Cancer Health Self-Efficacy Improvement in a Randomized Controlled Trial

Abstract

Background

eHealth interventions can help cancer survivors self-manage their health outside the clinic. Little is known about how best to engage and assist survivors across the age and cancer treatment spectra.

Methods

The American Cancer Society conducted a randomized controlled trial that assessed efficacy of and engagement with Springboard Beyond Cancer, an eHealth self-management program for cancer survivors. Intent-to treat analyses assessed effects of intervention engagement for treatment (on-treatment vs. completed) overall (N=176; 88 control, 88 intervention arm) and separately by age (<60 vs. older). Multiple imputation was used to account for participants who were lost to follow-up (n=41) or missing self-efficacy data (n=1) at 3 months follow-up.

Results

Self-efficacy for managing cancer, the primary outcome of this trial, increased significantly within the intervention arm and for those who had completed treatment (Cohen’s d=0.26, 0.31, respectively). Additionally, participants with moderate-to-high engagement in the text and/or web intervention (n=30) had a significantly greater self-efficacy for managing cancer-related issues compared to the control group (n=68), with a medium effect size (Cohen’s d=0.44). Self-efficacy did not differ between the intervention and control arm at 3-month post-baseline.

Conclusions

Study results suggest that cancer survivors benefit variably from eHealth tools. To maximize effects of such tools, it is imperative to tailor information to a priori identified survivor subgroups and increase engagement efforts.

Precis:

This randomized controlled trial assessed the efficacy of the Springboard Beyond Cancer eHealth program in a sample of 176 cancer survivors. Results demonstrated significant improvement in self-efficacy for managing cancer issues among the intervention group and no significant change among the control group.

As of January 2019, the US had more than 16.9 million cancer survivors, a number projected to exceed 22.1 million by 2030. These impressive gains in survival result from improved screening and early detection, better multi-modal cancer treatments, and population aging^1,2. Such gains come with a cost; many survivors report challenges from long-term and late effects of treatment or in managing pre-existing chronic conditions. Specific challenges include physical symptoms, (e.g., pain and fatigue), functional limitations; emotional issues, like fear of recurrence and depression; and financial concerns^3,4. Information alone is not enough to drive behavior change⁵. Indeed, survivors often need to learn new strategies to effectively manage cancer-related concerns and to continue or adopt healthy behaviors, like engaging in recommended levels of physical activity^6,7.

Interventions are often designed to initiate and support actionable strategies to enhance patient self-management of symptoms and to promote changes in health behaviors to improve survivor health outcomes^8,9. Chronic disease self-management programs commonly help patients choose appropriate management strategies, set measurable goals for behavior change, self-monitor their progress, and boost health self-efficacy (confidence in one’s ability to manage various aspects of one’s disease(s)) and a primary objective of self-management^10,11. Comprehensive programs tailored for cancer-specific issues during and after treatment are needed to fill gaps in currently offered self-management programs. Most programs that are available for cancer survivors focus on one specific cancer type^12,13 or symptom¹⁴ (e.g., fatigue, psychological adjustment), narrowing their ability to comprehensively address the complexity of issues faced by cancer survivors’¹⁵. Further, many interventions are only available as face-to-face programs, which are costly and reach limited numbers of patients. Self-management programs should have universal and equitable reach through highly engaging programs delivered at the point of need (e.g., in survivors’ homes), rather than only at the point of care in clinicians’ offices⁸. eHealth programs have the potential to reach many cancer survivors at a time and place convenient to survivors. Despite the potential positive impact of online self-management programs, few free, comprehensive, and empirically supported programs have been specifically designed for cancer survivors.

To address this gap, the American Cancer Society (ACS) and the National Cancer Institute (NCI) began a joint venture to build a free, comprehensive eHealth program for cancer survivors, Springboard Beyond Cancer (SBC) (http://web.archive.org/web/20200417092703/https:/survivorship.cancer.gov/)¹⁶.

