Evaluation of an Online-Based Self-Help Program for Patients With Panic Disorder: Randomized Controlled Trial

Original Paper Evaluation of an Online-Based Self-Help Program for Patients With Panic Disorder: Randomized Controlled Trial Christopher Lalk1, MSc; Teresa Väth1, MSc; Sofie Hanraths1, MSc; Luise Pruessner2, PhD; Christina Timm2, PhD; Steffen Hartmann2, MSc; Sven Barnow2, PhD; Julian Rubel1, PhD 1Clinical Psychology and Psychotherapy of Adulthood, Institute of Psychology, University Osnabrück, Osnabrück, Germany 2Clinical Psychology and Psychotherapy, Institute of Psychology, Heidelberg University, Heidelberg, Germany Corresponding Author: Christopher Lalk, MSc Clinical Psychology and Psychotherapy of Adulthood Institute of Psychology University Osnabrück Lise-Meitner-Straße 3 Osnabrück, 49076 Germany Phone: 49 541 969 76 Email: christopher[email protected] Abstract Background: Panic disorder is an anxiety disorder marked by severe fear of panic attacks in the absence of causes. Agoraphobia is a related anxiety disorder, which involves fear and avoidance of specific situations in which escape or help may be difficult. Both can be debilitating and impair well-being. One treatment option may be internet-based cognitive behavioral therapy (iCBT), which allows large-scale application and may overcome treatment barriers for some individuals. Objective: This study aimed to evaluated the effectiveness of a novel online self-help intervention for panic disorder with or without agoraphobia. As our primary hypotheses, we expected the intervention to improve panic and agoraphobia symptoms and well-being. Our secondary hypotheses entailed improvements in daily functioning, mental health literacy, working ability, and health care use in the intervention group. Methods: German-speaking patients (N=156) aged 18-65 years with internet access and a diagnosis of panic disorder with or without agoraphobia were recruited for this randomized controlled trial. The intervention group (n=82) received access to a 12-week online self-help program entailing psychoeducation, cognitive restructuring, exposure, and mindfulness elements. The control group (n=72) received care as usual during the study period and was offered the prospect of using the program after 12 weeks. The primary outcomes were assessed via the Panic and Agoraphobia Scale (PAS) and the WHO (World Health Organization)-5 Well-Being Index (WHO-5). Mixed effect models were computed using multivariate imputation by chained equation for the analysis of intervention effects. Results: In the intervention group, participants completed on average 7.3 out of 12 (60.8%) modules, and 27 out of 82 (32.1%) participants finished the whole course. Changes in PAS revealed a significant effect in favor of the intervention group (t110.1=–2.22, Padj=.03) with a small to moderate effect size (d=–0.37, 95% CI –0.70 to –0.04). No significant effect was found for the second primary outcome WHO-5 (t149.8=1.35, Padj=.09) or the secondary outcomes. Improvements were observed in anxiety (t206.8=–4.12; P<.001; Cohen d=–0.60, 95% CI –0.089 to –0.32) and depression (t257.4=–3.20; P<.001; Cohen d=–0.41 95% CI –0.66 to –0.16). No negative effects were associated with the intervention (t125=–1.14, P=.26). Conclusions: The presented online intervention can help reduce the core symptomatology of panic disorder and agoraphobia, as well as anxiety symptoms and associated depression. No effects were found for well-being and secondary outcomes. This may be due to higher illness burden in the intervention group and possibly the COVID pandemic, which caused unique challenges to patients suffering from panic disorder. Therefore, further research and intervention adaptations may be warranted to improve these outcomes. Trial Registration: German Clinical Trials Register DRKS00023800; https://drks.de/search/en/trial/DRKS00023800 J Med Internet Res 2025 | vol. 27 | e54062 | p. 1https://www.jmir.org/2025/1/e54062 (page number not for citation purposes) Lalk et alJOURNAL OF MEDICAL INTERNET RESEARCH XSL • FO RenderX (J Med Internet Res 2025;27:e54062) doi: 10.2196/54062 KEYWORDS internet-based CBT; agoraphobia; well-being; iCBT; internet-based intervention; panic disorder with and without agoraphobia; panic disorder; self-help; quality of life; effectiveness; online; self-help intervention; panic symptoms; well-being; daily functioning Introduction Background Panic disorder involves recurring sudden panic attacks and a constant fear of experiencing more episodes. These attacks come with physical symptoms like breathing issues, palpitations, sweating, and nausea, as well as