Randomized Trial of a Tailored Cognitive‐Behavioral Therapy Mobile Application for Anxiety in Patients with Incurable Cancer

Abstract

Background

The aim of this study was to test the efficacy of a tailored cognitive‐behavioral therapy (CBT) mobile application (app) to treat anxiety in patients with incurable cancer.

Materials and Methods

Patients with incurable cancers (n = 145) who reported elevated anxiety symptoms at two cancer centers were randomized to receive either the CBT mobile app for anxiety or a mobile health education program (control) delivered via tablet computers, which patients self‐administered over 12 weeks. To assess anxiety, depression symptoms, and quality of life (QOL), we used the Hamilton Anxiety Rating Scale (HAM‐A, primary outcome), Clinical Global Impression Scale, Hospital Anxiety and Depression Scale (HADS), Patient Health Questionnaire‐9, and Functional Assessment of Cancer Therapy‐General at baseline and 12 weeks. Analysis of covariance models were calculated to assess intervention effects on patient outcomes.

Results

Patients (73.8% female; 91.0% white; mean age = 56.45 years, SD = 11.30) in both study groups reported improvements in anxiety, depression symptoms, and QOL from baseline to postassessment, with no significant differences in any outcome measure between groups. Secondary analyses showed that, among the subgroup of patients with severe baseline anxiety, those randomized to the CBT app had greater improvements on the HAM‐A (Mean Difference = 7.44, standard error [SE] = 3.35, p = .037) and HADS‐Anxiety Subscale (Mean Difference = 4.44, SE = 1.60, p = .010) compared with the control group.

Conclusion

Both the tailored CBT app for anxiety and the health education program were associated with improvements in anxiety, mood, and QOL, but these outcomes did not differ between study groups. The CBT app was more beneficial than health education for patients with severe baseline anxiety.

Cognitive‐behavioral therapy, Anxiety, Advanced cancer

Implications for Practice

A cognitive‐behavioral therapy mobile application tailored to treat anxiety in patients with advanced cancer helps improve access to evidence‐based supportive care in a convenient, private, and timely manner.

Introduction

Up to one third of patients with advanced cancer report experiencing severe anxiety symptoms [1]. Such anxiety is often not only comorbid with depression and physical symptoms such as fatigue and pain 2,3 but also associated with worse quality of life (QOL) 4,5, poor coping [6], suboptimal adherence to chemotherapy [7], and increased health care use 3,8. A first‐line treatment for anxiety is cognitive‐behavioral therapy (CBT), which investigators have demonstrated to be efficacious in the general population [9–11]. However, research is needed to discern the benefit of CBT for patients with incurable cancer, who are confronting very real worries about foreshortened future, debilitating symptoms, and loss of functioning over time.

We previously developed a brief CBT intervention tailored to the specific anxiety concerns of patients with incurable cancer [12]. This manualized treatment incorporated skills for coping with cancer‐related worries and progressive disability. Results from the pilot trial of this intervention showed that patients assigned to CBT reported a significant reduction in anxiety symptoms compared with the wait‐list control group, with corresponding large effect sizes [13]. Yet, the majority of patients receive their cancer care in community settings with minimal access to CBT because of the lack of available clinicians providing evidence‐based psychotherapies [14]. Internet technologies for treating anxiety may offer a promising solution for insufficient access to trained clinicians [15], and preliminary study suggests that web‐based CBT platforms are feasible and acceptable for patients with advanced cancer [16]. However, further work is needed to examine the benefit of such technology in delivering CBT for patients with incurable cancer and anxiety.

We adapted our successful CBT intervention to a mobile application (app) that patients could self‐administer using a tablet computer. The study aim was to test the efficacy of the CBT mobile app intervention in a multisite randomized controlled trial. We hypothesized that patients with incurable cancer and elevated anxiety assigned to receive the intervention would report significant reductions in anxiety symptoms compared with a control group who received a health education program similarly administered on a tablet. We also hypothesized that patients receiving the CBT mobile app would report improved depression symptoms and QOL compared with the control group.

Materials and Methods

Study Design and Setting

For this randomized controlled trial (ClinicalTrials.Gov: NCT02286466), we recruited patients from Massachusetts General Hospital (MGH) Cancer Center in Boston, Massachusetts, and Lee Memorial Regional Cancer Center (LMRCC) in Fort Myers, Florida. The study was IRB‐approved at both the Dana‐Farber/Harvard Cancer Center (DF/HCC) and LMRCC.

