Introduction
Depressive and anxiety disorders represent one of the greatest burdens among human diseases worldwide (World Health Organization, 2017). These emotionally difficult conditions often manifest as comorbidities (Kessler et al., Reference Kessler, Berglund, Demler, Jin, Merikangas and Walters2005; Lamers et al., Reference Lamers, van Oppen, Comijs, Smit, Spinhoven, van Balkom and Penninx2011). Advances in understanding the genetic, biological, neurological, and psychological aspects of psychopathologies have indicated that transdiagnostic or higher-order features are shared across a variety of mental disorders (Caspi & Moffitt, Reference Caspi and Moffitt2018; Grisanzio et al., Reference Grisanzio, Goldstein-Piekarski, Wang, Rashed Ahmed, Samara and Williams2018; Insel et al., Reference Insel, Cuthbert, Garvey, Heinssen, Pine, Quinn and Wang2010; Krueger et al., Reference Krueger, Kotov, Watson, Forbes, Eaton, Ruggero and Zimmermann2018; McGorry, Hartmann, Spooner, & Nelson, Reference McGorry, Hartmann, Spooner and Nelson2018; Sloan et al., Reference Sloan, Hall, Moulding, Bryce, Mildred and Staiger2017; Watson, Reference Watson2005). In parallel with the transdiagnostic understanding of psychopathology across ‘emotional disorders’ [i.e. major depressive disorder (MDD), generalized anxiety disorder (GAD), panic disorder (PD), social anxiety disorder (SAD), other anxiety-related disorders, post-traumatic stress disorder (PTSD), and obsessive-compulsive disorder (OCD)], accumulating evidence indicates that the transdiagnostic approach for treating these disorders is safe, feasible, and efficient (Newby, McKinnon, Kuyken, Gilbody, & Dalgleish, Reference Newby, McKinnon, Kuyken, Gilbody and Dalgleish2015). However, non-negligible methodological issues have been observed in the clinical trials of transdiagnostic psychotherapies (Fusar-Poli et al., Reference Fusar-Poli, Solmi, Brondino, Davies, Chae, Politi and McGuire2019).
The unified protocol of transdiagnostic treatment for emotional disorders (UP) is an emotion-focused treatment consisting of the intervention modules derived from cognitive behavioral therapy (CBT) (Barlow et al., Reference Barlow, Farchione, Fairholme, Ellard, Boisseau, Allen and Ehrenreich-May2011). These treatment modules target the patients' negative emotionality and aversive reactions to emotions when they occur. A large clinical trial (n = 225) showed equivalent efficacy of the individual format of UP compared to established disorder-specific CBT and superior efficacy compared to wait-list conditions in patients with PD, SAD, GAD, and OCD (Barlow et al., Reference Barlow, Farchione, Bullis, Gallagher, Murray-Latin, Sauer-Zavala and Cassiello-Robbins2017). The evidence for the feasibility of UP has accumulated for multiple formats (e.g. individual, group, and online) among diverse clinical populations (e.g. patients with anxiety, obsessive-compulsive, trauma-related, depressive, bipolar, borderline personality, and eating disorders) in various settings (e.g. public health services, ambulatory care, and inpatient care) and among different countries (e.g. the US, Japan, Brazil, Spain, Iran, Denmark, and Romania) (Barlow & Farchione, Reference Barlow and Farchione2017; Cassiello-Robbins, Southward, Tirpak, & Sauer-Zavala, Reference Cassiello-Robbins, Southward, Tirpak and Sauer-Zavala2020; Sakiris & Berle, Reference Sakiris and Berle2019).
