Feasibility and Acceptability of a Web-Based Dietary Intervention with Text Messages for Colorectal Cancer: A Randomized Pilot Trial

Over 1.2 million Americans currently live with colorectal cancer in the United States (1). Although these individuals are living longer, colorectal cancer remains the second-leading cause of cancer death in the United States (2).

Diet after colorectal cancer may impact survival (3). In particular, adherence to the American Cancer Society (ACS) guidelines, including a diet rich in fruits, vegetables, and whole grains and low in red and processed meats (4), after colorectal cancer diagnosis is associated with lower risk of colorectal cancer-specific and all-cause mortality (5, 6). Unfortunately, adherence to these recommendations is low among individuals with colorectal cancer (5, 6). Thus, interventions are needed to help people with colorectal cancer adopt dietary recommendations.

Growing data suggest that behavioral interventions using Web and mobile technology are acceptable and may modify dietary behavior (7–12). Because these interventions can be largely automated, they may be more scalable in cancer survivorship care compared with in-person or telephone counseling (13, 14). However, the feasibility of a remotely delivered technology-based dietary intervention among colorectal cancer survivors is not known.

Identifying a feasible and acceptable dietary intervention that is also scalable is a critical step toward the long-term goal of determining the effect of dietary change on colorectal cancer survival. Thus, we developed a website with a simple diet tracking tool and a 12-week text messaging program to help individuals with colorectal cancer make recommended dietary changes. We conducted a two-arm randomized controlled pilot study to determine the feasibility and acceptability of the intervention. To inform the design of a larger definitive study, we also explored change in diet, anthropometry, and circulating biomarkers of cardiometabolic health.

We conducted a pilot study of a remotely delivered dietary intervention among 50 individuals with colon or rectal cancer (Survivor Choices for Eating and Drinking; SUCCEED). At enrollment, participants were randomized 1:1 to receive the intervention for 12 weeks or wait-list control. We included a wait list control to facilitate enrollment and maximize feedback on the intervention. All participants were asked to complete 4-day diet records and surveys at 0, 12, and 24 weeks. Written informed consent was obtained from all participants and the study was conducted in accordance with recognized ethical guidelines (e.g., Declaration of Helsinki, CIOMS, Belmont Report, U.S. Common Rule) and approved by the institutional review board of the University of California, San Francisco (UCSF).

To be eligible, individuals must have been diagnosed with colon or rectal adenocarcinoma; not be actively undergoing chemotherapy; be considered disease-free or have stable disease; be able to speak and read English; have access to a mobile phone with Internet and text messaging capabilities; and able to navigate websites, fill out forms on the Web, communicate by email, and have regular access to the Internet. We excluded individuals who were already meeting four or more of the six target dietary behaviors at enrollment (based on self-reported intake of vegetables, whole grains, fish, processed meat, sugar-sweetened beverages, alcohol; Table 1).

Participants were identified through the gastrointestinal oncology clinic at UCSF between April 2017 and May 2018. We also advertised at local clinics, support groups, and a national conference. If interested, individuals were asked to complete a screening survey online and have their provider complete a form to verify clinical information (e.g., diagnosis, stage, treatment). The screening survey asked participants how many servings of vegetables, whole grains, processed meat, fish, sugar-sweetened beverages, and alcohol they usually consume. A definition of the food group as well as serving size was provided. For vegetables, participants were asked to report how many servings they consume each day, with nine options ranging from zero to nine or more servings/day. For the other food groups, participants were asked “on average, how many servings of [food group] do you eat?” and were provided with nine options, ranging from never to four or more servings/day.

After an individual was determined to be eligible, they were sent a consent form via mail, e-mail, or fax. A research coordinator scheduled a time to review the consent form over the phone. At the conclusion of the call, the research coordinator asked the participant to confirm that they understood the consent form, and if so, asked them to sign a paper or electronic consent form.

Following consent, participants were asked to complete 4-day diet records using the NCI's Automated Self-administered Dietary Assessment Tool (ASA24) and surveys using REDCap electronic data capture tools hosted at UCSF (15–18). Participants repeated the 4-day diet records and surveys at 12 and 24 weeks. Participants were also asked to go to a LabCorp for a fasting blood draw and anthropometric assessment at enrollment and 12 weeks.

Eligible, consented participants who completed enrollment assessments were randomly assigned in blocks of two and four to intervention or control with allocation weight of 1:1. A study biostatistician (LZ) generated the randomization scheme prior to the start of the study and this was uploaded into REDCap. The Randomization module in REDCap was then used to determine participants' intervention assignment.