Participants

Eligible patients included those with (a) a diagnosis of an incurable solid tumor (stage IV or metastatic disease not being treated with curative intent per oncologist documentation in the electronic health record [EHR]) and (b) clinically significant anxiety symptoms (Hospital Anxiety and Depression Scale [HADS]‐Anxiety Subscale score >7). Additionally, patients had to be ≥18 years of age; at least 4 weeks postdiagnosis; ambulatory and capable of self‐care (Eastern Cooperative Oncology Group [ECOG] Performance Status between 0 and 2) [17]; and able to complete questionnaires in English. Patients with delirium, dementia, or an active serious mental illness interfering with study participation were ineligible.

Procedures

Research staff identified potentially eligible patients by querying the EHR. Following written or verbal approval from the patients’ oncologists, staff approached patients who met initial inclusion criteria at their upcoming clinic visit. Those who screened positive with the HADS were invited to participate and signed an informed consent form. Enrolled patients completed baseline self‐report assessments and a clinician‐administered interview prior to randomization. The study statistician developed a computer‐generated randomization scheme stratified by study site and cancer type. The DF/HCC Office of Data Quality randomly assigned participants 1:1 to either the CBT mobile app intervention group or the mobile health education program control group. The postassessment took place approximately 12 weeks following baseline and included the same self‐report measures and clinician‐administered interview. Participants completed questionnaires on paper or via Research Electronic Data Capture, a Health Insurance Portability and Accountability Act‐compliant web‐based survey tool [18], receiving a $25 gift card for each assessment.

Study Groups

CBT Mobile App Intervention

Using the treatment manual from our prior trial 12,13 and qualitative interviews with patients with incurable cancer and elevated anxiety, we adapted the CBT intervention to a mobile app platform in collaboration with the Partners Center for Connected Health, an MGH‐affiliated software developer. The first phase of this research involved an extensive, iterative process of app development with numerous rounds of user testing and review by the study team consisting of clinical researchers in psychology, psychiatry, and oncology. The finalized app incorporated patient‐centered, interactive, and personalized features, including videos simulating patient and therapist interactions, aimed to teach skills for managing cancer‐related anxiety. These skills were delivered over six sessions (approximately 20–30 minutes each), with corresponding homework exercises (approximately 10–15 minutes each), and a seventh session to review material from prior sessions. The sessions were ordered as follows: (a) Mobile App Introduction and Psycho‐education; (b) Relaxation Training; (c) Activity Planning and Pacing; (d) Problem‐Solving Communication; (e) Staying Present; (f) Creating New Thoughts; and (g) Summary and Review. Participants randomized to the CBT intervention were loaned a study tablet with the CBT app predownloaded. Research staff instructed participants to complete the sessions over 10–12 weeks, noting that the sessions were to be completed in order and that each subsequent session would become available when the participant completed the preceding one. We monitored adherence to the CBT mobile app protocol by tracking the number of sessions and homework exercises that participants accessed.

Health Education Program Control

Patients assigned to the control group received a health education program similarly administered over a tablet during the 12‐week study period. We adapted the content of this program from a health education intervention used as the control condition for an NIH‐funded trial of a cognitive‐behavioral intervention for QOL in patients with advanced prostate cancer [16]. The program consisted of the same number of sessions (approximately 20–30 minutes each) as the CBT mobile app intervention but had no tailored programming logic, videos, or homework assignments. The health education sessions consisted of slide decks with voiceover accompaniment, detailing general information about the following topics: side effects of cancer treatment, exercise, nutrition, memory and cognition, sexual health, and QOL.

Measures

Sociodemographic and Clinical Factors

Patients reported their age, gender, race, ethnicity, marital status, and educational level on a demographic questionnaire. Additionally, participants reported current use of psychotropic medications and other mental health services, such as meeting with a support group, social worker, psychiatrist, psychologist, or other psychotherapist. Research staff collected information on cancer type and treatments, ECOG performance status, and date of diagnosis of metastatic disease from the EHR.

Primary Outcome: Clinician‐Administered Assessments

An independent evaluator (a trained clinical psychology postdoctoral fellow) blinded to group assignment conducted the following interview over the telephone with all participants:

Hamilton Anxiety Rating Scale [19]

To evaluate the severity of anxiety symptoms in the past week, we administered the Hamilton Anxiety Rating Scale (HAM‐A), which consists of 14 items scored from 0 (not present) to 4 (very severe). Higher total scores indicate more severe anxiety symptoms, which can also be categorized as follows: ≤17 = mild anxiety, 18–24 = moderate anxiety, and ≥25 = severe anxiety. Study evaluators used the Structured Interview Guide for the HAM‐A, which is a validated assessment tool [20]. Given the telephone administration, we omitted the final item of the measure, which requires a direct observation of the respondent's behavior. Interviews were audio‐recorded, with a subset reviewed monthly by the principal investigator and the independent evaluators to assess inter‐rater agreement and prevent rater drift.