Although UP is hypothesized to be transdiagnostically effective for a range of mental disorders associated with deficits in emotion regulation, evidence on the efficacy of UP for patients with a principal diagnosis of depressive disorder is limited (Sauer-Zavala et al., Reference Sauer-Zavala, Bentley, Steele, Tirpak, Ametaj, Nauphal and Barlow2020). Only two randomized controlled trials have reported the efficacy of individual format of UP for patients with a principal diagnosis of depressive disorders (total n of UP sample = 25) (Bameshgi, Kimiaei, & Mashhadi, Reference Bameshgi, Kimiaei and Mashhadi2019; Marnoch, Reference Marnoch2014), whereas seven randomized controlled trials have reported the efficacy of individual format of UP for patients with a principal diagnosis of anxiety-related disorders (total n of UP sample = 251) (Cassiello-Robbins et al., Reference Cassiello-Robbins, Southward, Tirpak and Sauer-Zavala2020; Sakiris & Berle, Reference Sakiris and Berle2019). These two depression trials were conducted with nonblinded assessment and no fidelity rating of the intervention (Bameshgi et al., Reference Bameshgi, Kimiaei and Mashhadi2019; Marnoch, Reference Marnoch2014). Moreover, these two trials represented limited clinical populations (i.e. depressed women due to marital problems and people aged ⩾60 years). To date, no clinical trial has been conducted with a rigorous design (e.g. blinded independent evaluator assessment, allocation concealment, large sample size, and fidelity assessment) including both patients with a principal diagnosis of depressive disorder and those with a principal diagnosis of anxiety disorders. The randomized controlled trial of a group format of UP (n = 291) showed promising efficacy on well-being for patients with a principal diagnosis of MDD, SAD, and PD (Reinholt et al., Reference Reinholt, Hvenegaard, Christensen, Eskildsen, Hjorthøj, Poulsen and Arnfred2021). Along with the transdiagnostic theorization of UP, we hypothesized that the individual format of UP is effective for patients with either a principal diagnosis of depressive or anxiety disorders.
From a global mental health perspective, it is important to test the efficacy of transdiagnostic psychotherapy outside of Western countries. Efficacy study into psychosocial interventions has predominantly been limited to North America, Western Europe, and Australia (Patel et al., Reference Patel2007; Saxena, Thornicroft, Knapp, & Whiteford, Reference Saxena, Thornicroft, Knapp and Whiteford2007). The lack of evidence from other regions may be among the reasons for the scarcity of psychotherapy dissemination in other parts of the world. Transdiagnostic psychotherapies, such as UP, have several potential advantages over disorder-specific treatment, especially for those countries with limited recourses and sparse efficacy data (Martin, Murray, Darnell, & Dorsey, Reference Martin, Murray, Darnell and Dorsey2018; Murray, Metz, & Callaway, Reference Murray, Metz, Callaway, Stein, Bass and Hofmann2019). Potential strengths of the transdiagnostic approach include applicability to diverse disorders, effective and efficient strategies for dealing with comorbidities, simplification of treatment models for multiple emotional disorders, ease of learning and training for novice therapists, reduction of confusion around what evidence-based treatment to choose, capitalization on the similarities across diagnoses and treatments within existing disorder-specific treatments, and cost reduction in comparison to training practitioners in multiple treatment models (McHugh & Barlow, Reference McHugh and Barlow2010; Murray et al., Reference Murray, Metz, Callaway, Stein, Bass and Hofmann2019). Owing to these potential benefits, the transdiagnostic approach is expected to improve mental healthcare in low, middle, and high-income countries (Martin et al., Reference Martin, Murray, Darnell and Dorsey2018; Murray et al., Reference Murray, Metz, Callaway, Stein, Bass and Hofmann2019).
Hence, we aimed to test whether the addition of UP to treatment-as-usual (UP-TAU) would be more efficacious than wait-list with treatment-as-usual (WL-TAU) for reducing depressive symptoms, as assessed by the GRID-Hamilton depression rating scale (GRID-HAMD) (Williams et al., Reference Williams, Kobak, Bech, Engelhardt, Evans, Lipsitz and Kalali2008), in Japanese patients with a principal diagnosis of depressive or anxiety disorders.
Materials and methods
Design
This study was designed as a single-centre, assessor-blinded, randomized, parallel-group superiority trial with a target sample of 104 patients with depressive and/or anxiety disorders with a primary time point of 21 weeks and a secondary follow-up period of 43 weeks. The participants were recruited from December 2013 to March 2018. The detailed study protocol, including the rationale of the comparison group, a detailed description of study settings, and the procedures, has been previously published (Ito et al., Reference Ito, Okumura, Horikoshi, Kato, Oe, Miyamae and Ono2016). This trial was originally registered at ClinicalTrials.gov (NCT02003261) and later at the UMIN clinical trial registry (UMIN000030708) to comply with the change in the Japanese Ethical Guideline.
