Colorectal cancer (CRC) is the fourth most diagnosed cancer and the second leading cause of cancer-related deaths in the United States [1]. Prospective studies suggest that physical activity after CRC diagnosis is associated with longer survival, including in patients with advanced or metastatic disease [2-5]. Moreover, patients with nonmetastatic CRC who engage in less physical activity after diagnosis have a 32% increased risk of CRC-specific mortality compared with patients who maintain their prediagnosis levels of activity [6]. Given that physical activity tends to decline during treatment [7], interventions that help patients with CRC to maintain their physical activity levels during treatment may be important adjuncts to standard oncological therapies.

Several interventions are being evaluated for their impact on physical activity in patients with CRC [8]. The Colon Health and Life-Long Exercise Change (CHALLENGE) and Focus on Reducing Dose-limiting Toxicities in Colon Cancer with Resistance Exercise (FORCE) trials are 2 examples of such interventions [8,9]. CHALLENGE is an active randomized controlled trial examining the effects of a structured exercise program on disease-free survival among patients with high-risk stage 2 or 3 colon cancer who have completed adjuvant chemotherapy [8]. FORCE is an open randomized controlled trial examining the effects of resistance training on relative dose intensity and chemotoxicities in patients with nonmetastatic colon cancer receiving adjuvant chemotherapy [9]. Notably, CHALLENGE was designed as a supervised program, and FORCE focused on resistance training; both studies enrolled only patients with nonmetastatic colon cancer. Indeed, most studies to date have focused on people with nonmetastatic disease and those who have already completed treatment. There remains a need to determine the feasibility of physical activity interventions for patients with CRC during active treatment. Moreover, participation in supervised exercise intervention programs for patients with cancer may be limited by time, expense, and access to treatment centers offering exercise services. Thus, remotely delivered interventions may increase the accessibility of exercise interventions for patients with CRC.

Digital health tools, such as physical activity trackers, SMS text messaging, and apps, offer low-cost approaches to increase physical activity [10]. One study evaluated adherence to wearing a Fitbit in patients with early breast cancer on chemotherapy and concluded that additional intervention components, such as phone calls, SMS text messages, or other reminders, are needed to maintain adherence to wearing the Fitbit [11]. Few studies have evaluated similar intervention components in patients with CRC, especially those undergoing chemotherapy. A review of consumer wearable health intervention studies with survivors of breast cancer, prostate cancer, and CRC identified 8 randomized controlled trials conducted among people with these cancers; only one of these trials (Smart Pace I), conducted by our team, focused exclusively on survivors of CRC [12]. In that study, we reported that digital health tools, including a Fitbit Flex and SMS text messages, were feasible, were acceptable, and may increase physical activity among survivors of CRC after completion of chemotherapy [13].

In this study (Smart Pace II), we aim to determine whether a digital health physical activity intervention is feasible and acceptable during chemotherapy, with the goal to prevent the decline in physical activity that often occurs during treatment for CRC. We conducted a 12-week pilot 2-arm randomized controlled trial with patients with CRC receiving chemotherapy. Our primary objective is to evaluate the feasibility and acceptability of a digital physical activity intervention in this patient population. In addition, we sought to estimate the effect of the intervention on physical activity, cardiorespiratory fitness estimated through the 6-minute walk test distance, body weight, and blood pressure from enrollment to 12 weeks.

Descriptive statistics, including counts, percentages, means, SDs, medians, and ranges were used to describe participant characteristics and reports of adverse events. All statistical analyses were conducted using R [23].

We conducted 1-sample Z tests to determine whether the observed adherence was significantly less than the a priori cutoff of 70%. We also used 1-sample Z tests to determine whether the proportion of the study participants (overall and by group) that completed a 12-week follow-up survey was significantly less than a prior cutoff of 80% or more. Fisher exact test was used to compare attrition between the 2 arms. We reported the participants’ responses to the feedback questionnaire using descriptive statistics.

The secondary effects of the intervention from baseline to 12 weeks within and between the intervention and control arms were estimated using weighted t tests for physical activity measures and Mann–Whitney tests for body weight, blood pressure, and the 6-minute walk test.