Clinical Global Impression Scale [21,22]

The Clinical Global Impression Scale is a single‐item measure of overall illness severity and impairment. The study evaluators completed this scale at the end of the interview to denote the participant's anxiety level. The scale ranges from 1 (not at all ill) to 7 (extremely ill), with higher scores indicating greater illness severity.

Secondary Outcomes: Self‐Report Assessments

HADS [23]

The HADS is a well‐validated instrument comprising two subscales of seven items each measuring anxiety and depression symptoms in the past week. Scores range from 0 (no distress) to 21 (maximum distress) on each subscale, with higher scores indicating worse symptoms.

Patient Health Questionnaire‐9 [24,25]

The Patient Health Questionnaire‐9 is a nine‐item measure designed to assess depressive symptoms in the past 2 weeks based on the criteria of the Diagnostic and Statistical Manual of Mental Disorders [26]. Total scores range from 0 to 27, with higher scores indicating worse symptoms.

Functional Assessment of Cancer Therapy‐General [27]

The Functional Assessment of Cancer Therapy‐General is a 28‐item instrument with strong psychometric properties, consisting of four subscales to evaluate physical, functional, emotional, and social wellbeing during the past week in patients with cancer. Higher total scores indicate better QOL.

Statistical Analyses

We conducted analyses with SPSS v.25 (IBM, Armonk, NY) beginning with descriptive summaries of the sample characteristics and mobile app use. To assess the effects of the intervention (and intervention dose) on the continuous primary and secondary outcomes with all available cases, we used analysis of covariance (ANCOVA) with randomization group as the independent variable and change scores for each outcome as the dependent variables, controlling for baseline values of the criterion outcomes and psychotropic medication use (which was imbalanced between study groups). We also calculated Cohen's d ([Mean_post‐Mean_baseline]/SD_diff) to estimate the within‐group effect sizes of the intervention for each outcome, with 0.2, 0.5, and 0.8 corresponding to small, medium, and large effect sizes, respectively [28]. To conduct post hoc moderation analyses, we used the PROCESS macro for SPSS [29], calculating interaction terms between randomization group and baseline HAM‐A scores in linear regression to predict study outcomes. All tests were evaluated at a two‐sided .05 significance level. To account for missing data at postassessment, we repeated all analyses using multiple imputation with 10 pooled data sets. Finally, for our power analysis and sample size calculation, we aimed to recruit up to approximately 150 patients to have at least 128 study completers, allowing for 80% power to detect a between‐group difference in the primary HAM‐A outcome of at least 0.5 standard deviations.

Results

Baseline Characteristics

From February 2015 to October 2016, 154 patients enrolled in the study (89.0% of eligible patients who screened positive on the HADS‐Anxiety Subscale), of whom 145 completed baseline assessments and were randomly assigned to the two study groups (Fig. 1). The mean age of the sample was 56.45 years (SD = 11.30), with the majority self‐identifying as white (91.0%), female (73.8%), and having a spouse/partner (75.9%). The most common cancer types were advanced gastrointestinal, gynecological, lung, and breast malignancies, with a median time from diagnosis of more than 7 months (Table 1). The study groups were well balanced with respect to baseline characteristics except for rates of psychotropic medication use, which were higher in the intervention group.

Figure 1

Participant flow diagram.

Abbreviations: CBT, cognitive‐behavioral therapy; HADS, Hospital Anxiety and Depression Scale; HADS‐A, Hospital Anxiety and Depression Scale‐Anxiety Subscale.

