Feasibility and preliminary efficacy of a physiotherapy-led remotely delivered physical activity intervention in cancer survivors using wearable technology. The IMPETUS trial

ABSTRACT Background Physical activity levels are low in cancer survivors. Remotely delivered programs which harness wearable technology may potentially be beneficial. Objective To evaluate the feasibility and preliminary efficacy of a remotely delivered, physical activity intervention which harnessed wearable technology. Methods This single arm pre-post longitudinal study included cancer survivors who had completed treatment in the preceding 3 years. Participants were supplied with a Fitbit One® or Flex® for 12 weeks. Physical activity goals were discussed during support phone calls. Outcome measures, assessed at baseline (T1), 12 weeks (T2), and 24 weeks (T3), included feasibility (recruitment, adherence, safety, acceptability) and efficacy [physical activity (Godin leisure time Index, ActiGraph GT3X+), quality of life (functional assessment of cancer therapy – general, short form 36 physical functioning component), functional capacity (six-minute walk test)]. Results Forty-five participants completed T1 assessments (10 males, 35 females). Thirty-nine (86.6%) of those underwent assessment at T2 and 31 (68.8%) at T3. The intervention was perceived positively with no adverse effects. There were increases in functional capacity (six-minute walk test, p = .002) between T1-T3, an increase in quality of life [short form 36 physical functioning measure (p = .0035), functional assessment of cancer total score (p = .02)] and self-report physical activity levels (p = .000123) between T1-T2, although effect sizes were generally low (d = 0.180 to d = 0.418). Objectively measured physical activity did not change. Conclusion A physical activity intervention including wearable technology was safe, feasible, and well received by cancer survivors. An intervention based on this proof of concept should be followed up in further studies.


Introduction
Cancer is a global problem and the incidence of cancer is increasing with an estimated 18.1 million new cases of cancer diagnosed globally in 2020 excluding nonmelanoma skin cancers (Sung et al., 2021).This reflects a global trend of increasing survival after cancer treatment.In January 2019, there were approximately 16.9 million people with a history of cancer alive in the United States (U.S.).That number is projected to surpass 22.2 million by 2030 (National Cancer Institute, 2019).The documented benefits associated with physical activity (PA) in cancer survivors include reduced treatment-related side effects, improved quality of life, and, in some cancer types such as post-menopausal breast cancer and colon cancer, fewer cancer recurrences, and better survival (Fong et al., 2012;Forbes et al., 2020;Kerr, Anderson, and Lippman, 2017;McTiernan et al., 2019;Speck et al., 2010).Despite these benefits, levels of PA in cancer survivors are low (Broderick, Hussey, Kennedy, and O'Donnell, 2014) and drop significantly within the first year after diagnosis (Irwin et al., 2003).
Traditional models of PA intervention are expensive, provide limited access, and are rarely scalable, meaning there is a challenge to reaching people and places (Phillips et al., 2018;Reis et al., 2016).The challenge therefore remains to explore alternative means of PA and exercise program delivery.In the United States approximately 85% of people own a smartphone.A key focus of the smartphone application market is health and fitness.People with cancer can be receptive to health and fitness smartphone apps which promote PA (Short, Finlay, Sanders, and Maher, 2018).While applications are often freely available and convenient, they are frequently not underpinned by behavior change theories and their efficacy is questionable (Harris, Cheevers, and Armes, 2018).A recent study (Martin Payo, Harris, and Armes, 2019) showed that the content of only one-fifth of freely available exercise and fitness apps studied were deemed suitable for people with cancer, with the app and behavioral content evaluated using the Behavioral Change Techniques Taxonomy.However, it must be acknowledged that suitability for people with cancer and other clinical conditions was not necessarily the intention of developers of these commercially available PA/fitness apps.A further problem with internet applications is digital preservation, as they are highly likely to change or disappear over the years (Eysenbach, 2011).
Recent research has suggested that some cancer survivors are receptive to remotely delivered technologysupported exercise programs (Pope et al., 2018) such as mHealth PA interventions, using wearable activity monitors (Lynch et al., 2019).The National Cancer Institute has nominated the development and testing of mHealth interventions in cancer as a research priority (Alfano et al., 2016).Wearable PA trackers, used in isolation, are unlikely to generate meaningful behavioral change (Alfano et al., 2016).However, they provide objective PA monitoring which is bidirectional, so that an end-user and health/exercise professional can both view activity levels, which potentially could encourage collaborative goal-setting and incorporate recognized behavioral change techniques such as social support, self-monitoring, and goal setting.The objective of the present study was to evaluate the feasibility and preliminary efficacy of a physiotherapy-led remotely delivered technology enhanced PA intervention for cancer survivors.Primary feasibility outcome assessment included: 1) evaluation of recruitment capability and resulting sampling characteristics; 2) evaluation of adherence; 3) safety; and 4) and evaluation of participant responses to intervention.Secondary efficacy outcome assessment included assessment of the following: 1) PA; 2) body anthropometrics; 3) quality of life; and 4) functional capacity.

