less time spent in Matthew's moderate activity at baseline (p=0.05) predicted worse cerebral blood flow volume
Daily step count: 0-2,499 steps: sedentary 2,500-4,999 steps: limited
physical activity 5,000-12,000 steps: physically
active
Average number of minutes in each of the following activity levels: Sedentary: <100 counts/min Light intensity: 100-760
counts/min Matthews' moderate intensity:
760-5,724 counts/min
Freedson's moderate intensity: 1,952-5,724 counts/min
Vigorous intensity: >5,724 counts/min
Alosco et al. 2015 (38)
Study period not reported
Prospective cohort study
Good quality
Examine if changes in physical activity predicts cognitive changes over 12 weeks in older adults with CHF
145 patients recruited, 57 patients included in analysis 69.7±10.3 years old 59.6% male
GT1M accelerometer (Actigraph, Pensacola, FL)
Right waist
7 days Cognitive function: attention/executive function, episodic memory, language function
Baseline and 12-weeks Excluded patients due to attrition, missing data and invalid accelerometer data due to invalid wear or mechanical issues
High rates of physical inactivity at baseline with an average of:- 597.8±75.9 minutes/day being sedentary (<100 counts/min)- 46.0±34.6 minutes/day in moderate-vigorous activity (>760 counts/min)
High rates of physical inactivity at 12 weeks:- 583.3±62.9 minutes/day being sedentary- Significant decline in daily step count
Physical activity decline predicts changes in attention/executive function
See Alosco et al. 2014 (37)
Alosco et al. 2012 (39)
Study period not reported
Prospective cohort study
Good quality
Examine the association of low physical activity with adverse outcome measures in CHF patients
123 patients recruited, 96 patients included in analysis 69.8±8.79 years old 63.5% male
GT1M accelerometer (Actigraph, Pensacola, FL)
Right waist
7 days Quality of life: SF-12
Cognitive function: MMSE and Trail Making Test A and B
Baseline Generally low levels of physical activity:- Average 3,677±2,121.16 steps/day (32.3% inactive, 45.8% limited physical activity, 21.9% physically active)- Minimal time in Freedson's moderate (1,952-5,724 counts/min) (7.50±11.5min/day), Matthew's moderate (760 -5724 counts/min) (48.2±37.2min/day) or vigorous (>5,724 counts/min) (0.31±1.55min/day) intensity
Sedentary group significantly different from:- Both limited physical activity and physically active group on the SF-12 physical composite score- Limited physical activity group on the MMSE
See Alosco et al. 2014 (37)
Fulcher et al. 2014 (40)
Study period not reported
Examine the effects of different aspects of physical activity on cognitive function in older
GT1M accelerometer (Actigraph, Pensacola, FL)
7 days Cognitive function: global function, attention/executive function, episodic memory
Baseline 65 patients excluded due to incomplete actigraphy data due to mechanical issuesNot reported
Prospective cohort study
Good quality
adults with CHF
159 patients recruited, 93 patients included in analysis 60.7±15.4 years old 52% male
Right waist Generally low levels of physical activity (46.3% inactive, 27.5% limited physical activity, 26.3% physically active) with large proportion of time being sedentary (587±75 minutes/day)
Average daily step count independently predicted global cognitive function (p<0.026), attention/executive function (p<0.001), processing speed (p<0.032)
Lower physical activity predicts poorer global cognitive function (p<0.022) and attention/executive function (p<0.001) in CHF but not memory
Conraads et al. 2014 (41)
2005 to 2009
Prospective cohort study
Good quality
Determine the extent that early daily physical activity measured by implanted devices is related to CHF patient outcomes
836 patients recruited, 731 patients included in analysis 65±10 years old 85% male
ICD/CRT-D (InSync Sentry, Concerto and Virtuoso) with the OptiVol feature (Medtronic Inc, Minneapolis, MN)
Implanted device
15-18 months Death or HF hospitalisationContinuous measurement 5% relative risk reduction for death or CHF
hospitalisation for each 10 minutes/day additional activity- HR=0.92 for death- HR=0.97 for CHF hospitalisation