This manuscript presents the results of a randomized controlled trial (RCT) to compare self-efficacy for managing cancer-related issues among cancer survivors randomized to receive the enhanced SBC intervention (website and text message components) versus a website covering the same topics without dynamic and interactive features (e.g., images, videos, links, and personalized lists). This control website more closely mimicked a traditional printed brochure available from many cancer centers and cancer nonprofit organizations that includes information and education about common survivorship issues, such as fatigue and pain, to represent standard care. A priori analyses investigated differences in change of self-efficacy by age, treatment status, and engagement with the intervention. Self-efficacy was used as the outcome variable given its essential role as a precursor to behavior change to self-manage multiple outcomes (e.g. fatigue, depression, health behavior improvement).

Methods

This study received expedited human subjects’ approval by both the Morehouse School of Medicine IRB and the ICF IRB. ICF is a consulting firm that provided digital and communication services to develop the SBC website and text program.

Sampling

Eligibility criteria included: (1) history of cancer; (2) diagnosis at age 18 or older; (3) regular Internet access; (4) valid email address; and (5) self-reported ability to read English “well” or “very well” (compared to “not well” or “not at all”).

Recruitment was conducted through multiple channels. Direct email invitations were sent to all potentially eligible participants in the constituent database (900 men and 3,600 women), though it was not possible to track how many email invitations were successfully delivered to a valid email address or were opened. Based on feedback from ACS marketing, the risk for undeliverable emails and emails routed to a junk folder were quite high. For online recruitment, study advertisements were on the ACS Cancer Survivors Network and the ACS Twitter handle. Flyers were available at ACS Hope Lodges and through the ACS Patient Navigator program. It is unclear how many people were reached using these recruitment measures.

Study design

Full study flow is shown in a Consolidated Standards of Reporting Trials (CONSORT) diagram, Figure 1. Potential participants were directed to an online screening questionnaire to assess eligibility for the study. Among 309 survivors who completed the screener, 133 were ineligible for a variety of reasons, such as incomplete screener (N=54) or baseline survey (N=61). Eligible participants were directed to the study web page, which contained additional study details and the online consent form. Individuals who completed the consent form were directed to the online baseline survey. Upon completing the baseline survey, participants were randomized to either the intervention group (N=88) or control group (N=88) and were directed to a webpage with instructions for control or intervention website access. An error directing one participant to the assigned webpage led to one person in the control group receiving the text message intervention. As analyses were intent-to-treat, this individual was kept as a control participant in the main analyses.

Figure 1.

CONSORT trial flow diagram.

Intervention group

The intervention group was assigned a research version of an updated version of SBC that included a log-in allowing for participant activity tracking. SBC was guided by prior digital health interventions, specifically through the inclusion of effective features of Smokefree.gov^17,18, and underwent rigorous useability testing, described in detail elsewhere¹⁶ that followed principles published by www.usability.gov. Participants in the intervention arm were also enrolled in the newly created 4-week self-management text message program that helped them develop an action plan tailored to their own cancer-related issues. The text message program lasted for 28 days starting on the day of enrollment. Between 1 and 5 messages per day were sent (73 messages total).

Control Group

The control group received access to a website, built for this study, that included content from approximately 40 cancer.gov pages relevant to cancer survivors (e.g., coping with cancer, dealing with treatment-related side effects, and survivorship). As the control arm mimicked a brochure, it was not “sticky,” did not include features to keep participants engaged; and did not include any self-management intervention strategies or dynamic content. The website control increased internal validity because participants did not know if they were assigned to the intervention or control group.

Data collection

All data collection took place online. Participants who did not respond to the initial survey invitation were sent reminder emails to complete the survey at 7, 10, and 14 days following initial invitation to the survey. The baseline survey, which took approximately 25 minutes to complete, and follow-up surveys at 1-month and 3-months post-randomization tracked changes in self-efficacy, other secondary outcomes, and obtained feedback on perceived utility of the websites. The same reminder procedures were used for follow-up surveys.