Participants
The IMPETUS trial (https://clinicaltrials.gov/ct2/show/ NCT03036436) was a prospective, single-arm feasibility trial which was approved by the Joint Research Ethics Committee of St. James's Hospital/Tallaght University Hospital.All participants provided written, informed consent.A detailed description of the IMPETUS trial has been published previously (Haberlin et al., 2019).
In brief, this study was conducted in a single designated cancer center St. James's Hospital, Dublin, Ireland.Potential participants were screened during cancer out-patient clinics by their treating cancer clinicians.Inclusion criteria were: 1) agreement of treating clinician that patient can participate; 2) age ≥18 years; 3) completed chemotherapy or radiotherapy with curative intent within the preceding 3 years (i.e.participants may have had chemotherapy or radiotherapy as the sole treatment for cancer, with adjunctive surgery, but not surgery alone, participants who were still on adjuvant hormone therapy and/or adjuvant HER2-directed therapy were also eligible, with physician agreement as above); 4) ability to understand English; and 5) owned or had access to a device compatible with the Fitbit® app.Exclusion criteria were as follows: 1) chronic medical and orthopedic conditions that preclude exercise (e.g. an acutely inflamed hip or knee or awaiting joint replacement surgery); 2) confirmed pregnancy; 3) cognitive impairment or psychiatric illness that would preclude ability to participate in the study; 4) incomplete hematological recovery after chemotherapy (i.e.white cell count <3, hemoglobin <10 or platelets <100); 5) evidence of active cancer; and 6) diagnosis of prostate cancer or upper gastrointestinal cancer.The last exclusion criterion was to ensure that the participants were not enrolled in other ongoing disease-specific exercise studies in our center.

Procedure
The study consisted of two 12-week phases; an intervention phase followed by a maintenance phase.Study assessments were conducted in the Clinical Research Facility of St. James's Hospital, Dublin, Ireland, at the following time-points; T1 (baseline, week 0), and reassessment visits T2 which took place after the intervention (12 weeks ± 2 weeks) and T3 to assess maintenance (24 weeks ± 2 weeks).
The lead researcher met potential participants who had been screened by their treating oncology clinician and gave them a patient information leaflet and consent form.If participants signed the consent form, they proceeded to study entry.The participant wore an Actigraph accelerometer for 7 days before attending a face-to-face baseline session.That session (T1) consisted of the following; baseline study measurements, allocation of a Fitbit®, training on Fitbit® use, synchronization of the device with the participants' Fitbit® application on their phone or other compatible device, and PA goal-setting.A motivational interviewing approach (Miller and Rollnick, 2012) was used to set PA goals jointly with the participant, based on self-reported PA levels using the Godin leisure time index questionnaire (Godin and Shephard, 1985) as well as expressed preferences and goals.The baseline accelerometry data was not analyzed before that session nor used to set PA goals.