Participants
The inclusion criteria were as follows: (1) a DSM-IV diagnosis of MDD, dysthymia, depressive disorder not otherwise specified, PD with agoraphobia, PD without agoraphobia, agoraphobia without history of PD, SAD, OCD, PTSD, GAD, or anxiety disorder not otherwise specified, as assessed by a Structured Clinical Interview for DSM-IV-TR Axis I Disorders (SCID-IV-TR) (First, Spitzer, Gibbon, & Williams, Reference First, Spitzer, Gibbon and Williams1997); (2) mild or severe depressive symptoms (GRID-HAMD ⩾8); (3) aged 20–65 years old; and (4) provision of full informed consent to participate in the study. The exclusion criteria were as follows: (1) alcohol or substance abuse disorder in 6 months prior to the baseline; (2) current manic episode, schizophrenia, or other psychotic disorder; (3) serious suicidal ideation; (4) life-threatening, severe, or unstable physical disorders or major cognitive deficits; (5) evidence of an inability to participate in half or more of the intervention phase; (6) structured psychotherapy; and (7) other relevant reasons. We used the SCID-IV-TR to assess diagnostic status because the Japanese version of the DSM-5 and its structured instrument (e.g. SCID-5) had not been published. We included structured psychotherapy (i.e. psychotherapy following a specific pre-determined protocol, such as CBT) among the exclusion criteria to exclude possible effects due to other treatments.
Six independent evaluators, who were blinded to the allocation, conducted the SCID-IV-TR, GRID-HAMD, and other rating scales. All evaluated assessments were recorded, and 20% of the completed assessments were randomly selected to calculate the intraclass correlation coefficient using a two-way random-effects model for absolute agreement. Independent evaluators were unaware as to which interview sessions were to be re-evaluated for inter-rater consistency.
Randomization and blinding
We employed central randomization using the Allocation and Registration Control System computer software, which was developed, set up, and managed by an independent institution (i.e. Keio University). A random sequence was generated using minimization, with a ratio of 1:1 to balance stratified factors (i.e. the principal diagnosis of depressive v. anxiety disorder). Allocation was implemented by the primary investigator (MI) or research coordinator via the internet using a laptop in front of the eligible participant. The independent evaluators responded to the modified version of the Independent Evaluator Knowledge of Treatment scale at mid- (10 weeks), post- (21 weeks), and follow-up assessment (43 weeks).
Intervention
UP
UP is an emotion-focused CBT intervention that targets core mechanisms shared across depressive and anxiety disorders (Barlow et al., Reference Barlow, Farchione, Fairholme, Ellard, Boisseau, Allen and Ehrenreich-May2011). Standard administration comprises 12–18 weeks of 50–60 min weekly sessions. Participants learn cognitive behavioral skills to regulate their emotions by practicing the core components of UP, such as monitoring their emotional experience, mindful emotional awareness, cognitive flexibility, countering emotional behaviors, and interoceptive and emotional exposures. Four clinical psychologists conducted the individual format of UP (two females and two males, all with 4–6 years of clinical practice after certification). Treatment adherence was monitored by weekly group supervision and assessed using the Treatment Adherence Scale for UP (Barlow et al., Reference Barlow, Farchione, Bullis, Gallagher, Murray-Latin, Sauer-Zavala and Cassiello-Robbins2017). To assess treatment adherence and fidelity, we selected 25% of the total expected sessions. This random sampling was blinded to the therapists. The overall adherence was calculated as the proportion of non-deviation from the treatment procedure in the assessed session. The treatment fidelity was assessed by the overall session rating, which ranged from poor to excellent (0–5). The intervention period was set at 20 weeks to allow participants to skip sessions in case of unforeseen circumstances (e.g. catching a cold) and to allow them to practice the UP skills independently with longer intervals between sessions later in the treatment period.
TAU
The Japanese Society of Mood Disorders has published treatment guidelines for MDD that recommend pharmacotherapy for moderate or severe MDD patients and pharmacotherapy and psychotherapy for mild MDD patients (Japan Society for Mood Disorders Treatment, 2013). However, there are no treatment guidelines for anxiety disorders specific to a Japanese clinical setting. In this clinical trial, we defined TAU as any pharmacological or psychological intervention and clinical management except structured psychotherapy and electroconvulsive therapy. For ethical reasons and to maintain clinical relevance specific to the Japanese clinical setting, all participants received TAU without any restriction on routine psychiatric treatment, including changes to the dose and types of psychotropic medicines. The primary psychiatrists who provided TAU did not blinded to the treatment allocation.