The characteristics of the intervention and control arms are listed in Table 1. Most participants (29/44, 66%) were enrolled at the start of their first line of chemotherapy, and 6 were receiving their third or more line of chemotherapy (6/44, 14%). The individuals enrolled with an initial diagnosis of stage 1 or 2 disease were receiving neoadjuvant chemotherapy (1 person), adjuvant chemotherapy to reduce the risk of recurrence (3 people), or chemotherapy for recurrent disease (1 person). The intervention and control groups had a similar median age at enrollment and similar gender and cancer site, stage, and treatment distributions. However, by chance owing to the small sample size, a higher proportion of the control group were patients from UCSF who identified as Asian American or Pacific Islander; the median BMI of this group was also lower than that of the intervention group.

Participants randomized to the intervention arm wore their Fitbits for a median of 67 out of 84 study days (IQR 53-80 days). A total of 2 participants never wore the Fitbit, and 2 participants had <10 days of wear time. A total of 6 participants had >80 days of wear time. Fitbit use trended down slightly over time (Figure 2). There was no correlation between age and gender of the participants and wear time. Participants with stage 4 cancer had a median Fitbit wear time of 56 days (IQR 47-76 days) compared with a median of 77 days among participants with stage 1 to 3 disease (IQR 56-82).

Overall, participants in the intervention arm responded to a median of 17 out of 27 SMS text messages that asked for a reply (63%; IQR 12-23; range 1-26). SMS text message response rates fluctuated over time (Figure 3). SMS text messages sent on days 15, 36, and 62, which queried whether participants had achieved the goals they were asked to set at the beginning of the study on day 8, were among the messages with the lowest response rates. No patterns were observed regarding the content of SMS text messages that received the highest response rates. SMS text message response rates did not vary by age, gender, or cancer stage.

Most participants reported that the intervention was acceptable (Table 2 and Table 3). Out of the 22 participants in the intervention arm, 19 (86%) completed the feedback questionnaire. Among the respondents, 63% (12/19) reported satisfaction with the SMS text messages overall and 89% (17/19) reported satisfaction with the Fitbit and an expectation that they would continue to wear the Fitbit after the study ended.

When asked about specific features (Table 3), 68% (13/19) agreed that the SMS text messages motivated them to exercise and that the content was interesting; 74% (14/19) said that the frequency of the messages was ideal (1 every 1-3 days), and 79% (15/19) said that the timing of the messages was ideal (morning and evening). The most frequent recommendation for improvement was to improve the personalization of messages (Multimedia Appendix 2). Regarding the Fitbit, 2 participants said that they did not like using a wearable device or did not feel the need to track their activities daily. Additional feedback from participants included difficulty adhering to the intervention because of treatment-related fatigue and restrictions imposed during the COVID-19 pandemic.

Physical activity levels measured by the accelerometer at enrollment and 12 weeks for participants in the intervention and control groups are shown in Multimedia Appendix 3. No patient recorded any vigorous physical activity at any time point. Overall, on average, the participants accumulated 110 minutes per week (SD 103 minutes per week) of moderate-intensity activity in bouts of 10 minutes or longer at enrollment. By chance, the intervention arm recorded more time in moderate activity bouts compared with controls at enrollment (mean 141.5, SD 115.5 minutes per week and mean 80.7, SD 83.5 minutes per week, respectively).

When examining changes from 0 to 12 weeks, both the intervention and control groups decreased their physical activity on average over the 12-week study period. The intervention arm had a mean reduction in moderate activity accumulated in 10-minute bouts of 21.3 minutes per week (SD 144.8 minutes per week); the control arm had a mean reduction in moderate activity accumulated in 10-minute bouts of 16.3 minutes per week (SD 121.2 minutes per week). There was no difference in the change in moderate activity accumulated in bouts of 10 minutes when comparing the 2 groups (mean difference 0.2, SD 6.2 minutes per week). When examining individual changes, 47% (8/17) of participants in the intervention arm and 35% (7/20) in the control arm with data at both time points increased bouts of moderate activity from enrollment to 12 weeks by at least 1 minute. Notably, when examining total activity throughout the day (not specifically ≥10-minute bouts), participants in the intervention arm reduced moderate activity, light activity, and steps more than the control arm. Finally, the total time moving at a moderate intensity was high in both the intervention and control arms and the change in activity between time points was highly variable with wide SDs.