Intervention
Participants were provided with a Fitbit® (Fitbit Inc., San Francisco, CA) and were instructed to wear it during all waking hours, except in bath, shower, or other waterbased activities.The purpose of the Fitbit® was to support self-monitoring of PA by participants and to provide real-time monitoring of PA information to the lead researcher to inform physiotherapist-led goal-setting and phone call guidance.Participants uploaded their PA data remotely using an anonymized study-specific number for review by the lead investigator.
Participants received 14 support phone calls in total over the course of the intervention phase.Support phone calls were conducted by the lead researcher who was a physiotherapist.Call frequency was tapered over the course of the program as follows: weeks 1-4, two calls per week, weeks 5-8, one call per week, weeks 8-12, one call every second week.The goals prescribed by the lead researcher were collaborative in nature, with participants encouraged to provide feedback about their ability to achieve their goals and any changes they would like to these goals moving forward.The call consisted of a review of the previously set goal, identification of updated PA goals, identification of barriers and solutions, discussion of modifications (e.g.step goal usually increased by 10% per week) and specific PA plans needed to address specific goals.If any technological issues had arisen, these were addressed during the call.Participants self-monitored their step counts and levels of moderate to vigorous PA using the Fitbit®.Both participant and the lead researcher had access to the Fitbit® data throughout the intervention phase.
The content of the calls was individualized and designed to support and motivate participants, by introducing specific PA goals.Phone calls in the first 4 weeks were divided into a 'Goal' phone call and a 'Check-in' call.During the 'Goal' phone call, the lead investigator discussed the past week of PA with the participant.
Figure 1 provides more detail of the content of 'Checkin' calls and 'Goal' calls.The participant was also invited to provide feedback regarding their perceived progress.This phone call also provided the participant with their updated PA goal, which was prescribed using FITT (i.e.frequency, intensity, type, and time) principles.Goals for this study included daily step goals and also weekly moderate intensity exercise goals, and were tailored to each participant.The delivery of the goal phone call by the physiotherapist was underpinned by key elements of motivational interviewing, of which the lead investigator had previous training.Motivational interviewing is an approach to behavioral change which is patientcentered, collaborative, and focuses on the promotion of autonomy in an individual to enable them to evoke their own change (Shingleton and Palfai, 2016).The 'Check-in' phone calls were designed as a means to provide advice and a reminder on uploading Fitbit® data.Ongoing technological support regarding the Fitbit® was also provided to participants in phone calls if required.If participants informed the study team that they would be uncontactable by telephone (e.g.abroad) for some of these scheduled phone calls, we requested permission from the participants to contact them during those times via a secure messaging service, to allow for an uninterrupted schedule of reminders.If a participant was unwilling to complete the intervention, and requested to be discontinued, the lead researcher terminated the intervention for this particular participant.

Outcome measures
All outcomes were assessed by the lead researcher.Feasibility outcomes were generated on completion of the intervention.Secondary efficacy outcomes were assessed by the lead researcher at T1, T2, and T3.Primary feasibility outcome assessment included: 1) evaluation of recruitment capability and resulting sampling characteristics; 2) evaluation of adherence and FitBit® synchronization compliance; 3) safety; and 4) and evaluation of participant responses to intervention.Secondary efficacy outcome assessment included assessment of the following: 1) PA; 2) body anthropometrics; 3) quality of life; and 4) functional capacity.Feasibility was assessed by tracking recruitment capability and adherence data.Recruitment capability was measured by the number of new participants recruited to the trial each month.Participant satisfaction was assessed using a questionnaire generated for the purpose of this study.In this questionnaire, participants were asked: 1) if they felt the program increased their PA levels answered by a 5-point Likert scale from 'strongly agree' to 'strongly disagree'; and 2) about aspects of the program they enjoyed, any they did not enjoy, and aspects which could be improved.Compliance was defined as synchronization of the Fitbit® with the mobile application each day.'Good' compliance was considered to be >70%.Safety was measured by recording any adverse effects such as injuries.Satisfaction was measured by asking participants to complete a survey with the following three questions which was used successfully in a previous study (Broderick et al., 2013): 1) What did you like about this intervention; 2) What did you not like about this intervention; and 3) Have you any suggestions for change?
PA was measured in the following ways.Firstly, it was measured objectively using an accelerometer (Actigraph GT3X+; Actigraph, Pensacola, FL, worn for 7 days at T2 and T3).This has shown to be a reliable (Aadland and Ylvisåker, 2015) and accurate (Santos-Lozano et al., 2012) method of measuring free-living PA.Accelerometer data evaluated in this study was daily time in the following domains of activity; sedentary behavior and moderate-to-vigorous activity.PA was measured subjectively using the Godin Leisure Time Questionnaire, which is a valid and reliable self-report measurement of PA (Godin and Shephard, 1985) although this has not been specifically validated in cancer survivors.Mean weekly Fitbit® steps, and percentage of participants achieving >10,000 steps per day were also calculated.Fitbit® devices have been shown to be partially accurate for measuring step count (Feehan et al., 2018) but this was used more of a goal setting and motivational tool in this study.Height was measured using a Seca stadiometer (Seca, Hamburg, Germany) and weight/body fat percentage using a Tanita® MC-78OU Bioelectric Impedance Technology Monitor.Quality of life was measured using two questionnaires: 1) Functional Assessment of Cancer Therapy (FACT-G) scale (general) (Cella et al., 1993) and Short-form 36 (SF-36) (physical functioning subset) (SK, Kosinski, and Gandek, 1993).The validity of the FACT-G has been demonstrated in a mixed population of people with cancer (Cella et al., 1993) and psychometric properties of the SF-36 have been well established (SK, Kosinski, and Gandek, 1993).Functional capacity was measured using the six-minute walk test (American Thoracic Society, 2002).The validity of the six-minute walk test has been previously demonstrated in patients with cancer (Schmidt et al., 2013).