Participants received $10, $20, and $35 e-gift cards for completing the baseline, 1-month, and 3-month surveys, respectively. As shown in the CONSORT diagram (Figure 1), among the 176 participants randomized after a complete baseline survey, 146 (83%) completed the 1-month survey (n=70 intervention; n=76 control) and 135 (77%) completed the 3-month survey (n=65 intervention; n=70 control).

Measures

Self-Efficacy for Managing Cancer

Self-Efficacy for Managing Cancer, the primary outcome of this study, hereafter referred to as self-efficacy, was measured using a modified version of the Stanford Self-efficacy for Managing Chronic Disease Scale¹¹. This 6-item scale uses a 10-point response to assess participants’ confidence for performing certain activities, like managing the long-term physical effects of cancer treatment, managing everyday activities, and asking a doctor about concerns. The self-efficacy score was calculated by averaging the point responses for each item if no more than two items were missing.

Cancer history

Participants self-reported the type, stage, month and year of their first cancer diagnosis. In addition, the treatment status for their most recent cancer was also assessed.

Demographics

Participants self-reported their age, gender, marital status, employment status, education, household income, race, and ethnicity.

Engagement among the intervention group

How often and how long participants interacted with the website was up to the participant. Program usage data among the intervention participants were assessed to explore patterns of use and whether specific patterns (e.g., time spent on site, # of visits, pages visited, functionality used, text messages participants sent in response to text probes) were associated with health outcomes. Engagers were defined as survivors with 10 or more text interactions (defined as number of times each participant responded to prompted question, clicked on the link from a prompted text, responded with a keyword, or clicked link from a keyword text), 10 or more minutes total of web duration, and/or 6 or more web pages viewed throughout the study period that ended at the 3-month survey. Distributions of web and text engagement among intervention participants determined engagement thresholds.

Statistical analysis

Initial group comparisons were conducted on participant characteristics using chi square tests and t-tests, as appropriate (see Table 1), to examine whether randomization was successful. To examine potential bias introduced through selective attrition, participants who completed both the baseline and 3-month online surveys (n=135) were compared with participants who only completed the baseline survey (n=176) on all variables, including baseline self-efficacy. Participants who dropped out were not significantly different than those who completed the 3-month survey on any variables (data not shown).

Table 1.

Demographic, lifestyle, and cancer-related variables across intervention and control groups