We designed a 12-week Web-based dietary intervention using human centered design methodology with the UCSF School of Medicine Technology (SOM Tech; ref. 19); a patient advisory board (PAB) of five colorectal cancer survivors; and a registered dietician at the UCSF Helen Diller Family Comprehensive Cancer Center. The PAB participated in intervention development and tested a prototype for 1 week. We then conducted semistructured interviews with PAB members and revised the intervention.

The goal of the intervention was to increase vegetables, whole grains, and fish and decrease processed meats, sugar-sweetened beverages, and alcohol. These goals were based on the ACS guidelines plus literature on diet and colorectal cancer survival (4, 20–23). The intervention included print materials, a personalized report, one 15 to 30 minute session with a study coordinator in person (22%) or by phone (78%), a study website, and text messages for 12 weeks. The personalized report indicated whether the participant currently met, almost met, or did not meet the six target dietary recommendations at enrollment (Table 1). The orientation session focused on how to use the study website; no dietary counseling or health coaching was performed. The intervention incorporated theoretical constructs from the Theory of Planned Behavior and Social Cognitive Theory and addressed outcome expectations, self-efficacy, goal setting, self-monitoring, and social support (24–26).

The website was designed to be used on a computer, tablet, or phone and built on the UCSF Drupal 8 infrastructure. We collected and stored data from participants securely with a username and password. Data collected through the website were encrypted and transmitted to a HIPAA-compliant secure server. The website included goal setting, daily tracking of target food groups, visual summaries of tracked dietary intake and progress toward goals, recommendations, recipes, meal planning, frequently asked questions, and a profile page. Participants could contact the study team with questions through the website. In the profile page, participants could turn tracking on or off. For example, if an individual abstained from alcohol, they could turn it “off” and alcoholic drinks would not show in the goals, tracking, or progress pages. Prompts to visit the website were included in 13 text messages, and up to an additional five messages depending on participants' responses. No reminders were sent to participants based on their engagement with the website. Data on website use by participant were obtained through logins, tracking data in the study website (which recorded responses by study ID, date, and time), as well as google analytics (see Supplementary Fig. S1A to S1D, for screenshots of the website).

We sent text messages to participants via Twilio (www.twilio.com). The messages included educational content (e.g., how to estimate serving size, what is a whole grain); prompts to set goals, track, meal plan, try recipes; motivational messages; quizzes; and challenges. Participants could set the time of day (i.e., morning, afternoon, evening) that they received messages and turn the messages on/off through the website. Twenty-one of the 84 text messages (25%) asked the participants for a reply (see Supplementary Table S1, for the first 14 text messages).

Participants randomized to the control arm received print materials on diet after colorectal cancer at enrollment. After completion of the 12-week assessments, controls had the option to receive the intervention from 12 to 24 weeks; 21 of the 25 control participants received the intervention. These individuals received access to the study website, had a brief orientation session on how to use the website, and 12-weeks of daily text messages.

We assessed the intervention's feasibility and acceptability at 12 weeks in the intervention arm and 24 weeks in the control arm (among individuals who received the intervention from 12 to 24 weeks). To assess the intervention's feasibility, we evaluated adherence (frequency of using the study website and response rates to text messages that asked for a reply) and attrition (proportion of participants who completed the 12-week diet records). A priori, we stated that the intervention would be considered feasible if we achieved ≥70% adherence and ≤20% attrition in the intervention arm (27, 28).

We explored the acceptability of the intervention through a self-administered investigator-developed online questionnaire. Participants were asked to what degree they agreed (strongly agree, agree, undecided, disagree, strongly disagree) with nine statements about the text messages. For the study website, participants were asked to rate the quality of the website, overall and each page, from poor (0) to excellent (100). Participants were asked how satisfied they were overall and could provide open-ended feedback.

Participants were asked to complete 4-day diet records using ASA24 at 0, 12, and 24 weeks (29). Participants received instructions with screenshots to guide them through completing a diet record on ASA24. Participants who had trouble using the online system were allowed to submit paper diaries, which were entered into ASA24 by one trained researcher blinded to the participants' assigned study arm. If a participant submitted a paper diary at enrollment, they completed paper diaries at subsequent time points. In total, 11 (22%) enrollment diaries, 16 (36%) 12-week diaries, and 14 (35%) 24-week diaries were entered by the researcher. All of the days recorded had acceptable calorie intakes, defined as between the 5th and 95th percentile of calorie intakes based on ASA24 users in the National Health and Nutrition Examination Survey (600–4,400 kcal/day for women; 650–5,700 kcal/day for men; ref. 30).