Statistical analysis
The difficulties of generating accurate sample size calculation for feasibility studies are well known.For feasibility studies, sample sizes between 24 and 50 have been recommended (Sim and Lewis, 2012).Based on this, we proposed to recruit a sample size of 60 to allow for a 20% dropout.Descriptive statistics of participant demographics were reported as numbers and percentages.Free text responses from the participant satisfaction questionnaire were reported and organized into topic areas by the lead researcher.Data from T1, T2, and T3 were analyzed.Only within-subject changes were considered as there was no control group.The Shapiro-Wilk normality test and Q-Q plots were used to evaluate whether continuous data was normally distributed.Homogeneity of variance was tested using Maulchly's Test of Sphericity.Friedman's test was used to compare non-normally distributed data, and one-way ANOVA (analysis of variance) repeated measures parametric tests were used to compare normally distributed data at each time point with Tukey post hoc tests.Effect sizes (partial eta squared, ηp) were also calculated for each efficacy outcome, and interpreted using Cohen's d with: small interpreted as d = 0.2; medium d = 0.5; and large d = 0.8 (Cohen, 1988).A complete case analysis was carried out because of the feasibility focus of the study.IBM SPSS Statistics package (V24) was used to analyze data.For all analyses, p-value <0.05 (two tailed) was considered statistically significant.

Evaluation of recruitment capability and resulting sample characteristics
In total, 62 participants expressed interest in the study, signed a consent form and were invited to assessment, over a 15-month period from January 2017 -March 2018, of whom 45 participated.The flow of participants through the study is shown in Figure 2. Baseline characteristics of participants are presented in Table 1.Mean age of participants (10 males and 35 females) was 50.7 (SD = 11.8) years of age (range 20-71 years) and had a mean BMI of 27.1 kg.m 2 .A diverse mix of cancer diagnoses was represented; hematological, gynecological, colorectal, and breast cancers predominated.

Evaluation of feasibility, adherence, compliance, and safety
Duration of outcome measure testing at each time point (T1, T2, and T3) was approximately 45 minutes and phone calls were <10 minutes in duration.From 62 potential participants who met inclusion criteria and indicated willingness to participate in the trial, the uptake rate was 72.6%, with 45 participants proceeding to trial entry and assessed at T1.The recruitment rate averaged at four new participants per month.At T2 and T3, 62.9% (n = 39) and 50% (n = 31) were assessed.Most participants [84% (n = 38)] who commenced the study complied with ≥70% of phone calls.In total, 39 participants synchronized and recorded Fitbit® data spanning the entire intervention.Mean daily wear and synchronization compliance for these participants was high at 92.3% (n = 36).The intervention was safe with no adverse effects reported.

Evaluation of participant responses to intervention
In total, 75% (n = 27) of participants 'strongly agreed' that they found the intervention helpful toward the goal of increasing PA levels.Eight participants (n = 8) answered 'agree,' while the remaining participant answered 'strongly disagree.'Aspects of the program enjoyed by participants were organized into the following topic areas (i.e.goals and achievements, motivation via Fitbit, improvement of fitness, creating good habits, monitoring, support, and feedback and development of PA knowledge) as detailed in Supplementary File 1 and free text responses were reported.The majority of participants (n = 26, 72%) responded that they did not identify any aspects of the program that they did not enjoy.A small number of program aspects identified as less enjoyable were: not reaching goals was sometimes disappointing, however this was very rare; wearing (the) actigraph can be a pain, hard to hide it; Trying to mind the Fitbit!; taking off Fitbit® while swimming; yes, wet weather; not reaching my goals; get out but once I was out I enjoyed been out; the guilt on myself for not making myself get up and walk; remembering to recharge battery or keeping an eye on charge level; and feeling guilty when CH phoned.
In total, 94% (n = 34) of participants felt this program increased their PA levels.The majority of participants 89% (n = 32) reported that using the Fitbit® was a positive experience.Suggestions for change to aspects of the intervention are reported in Supplementary File 2.