	Full Sample N=176
	Intervention (n=88)	Control (n=88)
	n(%)	n(%)	p-value
Race			0.21
Non-Hispanic White	71(80.7)	67(76.1)
Other	15(17.0)	21(23.9)
Missing	2(2.3)	0
Gender			0.52
Male	7(8.0)	9(10.2)
Female	81(92.0)	78(88.6)
Missing	0	1(1.1)
Education			0.75
HS or less	14(15.9)	11(12.5)
Some college	24(27.3)	25(28.4)
College graduate	30(34.1)	27(30.7)
Post-college graduate	20(22.7)	24(27.3)
Missing	0	1(1.1)
Marital status			0.43
Married	56(63.6)	50(56.8)
Not married	32(36.4)	37(42.0)
Missing	0	1(1.1)
Employment			0.75
Full/part time/self-employed	41(46.6)	40(45.5)
Disabled	22(25.0)	18(20.5)
Retired	21(23.9)	23(26.1)
Other	4(4.5)	6(6.8)
Missing	0	1(1.1)
Income			0.68
0 to <40k	27(30.7)	31(35.2)
40 to <100k	38(43.2)	31(35.2)
100k+	20(22.7)	21(23.9)
Missing	3(3.4)	5(5.7)
Smoking status			0.89
Never	54(61.4)	57(64.8)
Former	30(34.1)	27(30.7)
Current	4(4.5)	4(4.5)
BMI			0.19
<25	20(22.7)	32(36.4)
25–29.9	30(34.1)	25(28.4)
30+	37(42)	31(35.2)
Missing	1(1.1)	0
Number of physical comorbidities			0.17
None	31(35.2)	40(45.5)
1	26(29.6)	16(18.2)
2+	31(35.2)	32(36.4)
Cancer type			0.94
Breast	43(48.9)	47(53.4)
Female reproductive	9(10.2)	12(13.6)
Blood/Lymph	12(13.6)	9(10.2)
Colon	2(2.3)	3(3.4)
Head/Neck	4(4.5)	2(2.3)
Lung	4(4.5)	4(4.5)
Prostate	1(1.1)	1(1.1)
Other/Unknown	13(14.8)	10(11.4)
Stage			0.50
Localized	31(35.2)	41(46.6)
Regional	39(44.3)	32(36.4)
Distant	14(15.9)	12(13.6)
Unknown	4(4.5)	3(3.4)
Cancer treatment status			0.81
Completed treatment	64(72.7)	62(70.5)
In-treatment	20 (22.7)	20(22.7)
Other	4(4.6)	6(6.8)
Cancer-related symptoms in past month
Fatigue	84(95.5)	80(90.9)	0.23
Difficulty concentrating	82(93.2)	76(86.4)	0.14
Difficulty sleeping	82(93.2)	84(95.5)	0.51
Feeling nervous/worrying	81(92.1)	79(89.8)	0.60
Sexual problems	66(75.0)	71(80.7)	0.47
Shortness of breath	55(62.5)	58(65.9)	0.64
Feeling sad	76(86.4)	73(83.0)	0.53
Physical pain	75(85.2)	70(79.6)	0.32
Incontinence	39(44.3)	38(43.2)	0.60
Diarrhea/IBS	56(63.6)	50(56.8)	0.36
Lymphedema	39(44.3)	35(39.8)	0.54
Numbness/tingling	63(71.6)	60(68.2)	0.88
	M(SD)
Age	54.9(11.9)	55.0(12.6)	0.98
Age at diagnosis	52.3(12.5)	52.7(12.6)	0.85
Years since diagnosis	2.6(4.1)	1.9(2.2)	0.17

Missing data were handled using multiple imputation with 25 iterations and imputing the self-efficacy score rather than individual scale items based on the small sample size^19,20, resulting in 176 survivors; n=88 intervention; n=88 control. Pooled two-sample t-tests across the multiply-imputed datasets were used for the primary, intent-to-treat analysis to compare differences in self-efficacy change between intervention group and control group participants from baseline to 3-months follow-up.

To further explore individual variation in benefitting from the intervention, two additional a priori planned exploratory analyses were conducted. First, an intent-to-treat pooled paired t-test across the multiply-imputed datasets was used to compare the change in self-efficacy within the intervention group and within the control group participants. Effect sizes were calculated for the between- and within-groups analyses as well as for stratified subanalyses by age (<60 vs. 60+) and cancer treatment status (completed treatment vs. receiving treatment). Second, an engagement subanalysis with 98 survivors was conducted with complete outcome data at 3 months between engagers in the intervention arm (n=30) and the full control group (n=68) to assess the effect of the intervention on self-efficacy of those who actually engaged with the intervention relative to those in the control condition.

Power calculations for this initial RCT study indicated that, with a sample of 200 survivors (100 per arm), we had 80% power to detect an effect size of .15 (small effect) in the primary outcome, self-efficacy, from baseline to 3 months, assuming an estimated 30% attrition rate, which is typical of prior web-based surveys^21,22.

Results

Baseline characteristics of the 176 participants are shown for the intervention and control groups in Table 1. No significant differences were found between those randomly assigned to the intervention or control arm of the study on any sociodemographic or cancer-related variable. Table 1 also shows participants were experiencing significant ongoing cancer-related symptoms and needed management strategies. A significant proportion reported experiencing symptoms including fatigue (93%), difficulty sleeping (94%), pain (82%), and sexual dysfunction (78%).

