Participants were recruited from a sample (N=415) of breast, prostate, and colorectal cancer survivors living in Nova Scotia who had previously taken part in a survey study and had indicated an interest in future studies [11]. The sample was contacted via email, mail, or telephone with an invitation to participate that included an information sheet from the investigators explaining the purpose of the study and instructions on how to proceed if interested, a consent form, and a copy of the primary publication from the previous survey. Eligibility criteria were (1) being able to speak and read English, (2) having access to the Internet, and (3) being able and interested in an Internet-delivered program designed to increase weekly PA levels.

This study was a pilot 2-group randomized controlled trial to compare a usual care group (no intervention) with an Internet-delivered behavior change group. The focus of the behavior change program specifically was to increase PA in the form of steps or minutes. Eligible participants provided informed consent and completed a baseline questionnaire to gather demographic, behavioral, and PA information prior to randomization.

A rolling blocked randomization was completed after baseline measures were collected to ensure participants did not have an overly long wait to start the intervention. Participants were randomly allocated to one of two groups using a computer generated random numbers list. The 2 groups were the control group—also called usual care (UC)—and the intervention group (UCAN), which consisted of membership in a private online community called Active Nova Scotia housed on the PA tracking website UWALK [31] and modified for cancer survivors. Group assignments were generated by a research assistant and assigned after blocks of baseline measures were received to eliminate bias in group allocation. Participants were then notified of their group assignment via email.

Those randomized into the UCAN group were given access to a 9-module behavior change program which was developed using previous print materials as a template [13,32]. The modules were published sequentially on the site as the intervention progressed to increase retention. Information module topics were developed from survey results of the same group [11,27] and were as follows: (1) welcome, general information about the site, types of exercise, and how to gauge intensity; (2) exercise myths, dispelling common exercise myths; (3) exercise safety, tips on how to exercise smart and safe; (4) goals and planning, how to plan and make SMART goals; (5) exercise benefits, specific benefits of exercise for cancer survivors; (6) make it fun, tips on how to keep exercise fun; (7) exercise barriers, tips on how to overcome the most common barriers identified; (8) support network, how others can help you exercise; and (9) relapse, strategies on how to avoid and deal with relapse. Each module remained available to review after the week was concluded. In addition, each module included a video relevant to the current topic featuring the first author to foster a connection and simulate face-to-face interactions.

Aside from the behavior change program, the UCAN group was able to use the UWALK website to track their PA in steps, moderate or vigorous minutes, and flights of stairs. Participants were able to see the progress of other group members as well as their own progress over time. Participants in the UCAN group also received weekly email updates informing them of new information posts as well as a brief summary of their previous weeks PA levels. Emails were developed to offer encouragement to those who were not meeting the guidelines and congratulate those who were sufficiently active. Upon being informed of their group assignment, the UC group was asked to keep their regular exercise routine over the intervention period and they would receive access to the website and the behavior change program once the follow-up questionnaire was completed.

All analyses were performed using PASW Statistics 22 (PASW Inc., Chicago, IL, USA). Feasibility was assessed using recruitment rate, website satisfaction, and usage statistics gathered from UWALK and Mixpanel. Chi-square and analyses of variance (ANOVAs) were performed to determine the differences between the intervention groups for PA behavior and QOL. Analyses of covariance (ANCOVAs) were also conducted to adjust for baseline value when comparing intervention groups. Results were interpreted for statistical trends as well as for potential clinical significance. Using a two-tailed alpha of P≤.05, the study had 80% power to detect medium standardized effects (d=0.50) after adjustment for covariates with 45 participants per condition. Trends were defined as P<.10 and potential clinical significance as a standardized effect size of d≥0.33 [41]. Intention-to-treat protocol was adhered to for all analyses. Responders and nonresponders were compared to determine any differences. Based on the higher than expected number of participants meeting PA guidelines at baseline, subgroup analyses were conducted for those with less than 150 minutes versus 150 minutes or more of total exercise.

Detailed weekly Web statistics are shown in Figures 2-5. The overall average number of logins was 10.3 for the 9-week duration of the intervention. There were 2293 individual PA events logged over 1085 days (average 23 days per participant) and 4319 page views recorded. The most frequently visited page was the log page where participants entered their PA data. The modules were visited 213 times over the length of the study with an overall read rate of 26%. Moreover, 94% (45/48) of participants logged in at least once, 85% (41/48) recorded PA at least once, and 67% (32/48) viewed the modules at least once.

With regard to the intervention program, 73% (30/41) said they enjoyed the Active Nova Scotia program, 63% (26/41) would be willing to continue participating, 46% (19/41) indicated they increased their PA because of this program, 71% (29/41) said they were more aware of the amount of PA they get each day, and 73% (30/41) thought the information in the weekly modules was useful and relevant. About half of the participants (51%, 21/41) liked the video posts and felt they were not too burdensome on their computer. When evaluating the website, 68% (28/41) were able to easily navigate and enter PA information on the site. When asked if they would recommend the site to others, 64% (26/41) indicated yes and 39% (16/41) said they would continue using the site after the study had finished.