Dietary outcomes of interest included change in vegetables, whole grains, fish, processed meat, sugar-sweetened beverages, and alcohol from enrollment to 12 and 24 weeks. We also examined change in refined grains, to explore if participants replaced refined grains with whole grains or added the recommended servings of whole grains without decreasing refined grain intake. We measured diet at 24 weeks in both arms to explore whether changes that occurred in the intervention arm from 0 to 12 weeks were maintained at 24 weeks, and whether individuals in the control arm who received the intervention from 12 to 24 weeks made any changes.

Participants were asked to go LabCorp, an accredited clinical laboratory, at 0 and 12 weeks. A trained technician measured participants' height, weight, waist circumference, and blood pressure and took a fasting blood sample. The technician was an employee of LabCorp and blinded to the participants' assigned group. LabCorp visits were optional, and two participants chose not to complete them at enrollment. Assessments were done according to the LabCorp standard operating procedures. From the fasting blood sample, LabCorp measured C-reactive protein, hemoglobin A1c, glucose, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, and triglycerides.

We powered the study based on the proportion of participants who adhered to the intervention, defined as using the website and/or engaging with text messages. Using a one-sided one-sample binomial test with α = 0.1, 20 participants who complete the trial in the intervention arm, and a null hypothesis that 70% would adhere to the intervention, we estimated that we would have 80% power to reject the null hypothesis if the observed adherence was 47% or less. We randomized 50 participants to account for up to 20% drop-out.

Enrollment characteristics were summarized by arm and overall. Frequency distributions and percentages were used to summarize categorical measures and medians (interquartile range, IQR) were used to describe continuous measures. Point estimates of adherence (frequency of using the study website, response rate to text messages that asked for a reply) and attrition (completion of diet records) were calculated by arm.

To estimate the effect of the intervention on diet, we calculated each participant's average servings per day for the target food groups based on the 4-day diet records at enrollment, 12, and 24 weeks. We then calculated the median (IQR) intake of each target food group at each time point by arm. Because alcohol has different recommendations by sex, and to explore the effect of the intervention separately in men and women, we reported each food group by sex. We described the absolute change in intake from 0 to 12 and 0 to 24 weeks using medians and IQR. We also described the difference in changes from enrollment to 12 weeks between the two groups using means and 95% confidence interval (CI). We adhered to the CONSORT guidelines for reporting pilot trials and did not perform statistical tests (31). Analyses were conducted using SAS v. 9.4.

As shown in Fig. 1, we screened 94 individuals for eligibility. Of these, 19 (20%) were ineligible, 14 (15%) met 4 or more of the 6 target dietary behaviors, 4 were receiving chemotherapy, and 1 did not have colorectal cancer. Among the remaining individuals, 11 decided not to consent after completing the screening survey and 14 did not complete enrollment procedures. After screening, consent, and completion of enrollment procedures, 50 individuals were randomized 1:1 to intervention (n = 25) or wait-list control (n = 25). All participants received their assigned intervention. Follow-up based on completion of the diet record was 90% at 12 weeks and 84% at 24 weeks. By arm, these values were 22 (88%) in the intervention arm and 23 (92%) in the control arm at 12 weeks and 22 (88%) in the intervention arm and 20 (80%) in the control arm at 24 weeks.

Characteristics of the study sample are presented in Table 2. The intervention versus control group was slightly older (median: 57 years vs. 54 years), had a higher BMI (median: 26.9 kg/m² vs. 25.5 kg/m²), and were more likely to be female, have rectal cancer, and have elevated cholesterol.

Responses to text messages were relatively stable over the 12 weeks in the intervention arm (Fig. 2A). For any given text message, the median number of responses was 15 (60%; IQR = 15, 16). When examining text message responses by participant, those in the intervention arm responded to a median of 15 (71%) of the 21 text messages that asked for a reply (IQR = 8, 19).

The response rate to text messages among people initially assigned to the control arm was lower and declined over time (Supplementary Fig. S2A). For any given text message that asked for a reply, the median number of responders was 12 (57%; IQR = 10, 13). When examining text message response by participant, those in the control arm responded to a median of 9 (43%) of the 21 text messages that asked for a reply (IQR = 7, 17).