Primary, intent-to-treat analysis: As shown in Table 2, the intervention group (n=88) had a mean change in self-efficacy of 0.39 (SE=0.03) and the control group had a mean change in self-efficacy of 0.12 (SE=0.04). The 0.28 (SE=0.26) difference of mean changes between groups was not statistically significant in the intent-to-treat between-groups analysis (p =0.29, Cohen’s d = 0.17). Subanalyses showed no significant differences in pre-post changes in self-efficacy between intervention and control participants among subgroups of participants by age or cancer treatment status.

Table 2.

Differences in cancer health self-efficacy between intervention and control groups at 3 months (n=176)

	Intervention	Control	Intervention	Control	Difference	t-value	p-value	Cohen’s d
	N	N	Mean (SE)	Mean (SE)	Mean (SE)
Total	88	88	0.39 (0.03)	0.12 (0.04)	0.28 (0.26)	1.06	0.29	0.17
Baseline age
<60 years	55	51	0.37 (0.04)	0.10 (0.04)	0.26 (0.33)	0.79	0.43	0.18
60+ years	33	37	0.44 (0.05)	0.13 (0.06)	0.31 (0.43)	0.71	0.48	0.19
Treatment status
Completed	64	62	0.44 (0.04)	0.11 (0.04)	0.33 (0.30)	1.11	0.27	0.22
Current	20	20	0.19 (0.07)	0.14 (0.07)	0.04 (0.56)	0.07	0.94	0.03

Note: Participants who reported not having started treatment or not sure were not included in the treatment status stratification (n=5)

Within-groups, intent-to-treat analyses (Table 3): The intervention group had a statistically significantly improved self-efficacy from pre-post intervention, t(88)=2.32, p=0.02, with a small to medium effect size (Cohen’s d=0.26); whereas, the control group had no statistically significant improvement in self-efficacy, t(88)=0.58, p=0.56. Among participants who had completed cancer treatment, those who received the intervention experienced a statistically significant self-efficacy improvement (p=0.02) with a small to medium effect size (Cohen’s d=0.31). This was not found for control participants. There were no differences by age.

Table 3.

Differences in cancer health self-efficacy between baseline and 3-month follow-up separately for intervention and control stratified by age and treatment status (n=176)

	Intervention	Baseline	3-month	Difference	t-value	p-value	Cohen’s d
	N	Mean (SE)	Mean (SE)	Mean (SE)
Total	88	7.20 (0.03)	7.59 (0.04)	0.39 (0.17)	2.32	0.02	0.26
Baseline age
<60 years	55	7.10 (0.04)	7.47 (0.05)	0.37 (0.22)	1.64	0.10	0.25
60+ years	33	7.37 (0.05)	7.81 (0.06)	0.44 (0.29)	1.52	0.13	0.29
Treatment status
Completed	64	7.31 (0.04)	7.76 (0.04)	0.44 (0.19)	2.30	0.02	0.31
Current	20	7.05 (0.07)	7.23 (0.08)	0.19 (0.42)	0.44	0.66	0.11
	Control	Baseline	3-month	Difference
	N	Mean (SE)	Mean (SE)	Mean (SE)	t-value	p-value	Cohen’s d
Total	88	7.16 (0.03)	7.28 (0.04)	0.12 (0.20)	0.58	0.56	0.07
Baseline age
<60 years	51	7.04 (0.05)	7.14 (0.05)	0.10 (0.25)	0.41	0.68	0.06
60+ years	37	7.33 (0.04)	7.46 (0.06)	0.13 (0.32)	0.42	0.67	0.08
Treatment status
Completed	62	7.29 (0.04)	7.40 (0.05)	0.11 (0.23)	0.48	0.63	0.07
Current	20	6.66 (0.07)	6.80 (0.10)	0.14 (0.41)	0.35	0.72	0.09

Engagement analysis among intervention participants: Engagement metrics for the intervention group are available in the supplementary table. Among participants with 3-month outcome data (n=65), 46% (n=30) were categorized as engagers. Seventeen survivors did not participate in the web or text portion, six engaged only in the web portion, and thirty engaged only with the text portion. Twenty-three participants never enrolled in the text portion due to not providing a cell phone number or not having texting within their cellular telephone plan.