The differences in PA behavior between the UC and UCAN groups at baseline and postintervention are described in Table 2. Overall, the adjusted between-group mean change scores favored the UCAN group; however, there were no significant differences between the groups in any PA measure. The adjusted between-group difference for total exercise minutes was 42 (95% CI -65 to 150; P=.44, d=0.17) in favor of UCAN. The adjusted between-group difference for strength training frequency achieved a meaningful difference of 0.5 (95% CI -0.2 to 1.1; P=.14, d=0.34). The subgroup analysis of the changes in total exercise minutes by baseline PA levels is illustrated in Figure 6. Among those not meeting guidelines at baseline (54%, 51/95), the UCAN group (54%, 26/48) increased their PA levels by 52 minutes (95% CI -74 to 178) while the UC group (53%, 25/47) decreased by 15 minutes (95% CI -140 to 109); whereas among those meeting guidelines, the UCAN (46%, 22/48) and UC group (47%, 22/47) increased PA by 88 (95% CI -55 to 230) and 65 minutes (95% CI -57 to 186), respectively.

The general and cancer-specific QOL measures at baseline and postintervention are summarized in Tables 3 and 4. Change in the SF-36 measure of mental health favored the UC group with a mean change score of -2.9 (95% CI -5.1 to -0.6; P=.014, d=0.37). All other measures were nonsignificant. No measures met the point difference that indicates clinical significance (SF-36=3 point difference; FACT subscales=2-3 point difference).

Our study is one of the first to use an online platform to deliver a theory-based PA behavior change program to cancer survivors, and the first to target Nova Scotian cancer survivors. Based on recruitment, retention, and participant evaluation of the program we believe that the program is feasible. In addition, there were trends suggesting the potential effectiveness of the program for promoting PA, especially in cancer survivors who were inactive at baseline. Engagement in the program and influencing QOL, however, remain a challenge for distance-based program delivery.

Our expression of interest rate (29%, 120/415) and recruitment rate (23%, 95/415) was similar to other studies [22,28]. The previous study among cancer survivors resulted in a 17% recruitment rate but used community- and clinical-based recruitment methods that were unable to track the initial reach of the invitation [28]. Our postintervention retention (88%, 84/95) was higher than the majority of previous studies using the Internet as a delivery method [19,28,42]. Large attrition is common among Internet-based interventions [18,20-22,42] and like previous research we had slightly higher attrition in the intervention group (15% vs 9%) despite the high satisfaction ratings [22]. It is difficult to pinpoint the reason for such high dropout rates in Web-based studies but previous research indicates it is easier for participants to disengage from Web-based interventions [43]. Using strategies to increase the contact between user-to-user and user-to-researcher may help increase the connection and make the intervention meaningful to the participant [23].

Engagement in our study was fairly low compared to other Internet-based studies [22,28]. The modules had a completion rate of 26% (111/432 potential completions). As with logins, the number of completed modules dropped after the first few weeks. Our average number of logins was 10.3 per person. This equals about once per week per person which may be insufficient to induce PA behavior change. This is similar to other studies using Internet delivery [24,42]. A meta-analysis by Davies et al [19] found the average number of logins per-person-per-week was 3.08 across 11 studies. One potential reason for our lower login average is that the website was able to automatically pull data from devices such as the FitBit without the participants having to login. One recent suggestion for increasing user engagement is to allow user-generated content (eg, creating a post to add to the newsfeed) [44] which may increase user “buy-in.” This method, however, requires close monitoring as information would need to be vetted to ensure accuracy and relevance. Retaining and engaging participants remain an issue among Internet-delivered behavior change programs.

Overall, the program was very well received among participants in the UCAN group despite the low usage numbers. This is similar to other Internet-based PA programs [22,28,42,45]. Most participants felt that the information provided was useful and relevant and they indicated that they were more aware of their level of daily activity. They also indicated they liked the weekly posts and videos and would be interested in continuing with the ANS program. Participants evaluated the website favorably and said they would recommend it to a friend but the majority indicated they would not continue using the site with the program finished. Very few participants contacted the study coordinator with issues related to using the website.

Engagement seems to be the biggest hurdle in testing and implementing Internet-delivered interventions. Vandelanotte et al [23] evaluated freely accessible websites that promote PA and found that many did not use tools such as self-monitoring, goal setting, and targeted feedback despite the supporting evidence [21,24,25,46]. An aspect found to be useful that our study lacked is a method of users generating their own content. Despite having a “news feed,” our users were not able to directly message other participants which has been shown to increase effectiveness of Web-based interventions [25]. Standardizing the components of behavior change websites and thoroughly testing them will allow researchers to determine which are most effective among various populations.