Measures
Primary outcome
The primary outcome was depressive symptoms assessed by the 17-item version of GRID-HAMD at 21 weeks. We used GRID-HAMD because it assesses not only depressive symptoms (i.e. depressed mood, guilt, suicide, insomnia, difficulties in work and activities, psychomotor retardation or agitation, loss of appetite, sexual interest, weight, or insight) but also anxiety-related symptoms (i.e. psychic and somatic anxiety, general somatic symptoms, and hypochondriasis), which reflect common emotional symptoms across depressive and anxiety disorders (range: 0–52). The GRID-HAMD has been reported to exhibit excellent inter-rater reliability with an intraclass correlation coefficient of 0.95–0.99 and acceptable internal consistency as shown by a Cronbach's alpha of 0.78 (Williams et al., Reference Williams, Kobak, Bech, Engelhardt, Evans, Lipsitz and Kalali2008). The excellent inter-rater reliability has been previously reported for the Japanese clinical population (Tabuse et al., Reference Tabuse, Kalali, Azuma, Ozaki, Iwata, Naitoh and Furukawa2007).
Secondary outcomes and other measures
Secondary outcomes included the severity of anxiety assessed by the Structured Interview Guide for Hamilton Anxiety Rating Scale (SIGH-A) (Shear et al., Reference Shear, Vander Bilt, Rucci, Endicott, Lydiard, Otto and Frank2001), overall severity and improvement rated by the clinical global impression (CGI-S, CGI-I) (Guy, Reference Guy1976), responder status defined by a ⩾50% reduction in the GRID-HAMD score, remission of symptoms defined as a GRID-HAMD score of <8 (Frank et al., Reference Frank, Prien, Jarrett, Keller, Kupfer, Lavori and Weissman1991), and loss of a principal diagnoses as assessed by SCID-IV-TR at baseline. The reliability and validity for the SIGH-A have been previously demonstrated for the Japanese clinical population (Yamamoto, Aizawa, Inagaki, & Inada, Reference Yamamoto, Aizawa, Inagaki and Inada2012). These secondary measures (except for loss of diagnosis) were assessed at 10, 21, and 43 weeks. We cautiously assessed any adverse events using the Japanese version of the Common Terminology Criteria for Adverse Events (CTCAE v4.0) (US Department of Health & Human Services, 2009). The therapists or independent evaluators actively solicited patients at each visit to assess the occurrence or exacerbation of any adverse symptoms as well as their severity, duration, and relation to the study. A ‘serious’ adverse event included death, life-threatening, admission to the hospital, prolongation of hospitalization, disability, permanent damage, or a congenital anomaly/birth defect. The Credibility/Expectancy Questionnaire was administered at session 2 to assess the treatment expectancy (Devilly & Borkovec, Reference Devilly and Borkovec2000). The detailed psychometric properties of these measures for Western and Japanese populations have been described in a study protocol (Ito et al., Reference Ito, Okumura, Horikoshi, Kato, Oe, Miyamae and Ono2016).
Sample size estimation
We initially set our minimum sample size to a total of 54 based on an assumed effect size of −0.85 for UP-TAU v. WL-TAU for the reduction of the primary outcome assessed by GRID-HAMD [see detailed discussion by Ito et al. (Ito et al. Reference Ito, Okumura, Horikoshi, Kato, Oe, Miyamae and Ono2016)]. However, in the middle of the ongoing trial, the original developer laboratory of the UP at Boston University reported that the effect size of the UP against the wait-list condition on the HAMD at post-treatment was −0.69 (95% CI −1.06 to −0.31) in the definitive clinical trial of UP (Farchione et al., Reference Farchione, Ghallager, Sauer-Zavala, Latin, Boswell and Bentley2016). In response to this report, we re-estimated our sample size to detect a more conservative between-group effect size of −0.60 with a statistical power of 80% and a significance level of 0.05 with the two-tailed test. This resulted in a required sample size of 45 for each group. Considering the reported dropout rate of 15.39% (Farchione et al., Reference Farchione, Fairholme, Ellard, Boisseau, Thompson-Hollands, Carl and Barlow2012), we required at least 52 participants in each group to test the primary hypothesis of this study. This protocol amendment without any interim analysis was discussed with an ethicist in medicine, a biostatistician, and a psychiatrist with expertise in clinical research, all of whom belonged to the Department of Clinical Research Support at the NCNP, and was approved by the institutional review board (original approval number: A2013-092; approval number for the modified protocol: 2016-065), and timely changed in the records of ClinicalTrial.Gov (NCT02003261).