Intervention arm participants went to the website throughout the 12-week study, although website use declined over time (Fig. 2B). On any given study day, the median number of participants who visited the study website was 6 (24%; IQR = 4, 7). When examining website use by participant, those randomized to the intervention arm went to the website a median of 13 (15%) of 84 days (IQR = 1, 33 days). Five participants did not visit the website and three visited once during the 12 weeks. In contrast, five participants went to the study website on more than half of the 84 study days; the most frequent user visited the website on 63 days. Twenty (80%) intervention participants tracked a food on the website at least once (Fig. 2C). Among trackers, the median number of days with at least one food group tracked was 48 days (57%; IQR = 25, 82 days). This value was higher than website visits because participants often recorded multiple days of intake at each visit. When an individual tracked, they entered data for all six target food groups 53% of the time.

Website use among those initially assigned to the control group was lower compared with the intervention arm (Supplementary Fig. S2B). On any given study day between week 12 and 24, the median number of control participants who visited the study website was 3 (14%; IQR = 2, 5). When examining website use by participant, those randomized to the control group went to the study website a median of 4 (5%) of 84 days (IQR = 0, 21 days). Seven control participants did not visit the website and two went on 1 day. In contrast, three control participants went to the study website on more than half of the 84 days; the most frequent control user went to the website on 72 days. Seventeen (81%) control participants tracked a food group on the website at least once (Supplementary Fig. S2C). Among trackers, the median number of days with at least one food group tracked was 33 days (39%; IQR = 22, 78 days). As with the intervention arm, this value was higher than website visits because participants recorded multiple days of intake per visit.

On the basis of self-report, participants accessed the website from an iPhone (24, 57%), computer (23, 55%), android phone (9, 21%), iPad (7, 17%), and/or android tablet (2, 5%). The most used page was Tracking (69% reported they used this page frequently) followed by recipes (40%), view progress (37%), set goals (23%), recommendations (17%), meal planning (14%), and FAQ (6%).

The feedback questionnaire was completed by 22 of the 25 intervention participants at 12 weeks and 20 of the 21 control participants who received the intervention at 24 weeks. Of these, 31 (74%) were satisfied or very satisfied with the text messages, 27 (64%) said the content of the messages was interesting, and 33 (79%) said the frequency of the messages (1 per day) was ideal. Feedback on the text messages was generally positive. However, some participants commented that they would have liked additional personalization. For example, “… it would be great to tailor some of the messages if I select what I am already good at” and “the language was impersonal” (see Supplementary Table S2, for participant quotes).

Overall, 28 (64%) participants were satisfied or very satisfied, 11 (26%) were neutral, and 3 (10%) were dissatisfied or very dissatisfied with the intervention. The one participant who responded very dissatisfied was in the intervention arm and stated that the website's simplified tracking “cancelled out a lot of things on my diet (ex. [I didn't know where to add things such as miso soup).” The two participants who responded “dissatisfied” were both in the control arm; one of whom experienced a recurrence during the first 12 weeks of the study prior to receiving the intervention.

Intake at enrollment and absolute change from enrollment of vegetables, whole and refined grains, fish, processed meats, sugar-sweetened beverages, and alcoholic drinks at 0, 12, and 24 weeks are presented in Table 3. The intervention arm had higher intake of whole grains at 12 weeks compared with controls (difference in means comparing the intervention to control group at 12 weeks: 0.9 servings/day; 95% CI, 0.1–1.6). This difference appeared to be driven by women. In addition, the intervention arm's increase in whole grains appeared to be maintained at 24 weeks. Controls had no clear improvements in diet from 0 to 12 or 24 weeks.

We calculated the percentage of participants who met our target recommendations at each time point by arm (Fig. 3). Overall, adherence to the recommendations was low. The recommendation to avoid or limit alcohol to moderate levels had the highest adherence rates at 80% and 76% in the intervention and control arms, respectively, at enrollment. Only one participant in the intervention arm met the recommendation to consume 5 servings/day of vegetables at enrollment; no participant achieved this target at 12 or 24 weeks. We observed modest improvements in the percentages of people meeting recommendations for whole grains (0, 12, and 24 week percentages; intervention: 0%, 16%, 12%; control: 4%, 0%, 16%); fish in the intervention arm (56% at enrollment and 68% at 12 weeks); and sugar-sweetened beverages in the intervention arm (36%, 48%, and 52% at 0, 12, and 24 weeks).