As expected based on the small sample size of this pilot study, there were no significant between-group differences in any PA measure, one component we used to determine efficacy. Nevertheless, after adjusting for baseline measures, the UCAN group increased total exercise by 42 minutes more than UC (29 aerobic minutes plus 12 strength minutes) which translated into a small standardized effect size of d=0.17. This is slightly higher than the overall effect size of d=0.12 found by Davies and colleagues [19]. Moreover, the largest effect of the intervention was for strength training frequency where the UCAN group added a half day per week compared to the UC group (d=0.34). Despite the majority of PA measures showing nonsignificant increases favoring the UCAN group, the percentage of participants meeting guidelines, based on the standard cutpoint of 150 minutes, showed a nonsignificant potentially meaningful change favoring the UC group. This finding is somewhat arbitrary because it is dependent on the baseline level of PA (ie, where participants start). There were more UC participants in the “insufficiently active” category than UCAN at baseline; consequently, the smaller increase in minutes per week may have been enough to result in a larger proportion meeting guidelines.

The previous research among cancer survivors [28] and the meta-analysis by Davies and colleagues [19] found that computer-tailored PA programs had positive effects on PA. Previous reviews also indicate that Internet-delivered interventions have positive effects on PA levels [47-49]. One possible explanation for the modest effect of our intervention is the relatively high percentage of participants meeting PA guidelines at baseline (46%, 44/95). Our invitation was to any cancer survivor who wanted to increase his/her PA with the assumption that only less active people would volunteer for such a study. Moreover, we included those meeting the guidelines because research has shown that even more health benefits can be gained by increasing activity levels to 300 or more minutes per week [36,37].

After performing an exploratory subgroup analysis we found a suggestion that the program may be more effective for those who were not meeting guidelines at baseline. Among those not meeting guidelines at baseline, the UCAN group increased their PA levels by 52 minutes while the UC group decreased by 15 minutes (Figure 3). Among those meeting guidelines, the UCAN and UC groups increased PA by 88 and 65 minutes, respectively. The suggestion that PA behavior change programs are most beneficial to those least active is similar to previous research [19]. Targeting specific populations that have lower than average PA levels (ie, cancer survivors, inactive population) may have an even larger effect on clinical and public health outcomes [19].

Not surprisingly, our study did not find any beneficial changes in QOL measures, the second component used to determine efficacy. In fact, the only significant finding was a negative effect on mental health (P=.014, d=0.37). It is common to find no significant benefits to QOL among distance-based PA interventions for cancer survivors even when PA increases are noted [15,50]. Similar to the PA measures, some studies have found significant improvements in aspects of QOL at postintervention that were not sustained when assessed at follow-up [32,51-57]. Over the course of the study intervention, 14 (29%, 14/48) intervention participants contacted the study coordinator indicating they were having physical or personal issues, which may be a possible explanation for the negative trend in QOL evident in this study. Based on qualitative comments left by participants at the postintervention survey, many felt that the QOL measures used did not apply to them as it had been so long since diagnosis. Approximately 85% (81/95) of the study sample was over 5 years since diagnosis. It may be that the measures used to assess QOL are more applicable to patients on treatments. Despite our inclusion of the generic SF-36, it may be beneficial to include long-term cancer-specific QOL measures for studies among long-term cancer survivors to see if they would be more applicable.

This is the first study to deliver a computer-tailored, Web-delivered PA behavior change intervention to Nova Scotian cancer survivors, and one of the first in any cancer survivor group. This study showed that some cancer survivors are interested and willing to receive PA information through the Internet; however, modifications to the website are necessary to optimize the effectiveness. Limitations of this study are the use of self-report data, selection bias toward those more motivated and Internet savvy, the low usage rate overall, and the decline in usage over the intervention period. Despite the user-friendly website we used to pilot this program, there were comments about confusion on how to use the site and find our information. In the future, we would recommend ensuring a separate site that would be able to house the information in a more prominent position.

Our original study [11] invited people to participate in a PA survey, leading to a selection bias for those motivated to engage in PA. It is common in nonblinded studies to have self-selection bias among participants. Despite this, we were still surprised at the number of participants in the intervention meeting guidelines (46%, 44/95). It may be that the most motivated and active of the previous highly motivated and active survey sample were the ones to come forward for this intervention. In addition, our participants were more likely to prefer receiving information via the Internet which may also bias results. When being asked to participate in an online study, those who prefer this method are more likely to come forward. However, if we were to exclude those already active, our sample size would have been reduced by almost half. Previous preference research found that those who preferred Web-based interventions were more likely to have higher Internet use and higher PA participation [58]. More research into preferences for Internet delivery PA interventions should be explored.

In conclusion, using a Web-based platform to deliver a PA behavior change intervention to cancer survivors may be a feasible alternative to other methods of information delivery. There was a trend toward increased activity in the UCAN group when compared to the UC group, especially among inactive cancer survivors, although no significant differences were found. User engagement remains a challenge and future research should incorporate as many of the tools previously found to be effective among Web-based interventions to increase engagement and maintain PA behavior.