Table 1

Baseline demographic and clinical characteristics

Variable	Health education control group (n = 73)	CBT mobile app intervention (n = 72)
Age, years, mean (SD)	57.03 (12.42)	55.86 (10.08)
Gender, n (%)
Female	52 (71.2)	55 (76.4)
Male	21 (28.8)	17 (23.6)
Race, n (%)
American Indian or Alaska Native	0 (0)	0 (0)
Asian	3 (4.1)	1 (1.4)
Black	2 (2.7)	3 (4.2)
White	65 (89.0)	67 (93.1)
Multiracial	1 (1.4)	0 (0)
Unknown/not reported	2 (2.7)	1 (1.4)
Ethnicity, n (%)
Hispanic or Latino	2 (2.7)	2 (2.8)
Not Hispanic or Latino	68 (93.2)	65 (90.3)
Unknown/not reported	3 (4.1)	5 (6.9)
Relationship status, n (%)
Married/partner	54 (74.0)	56 (77.8)
Single	10 (13.7)	7 (9.7)
Divorced/separated	5 (6.8)	8 (11.1)
Widowed	4 (5.5)	1 (1.4)
Education, n (%)
High school or less	15 (20.5)	16 (22.2)
Some or completed college	34 (46.6)	38 (52.8)
Graduate school	24 (32.9)	18 (25.0)
Cancer type, n (%)
Gastrointestinal	18 (24.7)	23 (31.9)
Gynecological	14 (19.2)	10 (13.9)
Lung	15 (20.5)	9 (12.5)
Breast	8 (11.0)	8 (11.1)
Melanoma	5 (6.8)	10 (13.9)
Sarcoma	7 (9.6)	7 (9.7)
Genitourinary	5 (6.8)	2 (2.8)
Peritoneal	0 (0)	3 (4.2)
Thyroid	1 (1.4)	0
Cancer treatment during study, n (%)
Intravenous chemotherapy	60 (82.2)	58 (80.6)
Oral chemotherapy	4 (5.5)	3 (4.2)
Immunotherapy	10 (13.7)	8 (11.1)
Radiation	5 (6.8)	5 (6.9)
Surgery	5 (6.8)	7 (9.7)
Time since cancer diagnosis, months, median (IQR)	7.69 (1.86–23.26)	7.49 (1.99–24.28)
ECOG PS, n (%)
0	27 (37.0)	26 (36.1)
1	44 (60.3)	39 (54.2)
2	2 (2.7)	7 (9.7)
Psychotropic medication use, n (%)
Antidepressant	19 (26.0)	30 (41.7)
Anxiolytic	46 (63.0)	43 (59.7)
Antipsychotic	5 (6.8)	17 (23.6)
Stimulant	5 (6.8)	6 (8.3)
Current mental health care, n (%)
Support group	3 (4.1)	1 (1.4)
Social worker	7 (9.6)	13 (18.1)
Psychologist/other therapist	6 (8.2)	6 (8.3)
Psychiatrist/nurse	4 (5.5)	9 (12.5)
Clinician‐administered measures, mean (SD)
HAM‐A	19.37 (8.98)	19.11 (7.59)
CGI	3.29 (0.99)	3.32 (0.96)
Self‐report measures, mean (SD)
HADS
Anxiety subscale	11.23 (2.44)	11.60 (3.00)
Depression subscale	6.79 (3.11)	6.97 (3.35)
PHQ‐9	9.16 (5.11)	9.31 (5.06)
FACT‐G	66.40 (12.24)	64.84 (16.26)

Abbreviations: CBT, cognitive‐behavioral therapy; CGI, Clinical Global Impression Scale; FACT‐G, Functional Assessment of Cancer Therapy‐General; ECOG PS, Eastern Cooperative Oncology Group Performance Status; HADS, Hospital Anxiety and Depression Scale; HAM‐A, Hamilton Anxiety Rating Scale; IQR, interquartile range; PHQ‐9, Patient Health Questionnaire‐9.

Intervention Adherence: CBT Mobile App Use

The percentage of participants assigned to the intervention group who used the CBT mobile app declined across the six skill‐based sessions (Fig. 2), although remaining above 50% for each session and homework assignment. The final review session had the lowest rate of participant use (38.9%, n = 28/72). Five participants assigned to the intervention withdrew or died during the study. Overall, 70.8% (n = 51/72) of intervention participants used the mobile app at least through session five, which fell below the a priori adherence criterion of 75% of sessions. Rates of homework use were lower than the sessions, and although strongly encouraged, the homework was not required to progress through the program.

Figure 2

Cognitive‐behavioral therapy mobile app use within intervention group (n = 72).

Intervention Effects on Primary and Secondary Outcomes

Both the CBT mobile app and health education study groups experienced significant improvements across most outcomes from baseline to postassessment, except for the change in HAM‐A scores within the control group (Table 2). The effect sizes for these improvements ranged from small to medium (d = 0.24–0.39) for the primary anxiety outcome and medium to large (d = 0.45–1.20) for the secondary self‐report outcomes. However, these mean changes did not differ between study groups for any outcome. To account for time and adjustment to illness, we reran the ANCOVA models also controlling for time since cancer diagnosis, which did not alter the results (data not shown).