psychological symptoms such as derealization and a fear of losing control or dying [1,2]. Agoraphobia, which affects 35%-65% of people with panic disorder [3], involves excessive fear and avoidance of situations where escape may be difficult, like public transport, and help may not be readily available in case of a panic attack. Both panic disorder and agoraphobia are closely related and often seen as part of a continuum [4]. In earlier versions of the DSM (Diagnostic and Statistical Manual of Mental Disorders) (ie, the third and fourth editions [DSM-III and DSM IV]) agoraphobia was defined as a feature of panic disorder (ie, panic disorder with agoraphobia or panic disorder without agoraphobia), which is why some research is organized around this conceptualization. Panic disorder, with or without agoraphobia, comes with functional impairments and reduced well-being, increasing the risk of other mental disorders [5-8]. It also imposes a significant economic burden, surpassing that of other anxiety, mood, or alcohol-related disorders [9]. These costs include hospital treatments, health care visits, and, most notably, absenteeism, which accounts for 60% of all expenses [10]. Thus, it is crucial to offer effective and timely treatment for panic disorder for both societal and individual well-being. Regarding psychotherapeutic treatments, recently cognitive behavioral therapy (CBT) and short-term psychodynamic therapy have been identified as treatments of choice [11]. CBT targets fear and avoidance behavior by psychoeducation, exposure therapy, cognitive restructuring, mindfulness, and acceptance interventions leading to large effect sizes compared with a waitlist group (hedges g=0.96; [12]). Also, effects remain superior to treatment as usual for 6 months of follow-up [13]. Psychopharmacology yields small to moderate effect sizes in comparison to placebo and similar to CBT [14]. Although, compared with pharmacotherapy, CBT shows longer-term treatment effects, better cost-efficacy, and higher patient acceptance [15]. An analysis of treatment barriers in anxiety disorders [16] showed that 63 out of 77 (81.8%) patients with panic disorder contemplate treatment. Still, only 40 out of 59 (67.3%) sought help at least once in their life, and this number was even lower in patients with agoraphobia (21/56, 36.9%). Frequently reported barriers to help-seeking include self-reliance, presumed ineffectiveness, high waiting periods, or problems with the practitioner [16]. Further, negative attitudes and lack of knowledge and appropriate beliefs about mental health are associated with less help-seeking behavior [17]. During treatment waiting periods, it is recommended to offer patients self-help programs based on CBT [18], since technology-based treatment alternatives can help surpass the aforementioned barriers [19]. Most of these alternatives are based on CBT, since it is well suited for online intervention delivery due to its highly structured, directive, standardized nature and its focus on psychoeducation and homework [20]. Of particular note is the benefit of internet-based CBT (iCBT) to the healthcare system, as it is a cost-effective treatment alternative with similar efficacy to face-to-face CBT [21,22]. Also, iCBT can help to bridge the waiting periods for face-to-face psychotherapy, which on average lasts several months in Germany and has further increased since the beginning of the COVID-19 pandemic [23,24]. A systematic review and meta-analysis including 27 studies on iCBT for panic disorder [25] showed high efficacy and effectiveness for reducing symptoms of panic disorder (hedges g=1.16) and agoraphobia (hedges g=0.91) compared with waitlist. Another meta-analysis of 13 RCTs for panic disorder and agoraphobia found no efficacy difference between unguided iCBT and face-to-face CBT in terms of panic and agoraphobia symptoms, comorbid depression and anxiety, as well as quality of life improvement [26] indicating that unguided iCBT may be comparably effective to face-to-face CBT. However, in a network meta-analysis of 74 efficacy trials, unguided iCBT was not superior to care as usual for the treatment of panic with or without agoraphobia [27]. In the same analysis, both guided forms of iCBT and face-to-face CBT were superior to care as usual, though not superior to unguided iCBT. Therefore, results regarding the efficacy and effectiveness of unguided iCBT for panic and agoraphobia are mixed, paradoxically indicating similar effects as face-to-face CBT while also indicating no improvement to care as usual. Also, little research can be found investigating broader effects on