Higher activity levels were associated with lower incidence of primary outcome (death or CHF hospitalisation):- High activity (>235 minutes/day): 12.5% primary outcome, 2.5% mortality- Medium activity (146-235 minutes/day): 17.5% primary outcome, 9.9% mortality- Low activity (<145 minutes/day): 30% primary outcome, 22.0% mortality
Adding physical activity to a validated risk score (CHARM) for all-cause death in CHF patients improved its predictive ability
Minute is considered active if threshold of number and magnitude of deflections in accelerometer signal is reached
Number of minutes a patient is active per day is recorded
Early physical activity is defined as the average daily activity over earliest 30-day period in study
Howell et al. 2010 (42)
Study period not reported
Prospective cohort study
Good quality
Determine the feasibility of continuous monitoring using actigraphy and associations between peak daily 6 minutes of activity with functional capacity measurements, intercurrent morbid events and its prognostic utility in CHF
60 patients
Actical actigraph device (Minimitter, Inc., Respironics, Bend, OR)
Non-dominant wrist
9 months Occurrence of nonelective intercurrent events, including: Deaths Hospitalisations Emergency department visits Intercurrent illness Outpatient procedures
Continuous measurement Most active 6-minute MET value of 22 was found to be the most robust actigraphy parameter
Most active 6-minute value was a significant predictor
Activity was recorded in 1-minute epochs generating activity count for each minute of the day
60.7±15.4 years old 52% male Average activity counts generated
for most active epochs (6 minutes, 15 minutes, 1 hour, 10 hours) and least active epochs (5 hours)
Activity counts were converted to METs using Actical software
of subsequent intercurrent events (HR=2.73; 95% CI=1.10-6.77; p=0.03) on multivariate analysis
Izawa et al. 2013 (43)
November 2002 to October 2010
Prospective cohort study
Good quality
Determine the relationship between peak VO2, VE/VCO2 slope, physical activity and mortality and cut-off values associated with a reduction in mortality in CHF patients
477 patients recruited, 201 patients met inclusion criteria, and 174 patients included in analysis 65.2±8.5 years old 77% male
Kenz Lifecorder uniaxial accelerometer (Suzuken Co. Ltd., Nagoya, Japan)
Waist (side not specified)
7 days Mortality from cardiac-related deathNot reported Multivariate analysis revealed only step count
≤4,889.4/day to be a significant strong and independent predictor of survival (HR=2.28)
Patients with ≤4,889.4 steps/day had a significantly higher mortality rate than those with >4,889.4 steps/day (88% v.s. 63%; p=0.0005)
Average daily number of steps taken over 7 days
Izawa et al. 2014 (44)
November 2006 to October 2011
Cross-sectional study
Poor quality
Determine self-reported mental health-related differences associated with PA and target values of PA for improved mental health in CHF outpatients
261 patients recruited, 243 patients included in analysis and divided into high mental health (n=148, 57.3±11.1 years old, 76.7% male) and poor mental health (n=95, 56.8±11.3 years old, 88.6% male) groups
Kenz Lifecorder uniaxial accelerometer (Suzuken Co. Ltd., Nagoya, Japan)
Waist (side not specified)
7 days Mental health as measured by SF-36Not reported PA was strongly positively correlated to mental health
in all patients for both steps (r=0.46, p<0.001) and energy expenditure (r=0.43, p<0.001)
Average daily number of steps taken over 7 days
Energy expenditure is calculated every 4 seconds using body weight and exercise index
Jehn et al. 2009 (45)
Study period not reported
Assess habitual walking performance in CHF patients and investigate if this information can be used to distinguish NYHA
Accelerometer (Aipermon® GmbH, Germany)
6 days NYHA functional class based on clinical data and self-reported exercise tolerance
Not reported Difference in mean total walking, walking and fast walking times was statistically significant between:- NYHA class II and III (p=0.001)
