Web-Based Stress Management for Newly Diagnosed Patients With Cancer (STREAM): A Randomized, Wait-List Controlled Intervention Study.

Purpose Being diagnosed with cancer causes major psychological distress; however, a majority of patients lack psychological support during this critical period. Internet interventions help patients overcome many barriers to seeking face-to-face support and may thus close this gap. We assessed feasibility and efficacy of Web-based stress management (STREAM [Stress-Aktiv-Mindern]) for newly diagnosed patients with cancer. Patients and Methods In a randomized controlled trial, patients with cancer who had started first-line treatment within the previous 12 weeks were randomly assigned to a therapist-guided Web-based intervention or a wait-list (control), stratified according to distress level (≥ 5 v < 5 on scale of 0 to 10). Primary efficacy end point was quality of life after the intervention (Functional Assessment of Chronic Illness Therapy-Fatigue). Secondary end points included distress (Distress Thermometer) and anxiety or depression (Hospital Anxiety and Depression Scale). Treatment effect was assessed with analyses of covariance, adjusted for baseline distress. Results A total of 222 of 229 screened patients applied online for participation. Between September 2014 and November 2016, 129 newly diagnosed patients with cancer, including 92 women treated for breast cancer, were randomly assigned to the intervention (n = 65) or control (n = 64) group. Adherence was good, with 80.0% of patients using ≥ six of eight modules. Psychologists spent 13.3 minutes per week (interquartile range, 9.5-17.9 minutes per week) per patient for online guidance. After the intervention, quality of life was significantly higher (Functional Assessment of Chronic Illness Therapy-Fatigue: mean, 8.59 points; 95% CI, 2.45 to 14.73 points; P = .007) and distress significantly lower (Distress Thermometer: mean, -0.85; 95% CI, -1.60 to -0.10; P = .03) in the intervention group as compared with the control. Changes in anxiety or depression were not significant in the intention-to-treat population (Hospital Anxiety and Depression Scale: mean, -1.28; 95% CI, -3.02 to 0.45; P = .15). Quality of life increased in the control group with the delayed intervention. Conclusion The Web-based stress management program STREAM is feasible and effective in improving quality of life.

RCT Entities:   Population Interventions Outcomes

INTRODUCTION

Diagnosis of cancer elicits high levels of distress in a majority of patients,[1] which is associated with decreased quality of life as well as diminished treatment tolerance[2,3] and potentially worse disease course.[4,5] Psychosocial support for patients with cancer is effective in alleviating distress and improving quality of life, including fatigue, the most common complaint of patients with cancer.[2,6] However, a majority of newly diagnosed patients with cancer lack psychosocial support because of constraints on the part of both providers and patients.[6,7] Use of the Internet, which has become an integral part of our lives, has the potential to change this. At least 70% of patients with cancer use the Internet as a source of information shortly after diagnosis,[8] making it a powerful platform for reaching these patients. Recent approaches to integrating the Internet into patient care range from patient forums to information sites and even therapeutic games.[9,10] Internet programs based on cognitive behavioral techniques with patient guidance via regular online contact with a health care professional (ie, therapist-guided programs or guided self-help) have emerged as particularly effective options. For a range of psychological disorders, including anxiety disorders and depression in those without cancer,[11,12] therapist-guided online interventions seem similarly effective as face-to-face interventions.[13] The success of Web-based guided self-help in psychological disorders[9,13] coupled with the need to further improve access to psychosocial support for patients with cancer, especially outside of inner cities with large cancer centers,[6] has boosted interest in online interventions in oncology. Numerous piloted and ongoing trials in patients with cancer seek to define suitable indications, formats, and settings.[14] The few larger published randomized controlled trials[15-17] show encouraging results, with improvement in a number of relevant psychosocial domains, including coping with cancer,[15] sexual functioning,[16] and distress[17] in breast cancer survivors.

We designed the STREAM (Stress-Aktiv-Mindern) intervention specifically for the particularly vulnerable period immediately after first diagnosis of cancer.[18] The rationale behind this early intervention was three-fold. First, distress in patients with cancer peaks shortly after diagnosis,[18] irrespective of cancer type. Second, the time after diagnosis is busy with appointments for diagnostics and treatment. Therefore, the self-management of time and location allowed by Web-based interventions[9] might be of particular value. Third, successful early psychosocial interventions have shown potential to affect disease course beyond psychosocial outcomes.[19] We assessed feasibility and efficacy of our therapist-guided Web-based stress management program STREAM for newly diagnosed patients with cancer receiving first-line treatment.