Statistical analysis
All analyses for testing the efficacy with the primary and secondary measures were analyzed based on the intent-to-treat principle using a linear mixed model (LMM). For the primary outcome, the dependent variable was the GRID-HAMD score and the independent variables were assignment (i.e. UP-TAU v. WL-TAU), time (i.e. 0, 10, and 21 weeks), and interaction between the assignment and time as fixed-effect variables and participants as a random effect variable. We constructed a conditional growth model (Singer & Willett, Reference Singer and Willett2003) using a restricted maximum likelihood estimation method to compare changes in depression severity between groups from baseline to 21 weeks. The assessment period comprised the measurement time for the growth model (i.e. 0, 10, and 21 weeks). We used unstructured error covariance for this model. To test robustness, we conducted the same LMM, including a stratified variable (depressive v. anxiety disorder) as a covariate, and other sensitivity analyses taking into account the difference between intention-to-treat and per-protocol analysis (Thabane et al., Reference Thabane, Mbuagbaw, Zhang, Samaan, Marcucci, Ye and Goldsmith2013). Continuous secondary outcomes (SIGH-A, CGI-S, and CGI-I) were analyzed in the same way as primary outcomes. The LMM were conducted using nlme package (Version: 3.1-142) in R version 3.6.2. Standardized effect size was calculated because it is preferable in comparison to simple effect size when comparing conceptually similar effects using different units of measurement (Baguley, Reference Baguley2009). Dichotomous secondary outcomes, such as responder status, remission status, and loss of principal diagnosis were analyzed using the incidence proportion difference and ratio.
Missing data were imputed with multiple imputations by chained equation using mice package (3.7.0) in R version 3.6.2 (van Buuren & Groothuis-Oudshoorn, Reference van Buuren and Groothuis-Oudshoorn2011). The percentage of missing values for GRID-HAMD, SIGH-A, CGI-S, CGI-I, and loss of diagnosis were between 6.5% and 10.1%. All except one missing value was observed due to patient dropout. We did not observe any relationship between worsened or improved outcomes and missing data. Based on the assumption of missing at random, the results across 100 imputed datasets were combined by following Rubin's rule. We constructed multiple regression models, which included 15 demographic and clinical variables (e.g. gender, current occupation, principal diagnosis, months from first psychiatric appointment, and previous psychiatric hospitalizations). For all analyses, p < 0.05 was considered statistically significant. Maintenance of and/or continuing improvement were examined by calculating the within effect size from post-treatment (at 21 weeks) to follow-up (at 43 weeks). We calculated the occurrence of adverse events during the intervention and follow-up period. These analytic strategies were described in the published study protocol (Ito et al., Reference Ito, Okumura, Horikoshi, Kato, Oe, Miyamae and Ono2016); however, the statistical analysis plan was not pre-registered in any open repository.
Blinded data interpretation
We employed the blinded interpretation procedure to avoid interpretation bias for trial results (Järvinen et al., Reference Järvinen, Sihvonen, Bhandari, Sprague, Malmivaara, Paavola and Guyatt2014). In brief, the epidemiologist who was blinded to randomization conducted statistical analysis for the primary and secondary outcomes. The results that used the labels group ‘U’ and group ‘P’ were provided to the principal investigator. The steering committee conducted a blinded interpretation meeting. The consensus was documented in June 2020. Then, an external reviewer, who had experience in randomized controlled trials for psychotherapy and was not involved in any aspect of this study, examined the documents. After this external validation, the data manager broke the randomization code. Records of the blinded data interpretation meeting are provided in the online Supplementary materials.
Results
Description of the participants
Eligibility was assessed for 125 participants, of which 11 did not meet the inclusion criteria, and 10 met the exclusion criteria (Fig. 1). Table 1 shows the demographic and clinical characteristics at baseline. The mean age of the eligible participants was 37.4 (s.d. = 11.5). More than half of the participants (n = 62, 59.6%) met ⩾1 comorbid diagnoses. Furthermore, 26 out of the 54 participants (48.2%) with a principal diagnosis of depressive disorder had comorbidity with any anxiety disorder, whereas 26 out of the 50 participants (52.0%) with a principal diagnosis of anxiety disorder had comorbidity with any depressive disorder. The time range from the first psychiatric appointment was 2–498 months (mean = 93.8, s.d. = 93.0, first quartile = 16.75, median = 77, third quartile = 133.25).
