spiral.imperial.ac.uk€¦ · web viewcan activity monitors predict outcomes in patients with...

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

mailto:[email protected]

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




Good quality

Determine whether baseline levels of physical activity predict cognitive function and cerebral perfusion at 12 months in CHF patients



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




Good quality

Examine if changes in physical activity predicts cognitive changes over 12 weeks in older adults with CHF



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)




Good quality

Examine the association of low physical activity with adverse outcome measures in CHF patients



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)


Examine the effects of different aspects of physical activity on cognitive function in older


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


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


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)



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


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)


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)


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)


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


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)


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


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)


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)


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


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)


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)



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

spiral.imperial.ac.uk€¦ · web viewcan activity monitors predict outcomes in patients with...

Documents