Between groups engagement subanalysis (see Table 4a) for those who engaged in the text and/or web intervention (n=30) demonstrated significantly greater self-efficacy improvement compared to the control group (n=68), p=.05 with a medium effect size (Cohen’s d=0.44). Within groups engagement subanalysis (see Table 4b) found participants who completed treatment experienced a significant self-efficacy improvement (p=0.01) with a medium effect size (Cohen’s d=0.53). Control participants did not improve their self-efficacy.

Table 4a.

Differences in cancer health self-efficacy between intervention and control groups at 3 months among text and/or web engaged participants (n=98)

Intervention (n=30)	Control (n=68)	Difference
Mean (SE)	Mean (SE)	Mean (SE)	t-value	p-value	Cohen’s d
0.82 (0.29)	0.10 (0.20)	0.72 (0.36)	2.01	0.05	0.44

Table 4b.

Differences in cancer health self-efficacy between baseline and 3-month follow up among text and/or web engaged participants within intervention arm and control arm (n=98)

Intervention (n=30)
Baseline	3-month	Difference
Mean (SE)	Mean (SE)	Mean (SE)	t-value	p-value	Cohen’s d
7.10 (0.27)	7.93 (0.29)	0.82 (0.29)	2.89	0.01	0.53
Control (n=68)
Baseline	3-month	Difference
Mean (SE)	Mean (SE)	Mean (SE)	t-value	p-value	Cohen’s d
7.36 (0.19)	7.47 (0.24)	0.10 (0.20)	0.52	0.61	0.07

Discussion

After conducting usability testing and preliminary evaluation of SBC¹⁶, ACS and NCI developed an enhanced version of SBC, with expanded web offerings and a new text component. ACS assessed intervention effectiveness on self-efficacy for managing cancer. The primary trial analysis indicated a non-statistically significant between groups effect for the intervention compared to the control group, potentially due to low power. The power analysis for the trial indicated adequate power based on a sample of 200 survivors, but the intent-to-treat sample was only 176 participants due to recruitment challenges during the funding window, a common issue for eHealth interventions with cancer survivors^23,24. Self-efficacy change was potentially significant if the study had a larger sample size; the within group analysis demonstrated significant improvement in the intervention group with an effect size of 0.26 while the control group did not show significant self-efficacy improvement.

A surprising and notable trial finding was lower engagement in the intervention than anticipated. Qualitative evaluation of the initial version with cancer survivors¹⁶ suggested the eHealth tool would be well-received and highly used. However, survivors in the initial study also suggested that eHealth tools are more helpful for survivors who are closer to active treatment.¹⁶ In this study we found approximately half of participants never logged into the web intervention and overall usage was relatively low, challenges similar to other eHealth studies^{21–23,25,26}. Those in the intervention group were further away from active treatment than controls (2.6 vs 1.9 years out). Further, 34% only engaged in the text intervention. Limited logins to the web intervention might have been driven by a combination of perceived limited need, confusion over how to access the tools through the provided unique link, and by incentivizing responses to the surveys rather than accessing and using the intervention.

An engagement subanalysis was conducted comparing only those who were engaged in one or both components of the intervention compared to the full control group. For engaged survivors, the intervention had a significant, between groups effect with a moderate effect size (d=0.44). This engagement subanalysis suggests that using the intervention as designed leads to improved self-efficacy. These findings are similar to another web-based study focused on cancer-related fatigue where more engaged participants showed greater improvement in self-efficacy¹⁴. This points to the need to figure out how to make engagement with eHealth tools for long-term survivors relevant as well as appealing outside of a research context.