PATIENTS AND METHODS

Details are provided in the Appendix (online only) and the published protocol.[20] We included adult patients (age ≥ 18 years) with newly diagnosed cancer who started first-line treatment (either systemic treatment, including chemotherapy, hormonal treatment, or targeted therapy, or radiotherapy) no longer than 12 weeks before study registration. Patients were required to provide written informed consent, read and write in German, and have Internet access as well as basic computer skills. The ethics committee approved the study (EKNZ339/13). Patients were recruited online via the STREAM Web site of STREAM. We randomly assigned eligible patients at a ratio of one to one using blocked randomization with randomly selected block sizes to an intervention group or a wait-list control group (Fig 1). Patients were stratified according to baseline distress using an internationally accepted cutoff of ≥ 5 points on the 10-point visual analog scale (VAS) of the Distress Thermometer (DT).[21]

Fig 1.

Trial design. DT, Distress Thermometer; FACIT-F, Functional Assessment of Chronic Illness Therapy–Fatigue; STREAM, Stress-Aktiv-Mindern.

Intervention

We developed the Web-based intervention STREAM[20] based on established stress management intervention manuals[22] that incorporate cognitive behavioral– and mindfulness-based stress reduction techniques, which we adapted to the Web context. STREAM consists of eight modules (Appendix Table A1, online only), which can be completed in 60 to 90 minutes each. Daily use of downloadable audio files with relaxation and guided-imagery exercises was encouraged. Participants were asked to complete one module per week. Our therapists provided weekly written feedback via integrated secured e-mail.

Table A1.

Content of Web-Based Stress Management Program STREAM

Patients in the control group underwent their cancer treatment locally as planned and were recontacted by the study team 8 weeks after random assignment (T2; Fig 1). After T2 assessments, they received access to the online program. For patients in both groups, cancer treatment was determined locally, and supportive care according to local standards may also have included face-to-face psychosocial support and psychotropic drugs.

Assessments

Assessments were conducted electronically directly within the Web-based program via the open source application LimeSurvey at baseline (T1) and after the intervention or waiting period (control group), respectively (T2). In addition, 2-month follow-up (T3) was performed in both groups.

Efficacy End Points

Primary end point was quality of life at T2, assessed using the validated German version of the Functional Assessment of Chronic Illness Therapy–Fatigue (FACIT-F) questionnaire.[23] Minimal clinically meaningful differences are not well defined but have previously been set between 7 and 9 points, both as intraindividual changes and differences in groups.[19,24]

Secondary efficacy end points were assessed at the same points in time and evaluated psychological distress and anxiety or depression using the validated German versions of the National Comprehensive Cancer Network DT[21] and the Hospital Anxiety and Depression Scale (HADS),[25] respectively. Effect sizes are expressed as partial eta squared (η2p),[26] with the following cutoffs to categorize effect sizes into small (0.01), medium (0.06), and large (0.14), as suggested by Cohen.[27]

Assessments During Intervention

Usability was evaluated after the first and last module with the System Usability Scale; scores > 70 represent good usability.[28] Therapeutic alliance between patients and the online therapist was assessed using the Working Alliance Inventory in its short form (12 items)[29] after each module. Total score ranges from 0 to 5, and scores > 3.5 have been rated as good working alliances.[30]

Statistical Analyses and Sample Size Calculation

All analyses were performed in the intention-to-treat (ITT) population defined as all patients who were randomly assigned. The per-protocol (PP) population included all patients who underwent the program in the intended timeframe (ie, the time between random assignment and T2 assessments did not exceed 16 weeks, which is twice the minimal duration of the program). To demonstrate a 9-point difference[31] in FACIT-F total score between baseline and T2 (after 8 weeks) in the intervention group with a statistical power of 0.80 at a significance level of .05 (two sided), 60 participants were needed in each of the two conditions.