Secondary outcomes: efficacy outcomes
The effect of the intervention on efficacy outcomes of PA, functional capacity, anthropometric measures, and quality of life are shown in Tables 2 and 3. Objective PA data was available for: 96% (n = 43); 76% (n = 34); and 40% (n = 18) at T1, T2, and T3, respectively.At baseline, participants spent a median [interquartile range (IRQ)] of 8.3 (2.4) hours sedentary per day, 4.4 (1.8) hours in light activity and 29.1 (28.9) minutes in moderate and vigorous PA.PA and sedentary time did not change over the course of the intervention (p > .05).
On the other hand, self-report PA increased significantly over the course of the intervention (p = .000123)with changes between T1 and T2 (p = .005)and between T1 and T3 (p = .006)with a moderate effect size (d = 0.418).Median (IQR) Fitbit® recorded steps were 7967 (4247) for week 1, 8405 (4799) for week 6 and 9172 (5843) for week 12, although a drop in week 7 steps was noted.Mean weekly Fitbit® steps are shown in Figure 3.The difference in weekly Fitbit® steps between time points did not reach statistical significance (p > .05).The percentage of participants reaching over 10,000 steps per day was 25% at week 1, 30% at week 6 and 27% at week 12.
Waist circumference increased significantly (p = .002)from T1 to T3 (p = .008)with a mean (95%) difference of 2.8 cm (−0.6 to 5.1).There was no change to other anthropometric measures noted.Total FACT-G quality of life scores improved significantly (p = .02)with changes occurring between T1 to T2 (p = .0001).
Significant changes were also shown in the physical functioning component of the SF-36 (p = .035)between T1 and T2 (p = .001).However, the effect sizes for the FACT-G (d = 0.250) and physical functioning component of the SF-36 (d = 0.180) were small.
Distance covered during the baseline six-minute walk test was 557.4 (70.5) meters which increased significantly (p = .002)over the course of the study, although

Discussion
This study demonstrates the preliminary feasibility and acceptability of this remotely delivered, technologyenhanced PA intervention in a cancer survivor population.The intervention capitalized on commercially available Fitbit® technology paired with physiotherapist support to deliver a behaviorally sound, potentially scalable PA program, eliminating the need for repeat face-to -face visits to a center.Participants were positive about the intervention and no adverse effects were reported, providing initial reassurance that this remotely delivered exercise program for cancer survivors was safe.

Primary outcome: feasibility
Recruitment averaged about four participants per month, consistent with another study using smart scales and activity trackers in African American breast cancer survivors (Valle, Deal, and Tate, 2017).Despite the permissive entry criteria and the flexible nature of the intervention, recruitment was slow relative to the number of potential participants in our center, similar to reports of one study which incorporated a PA smartphone app (Ormel et al., 2018) and others which incorporated PA wearables (Lynch et al., 2019;Van Blarigan et al., 2019).A feature of our population was the high proportion of participants who had received allogeneic bone marrow transplant.They are underrepresented in other studies of PA, are universally deconditioned after their physiologically challenging therapy, and in our center attend a specialist clinic dedicated to their survivorship issues.Relative success in their recruitment is   PHYSIOTHERAPY THEORY AND PRACTICE likely explained by patient and clinician awareness of multifaceted survivorship needs.More than 1 in 4 dropped out before participation.Taken together, this suggests that improving recruitment for future studies of PA interventions must involve enhancing physician, nursing, and patient awareness of the need to improve PA after anti-cancer therapy and specific measures to ensure continued motivation between consent and intervention start.Although a small number (n = 4) who presented for assessment at T1 received a Fitbit® which was never synchronized and did not engage with phone calls, in general, among those who commenced the intervention, there were high levels of engagement and compliance with the technology.
Concerns about technological issues expressed in a preceding focus group study (Haberlin, O'Donnell, Moran, and Broderick, 2020) appeared largely unfounded as 77% reported they had no difficulty using the Fitbit® or paired applications.For example, in those who completed the 12-week intervention, mean percentage wear time was 92.6%, similar to values of 90% quoted in other Fitbit® studies in cancer survivors (Van Blarigan et al., 2019).This may have been thanks to the support phone calls.Compliance with support phone calls was also good, suggesting that the human interaction aspect was valued by participants.Another recent study (Lynch et al., 2019) examining the effect of a wearable intervention on PA in cancer survivors also reported high compliance with phone calls, although there were only five phone calls in that study compared to a 14 in ours.The optimum number of phone calls to support remotely delivered programs such as these needs to be elucidated further.
The intervention was well received, likely due to its design based on user perspectives at the pre-trial phase, which represents a major strength of this study.Almost all participants (97%) reported subjectively finding the intervention helpful in improving their PA levels.Our findings, in a heterogenous group of cancer survivors, agree with a published study research (Phillips et al., 2017) which reported that breast cancer survivors were interested in using technology to self-monitor PA and to receive coaching from a distance with personalized feedback.