Table 2

Within‐ and between‐group differences in primary and secondary study outcomes

Outcome measure	Health education control within‐group mean change over time			CBT mobile app intervention within‐group mean change over time			Between‐group mean differences
	Adjusted mean change (SE)	95% CI	Cohen's d	Adjusted mean change (SE)	95% CI	Cohen's d	Adjusted mean difference (SE)	p value
Primary
HAM‐A (n = 128)	−1.85 (0.98)	−3.79, 0.08	0.24	−2.99 (0.96)	−4.89, −1.09a	0.39	1.14 (1.38)	.412
CGI (n = 128)	−0.62 (0.13)	−0.88, −0.36a	0.60	−0.66 (0.13)	−0.92, −0.41a	0.63	0.04 (0.18)	.814
Secondary
HADS‐A (n = 135)	−3.55 (0.44)	−4.41, −2.68a	0.98	−4.33 (0.44)	−5.20, −3.45a	1.20	0.78 (0.63)	.215
HADS‐D (n = 135)	−1.41 (0.38)	−2.16, −0.65a	0.45	−1.89 (0.39)	−2.65, −1.13a	0.59	0.48 (0.55)	.379
PHQ‐9 (n = 135)	−2.14 (0.57)	−3.26, −1.02a	0.46	−2.29 (0.57)	−3.42, −1.17a	0.49	0.15 (0.81)	.852
FACT‐G (n = 133)	7.87 (1.77)	4.37, 11.38a	0.54	7.88 (1.76)	4.41, 11.36a	0.55	−0.01 (2.51)	.997

Note: Results from analysis of covariance models controlling for baseline values of criterion outcomes and psychotropic medication use.

^ap values <.05.

Abbreviations: CBT, cognitive‐behavioral therapy; CGI, Clinical Global Impression Scale; CI, confidence interval; Cohen's d, (Mean_post‐Mean_baseline)/SD_diff with 0.2, 0.5, and 0.8 corresponding to small, medium, and large effect sizes, respectively; FACT‐G, Functional Assessment of Cancer Therapy‐General; HADS‐A, Hospital Anxiety and Depression Scale‐Anxiety Subscale; HADS‐D, Hospital Anxiety and Depression Scale‐Depression Subscale; HAM‐A, Hamilton Anxiety Rating Scale; PHQ‐9, Patient Health Questionnaire‐9; SE, standard error.

Intervention Dose Effects on Primary and Secondary Outcomes

To determine the effects of intervention dose on outcomes, we first examined differences between CBT app intervention completers (≥6 sessions) versus the health education control group. ANCOVA results showed that the CBT mobile app completers reported a greater reduction in anxiety symptoms on the HADS‐Anxiety subscale compared with the control group (Adjusted Mean Difference = 1.67, standard error = 0.72, 95% confidence interval: 0.25–3.09, p = .021), but the between‐group differences in the other outcomes did not reach the threshold for statistical significance. Additionally, we examined differences within the CBT mobile app group alone, comparing intervention completers (≥6 sessions) versus intervention noncompleters (<6 sessions). The CBT mobile app completers reported significant improvements across all self‐report measures for anxiety, depression symptoms, and QOL compared with the CBT mobile app noncompleters (Table 3).

Table 3

Effects of intervention dose on primary and secondary outcomes

Outcome measure	Differences between CBT mobile app intervention completers (n = 40) vs. health education control group (n = 68)			Differences between CBT mobile app intervention completers (n = 40) vs. CBT mobile app intervention noncompleters (n = 27)
	Adjusted mean difference (SE)	95% CI	p value	Adjusted mean difference (SE)	95% CI	p value
Primary
HAM‐A	2.33 (1.59)	−0.82, 5.49	.145	3.39 (1.91)	−0.43, 7.22	.081
CGI	0.27 (0.22)	−0.16, 0.70	.217	0.46 (0.25)	−0.04, 0.97	.071
Secondary
HADS‐A	1.67 (0.72)	0.25, 3.09	.021	2.10 (0.87)	0.36, 3.83	.019
HADS‐D	1.18 (0.62)	−0.05, 2.41	.060	1.68 (0.76)	0.17, 3.19	.030
PHQ‐9	1.44 (0.84)	−0.22, 3.10	.087	3.00 (1.11)	0.77, 5.22	.009
FACT‐G	−4.47 (2.56)	−9.54, 0.61	.084	−10.65 (3.74)	−18.13, −3.18	.006

Note: Results from analysis of covariance models controlling for baseline values of criterion outcomes and psychotropic medication use.