well-being, functioning, mental health literacy, working ability, and health care use. These seem particularly relevant to assess effects on cost-effectiveness. Objectives This study aimed to evaluate a 12-week online self-help program (Selfapy) for patients with panic disorder with or without agoraphobia within the German health care system. For the analysis, we had 2 objectives: first, we compared the intervention with care as usual, since the superiority for unguided self-help toward care as usual has been questioned; second, we assessed effectiveness for a range of broader outcomes with limited research data (eg, well-being and functioning). J Med Internet Res 2025 | vol. 27 | e54062 | p. 2https://www.jmir.org/2025/1/e54062 (page number not for citation purposes) Lalk et alJOURNAL OF MEDICAL INTERNET RESEARCH XSL • FO RenderX Hypotheses Based on previous research, the primary hypotheses expect superior improvement on panic and agoraphobia symptoms and well-being in the intervention group (IG). The secondary hypotheses expect superior improvement of daily functioning, work capacity, mental health literacy, and the efforts and burdens of patients for the health care sector. Finally, changes in comorbid anxiety and depression symptoms as well as negative effects of the program beyond symptomatology, are being explored as exploratory hypotheses. All outcomes were tested against the control group (CG) after 12 weeks. Methods Study Design The study is reported according to the CONSORT-EHEALTH (Consolidated Standards of Reporting Trials of Electronic and Mobile Health Applications and Online Telehealth) eHealth guidelines (Multimedia Appendix 1). The parallel group trial was preregistered beforehand [28] and a study protocol was published [29]. Eligible patients were randomly assigned to the IG or CG in a 1:1 ratio. Patients in the IG could access the intervention immediately after randomization, while the CG could only access the intervention after a waiting period of 12 weeks. Interim and final evaluations occurred 6 (T2) and 12 (T3) weeks after the baseline assessment (T1). Participants Announcements for study participation were published in the whole of Germany via a university email newsletter, social media, and on flyers in clinics, pharmacies, and practices of medical doctors and psychotherapists. After an online prescreening, participants were invited to choose an appointment for a remote diagnostic interview assessing the inclusion and exclusion criteria. To this end, a structured diagnostic interview diagnostisches interview psychischer störungen-open access (DIPS-OA, [30,31]) was conducted with every participant via video calls. Regarding DSM-IV-TR (Diagnostic and Statistical Manual of Mental Disorders [Fourth Edition, Text Revision]) criteria, the DIPS-OA was found to have acceptable interrater (0.78) and retest (0.76) reliability for anxiety disorders [32]. Altogether 4361 people started the online screening, of whom 764 (17.5%) were deemed eligible to be scheduled for the diagnostic interview. The online screening consisted of short screening questionnaires of the inclusion and exclusion criteria. Inclusion and Exclusion Criteria Video calls were conducted with all individuals who checked the prescreening criteria to assess inclusion and exclusion criteria, during which eligibility was assessed via the DIPS-OA. Trained psychologists conducted all interviews under the supervision by a certified psychotherapist (CBT). Eligible individuals were those who (1) were between 18 and 65 years of age, (2) had sufficient knowledge of the German language, (3) had uninterrupted internet access, (4) provided electronic informed consent to participate in the study, and (5) met the criteria for a diagnosis of panic disorder with or without an additional diagnosis of agoraphobia. Individuals were excluded if they did not meet any of the inclusion criteria or met any of the following criteria: (1) past or current diagnosis of bipolar disorder, (2) past or current diagnosis of psychotic disorder, (3) current diagnosis of substance dependence, (4) current diagnosis of a severe major depressive episode, and (5) acute suicidality. The criteria were chosen because they could interfere with the successful implementation of the course. Intervention The online self-help program for the treatment of panic disorder with or without agoraphobia (Selfapy) was conducted as the intervention. The program is based on evidence-based methods and exercises derived from CBT (such as psychoeducation, cognitive restructuring, or exposure), as well as elements from Mindfulness-Based Therapy (eg, [33,34]). The