Efficacy outcomes
Overall, effect sizes were generally small with the largest shown in self report PA, waist circumference, and the six-minute walk test.There was a mean daily increase in step count of 805 steps from baseline to T2, which is lower than mean increases of 1126 steps reported for PA interventions in healthy community-based adults (Chaudhry et al., 2020).Although, with baseline mean daily step counts of almost 8,000 steps per day and one quarter achieving >10,000 steps, this may indicate an already reasonably active group, so improvements would be more challenging to observe.There was a significant increase in self-report PA from baseline to the 12 week time point (p = .000123),which was preserved during the observation phase when the Fitbit® and professional support were withdrawn.However, objectively measured PA did not increase.Fewer than half of our participants completed the accelerometer assessment at T3 probably due to "tech fatigue," but the failure to find objective PA improvement at T3 is unlikely to have been more than slightly affected by that missing data.It corresponds with no increase in objectively measured PA at T2, and the modest but nonsignificant increase in Fitbit® step count measurements also suggest it is a true finding.
We have previously reported a discrepancy between self-report PA changes and objectively measured PA in cancer survivors (Broderick, Hussey, Kennedy, and O'Donnell, 2014).Self-report measures are subject to recall bias and therefore prone to inaccuracies (Gell et al., 2017;Prince et al., 2008).Our findings in this study reinforce the need to incorporate objective measures in PA interventions.Other studies which used wearable technology to enhance PA in cancer survivors (Pope et al., 2018) also failed to show an increase in objectively measured PA.There are a couple of possible explanations for why objectively measured PA did not increase.Firstly, we only had accelerometer data on 40% of participants at T3 so results must be very cautiously interpreted on such a small sample size.Secondly, we did not exclude participants who were already active.It appears that the range of baseline PA in our group was wide; accelerometer data shows that the median time spent in moderate or vigorous activity approximated to guidelines (Campbell et al., 2019) but with a very wide quartile range (Table 2).Over a quarter of our participants were recording over 10,000 steps per day.In those already active participants, the likely effect of any PA intervention is small and would require a study of greater numbers or higher power to detect change.Perhaps such participants should be excluded from a future study and the intervention targeted at those less active.Second, the majority of our participants had completed their anti-cancer therapy more than a year before study entry.It is likely that at that stage, they had returned to their normal lives and obligations and "plateaued." A significant increase in global quality of life as well as physical functioning was demonstrated in this study.
The change in mean score for the FACT-G total score (+6.5) from baseline to 24 weeks is within the range of clinically important mean change of 5-6 previously identified in a mixed cancer group (Cella, Hahn, and Dineen, 2002).The change score for the physical component summary of the SF-36 (+5) meets the threshold identified for clinically important difference (Ogura et al., 2020), although this was established in an orthopedic cancer population so the comparison may not be directly comparable.
Functional capacity, as measured by the 6-minute walk test also improved significantly (p = .002).It is not known if the corresponding mean change has clinical relevance, as the minimally important clinically different for a mixed group of cancer survivors would be difficult to establish and does not appear to have been previously investigated.However, the mean change established in this study of 29.9 m is within the range of 22 m-42 m established for lung cancer survivors (Granger, Holland, Gordon, and Denehy, 2015).It is likely though that exercise tolerance after lung cancer may not be comparable to many other forms of cancer.Notably, mean functional capacity levels in this study were broadly in line with established reference values for 'healthy' populations (Casanova et al., 2011).
Disappointingly, waist circumference increased over the course of the intervention, even though body mass index did not change.We can only speculate for the reason for the increase in waist circumference as we did not track dietary intake over the course of the study, although it is possible that with participants 'feeling' more physically activity they may have paid less attention to their calorific intake, resulting in an increased waist circumference.Other anthropometric variables did not change.The high rate (64%) of overweight or obesity at baseline in our cohort matches almost exactly that in the Irish adult population (Department of Health -Ireland, 2016).Some study participants suggested in feedback comments that weight loss strategies could be included in future iterations of this intervention.Obesity and low PA levels are both associated with increased risk of recurrence in some cancers.On the other hand, it is extremely difficult to evaluate the efficacy of individual interventions in multimodal programs incorporating PA and dietary components.Furthermore, data from large cohorts (Moholdt, Lavie, and Nauman, 2018) illustrate that exercise and PA exert a beneficial effect on cardiovascular risk regardless of weight loss.Therefore, finding an acceptable, effective, scalable intervention to increase PA in cancer survivors independent of weight loss is still an important objective.