Fig. 1. Participant flow diagram of the study.
Table 1. Baseline demographic and clinical characteristics
Data are presented as the number (proportion) of patients unless otherwise indicated. UP-TAU, Unified Protocol with Treatment As Usual; WL-TAU, Wait-list with Treatment As Usual; SSRI, Selective Serotonin Reuptake Inhibitors; SNRI, Serotonin Norepinephrine Reuptake Inhibitors; NaSSA, Noradrenergic and Specific Serotonergic Antidepressant.
Attrition, treatment expectancy, and treatment adherence
Of the 52 participants, 49 completed the UP intervention (94.2%), while 47 participants maintained the WL-TAU condition (90.4%). The mean number of completed sessions was 15.15 (s.d. = 3.10) for the UP-TAU group. The median treatment duration was 133.0 (first quartile = 125.8, third quartile = 140.0) days. The total number of sessions among the UP-TAU group was 788. Of these sessions, 309 assessed for adherence and 318 assessed for the fidelity; 217 sessions (87.4%) completely adhered to the treatment procedure. The mean treatment fidelity score was 4.01 (s.d. = 0.54). The mean total score of treatment expectancy in the UP-TAU group was 63.46 (s.d. = 18.19; possible range, 0–100).
Assessment integrity and blinding success
To examine the inter-evaluator reliability of GRID-HAMD, 86 out of 406 (21.18%) randomly selected assessments were used. The single-measure inter-rater intraclass correlation coefficient (Model 2.1) between two evaluators was 0.96 (95% CI 0.94–0.98) for the total GRID-HAMD score. The intraclass correlation coefficient at the item level of GRID-HAMD (n = 1462) was 0.95 (95% CI 0.94–0.95).
The true treatment assignment was accurately assumed in 64 out of 98 cases at 10 weeks (65.3% correct, χ2 = 9.87, p < 0.01, 6 missing), 67 out of 96 cases at 21 weeks (69.8% correct, χ2 = 16.48, p < 0.01, 8 missing), and 58 out of 86 cases at 43 weeks (67.4% correct, χ2 = 13.70, p < 0.01, 18 missing). The number of ‘not at all sure’ responses of the assessors' certainty rating for their assumption was 69 out of 98 cases (70.4%) at 10 weeks, 71 out of 96 cases (74.0%) at 21 weeks, and 56 out of 86 cases (65.1%) at 43 weeks.
Primary outcome
Table 2 shows the mean and standard deviation of GRID-HAMD at the baseline, 10-week mid-assessment, 21-week post-assessment, and 43-week follow-up. Figure 2 shows the improvement in the primary outcome measure. The mean GRID-HAMD scores in the UP-TAU and WL-TAU groups were 16.15 (s.d. = 4.90) and 17.06 (s.d. = 6.46) at baseline and 12.14 (s.d. = 5.47) and 17.34 (s.d. = 5.78) at 21 weeks. As shown in Table 3, the LMM analysis showed a significant difference in the GRID-HAMD score over the primary time point of 21 weeks between the UP-TAU and WL-TAU groups (estimate = −3.99; 95% CI −6.10 to −1.87). The estimated changes from baseline to 21 weeks were −4.06 (s.d. = 4.92) for the UP-TAU group and −0.32 (s.d. = 5.64) for the WL-TAU group. The between-group standardized mean difference was −0.70 (95% CI −1.11 to −0.28).
Fig. 2. Course of patient improvement as measured with the GRID-Hamilton depression rating scale. Error bars represent 95% confidence intervals. UP, unified protocol; TAU, treatment-as-usual; WL, wait-list.
Table 2. Descriptive statistics of primary and secondary outcome measures by time point and treatment group
Data are presented as mean (standard deviation) for GRID-HAMD, SIGH-A, CGI-A, and CGI-I and number (proportion) for responder status, remission status, and loss of principal diagnosis. Participants allocated to WL-TAU received UP during the follow-up period. UP-TAU, Unified Protocol with Treatment As Usual; WL-TAU, Wait-list with Treatment As Usual; GRID-HAMD, GRID-Hamilton Rating Scale for Depression; SIGH-A, Structured Interview Guide for Hamilton Rating Scale for Anxiety; CGI-S, Clinical Global Impression Severity; CGI-I, Clinical Global Impression Improvement; Dx, diagnosis.