The 6-minute walk test, body weight, and blood pressure at enrollment and 12 weeks for participants in the intervention and control groups are shown in Multimedia Appendix 3. Participants in both arms increased their 6-minute walk test distance by an average of 37 meters (SD 39 meters) in the intervention group and 46 meters (SD 59 meters) in the control group. For body weight, the intervention group had a mean change of −0.8 pounds (SD 5.7 pounds), whereas the control group had a mean change of 0.1 pounds (SD 9.5 pounds). When examining individual changes, we observed that 65% (11/17) participants in the intervention arm with data available at both time points lost weight from enrollment to 12 weeks, whereas 42% (8/19) participants in the control arm lost weight from 0 to 12 weeks. There were no significant changes in blood pressure within or between the 2 groups. The average difference in systolic blood pressure from 0 to 12 weeks for the intervention group was 6.4 mm Hg (SD 12.2 mm Hg); and the control group had a mean change of −0.4 mm Hg (SD 10.6 mm Hg). The average difference in diastolic blood pressure for the intervention group was −0.5 mm Hg (SD 9.6 mm Hg); the control group had a mean change of −5.8 mm Hg (SD 9.0 mmm Hg). When examining individual changes, 53% (9/17) participants in the intervention arm and 67% (12/18) participants in the control arm decreased their systolic and diastolic blood pressure from 0 to 12 weeks. As with the physical activity data, there was considerable variability in responses between participants.

The number of reported adverse events is presented in Table 4. There were no serious adverse events related to the intervention, and the intervention did not appear to increase reports of nonserious adverse events compared with baseline. A total of 4 participants in the control group reported hospitalizations during the study, and 1 participant in the control group passed away during the study because of cancer progression. There were no hospitalizations or deaths in the intervention group.

For nonserious adverse events, fatigue was the most reported adverse event during the study, but the number of times fatigue was reported was highest at enrollment and it did not increase during the intervention period. In addition, as described above, 4 of the SMS text messages in the intervention arm asked participants to rate how they felt (days 4, 31, 45, and 67). On day 4, 32% (7/22) of the participants in the intervention arm responded saying they were very tired, 27% (4/15) said they were very tired on day 31, 11% (2/19) said they were very tired on day 45, and 27% (4/15) said they were very tired on day 67.

Overall, we observed that a remotely delivered physical activity intervention that included a wristband for self-monitoring physical activity and SMS text messages during chemotherapy for CRC was feasible and acceptable. Although this study was not powered to detect changes in physical activity, our pilot data show a nonstatistically significant decrease in moderate activity accumulated in bouts of at least 10 minutes in both arms (16-21 minutes per week).

Notably, the findings from this study with participants who were actively receiving chemotherapy differed from our previous study in people who had previously completed treatment for CRC. In our previous study (Smart Pace I), we observed an average increase in physical activity in participants in the intervention arm [13]. The main difference in our SMS text message content for this study (Smart Pace II) was the addition of questions about how participants felt and tailored activity advice in response. This modification was based on our expectation that participants would feel fatigued during chemotherapy and need support or motivation to promote activity. Messaging to take it easy or build up slowly sent to a group of people who were active at baseline and felt tired on treatment may have unintentionally contributed to why the intervention arm decreased activity levels slightly more than the control arm, which did not receive SMS text messaging. Further research is required to evaluate whether such messages would have the intended beneficial effect in a sedentary population (encouraging those who feel tired to do a light activity vs nothing). In addition, delivering more nuanced messages that encourage active people to stay active during treatment even when they are tired, without pushing them too far, is a challenge for automated intervention approaches such as SMS text messaging.