Our study has a number of strengths.First, the intervention was co-designed by end-users.Second, it was underpinned by behavioral change theory (Michie et al., 2013) which is central to successful interventions (Husebø, Dyrstad, Søreide, and Bru, 2013;Turner et al., 2018).Aspects of behavioral change theory embedded in this intervention were goal setting, prompts, self-monitoring, and encouragement of independent exercise.Thirdly, it was remotely delivered and incorporated new technology, which may have greater reach than traditional programs.The study included a heterogeneous cancer population and robust objective measures to measure PA.
One obvious limitation is the lack of a control arm.However, the average length of a randomized control trial is 5.5 years from enrollment to publication (Ioannidis, 1998) and an ongoing large study of exercise and dietary intervention in ovarian cancer survivors has an 8 year projected recruitment time (Thomson et al., 2016).Therefore, given the risk of results becoming outdated due to newer technology iterations before study completion, this type of single-arm study is justified to test initial feasibility in a fast-developing area such as this (Pham, Wiljer, and Cafazzo, 2016).
Due to resource limitations, the lead researcher delivered the intervention and performed outcome assessments, rather than an independent assessor, which could have limited validity and influenced the participants' adherence to and completion of trial assessments in either direction.Other limitations were the single institution setting, and the small sample size which has limited power to detect differences, although the final sample T3 sample size of 31 exceeded the lower acceptable number for feasibility studies identified by Sim and Lewis (2012) of 24 participants.Also, the reasonably high proportion of participants after BMT may have influenced the external validity of this study, so the generalizability of results to other settings is not known.As there is no established consensus on what to do with missing data and its random nature in this study we elected to do a complete case analysis which may lead to biased parameter estimates.
The intervention requires access to the internet and may therefore not be generalizable to a group, a minority in our cancer center, who do not have internet access and/or to those who do not wish to try new technology.Our study was 12 weeks in duration, and assessed maintenance until the 24 week time-point, so longer term maintenance was not tested.Results of the ACTIVATE trial which incorporated wearable technology (Garmin Vivifit 2) indicated that wearable technology can facilitate a more active lifestyle beyond an active intervention phase (Lynch et al., 2019).Conflicting studies report that many people stop wearing activity monitors after a few weeks (Rowe-Roberts, Cercos, and Mueller, 2014) so the potential of these devices for longer term adaptation is not known.

Conclusion
In conclusion, this study provides initial cautious but encouraging proof of concept of the safety, feasibility, and user acceptability of this technology-supported PA intervention in cancer survivors, although no significant objective increase in PA levels was observed in our heterogenous cohort.This study showed that physiotherapists are well positioned to embed technology as an adjunct therapeutic tool among this patient cohort.While employing a tool such as a Fitbit® will not in any way replace physiotherapy, the application of this work means our advice about PA can be reinforced by technology, leveraging and maximizing the impact of our face-to-face sessions.Further exploration of this or similar physiotherapy interventions should focus on identifying cancer survivors, particularly those with low baseline PA levels, who are most likely to benefit.Potentially other patient groups outside of cancer could also find this type of remotely delivered exercise program feasible and acceptable.Physiotherapists should be at the forefront of investigating the possibilities of leveraging available technology to deliver innovative interventions at a remote distance.

Figure 1 .
Figure 1.Content of the goal phone call and check-in phone call.

Table 2 .
Effects of exercise intervention on non-normally distributed variables.

Table 3 .
Effects of exercise intervention on normally distributed variables.