Increasing engagement within this and other eHealth tools is vital. A prior systematic review²⁷ suggests that targeting those with pressing health issues enhances engagement with Internet-based interventions. Future work in this area may benefit from screening participants and recruiting only those with a high level of need for the intervention. Research suggests that building in push notification via smartphone or other mobile application²⁸, as well as more opportunities for social interaction within eHealth interventions, such as through online peer-to-peer communities^29,30, may enhance engagement and limit attrition. During the usability testing and qualitative evaluation of SBC’s beta version, participants indicated interest in more peer-to-peer interactions but it was out of budget for the enhanced SBC intervention. Future versions of SBC should incorporate more opportunities for social interaction based on this and prior study findings. Further, what survivors are willing to spend their own money and time on compared to what they will do in a compensated research setting is a critical implementation issue for eHealth programs. A prior study suggests that the intervention being “too intensive” was one reason participants dropped out of a web-based mindfulness intervention³¹. Having the intervention pair participants with a preferred type and intensity of digital consumption, such as personalizing the number of daily text messages or building in the option of adapting the level of text messages received over time, could likely lead to higher engagement and lower rates of opting out. These logistical issues with the web and text message components of the intervention likely dampened the effect of the intervention and required a larger sample in order to measure its full effect. Additional potential solutions to enhance engagement in eHealth tools outside of the research context include gamification of tools (badges, stars, upgrades, etc.), competition with other users, and peer-to-peer or provider contact via message boards^32–34. Formal evaluation of these methods are still needed.

This study provided initial, preliminary evidence of improvements in self-efficacy for those who engaged in the SBC web and text message intervention, especially for post-treatment survivors. Strengths of the trial include randomization and intent-to-treat analyses with multiple imputation to manage the impact of missing data at 3 months. Several limitations exist. The recruitment strategy led to a mostly educated, non-Hispanic white, female sample with half having a history of breast cancer. Therefore, results from the trial are limited in generalizability to broader survivor populations and emphasize the importance of broadening strategies to recruit more diverse samples, in terms of gender, cancer type, race, ethnicity, and education, in the future. In addition, the denominator of the sampling frame is unknown since it was not possible to know how many emails were delivered to a valid email address or opened by potential participants. Future studies should implement strategies to validate email addresses and verify receipt. Further, future studies may benefit by supplementing recruitment strategies with paid diverse survivor panels or Facebook targeted ads. Further, future studies may need to investigate if there are differences in language that might make survivors more or less interested in participating. Recruitment for the trial was lower than expected and engagement in the intervention was lower than anticipated. Based on the relatively high number of participants who never logged into the web component, for the control and the intervention arms, the instructions for how to log in were likely not adequate. The trial was also not powered to fully compare intervention effects by treatment status and future research is needed to examine efficacy for those in treatment, for those transitioning out of treatment, and for those post-treatment. Despite recruitment and engagement challenges, the rate of attrition at 23% was slightly lower than the anticipated rate (30%) for a web-based intervention³⁵. Additional enhancements to SBC are also needed to facilitate better engagement, including greater personalization of intervention intensity, ability to track change over time, more push notifications to bring survivors back to the online tools, and more opportunities for social interaction.

The results from this and prior eHealth RCTs suggest that eHealth tools can work for those who engage as designed but also do not work for everyone, particularly those who do not engage as designed. Future research must determine who benefits from specific components of eHealth interventions, such as web, app, or text, and who does not to facilitate the targeting and tailoring of eHealth to those most likely to benefit and identify individuals who need alternative interventions that are higher touch or in person³⁶. Future research also must identify methods to increase engagement with our eHealth interventions to maximize the benefit among those for whom eHealth is appropriate.