Efficacy Analyses

Efficacy outcomes were modeled with analysis of covariance (ANCOVA), using postscore (T2) as the dependent variable, prescore (T1) as the covariate, and group allocation (intervention v control) as the independent variable. ANCOVAs were further adjusted for the stratification factor distress (DT ≥ 5 v < 5). For the follow-up period, score changes from T2 to T3 were analyzed with paired t tests, separately for each group (no between-group comparisons). Multiple imputations (n = 99) by chained equations[32] using predictive mean matching[33] incorporating all variables of the linear models underlying ANCOVA were used to impute missing outcome values.[34] To assess the robustness of the results, sensitivity analyses were conducted for all outcomes in the PP population. In addition, sensitivity analyses were carried out using other methods for handling missing data; more specifically, complete-case analyses and last observation carried forward analyses, as specified in the protocol,[20] were computed for all outcomes in both the ITT and PP populations.

RESULTS

We screened 229 patients, of whom 129 were randomly assigned between September 11, 2014, and November 24, 2016 (Fig 2). All patients received first-line cancer treatment, which they started a median of 17 days (interquartile range [IQR], 6-22 days) and 14 days (IQR, 7-20 days) after signing informed consent in the intervention and control groups, respectively. Patients were residents of Switzerland (n = 64), Germany (n = 59), Austria (n = 5), and the United Kingdom (n = 1). Medical, psychological, and socioeconomic baseline characteristics are listed in Table 1 and were balanced between the groups. All 21 patients (control group, n = 10; intervention group, n = 11) who scored 1 point in the Beck Depression Inventory suicide item at baseline were immediately contacted by telephone, but they clearly distanced themselves from acute suicidal intent.

Fig 2.

Patient flow (CONSORT diagram). ITT, intention to treat; RCT, randomized controlled trial.

Table 1.

Baseline Demographic and Clinical Characteristics

The intervention was designed to be feasible within 8 weeks. However, median duration of the online intervention (between first login to module one and postintervention assessment at T2) was 11.7 weeks (IQR, 9.1-18.6 weeks). In the intervention group, 52 patients (80.0%) used at least six modules, and 49 (75.4%) worked with all eight modules. Our psychologists spent a median time of 165 minutes (IQR, 127-210 minutes) for administering the online intervention (ie, 13.3 minutes [IQR, 9.5-17.9 minutes] per patient each week). Usability of the program was rated high, with a mean System Usability Scale score of 87.5 (IQR, 81.2-95.0) after module one and of 90.0 (IQR, 82.5-95.0) after module eight. As a measure of the therapeutic relationship between patient and online therapist, patients reported a mean score in the Working Alliance Inventory questionnaire of 3.77 (IQR, 3.38-4.14), similar to that of previously reported online working alliances.[30]

Primary and secondary efficacy outcomes are listed in Table 2 and illustrated in Figures 3 and 4. Quality of life (FACIT-F) after the intervention (T2; the primary end point) was significantly higher in the intervention group as compared with the control group (ANCOVA P = .007; Table 2). With a mean increase in total FACIT-F score of 8.59 (95% CI, 2.45 to 14.73; P = .007) in the ITT population and of 10.71 (95% CI, 4.49 to 16.94; P = .001) in the PP population, changes were clinically meaningful.[19,24] Effect sizes were medium[27] (η2p = 0.063 and 0.114 in the ITT and PP populations, respectively; Table 2). Increased scores within the fatigue (4.51; 95% CI, 1.81 to 7.22; P = .002), physical well-being (2.01; 95% CI, 0.43 to 3.59; P = .01), and functional well-being subscales (1.53; 95% CI, 0.11 to 2.95; P = .04) were major contributors to the increase in total FACIT-F score, whereas social well-being and emotional well-being scores were not (Table 2.).

Table 2.

Efficacy Outcomes

Fig 3.

Treatment effects. Mean changes in scores (95% CIs) postintervention (T2) for (A) quality of life (Functional Assessment of Chronic Illness Therapy–Fatigue), (B) distress (Distress Thermometer), and (C) anxiety/depression (Hospital Anxiety and Depression Scale) for the intervention and control groups and their differences (treatment effects) based on analyses of covariance with prescores (T1) as covariates.

Fig 4.

Percentage of patients with any increase or decrease or no change in total scores between T1 and T2 in (A) quality of life (Functional Assessment of Chronic Illness Therapy–Fatigue), (B) distress (Distress Thermometer), and (C) anxiety/depression (Hospital Anxiety and Depression Scale), where blue represents amelioration and gray deterioration within the respective assessment tool.