online intervention consists of core modules, which include mandatory exercise content, and a subsequent set of optional specialization areas that are individually selectable (for a complete overview, [29]). Each module covers a specific topic, such as exposure, mindfulness, or problem-solving training. The modules contain informative texts, videos, audio, and interactive exercises and can be used via the web as well as on mobile devices. Participants completed the online program independently and without support. They were monitored for suicidality in which case they were messaged by a psychologist. Active communication only occurred for safety reasons. The IG had immediate access to the 12-week self-help treatment and were advised to spend 15-20 minutes daily on it. The CG The CG received no treatment for 12 weeks but could seek other assistance beyond this trial, such as medication or therapy, to mimic routine care. All concurrent treatments were self-reported. The CG received the study intervention after study completion (=after 12 weeks), since 12 weeks is a common waiting time for psychotherapy in Germany [24]. Outcomes Measures were conducted at 3 different points in time: Before the start of the intervention (T1, baseline), after 6 weeks (T2, during treatment), and 12 weeks after the beginning of the intervention (T3, post treatment). At each measurement time point, the primary, secondary and exploratory outcomes were assessed. Assessment was conducted via an online assessment platform [35]. Primary Outcome Measures The change in panic and agoraphobia symptoms was evaluated using the Panic and Agoraphobia Scale (PAS; [36]). The PAS consists of 13 items that are rated on a 5-point Likert scale. There scale contains 5 subscales: panic attacks, agoraphobic avoidance, anticipatory avoidance, disability, and worries about health. In our data, we calculated McDonald ω [37] as a reliability measure with ω=0.86 at T1. Well-being was assessed by the WHO (World Health Organization)-5 Well-Being Index (WHO-5; [38]). The WHO-5 contains items measuring positive mood, calmness, high energy levels, good rest, and interest in daily activities. J Med Internet Res 2025 | vol. 27 | e54062 | p. 3https://www.jmir.org/2025/1/e54062 (page number not for citation purposes) Lalk et alJOURNAL OF MEDICAL INTERNET RESEARCH XSL • FO RenderX Secondary Outcome Measures Functioning in daily life was measured by the Work and Social Adjustment Scale (WSAS; [39]) assessing professional and personal functioning. Reliability was good (ω=0.79) at T1. Work capacity was measured with the iMTA Productivity Cost Questionnaire (iPCQ) to assess the amount of lost working hours in the last 4 weeks due to absenteeism or distress-related impaired work capabilities (iPCQ [40]). Mental health literacy was measured with the Mental Health Literacy Scale (MHLS; [41]) with high reliability (ω=0.85 at T1). The extent of therapy-related efforts and burdens of patients (Client Sociodemographic and Service Receipt Inventory [CSSRI] [42]) was collected on three subscales: CSSRI-partly inpatient to assess partly inpatient treatment, CSSRI-complementary to assess complementary services (eg, self-help groups), and CSSRI-ambulant to assess outpatient services (eg, psychotherapy treatment and medical treatment). Additional Outcome Measures Adverse treatment effects were assessed with the Negative Effects Questionnaire (NEQ; [43]). The NEQ contains 32 items and showed high reliability (ω=0.89 at T3). Also, general symptoms of anxiety were assessed using the Beck Anxiety Inventory (BAI; [44]) with ω=0.89 at T1. Depressive symptoms were collected with the Patient Health Questionnaire-9 (PHQ-9 [45]; ω=0.74) at T1. Sample Size The between-group effect size estimate was based on meta-analytic evidence for effect sizes in unguided online psychological interventions for anxiety disorders Cohen (d=0.45; eg, [46]). This effect was used as the basis for sample size determination. For the planned mixed model with 2 measurement time points with a general correlation structure [47], a directed hypothesis, a group allocation of 1:1, a power of 0.80, and an α level of .025 after Bonferroni-Holm correction, a total of 156 patients (78 per group) were needed. The number of cases was calculated using the R-tool (Michael C. Donohue) longpower [48]. For the secondary outcomes, we calculated a minimal detectable effect size of Cohen d=–0.46 with 80% power and an α level of .0125 (Bonferroni-Holm adjustment) based on a post hoc power analysis of the WSAS with simr (Peter Green) [49]. Randomization and Blinding Randomization in 1:1 ratio without