Total times per day spent: Passively (not defined)
Prospective cohort study
Poor quality
functional class
50 patients 60.9±14.0 years old 78% male
Left waist Actively (not defined) Walking (0 to 80m/minute) Fast walking (83 to
115m/minute)
- NYHA class I and III (p=0.001)
Fast walking time was also a strong determinant for discriminating moderate heart failure (NYHA class III)
Only 4 days of monitoring required for fast walking time to have a sensitivity and specificity of 74%
Melin et al. 2016 (46)
May 2009 to June 2013
Prospective cohort study
Good quality
Assess the additive value of variability characterised by accelerometer data to other risk factors in a prognostic model for CHF patients
60 patients recruited, 56 patients included in analysis 70.3 years old 76.8% male
GT3X accelerometer (Actigraph, Pensacola, FL)
Waist (side not specified)
7 days All-cause mortalityBaseline 1, 3 and 12-hour skewness showed the most significant
contribution to mortality amongst the accelerometer variables
Addition of peak 3-hour skewness to the HFSS model significantly improved predictive ability (c-index improved from 0.71 to 0.74)
Activity was recorded in 1-minute epochs generating activity count for each minute of the day
Used to estimate: 1, 3- and 12-hour skewness,
kurtosis and IQR Total no. of minutes monitor
was worn Sedentary time: vertical axis
counts per minute (cpm) <100 Light activity time: vertical
axis cpm between 100-1,951 Moderate vigorous physical
activity time: vertical axis cpm >1,952
Waring et al. 2017 (47)
October 2014 to March 2015
Prospective cohort study
Good quality
Test the hypothesis that physical inactivity is a predictor of rehospitalisation in HF
61 patients recruited, 50 patients included in analysis 71±15 years old 46.0% male
ActiGraph wGT3X-BT or GT3X+ (Actigraph, Pensacola, FL)
Non-dominant wrist & ipsilat. waist
Non-dominant hand: 30 daysIpsilateral waist: 7 days
All-cause readmissions within 30 days
After discharge from intial hospitalisation to 30 days or readmission
13 patients (26%) had all-cause readmission within 30 days: - CHF readmission: 9 patients- Non-HF readmission: 4 patients (due to dehydration, COPD exacerbation, cellulitis, gastrointestinal bleed)
Outpatients v.s. patients with eventual readmission:- No significant difference in minutes of higher-intensity activity over week 1- Significant differences in minutes of higher-intensity activity over weeks 2, 3 and 4
Lower level of activity in specific weeks showed significantly higher odds ratio of readmission:- All-cause: weeks 1 (OR=5.0, p=0.02), 2 (OR=16.9, p=0.001), 3 (OR=4.8, p=0.047) and 4 (OR=16.1, p=0.003)
≥3,000 vector magnitude units (sum of movements in 3 planes per minute of device use) was defined as higher-intensity activity for each minute of valid wrist activity output
Patients with ≥60 minutes of higher-intensity activity were considered more active while those with <60 minutes were considered less active
- CHF-related: weeks 1 (OR=5.5, p=0.03) and 2 (OR=8.6, p=0.01)
Edwards and Loprinzi 2016 (48)
2003 to 2006
Retrospective cohort study
Fair quality
Examine the association of sedentary behaviour on HRQoL in CHF patients
190 patients 66.7 years old (95%
CI 63.6-70.0) 56.2% male
ActiGraph 7164 accelerometer (Actigraph, Pensacola, FL)
Waist (side not specified)
Up to 7 days HRQOL: CDC HRQOL measureNot reported Average amount of time spent in each activity level:
- Sedentary behaviour (<100 counts/min): 537.0 minutes/day - LIPA (100-2,019 counts/min): 261.6 minutes/day - MVPA (≥2,020 counts/min): 8.6 minutes/day
Sedentary behaviour was associated with worse HRQoL (β=0.004; 95% CI=0.0004-0.007; p=0.03) and was still significant when MVPA was added as a covariate
Trivariate model including all three activity levels resulted in no association between sedentary behaviour and HRQoL
Sedentary behaviour: <100 activity counts/minuteLIPA: 100-2,019 activity counts/minuteMVPA: ≥2,020 activity counts/minute
Loprinzi 2016 (49)
2003 to 2006
Retrospective cohort study
Good quality
Investigate the relationship between physical activity and mortality in CHF patients