Distress on the VAS (scored from 0 to 10) of the National Comprehensive Cancer Network DT was significantly lower at T2 in the intervention group as compared with the control (−0.85; 95% CI, −1.60 to −0.10; P =.03). As summarized in Table 2, anxiety and depression (HADS) after the intervention (T2) were not significantly lower in the intervention as compared with the control group (P = .15) in the ITT population. However, decrease in HADS score was statistically significant in the PP population (−2.09; 95% CI, −4.03 to −0.16; P = .03). All results were confirmed in the prespecified sensitivity analyses (Appendix Table A2). Figure 4 shows the percentage of patients who reported any changes in scores between baseline and T2 for all three assessment tools.

Table A2.

Sensitivity Analyses for Efficacy Outcomes

During the 2-month follow-up period of the intervention group (T2 to T3), quality of life (FACIT-F T2 to T3: mean, 4.69; 95% CI, −0.74 to 10.12; P = .09), distress (DT T2 to T3: mean, −0.29; 95% CI, −1.03 to 0.44; P = .4), and mood (HADS T2 to T3: mean, −0.82; 95% CI, −2.28 to 0.65; P = .27) did not change significantly. In the follow-up phase (ie, after T2), 51 (79.7%) of 64 patients randomly assigned to the control arm opted to start the STREAM program. For this group of patients, T2 represents the assessments immediately before and T3 the assessments immediately after the online program. In an ITT analysis (n = 64), quality of life increased significantly (FACIT-F T2 to T3: mean, 10.95; 95% CI, 6.18 to 15.71; P < .001) and distress decreased significantly (DT T2 to T3: mean, −1.25; 95% CI, −1.95 to −0.55; P = .001) between T2 and T3. Self-reported anxiety and depression were also lower (HADS T2 to T3: mean, −2.83; 95% CI, −4.29 to −1.36; P < .001). Again, results were confirmed in the prespecified sensitivity analyses (Appendix Table A3, online only). Data for individual patients and group means are shown in Appendix Figure A1 (online only).

Table A3.

Sensitivity Analyses for Follow-Up Assessments

Fig A1.

Individual patients’ scores and group means for all time points. DT, Distress Thermometer; HADS, Hospital Anxiety and Depression Scale; IQR, interquartile range; FACIT-F, Functional Assessment of Chronic Illness Therapy–Fatigue.

DISCUSSION

In this randomized controlled trial, newly diagnosed patients with cancer reported significantly better quality of life and lower distress on the DT after participating in the therapist-guided Web-based stress management program STREAM. Recruitment to the STREAM study via online channels was successful, and patients in three countries, corresponding to a geographic area twice as large as the United Kingdom, were reached. Thus, dissemination of psychosocial interventions beyond urban centers, where face-to-face psychosocial interventions are available,[35] can be facilitated by a Web-based approach. The STREAM intervention was feasible in our population of patients during a period of active treatment of different types of cancer with good adherence; 80% of patients worked with at least six of the eight modules.[16]

Although it is indisputable that quality of life matters, it is also inherently difficult to measure.[36] To ensure robust and clinically relevant data, we rely on well-validated and standardized questionnaires.[1,23,25] There is no clear cutoff for clinically meaningful increases in overall quality of life in the FACIT-F score. However, on the basis of previous studies, changes reported in the postintervention scores of this trial were in a range that is considered highly noticeable to patients.[19,24] Quality-of-life analyses are often complicated by a large number of missing data. In our study, the low number of missing data (90.7% of all randomly assigned patients completed the primary assessment at T2) and robustness of the sensitivity analyses (Appendix Tables A2 and A3), increase reliability of patient-reported outcomes.