Abbreviations: CBT, cognitive‐behavioral therapy; CGI, Clinical Global Impression Scale; CI, confidence interval; FACT‐G, Functional Assessment of Cancer Therapy‐General; HADS‐A, Hospital Anxiety and Depression Scale‐Anxiety Subscale; HADS‐D, Hospital Anxiety and Depression Scale‐Depression Subscale; HAM‐A, Hamilton Anxiety Rating Scale; PHQ‐9, Patient Health Questionnaire‐9; SE, standard error.

Intervention Moderation Effects by Baseline Anxiety Level

We observed that the effects of group assignment on the anxiety outcome measures varied by baseline HAM‐A scores at a threshold score ≥25, which corresponds to severe anxiety symptoms. Subgroup analyses examining intervention effects stratified by baseline HAM‐A scores revealed that, among patients with severe baseline anxiety symptoms, the CBT mobile app led to significantly greater reductions in anxiety symptoms, as measured by both the HAM‐A and HADS‐Anxiety Subscale, compared with the health education program (Table 4). Tests of interaction effects between baseline anxiety and group assignment on the secondary outcomes of depression and QOL were not significant.

Table 4

Moderation effects of baseline anxiety (HAM‐A scores) on primary and secondary outcomes

Outcome measure	Health education control within‐group mean change stratified by baseline HAM‐A scores		CBT mobile app intervention within‐group mean change stratified by baseline HAM‐A scores		Group X baseline anxiety (HAM‐A scores) interaction	Between‐group mean differences stratified by baseline HAM‐A scores
	Adjusted mean change (SE)	95% CI	Adjusted mean change (SE)	95% CI	p value	Adjusted mean difference (SE)	p value
HAM‐A					.044
HAM‐A ≥ 25 (n = 26)	−6.74 (2.13)	−11.16, −2.31	−14.18 (2.51)	−19.38, −8.98		7.44 (3.35)	.037
HAM‐A < 25 (n = 102)	−0.56 (1.09)	−2.73, 1.61	−0.50 (1.03)	−2.54, 1.53		−0.05 (1.53)	.972
CGI					.053
HAM‐A ≥ 25 (n = 26)	−0.54 (0.27)	−1.10, 0.02	−1.35 (0.32)	−2.01, −0.69		0.81 (0.43)	.071
HAM‐A < 25 (n = 102)	−0.66 (0.15)	−0.96, −0.36	−0.51 (0.14)	−0.79, −0.23		−0.15 (0.21)	.478
HADS‐A					.012
HAM‐A ≥ 25 (n = 29)	−1.67 (1.01)	−3.75, 0.42	−6.11 (1.14)	−8.44, −3.77		4.44 (1.60)	.010
HAM‐A < 25 (n = 105)	−4.10 (0.48)	−5.06, −3.15	−3.98 (0.47)	−4.91, −3.05		−0.13 (0.68)	.854
HADS‐D					.204
HAM‐A ≥ 25 (n = 29)	−0.69 (1.04)	−2.84, 1.46	−2.53 (1.17)	−4.94, −0.12		1.84 (1.65)	.276
HAM‐A < 25 (n = 105)	−1.75 (0.40)	−2.54, −0.95	−1.78 (0.39)	−2.55, −1.01		0.03 (0.57)	.959
PHQ‐9					.336
HAM‐A ≥ 25 (n = 29)	−0.62 (1.48)	−3.67, 2.43	−3.24 (1.65)	−6.65, 0.17		2.62 (2.30)	.266
HAM‐A < 25 (n = 105)	−2.40 (0.60)	−3.58, −1.21	−2.31 (0.58)	−3.46, −1.16		−0.09 (0.85)	.921
FACT‐G					.127
HAM‐A ≥ 25 (n = 27)	8.62 (4.55)	−0.80, 18.04	15.94 (4.74)	6.13, 25.75		−7.32 (6.89)	.299
HAM‐A < 25 (n = 105)	7.95 (1.95)	4.07, 11.82	6.07 (1.90)	2.31, 9.83		1.88 (2.77)	.499

Note: All analyses controlled for baseline values of criterion outcomes and psychotropic medication use.