stratification or blocks was done automatically by a computer-generated code. Participants were automatically informed of their group via email. Data collection, evaluation, and statistical analysis were carried out blindly. A team member not involved in the analysis coded the group variable, and analysis scripts were prepared before knowing the actual data. Statistical Analyses The statistical analyses were conducted following the study protocol [29]. The analyses were performed with R (version 4.2.0; R Core Team) [50]. Adhering to intention-to-treat principles, none of the enrolled participants were generally excluded. Missing values were replaced by multivariate imputation by chained equations (MICE; with n=5 imputations [51]) based on the control arm, using the variables age and gender as predictors in addition to the T1 outcome measurements. Four additional sensitivity analyses are reported in the online support material (OSM) in Multimedia Appendix 2 (OSM 1-4) for the primary outcomes: A completer analysis using only patient data with completed T1 and T3 measures, last-observation-carried-forward (LOCF), baseline-observation-carried-forward (BOCF), and a reference-based-multiple jump to reference imputation (J2R; [52]). In addition to these analyses, a “per-protocol” sample sensitivity analysis was defined for exploratory analyses, including all IG patients who completed at least 4 of the 12 modules. The confirmatory analysis of the primary endpoints consisted of calculating a mixed model with 2 measurement time points and a general correlation structure [47]. A random effect for the participant was calculated (random intercept), and 3 fixed effects (group, time, and the interaction of these 2 effects). The 2 measurement times were nested within participants. The treatment effect was estimated as the fixed interaction effect after the final (T3) assessment. To assess the magnitude of the treatment effects, the fixed interaction effect of time and group was divided by the root of the summed variances of the random effects [53]. Effect sizes can be roughly interpreted according to Cohen d: effect sizes of 0.20 are considered small, 0.50 moderate, and 0.80 large [54]. Secondary confirmatory outcomes were calculated only after success in the primary analysis to prevent alpha inflation, using the same mixed model with a random intercept for the participant. The CSSRI questionnaire was split into the 3 most relevant subscales (CSSRI-partially inpatient, CSSRI-outpatient, and CSSRI-complementary) that were also subject to Bonferroni-Holm adjustment by dividing by 3, 2, and 1, respectively. Due to highly skewed data, the iPCQ and the CSSRI partially inpatient scales were log10-transformed. The CSSRI-outpatient and CSSRI-complementary scales were dichotomized because of rare extreme outliers, in which case a transformation is not recommended [55]. For these dichotomized measures, the analysis was adapted to a mixed logistic regression model to stay as close as possible to the study protocol and odds ratios were calculated as the effect size. The additional outcome measures, BAI and PHQ-9,were analyzed using the same model as the primary and secondary outcomes. Independent ttests and chi-square tests were used to estimate differences between groups in pretreatment sample characteristics. Also, ttests were used to identify differences in adverse effects in the NEQ at T3. Data and Code All data and analysis code have been made publicly available at the OSF repository and can be accessed at [56]. Materials about the content of the online intervention are reported in the study protocol [29]. J Med Internet Res 2025 | vol. 27 | e54062 | p. 4https://www.jmir.org/2025/1/e54062 (page number not for citation purposes) Lalk et alJOURNAL OF MEDICAL INTERNET RESEARCH XSL • FO RenderX Ethical Considerations This study was approved by the ethics committee at Heidelberg University in adherence to institutional guidelines (AZ Prüß 2021 1/1). All participants provided informed consent prior to their study participation and were informed that they could withdraw consent anytime. During the study, participant data were pseudonymized by replacing data identifiers (eg, participants’ name) with a pseudonymous code. After study completion, all data were anonymized by deleting the data identifiers. All participants received an allowance of 30€ after completing the questionnaires (T1 and T3). Suicidal thoughts were assessed at multiple time points (T1, T2, and T3) using a Likert scale. If participants rated their thoughts as 1 or higher (indicating they had thoughts of wanting to harm themselves) for the