Table 3. Primary and secondary outcomes
UP-TAU, Unified Protocol with Treatment As Usual; WL-TAU, Wait-list with Treatment As Usual; GRID-HAMD, GRID-Hamilton Rating Scale for Depression; SIGH-A, Structured Interview Guide for Hamilton Rating Scale for Anxiety; CGI-S, Clinical Global Impression Severity; CGI-I, Clinical Global Impression Improvement.
Secondary outcomes
The descriptive statistics for secondary outcomes are shown in Table 2. As shown in Table 3, the LMM analysis showed significant differences in the changes of SIGH-A (estimate = −3.36; 95% CI −6.05 to −0.68), CGI-S (estimate = −0.87; 95% CI −1.27 to −0.48). The estimated change of SIGH-A and CGI-S from baseline to 21 weeks was −3.55 (s.d. = 6.60) and −0.98 (s.d. = 1.04) in the UP-TAU group, respectively. The LMM analysis showed a significant difference in CGI-I at 21 weeks (estimate = −0.80; 95% CI −1.45 to −0.16). The mean CGI-I was 2.98 (s.d. = 1.04) for the UP-TAU group at 21 weeks. The Cohen's d values (95% CI) for SIGH-A, CGI-S, and CGI-I were −0.50 (−0.90 to −0.09), −0.88 (−1.30 to −0.46), and −0.74 (−1.15 to −0.32), respectively.
For the dichotomous outcomes, eight out of 52 (15.4%) patients in the UP-TAU group and one out of 52 (1.9%) patients in the WL-TAU group exhibited a treatment response at 21 weeks, resulting in an incidence proportion ratio of 7.27 (95% CI 0.89–59.67). The remission status was met in 10 patients (19.2%) in the UP-TAU group and three patients (5.8%) in the WL-TAU group, resulting in an incidence proportion ratio of 3.02 (95% CI 0.83–10.95). A loss of the principal diagnosis was observed in one patient in the UP-TAU group and in no patients in the WL-TAU group, resulting in an incidence proportion ratio of 1.02 (95% CI 0.69–1.51).
Sensitivity analyses for controlling stratified variables and using per-protocol samples
The LMM analyses that included stratified variables (i.e. the principal diagnosis of depressive v. anxiety disorders) as covariates showed significant differences between the UP-TAU and WL-TAU groups regarding the changes in GRID-HAMD, SIGH-A, CGI-S, and CGI-I over the primary time point of 21 weeks (see online Supplementary materials, Table S1). The LMM analyses using per-protocol samples revealed consistent results, as significant differences were observed between the UP-TAU and WL-TAU groups regarding the changes in GRID-HAMD, SIGH-A, CGI-S, and CGI-I (see online Supplementary materials, Table S2).
Maintenance of treatment effect during the follow-up period
Descriptive statistics at follow-up are shown in Table 2. The post- to follow-up effect sizes of each outcome for the UP-TAU group were not significant for GRID-HAMD, SIGH-A, and CGI-I (see online Supplementary materials, Table S3). The significant within effect size in CGI-S showed further improvement in patients in the UP-TAU group (Hedges' g = −0.48; 95% CI −0.95 to −0.01).
Adverse events
Table 4 shows the occurrence of adverse events. No serious adverse events were observed for the UP-TAU or WL-TAU groups during the intervention period. However, eight out of 52 participants (15.4%) reported a total of 13 intervention-related adverse events during the intervention period of which 10 events were rated as mild, as defined by CTCAE (missing data = 3). The mean duration of adverse events was 24.15 (s.d. = 16.72; range: 5–70) days.
Table 4. Summary of adverse events
Participants allocated to WL-TAU received UP during the follow-up period. UP-TAU, Unified Protocol with Treatment As Usual; WL-TAU, Wait-list with Treatment As Usual; AE, Adverse Event; SAE, Serious Adverse Event.