Anthropometric and blood values at 0 and 12 weeks are listed in Supplementary Table S3. On average, participants did not lose weight or waist circumference during the study. Compared with controls, which had no improvements, the intervention arm appeared to have lower systolic blood pressure at 12 weeks (difference in means: −7 mmHg; 95% CI, −14 to 1).

We developed a remotely delivered Web-based dietary intervention for colorectal cancer survivors. The text message program met the prespecified criteria for feasibility (>70% adherence), but the study website with a simple diet tracking tool did not. Although use of the website was less than intended, intervention participants appeared to increase whole grain consumption.

Our study had a number of strengths. We targeted our intervention for colorectal cancer survivors, because diet may impact colorectal cancer outcomes and this cancer site has been underrepresented in past dietary interventions (3, 26). In addition, we enrolled a fairly diverse study population including 30% non-white and 34% men. Most of the research on lifestyle interventions for cancer survivors has been conducted in white women with breast cancer (26). However, women were still overrepresented in our study population compared with all people diagnosed with colorectal cancer. Our study population was also younger and had a high level of education, which may limit the generalizability of the findings. Future studies are needed to determine the intervention's feasibility and acceptability in populations with more men, older individuals, and/or less education.

The ACS recommends 5 servings/day of fruits and vegetables. We focused on vegetables only, due to literature suggesting that carbohydrates may be associated with an increased risk of colorectal cancer recurrence and death (32). However, given that only one participant met the goal of 5 serving/day of vegetables at enrollment and no participants met the goal at 12 or 24 weeks, a lower target for vegetable intake may be more reasonable for this population. A pilot trial of 17 overweight/obese adults suggest that a 12-week digital health intervention targeting only vegetable intake was successful for increasing vegetable consumption (33). It is possible sequential diet goals, where participants are only asked to change one thing at a time, would be more successful for increasing overall healthy diet in the long term.

We did not include recommendations on food quantity or calorie intake in our intervention. Data on body weight and survival after colorectal cancer are mixed, and it is not known if calorie restriction is beneficial in colorectal cancer survivors (34). Accordingly, we did not see a change in body weight in the intervention arm from 0 to 12 weeks, suggesting participants did not change their calorie intake while participating in the study.

Although the intervention appeared feasible and acceptable, a number of aspects could be improved. The optimal frequency of website use/diet tracking to achieve dietary change is not known (35, 36). It is possible that less frequent tracking is needed to change intake of food groups, such as whole grains, compared with managing calorie intake for weight loss. In addition, a key area for improvement is personalization of text messages. Advances in technology, including artificial intelligence, hold promise for tailoring messages while maintaining scalability (37). In addition, it would be of interest to test the independent effect of text messages in a future study using a multiarm or factorial design.

In conclusion, we developed a Web-based dietary intervention for colorectal cancer survivors. Our pilot results suggest colorectal cancer survivors may engage more with text messages than a study website. Research to improve tailoring of text messages, while maintaining scalability, is needed.

No potential conflicts of interest were disclosed.

Conception and design: E.L. Van Blarigan, S.A. Kenfield, K. Van Loon, C. Miaskowski, Y. Fukuoka, A.P. Venook

Development of methodology: E.L. Van Blarigan, Y. Fukuoka

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.L. Van Blarigan, K. Van Loon, H. Chan, M.B. Savoie, A.G. Bocobo, L.X. Wong, A. Laffan, C.E. Atreya

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.L. Van Blarigan, S.A. Kenfield, J.M. Chan, A. Paciorek, L. Zhang, A.G. Bocobo, L.X. Wong, C. Miaskowski, J.A. Meyerhardt, A.P. Venook

Writing, review, and/or revision of the manuscript: E.L. Van Blarigan, S.A. Kenfield, J.M. Chan, K. Van Loon, A. Paciorek, L. Zhang, M.B. Savoie, V.N. Liu, L.X. Wong, C.E. Atreya, C. Miaskowski, Y. Fukuoka, J.A. Meyerhardt, A.P. Venook

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.L. Van Blarigan, V.N. Liu, L.X. Wong

Study supervision: K. Van Loon

We thank the patient advocates, Greta Macaire, RD, and the team at UCSF School of Medicine (SOM) Tech and Y3TI for their help developing the SUCCEED intervention. Research reported in this publication was supported by the NCI of the NIH under Award Number K07CA197077 to E. Van Blarigan. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.