Although the primary efficacy end point of better quality of life after the STREAM intervention was clearly met, the effect of the intervention on distress is less clear cut. The DT is an assessment tool that allows patients to summarize all subjective aspects of distress in a single number (VAS, 0 to 10). In its simplicity, the DT therefore has the advantage of covering various dimensions of distress, including physical, functional, social, socioeconomic, spiritual, and emotional distress.[21] However, the weight that patients assign, whether consciously or not, to each dimension is not discernible from the DT score. In contrast, the HADS questionnaire covers exclusively the emotional dimension of distress, but it does so in greater depth.[25] Whereas self-reported distress on the DT was lower after STREAM, with a small to medium effect size[27] (η2p = 0.043 and 0.069 in the ITT and PP populations, respectively), emotional distress as assessed by HADS did not change. This leads to the hypothesis that STREAM primarily affects dimensions of distress other than anxiety or depression. Of note, in our population, HADS scores at baseline were rather low (mean, 12; IQR, 7-17), whereas baseline DT scores were high (mean, 6; IQR, 5-8). It is therefore conceivable that a potential impact of STREAM on the emotional dimension of distress (anxiety and depression) cannot be assessed conclusively in our population. A study tailored specifically toward patients with high baseline levels of anxiety or depression would be more appropriate to answer this specific question.

Although STREAM was designed for and open to all newly diagnosed patients with cancer, women with breast cancer undergoing curative treatment represented the vast majority of the study population. This leaves uncertainty regarding generalizability of the results, particularly toward men and toward the palliative setting. Women with breast cancer are known to have the largest social media network in the cancer community, which likely allowed for effective online recruitment. The presence of other cancer groups in the Internet community is only emerging, with platforms such as that created by the Movember Foundation for men with cancer.[37] Such platforms may allow for integrating more men into future studies. If targeted specifically, men with prostate cancer also seem to be reachable via the Internet, as shown by an Australian self-help online program, which integrated a patient forum called My Road Ahead.[38]

At baseline, before random assignment, more patients in the control group reported face-to-face psychological support and use of psychotropic drugs than in the intervention group, although the number was not statistically significant. Data on the amount of time spent face to face with local psychologists during the course of the trial were not collected; hence, potential bias cannot be quantified. In contrast, attention bias toward the intervention group, possibly introduced by the time our STREAM psychologists spent online with the patients, may have affected outcome inversely. Because we opted for a care-as-usual (ie, wait-list) rather than active control, this will need to be differentiated in future studies.

A wait-list controlled design is generally accepted to control for the effect of time on the outcome of interest. However, the duration of the wait and consequently the timing of assessments (T2) for the control group are prospectively defined and rigid, whereas the timing of assessments (T2) in the interventions group is dependent on the duration of the intervention and therefore more variable. Hence, time sensitivity is only partially accounted for. This is also true for our study, where median time between T1 and T2 was 9.4 weeks (IQR, 8.6-12.1 weeks) for the intervention group but was shorter in the control group (median, 8.7 weeks; IQR, 8.3-9.3 weeks). Dynamic wait-list controlled designs have been proposed to minimize this potential bias.[39]

Another shortcoming of our trial is that we only show a benefit in distress and quality of life for patients early after diagnosis, with a limited follow-up. It is conceivable, however, that such an early intervention[19] may be of particular importance to prevent chronification of distress.[40] Whether lower distress and increased quality of life after STREAM translate into better treatment tolerance and favorable disease course warrants additional studies.

The unique and common feature of study participants in this trial was a recent diagnosis of cancer. In contrast, the few reported randomized controlled trials on online support for patients with cancer have mainly focused on cancer survivors (ie, interventions later in the disease trajectory). In a randomized, wait-list controlled trial, breast cancer survivors (on average, 3 years after initial diagnosis) who participated in an online program in a similar therapist-guided format as presented here reported significantly improved sexual functioning (the primary end point of the trial) as compared with the wait-list control group.[16] Breast cancer survivors were also the target population in the randomized trial for the Coping With Cancer Workbook.[15] Women who participated in this Web-based self-help program reported better self-efficacy in coping with cancer. Overall quality of life was not reported. The BREATH (Breast Cancer eHealth) intervention,[17] a Web-based intervention based on cognitive behavioral techniques but without therapist guidance, led to reduced distress in breast cancer survivors; however, it was not sustained during the 10-month follow-up.

In conclusion, with digital natives approaching an age that places them at risk for developing age-associated diseases, including cancer, use of the Internet in the health care setting will likely further increase. In this randomized trial, we found that a Web-based, guided self-help intervention resulted in a clinically meaningful improvement in quality of life. Our results indicate that Web-based, guided self-help has potential to efficiently support newly diagnosed patients with cancer.