past 2 weeks, they were contacted by phone or email, and an emergency plan was established. For individuals contacted due to suicidality, their questionnaire participation was halted to prioritize immediate support. Suicidal incidents occurred 3 times, all within the CG. Results Participants Flow Recruitment occurred from February 12, 2021, to March 21, 2022. A total of 764 clinical interviews were conducted, leading to 292 participants being excluded based on inclusion and exclusion criteria (Figure 1). Further, 160 participants declined study participation and additional 156 participants joined another trial on generalized anxiety disorder [57]. The remaining 156 participants were randomized into the IG (n=82) and CG (n=74). Sociodemographic differences between the groups were not significant. Figure 1. CONSORT flow diagram. Deviations from the sample size occurred for some of the secondary and additional outcomes due to partial missingness. More IG participants (30/81, 35.8%) had a current diagnosis of social phobia compared with CG (11/73, 15.1%, P=.005). However, more CG participants had a past diagnosis of social J Med Internet Res 2025 | vol. 27 | e54062 | p. 5https://www.jmir.org/2025/1/e54062 (page number not for citation purposes) Lalk et alJOURNAL OF MEDICAL INTERNET RESEARCH XSL • FO RenderX phobia (4/73, 5.5%) than IG (0/81, 0%, P=.048). For both current generalized anxiety disorder (31/81, 38.3% vs 17/73, 23.3%, P=.06) and current major depression (8/81, 9.8% vs 1/73, 1.4%, P=.08) there was a trend toward statistical significance for higher occurrence in the IG. No other diagnosis differences were observed (OSM 6 in Multimedia Appendix 2). Current psychopharmacology (P=.81) and psychotherapy (P=.34) also did not differ between groups. The CG had lower anxiety levels at baseline for the BAI (t153.8=2.30, P=.02). No group differences were found in primary or secondary outcomes. Participants’Characteristics The sociodemographic characteristics of all participants are displayed in Table 1. Altogether, out of 194 participants, 59 (38.3%) were diagnosed with panic disorder, whereas 96 (61.7%) fulfilled the diagnostic criteria of panic disorder with additional agoraphobia. Regarding comorbidities, 47 (30.1%) participants also fulfilled the diagnostic criteria for generalized anxiety disorder, which was the most prominent comorbid disorder (OSM 6 in Multimedia Appendix 2). Comorbid depression rates were relatively low, with only 9 (5.8%) participants fulfilling the diagnostic criteria for major depressive disorder. It is important to note that the low comorbidity rate can be attributed to this study’s exclusion criteria, which specifically excluded individuals with severe depression. Table 1. Sociodemographic characteristics of the study cohort at baseline. Total sample (N=156)Control (n=74)Treatment (n=82)Characteristics Sex, n (%) 121 (77.56)56 (75.68)65 (79.27)Female 33 (21.15)17 (22.97)16 (19.51)Male 2 (1.29)1 (1.35)1 (1.22)Nonbinary 35.0 (11.3)35.0 (11.1)35.1 (11.5)Age in years (mean, SD) Health care use, n (%) 61 (39.10)26 (35.14)35 (42.68)Psychotherapy 50 (32.05)23 (31.08)27 (32.93)Pharmacotherapy Relationship statusa, n (%) 49 (31.21)26 (35.14)23 (27.71)Married 3 (1.91)0 (0)3 (3.61)Not living with partner 96 (61.15)45 (60.81)51 (61.45)Single 8 (5.10)3 (4.05)5 (6.02)Divorced 1 (0.64)0 (0)1 (1)Widowed Children, n (%) 31 (19.87)19 (25.68)12 (14.63)Yes 125 (80.13)55 (74.32)70 (85.37)No Current diagnoses, n (%) 59 (38.3)32 (43.8)27 (33.3)Panic disorder without agoraphobia 96 (61.7)41 (56.2)54 (66.7)Panic disorder with agoraphobia 48 (31.2)17 (23.3)31 (38.3)Generalized anxiety disorder 9 (5.8)1 (1.4)8 (9.8)Major depressive disorder 40 (26)4 (5.5)29 (35.8)Social phobia aMultiple options were possible. Missing Data Noncompletion rates were 16.0% (25/84) for the PAS (first primary outcome) and 18.6% (29/72) for the WHO-5 (second primary outcome) at T3. Therefore, 70 out of 82 (85.4%) participants in the IG and 61 out of 74 (82.4%) participants in the CG completed the PAS at posttreatment. Similarly, 68 out of 84 (82.3%) participants in the IG and 59 out of 72 (79.7%) participants in the CG completed the WHO-5 post treatment. Logistic regression analyses showed that participant dropout for the WHO-5 was not associated with any of the following baseline variables: group allocation, sex, age, work capacity, medication intake, or the baseline values of any of the primary or secondary outcomes. For the PAS, dropout was not associated with group allocation, age, work capacity, psychotherapy, or J Med Internet Res 2025 | vol. 27 | e54062 | p. 6https://www.jmir.org/2025/1/e54062 (page number not for citation purposes) Lalk et alJOURNAL OF MEDICAL INTERNET RESEARCH XSL • FO RenderX the baseline values of any primary or secondary outcomes. However, female sex (P=.02) and medication intake (P=.05) were associated with reduced dropout for the PAS. Since none of the baseline outcomes were associated with dropout, the empirical data supported missing at random instead of missing not at random. Therefore, as determined in the study protocol, MICE analysis was conducted as our primary analysis. Adverse Events During the trial, 3 participants from the CG reported suicidality and had to be excluded from the study. Trained psychologists immediately contacted them. They were excluded from further data collection, but all previous data were still used. Adherence Due to a technical problem, usage data were missing for 1 person in the IG. Based on data from 81 participants, the IG completed an average of 7.3 (SD 3.9) modules out of the total 12 modules, amounting to 7.3 out of 12 (60.8%) of all modules. Overall, 26 out of 84 (32.1%) participants completed the whole course, and 73 out of 84 (90.1%) participants completed the first 4 modules, which was chosen as a sensitivity analysis to assess a basic amount of engagement. Primary Outcomes For the primary outcome PAS (t110.1=–2.22, P=.03), we found a significant time×group interaction effect, but not for the primary outcome WHO-5 (t149.8=–1.35, P=.09), as shown in Table 2. Effect sizes were small to moderate for the PAS (d=–0.37, 95% CI –0.70 to –0.04) and small for the WHO-5 Cohen (d=0.22; 95% CI –0.10 to 0.53). Within-group T1-T3 effect sizes for PAS were moderate in the IG Cohen (d=–0.58, 95% CI –0.77 to –0.39) and small in the CG Cohen (d=–0.21, 95% CI –0.42 to 0.00). Within-group effects were small to moderate in the IG Cohen (d=0.30, 95% CI 0.06 to 0.53) and minimal in the CG Cohen (d=0.08, 95% CI –0.11 to 0.28) for WHO-5. All imputations and plots can be seen in OSM 1-4 in Multimedia Appendix 2. Table 2. Bonferroni-Holm adjustment for the primary outcomes (multivariate imputation by chained equations imputed). Effect size d(95% CI)Adjusted PvalueAdjustment factorPvalue (1-sided)ttest (df)Primary outcome –0.37 (–0.70 to –0.04).032.01–2.22 (110.1) PASa 0.22 (–0.10 to 0.53).091.091.35 (149.8) WHO-5b aPAS: Panic and Agoraphobia Scale. bWHO-5: WHO (World Health Organization)-5 Well-Being Index. On average, severe levels of panic and agoraphobia were reported at baseline (mean 34.56, SD 8.31; scores range from 13 to 66; Table 3). Moreover, low well-being was reported at baseline (mean 2.87, SD=0.97; scores range from 0.0 to 4.2; Table 3). Table 3. Intention-to-treat data for the Panic and Agoraphobia Scale and the WHO (World Health Organization)-5 Well-Being Index. T3T2T1Imputation Mean (SD)NMean (SD)NMean (SD)N ITTaPASb 29.9 (7.90)7032.3 (8.41)6635.1 (7.93)82Treatment 32.7 (8.77)6133.7 (7.84)6233.9 (8.73)74Control ITT WHO-5c 3.12 (1.06)683.01 (1.01)642.83 (0.94)82Treatment 2.99 (1.04)593.03 (1.02)602.92 (1.00)74Control aITT: intention-to-treat. bPAS: Panic and Agoraphobia Scale. cWHO-5: WHO (World Health Organization)-5 Well-Being Index. Minimal Clinical Important Difference The Reliable Change Index (RCI; [58]) was used to calculate reliable improvement or deterioration. Regarding the PAS, 33.6% of IG and 16.8% of CG patients improved reliably from T1 to T3. In contrast, 6.1% deteriorated in the IG and 9.2% in the CG. Therefore, a significant difference was found (P<.001). For the WHO-5, improvement (28.3% vs 11.4%) favored the IG. However, deterioration (12.7% vs 11.4%) was stronger in the CG. A significant difference was identified (P=.002). Sensitivity Analyses for Per-Protocol Sample (4 Completed Modules) In addition, per-protocol sensitivity analyses were conducted for both primary outcomes, including only the 73 (90.1%) participants who completed at least the first 4 modules. Because J Med Internet Res 2025 | vol. 27 | e54062 | p. 7https://www.jmir.org/2025/1/e54062 (page number not for citation purposes) Lalk et alJOURNAL OF MEDICAL INTERNET RESEARCH XSL • FO RenderX no hypotheses were specified, the effects are calculated as 2-tailed tests without alpha adjustment. A significant group×time interaction (t252=–3.25, P=.001) was found for the PAS with a moderate effect size Cohen (d=–0.50; 95% CI –0.80 to –0.20). Regarding the WHO-5, no significant interaction was found (t252=1.80, P=.07). Secondary Outcomes None of the interaction effects of the secondary outcomes were significant after the Bonferroni-Holm adjustment (Table 4). On average, impairment of daily functioning (WSAS) was moderate at baseline (mean 4.1, SD 1.7), while mental health literacy (MHLS) levels were very high (mean 4.4, SD 0.4). iPCQ scores and CSSRI outpatient treatment were log-transformed due to the right-skewed distribution (Table 5). Inpatient (CSSRI partly inpatient) and complementary (CSSRI complementary) treatment occurred for 15.3% and 12.1% of participants at baseline (Table 6). Table 4. Linear mixed model and Bonferroni-Holm adjustment for the secondary outcomes based on multiple imputation by chained equations imputation (T1-T3). Effects are odds ratios. The additional factors are due to the additional alpha adjustments of the client sociodemographic and service receipt inventory subscales. Effect size d(95% CI)Group×timeOutcome Adjusted PvalueAdjustment factorPvalue (1-sided)ttest (df) –0.22 (–0.48 to 0.03).174.04–.72 (105.1) WSASa –0.09 (–0.34 to 0.15).452.22–0.76 (231.6) MHLSb –0.21 (–0.51 to 0.09).263.09–1.37 (206.8) iPCQc 0.07 (–0.22 to 0.37)1.0 1×1e 1.00.48 (231.6) CSSRIdoutpatient 0.24f(0 to 30.16) .88 1×3e .29–0.56 (6.2)CSSRI partly inpatient 0.16f(0 to 28134) .78 1×2e .39–0.30 (6.5)CSSRI complementary aWSAS: Work and Social Adjustment Scale. bMHLS: Mental Health Literacy Scale. ciPCQ: iMTA Productivity Cost Questionnaire. dCSSRI: Client Sociodemographic and Service Receipt Inventory. eThe additional factors are due to the additional α adjustments of the CSSRI subscales. fEffects are odds ratios. J Med Internet Res 2025 | vol. 27 | e54062 | p. 8https://www.jmir.org/2025/1/e54062 (page number not for citation purposes) Lalk et alJOURNAL OF MEDICAL INTERNET RESEARCH XSL • FO RenderX Table 5. Intention-to-treat data for the secondary outcomes Work and Social Adjustment Scale, Mental Health Literacy Scale, iMTA Productivity Cost Questionnaire, and client sociodemographic and service receipt inventory outpatient. multiple imputations by chained equations. T3T2T1Outcome Mean (SD)NMean (SD)NMean (SD)N WSASa 3.55 (1.68)823.89 (1.72)824.16 (1.67)82Treatment 3.90 (1.86)743.87 (1.68)744.12 (1.73)74Control MHLSb 4.35 (0.45)824.37 (0.40)824.38 (0.39)82Treatment 4.38 (0.42)744.36 (0.39)744.38 (0.37)74Control iPCQclog 0.40 (0.65)820.45 (0.63)820.50 (0.67)82Treatment 0.48 (0.74)740.49 (0.68)740.44 (0.64)74Control CSSRIdoutpatient log 0.95 (1.14)821.07 (1.12)821.25 (1.13)82Treatment 1.15 (1.11)741.19 (1.09)741.53 (1.06)74Control aWSAS: Work and Social Adjustment Scale. bMHLS: Mental Health Literacy Scale. ciPCQ: iMTA Productivity Cost Questionnaire. dCSSRI: Client Sociodemographic and Service Receipt Inventory. Table 6. Intention-to-treat data for the secondary outcomes client sociodemographic and service receipt inventory partly inpatient and client sociodemographic and service receipt inventory complementary. multiple imputations by chained equations. T3 cases, n (%)T2 cases, n (%)T1 cases, n (%)Outcome CSSRIapartly inpatient dichotomized 82 (10.7)82 (18.3)82 (16.3)Treatment 74 (16.8)74 (14.9)74 (18.6)Control CSSRI complementary dichotomized 82 (9.3)82 (13.7)82 (15.9)Treatment 74 (9.7)74 (9.5)74 (13)Control aCSSRI: Client Sociodemographic and Service Receipt Inventory. Additional Outcomes For the BAI, a significant interaction was found at T3 (t206.8=–4.12, P<.001) with a moderate to large effect size Cohen (d=–0.60 95% CI –0.89 to –0.32). Within-group effects (T1-T3) were large in the treatment group Cohen (d=-0.82 95% CI –1.05 to –0.60) and small in the IG Cohen (d=–0.22 95% CI –0.39 to –0.05). For the PHQ-9, also a significant interaction was found (t257.4=–3.20, P<.001) with a small to moderate effect size Cohen (d=–0.41 95% CI –0.66 to –0.16). Within-group effects (T1-T3) were small Cohen (d=–0.25 95% CI –0.42 to –0.09) in the IG and showed minimal to small deterioration in the CG Cohen (d=0.15 95% CI –0.03 to 0.34). Regarding adverse effects (NEQ), no difference was found between the groups (t=–1.14, P=.26; Table 7 and OSM 5 in Multimedia Appendix 2). J Med Internet Res 2025 | vol. 27 | e54062 | p. 9https://www.jmir.org/2025/1/e54062 (page number not for citation purposes) Lalk et alJOURNAL OF MEDICAL INTERNET RESEARCH XSL • FO RenderX ©Christopher Lalk, Teresa Väth, Sofie Hanraths, Luise Pruessner, Christina Timm, Steffen Hartmann, Sven Barnow, Julian Rubel. Originally published in the Journal of Medical Internet Research (https://www.jmir.org), 02.04.2025. This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research (ISSN 1438-8871), is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included. J Med Internet Res 2025 | vol. 27 | e54062 | p. 16https://www.jmir.org/2025/1/e54062 (page number not for citation purposes) Lalk et alJOURNAL OF MEDICAL INTERNET RESEARCH XSL • FO RenderX