256 patients Alive at follow-up:
188o 65.2 years old
(63.0-67.3)o 56.7% male
(47.8-65.7) Deceased at follow-
up: 68o 75.1 years old
(72.8-77.5)o 58.0% male
(46.2-69.8)
ActiGraph 7164 accelerometer (Actigraph, Pensacola, FL)
Waist (side not specified)
Up to 7 days All-cause mortalityNot reported For every 60min/day increase in PA, CHF patients
showed 35% reduced risk of all-cause mortality (HR=0.65; 95% CI=0.50-0.85; p=0.003)
See Edwards and Loprinzi 2016 (48)
Miyahara et al. 2017 (50)
April 2014 to May 2015
Prospective cohort study
Good quality
Evaluate physical activity before and after hospital discharge and determine the impact of physical activity on post-discharge cardiovascular events
41 patients recruited 66-84 years old 53.7% male
Triaxial accelerometer HJA-3501T (Omron, Kyoto, Japan)
Waist
At least 7 days CHF rehospitalisation during 6-months post-dischargeContinuous measurement 20 patients classified as low activity, while 21 patients
were considered to have high activity
Rate of CHF rehospitalisation was significantly higher in patients with low activity than in patients with high activity (30% v.s. 5%)
Total physical activity was the only significant predictor of rehospitalisation on multivariable regression analysis
Output signal from accelerometer was processed using commercial software (Omron, Kyoto, Japan) to calculate METs: LIPA: 1.5-2.9 METs
MVPA: ≥3 METs
METs were expressed as MET-hours/day
(OR = 0.65, p = 0.03)
Light-intensity physical activity was the strongest predictor of CHF rehospitalisation (OR = 0.60, p = 0.03)
Snipelisky et al. 2017 (51)
Study period not reported
Ancillary study from a randomised controlled trial (NEAT-HFpEF)
Good quality
Understand the determinants of baseline physical activity in CHF and relationships to functional assessments, and observing the impact of changes in daily activity with isosorbide mononitrate to the same assessments
110 patients recruited, 99 patients included in analysis 69 years old 40% male
Kinetic Activity Monitor (Kersh Health, Plano, TX)
Waist
2 weeks NYHA class, HRQoL (Kansas City Cardiomyopathy Questionnaire, Minnesota Living with Heart Failure Questionnaire), 6 minute walk distance, NT-proBNP
Continuous measurement of baseline and after isosorbide mononitrate
Lower average daily accelerometer units and hours active per day: higher proportion of patients in NYHA class III/IV, lower HRQoL scores, higher NT-proBNP
No significant relationships between changes in average daily accelerometer units and hours active per day to changes in standard CHF functional assessments
Accelerometer units stored as 15-minute epochs for a total of 96 data points per day and averaged to give average daily accelerometer units
Hours active per day were calculated using the number of epochs with accelerometer units >50
van den Berg-Emons et al. 2005 (52)
Study period not reported
Prospective cohort study
Good quality
Investigate factors related to daily activity in CHF patients and if level of activity is associated with quality of life
36 patients recruited 59 years old 75% male
ADXL202 accelerometer (Temec Instruments, the Netherlands)
Waist (accelerometers attached to sternum and thigh)
2 days HRQoL: Minnesota Living with Heart Failure Questionnaire
Continuous measurement No relationship between movement related physical activity and HRQoLDuration of movement-related
activity expressed as a percentage of a 24-hour period
Abbreviations: CHARM, Candesartan in Heart Failure Assessment of Reduction in Mortality and Morbidity; CHF, congestive heart failure; CI, confidence
interval; COPD, chronic obstructive pulmonary disease; HFSS, Heart Failure Survival Score; HR, hazard ratio; HRQoL, health-related quality of life; LIPA,
light intensity physical activity; MET, metabolic equivalent task; MVPA, moderate-to-vigorous physical activity; NT-proBNP, N-terminal pro b-type
natriuretic peptide; NYHA, New York Heart Association; OR, odds ratio; PA, physical activity; SF-12, Short Form-12 Health Survey; VE/VCO2, minute
ventilation/carbon dioxide production; VO2, volume O2