Few other studies have conducted remote physical activity interventions in patients with CRC or survivors. Kim et al [24] reported that a home-based exercise intervention with weekly supervised components (counseling or training sessions) significantly increased self-reported moderate physical activity from 97 minutes per week at enrollment to 325 minutes per week at 12 weeks, with no change observed among the controls. These data are consistent with previous findings from our team in a study of men with prostate cancer. In the Community of Wellness study, we observed a modest change in self-reported physical activity but only among men who reported <90 minutes per week of activity at baseline and in the group that received one coaching call with an exercise trainer [25]. It is possible that these previous studies reported greater changes in activity compared with this pilot study because they used self-report rather than objective measures. Nonetheless, some degree of coaching or more personalized contact may be needed to help people with cancer assess their current level of activity and identify what changes are needed to meet the physical activity guidelines and optimize their cancer outcomes.

The baseline physical activity level measured at enrollment was high in both arms and particularly high in the intervention arm. This occurred despite the exclusion of prospective participants who self-reported ≥150 minutes per week of MVPA. However, physical activity measured using the accelerometers indicated that self-reported MVPA may have underestimated actual MVPA. It is also possible that participants engaged in higher than usual levels of activity when wearing the devices. Interestingly, when we analyzed moderate activity accumulated in bouts of ≥10 minutes, the participants’ activity levels were similar to self-report. Although logistically difficult, future studies should consider using accelerometer data to determine eligibility or set a lower cutoff point for self-report to ensure they enroll an inactive study population who may most benefit from the intervention.

In addition, our sample included highly educated participants and low enrollment of Black or Latinx CRC survivors, which may limit the generalizability of our findings. Given the high CRC incidence and mortality among Black people and rising rates of young-onset CRC in some Latinx populations, research is critically needed in these patient groups [26,27]. Self-identified race or ethnicity was not assessed in our study until after participants provided consent. Although Hispanic or Latinx patients comprise 17% of patients with CRC at our institution, it is possible many may have been excluded owing to the requirement for English proficiency. We encourage future studies to support translations into multiple languages and to track the race or ethnicity of all screened participants to identify and address potential barriers to enrollment and ensure future studies enroll representative patient populations.

Finally, the COVID-19 pandemic began while the last 7 participants were active in the study. Several of our SMS text messages provided tips for participants to find social support and exercise with others, which were perceived by participants as irrelevant or incompatible with social distancing guidelines imposed during the pandemic. Out of these 7 participants, 6 (85%) participants had paired accelerometer data available. Out of these 6 participants, 5 (83%) decreased their time spent in bouts of moderate activity at 12 weeks compared with enrollment; 1 participant increased their time spent in moderate activity bouts. Although the numbers are small, it is possible that the pandemic led to a slightly greater decrease in planned moderate activity from enrollment to 12 weeks in our study, on average, than would have been observed in a study conducted before the COVID-19 pandemic. Decreases in physical activity, on average, have also been reported among noncancer study populations during the pandemic [28].

Although the intervention was determined to be feasible and acceptable, there are aspects that could be improved in future studies. The main takeaway based on participant feedback was that the intervention, specifically the SMS text messages, needed to be more personalized. For example, several participants suggested that SMS text messages could be tailored to the data collected from the Fitbit. Future studies may be strengthened by having real-time access to activity data to determine appropriate messaging. Machine learning approaches, such as reinforcement learning, could also provide a platform to improve the tailoring of SMS text messages [29]. In addition, 4 participants in the intervention arm did not wear the Fitbit. Although the number is small, it is worth considering offering other mechanisms for self-monitoring physical activity, such as paper diaries, in future studies. Finally, studies are needed to determine the feasibility and acceptability of digital health physical activity interventions in individuals with lower levels of education, individuals with low English proficiency, and individuals who identify with minority racial or ethnic groups. Adaptation of digital health interventions and messaging into other languages and with attention to cultural contexts will be critical to improving access for a more diverse population.

Overall, this pilot study demonstrated that patients with CRC were interested in a remotely delivered, automated digital health physical activity intervention during chemotherapy. However, more tailored support is needed to further enhance participant satisfaction and possibly improve physical activity behavior.