Abbreviations: CBT, cognitive‐behavioral therapy; CGI, Clinical Global Impression Scale; CI, confidence interval; FACT‐G, Functional Assessment of Cancer Therapy‐General; HADS‐A, Hospital Anxiety and Depression Scale‐Anxiety Subscale; HADS‐D, Hospital Anxiety and Depression Scale‐Depression Subscale; HAM‐A, Hamilton Anxiety Rating Scale; PHQ‐9, Patient Health Questionnaire‐9; SE, standard error.

Missing Data Analyses

We repeated all tests using multiple imputation to account for the 11.7% (n = 17/145) and 6.9% (n = 10/145) of missing data in the clinician‐administered and self‐report assessments, respectively. These results generally confirmed the available case analyses showing that both study groups experienced significant improvements in anxiety, depressive symptoms, and QOL. We also observed similar intervention dose effects and moderation effects with higher baseline anxiety (supplemental online Tables 1–3).

Discussion

In this randomized trial of a tailored CBT mobile app for anxiety versus a health education program, we found that patients with incurable cancer in both study groups experienced improvements in anxiety, depressive symptoms, and QOL, with no outcomes differing significantly between groups. Although adherence to the CBT app was suboptimal, we observed salient intervention dose effects between study groups and within the intervention group. Finally, the subgroup of patients with severe baseline anxiety had greater reductions in anxiety symptoms with the CBT mobile app compared with the health education program.

To our knowledge, we are the first to report on the efficacy of a CBT mobile app intervention tailored for patients with incurable cancers and elevated anxiety. Studies of web‐based interventions for patients with cancer have primarily focused on survivors, especially with breast cancer [30], targeting a range of concerns including general distress, sleep disturbance, and physical symptoms [31–36]. Only a few web‐based CBT intervention trials have targeted anxiety symptoms, such as fear of recurrence and post‐traumatic symptoms in cancer survivors [39–41]. Although many of these investigations were pilot trials demonstrating promising findings, our study advances this field with an efficacy trial of a novel CBT mobile app for anxiety in an understudied population of patients with incurable cancers.

Patients assigned to the CBT mobile app did not complete the expected number of sessions, falling below the threshold of 75%. Yet, our intervention was completely self‐administered, and the sessions only became accessible when participants completed the preceding ones in order, which may have impacted adherence. Furthermore, web‐based interventions suffer high rates of attrition, with only 50% of participants adhering fully to such programs on average [40]. In our prior, in‐person CBT study with this population, 80% of participants adhered to the intervention, with 53% of sessions taking place on the same day as other medical appointments or via telephone, facilitating participation [13]. Adherence to e‐therapies tends to be associated with outcomes [41], and we observed that the CBT app completers experienced improvements in the self‐report outcomes compared with the control group and compared with intervention noncompleters, despite reduced statistical power. Such findings underscore the importance of not simply translating evidence‐based therapeutic interventions to mobile app formats for broader dissemination but also considering the personalized features that encourage patient use of such technology. For example, incorporating engaging design aesthetics that enhance positive affective states and gamification to promote learning of skills through educational games, as well as administering a hybrid intervention that includes a live therapist who has intermittent contact with the patient, may better foster skill acquisition and reinforce adherence to mobile app interventions.

Participants in both study groups reported improvements across outcomes over time, with the CBT app being more effective than the health education program in reducing anxiety symptoms in the subgroup of patients with severe baseline anxiety. Such differential effects raise questions about how to tailor interventions to patients’ needs. Access to online educational materials may be sufficient support for patients with mild to moderate anxiety symptoms, whereas an interactive CBT mobile app may be more powerful for those with severe anxiety.

Potential explanations for our findings include the possibility that the health education program was an active control that effectively improved outcomes; participants naturally regressed to the mean across the study measures; or participant expectancy effects led to the observed improvements. Although web‐based psycho‐educational interventions can have positive effects on distress in patients with cancer 42,43, concerns about regression to the mean or expectancy effects are mitigated. Participants were on average more than 7 months postdiagnosis of advanced cancer, well past the usual period of adjustment, and controlling for time since diagnosis did not affect the findings. Moreover, the intervention dose effects and between‐group differences in patients with severe baseline anxiety would be unlikely if due entirely to regression to the mean or patient expectations for the treatment. Thus, it is noteworthy that for the most distressed patients, the CBT mobile app was able to reduce the severity of anxiety symptoms.

As oncology clinics move toward comprehensive distress screening as part of standard care [44], online tools represent welcome opportunities for patients to access beneficial supportive care interventions, especially when mental health resources are scarce. While in‐person CBT may remain the gold standard for improving outcomes, many patients with cancer lack opportunities for such care. Indeed, the most difficult‐to‐reach patients, whether it be a result of geographic barriers, costs, or patient reluctance to seek help, are often those who need the most support. The tailored CBT mobile app therefore brings evidence‐based mental health treatment to patients in a convenient, private, low‐burden, and timely manner.

Strengths of our study include the rigorous randomized trial design, adaptation of an evidence‐based CBT intervention to mobile technology, use of independent evaluators for the primary outcome, and multimethod assessment with validated measures. Nonetheless, several limitations warrant consideration. First, the sample was relatively homogeneous with respect to racial and ethnic background as well as gender, limiting generalizability of findings. Using a health education program for the comparison group may have underestimated intervention effects on anxiety relative to a nonactive control. We also did not have the capacity to monitor participant adherence to the health education program, which would have allowed for a more direct comparison of intervention completers in both study groups. In addition, patients randomly assigned to the CBT mobile app group had a higher rate of psychotropic medication use, although controlling for this factor in all analyses did not alter the findings. Lastly, the results regarding the subgroup of patients with severe baseline anxiety must be interpreted with caution given the small samples.

Conclusion

Patients with incurable cancer experience considerable morbidity with often unmet supportive care needs. Mobile apps represent a promising approach for enhancing broader dissemination of essential psychosocial care in this population. Ongoing research is needed to optimize patient adherence to mobile app interventions through innovative design features, greater flexibility in how patients engage with intervention components, and perhaps intermittent opportunities for reinforcement of learning with live therapists. Our adaptation of a tailored, in‐person CBT intervention to a novel mobile app is a key first step not only in improving access to high‐quality supportive care for patients with incurable cancer but also in considering potential strategies for targeting interventions based on their level of anxiety.

Acknowledgments

This work was supported by the American Cancer Society Research Scholar Grant 124138‐RSG‐13‐019‐01‐CPPB (principal investigator: J.A.G.); additional author time was supported by NIH/National Institute on Drug Abuse 9K24DA040489 (principal investigator: S.A.S.).

Author Contributions

Conception/design: Joseph A. Greer, Steven A. Safren, William F. Pirl, Jennifer S. Temel

Provision of study material or patients: Joseph A. Greer, Steven A. Safren, William F. Pirl, Jennifer S. Temel

Collection and/or assembly of data: Jamie Jacobs, Nicole Pensak, James J. MacDonald, Charn‐Xin Fuh, Giselle K. Perez, Alina Ward

Data analysis and interpretation: Joseph A. Greer, Jamie Jacobs, Nicole Pensak, Giselle K. Perez, Colleen Tallen, Alona Muzikansky, Lara Traeger, Frank J. Penedo, Areej El‐Jawahri, Steven A. Safren, William F. Pirl, Jennifer S. Temel

Manuscript writing: Joseph A. Greer, Jamie Jacobs, Nicole Pensak, James J. MacDonald, Charn‐Xin Fuh, Giselle K. Perez, Alina Ward, Colleen Tallen, Alona Muzikansky, Lara Traeger, Frank J. Penedo, Areej El‐Jawahri, Steven A. Safren, William F. Pirl, Jennifer S. Temel

Final approval of manuscript: Joseph A. Greer, Jamie Jacobs, Nicole Pensak, James J. MacDonald, Charn‐Xin Fuh, Giselle K. Perez, Alina Ward, Colleen Tallen, Alona Muzikansky, Lara Traeger, Frank J. Penedo, Areej El‐Jawahri, Steven A. Safren, William F. Pirl, Jennifer S. Temel

Disclosures

The authors indicated no financial relationships.

For Further Reading: Kelly E. Irwin, Elyse R. Park, Lauren E. Fields et al. Bridge: Person‐Centered Collaborative Care for Patients with Serious Mental Illness and Cancer. The Oncologist first published on January 29, 2019; doi:10.1634/theoncologist.2018‐0488.

Implications for Practice: Serious mental illness affects 13 million U.S. adults who experience increased cancer mortality. To improve outcomes, new models of integrated oncology and mental health care are urgently needed. This study found that it was feasible to identify, enroll, and retain patients with serious mental illness and a new cancer in a trial of integrated mental health and cancer care (Bridge). Patients, caregivers, and oncologists reported that Bridge facilitated the initiation and completion of cancer care. Randomized trials are warranted to investigate the impact on cancer outcomes. Trial procedures may inform consent, engagement, and trial retention for patients with mental illness.

Author notes

Disclosures of potential conflicts of interest may be found at the end of this article.