Discussion
This is the first randomized controlled trial to test the efficacy of the UP that includes either a principal diagnosis of depressive or anxiety disorder. As hypothesized, the addition of the UP to TAU was more effective than waiting for the UP while receiving TAU with regards to improving the severity of depressive symptoms between baseline and post-treatment assessments. The UP-TAU group showed significant improvements compared with the WL-TAU group in the blinded assessor evaluation of anxiety, severity, and improvement of clinical global impression. In the UP-TAU group, improvements from the baseline to 21 weeks were maintained in all continuous outcomes at 43 weeks. The sensitivity analyses controlling the types of principal diagnosis and per-protocol sample analyses showed the robustness of these results. There was no significant difference regarding the treatment response, the remission status, and the loss of primary diagnosis at 21 weeks between the UP-TAU and WL-TAU groups. The proportion of dropout was low in the UP-TAU group (3/52, 5.8%). No serious adverse events and few non-serious adverse events occurred in the UP-TAU group during the intervention period.
The UP was effective for improving primary and most of the secondary outcomes. Compared to the largest clinical trial of the UP, our participants seemed to exhibit more severe depressive and anxiety symptoms (our trial: mean baseline GRID-HAMD = 16.15, s.d. = 4.90, mean SIGH-A = 21.12, s.d. = 5.97, Barlow et al., Reference Barlow, Farchione, Bullis, Gallagher, Murray-Latin, Sauer-Zavala and Cassiello-Robbins2017: mean HAMD = 11.55, s.d. = 7.02, mean SIGH-A = 17.06, s.d. = 8.50). Moreover, half of our participants exhibited comorbidity of both depressive and anxiety disorders. Our between-group effect sizes comparing wait-list at post-treatment (−0.70 for HAMD and −0.50 for SIGH-A) were consistent with Barlow's trial (−0.57 for HAMD and −0.53 for SIGH-A). Overall, our results extend the evidence regarding the efficacy of the UP not only to patients with a principal diagnosis of PD, SAD, GAD, and OCD but also to patients with a principal diagnosis of depressive disorders and with more severe depressive and anxiety symptoms. Our efficacy results were consistent with those of prior RCTs with similar designs (Bameshgi et al., Reference Bameshgi, Kimiaei and Mashhadi2019; Marnoch, Reference Marnoch2014). In addition, our results show that UP is effective for participants who have a prolonged history of receiving usual treatments. For example, approximately 6 years (median = 77 months) had passed since our participants had their first psychiatric appointment. Most of our participants had visited at least two other medical institutions before visiting our treatment site, and approximately 20% of them had a history of psychiatric hospitalizations [n = 19, 23.1% (UP-TAU), 13.5% (WL-TAU)].
Despite the efficacy evidence, our results suggest that the UP-TAU is not sufficient for most of our patients to fully recover from clinical conditions. The mean reduction in GRID-HAMD was 4.01, which is within the range of 1 s.d. at baseline and post-treatment. The mean CGI-S for the UP-TAU group at 43 weeks was 2.94, indicating that the patients were still ‘mildly ill.’ A CGI-S score of 3 means that the patient shows mild symptoms with minimal distress or difficulty in social and/or occupational function. Only one at 21 weeks and eight at 43 weeks out of 54 participants achieved the loss of principal diagnosis. Our participants in the UP-TAU group showed only 15.4% treatment response (i.e. ⩾50% GRID-HAMD score reduction) and 19.2% remission (i.e. a GRID-HAMD score of <8) at 21 weeks. Though these were improved at 43 weeks (26.9% and 34.6%, respectively), these were still lesser compared with the treatment response proportion of 41% (95% CI 38–43) and the remission (i.e. a GRID-HAMD score of <7) proportion of 26% (95% CI 20–33) in various psychotherapies (Cuijpers et al., Reference Cuijpers, Karyotaki, Ciharova, Miguel, Noma and Furukawa2021). However, our results of poor response proportion might partly be attributable to the participant characteristics in this study, because the proportion of response in WL condition was also poor compared with the results of meta-analysis (Our participant 5.5% v. mean of 92 trials 16%) (Cuijpers et al., Reference Cuijpers, Karyotaki, Ciharova, Miguel, Noma and Furukawa2021). Despite the relatively long treatment periods, these dichotomous outcomes were insignificant. Participants who showed a partial response to UP may require additional treatment for residual symptoms. Although our results support the efficacy of UP for depressive and anxiety disorders, there is room for improvement in the content or administration of UP to individual patients.
Our results provide additional evidence regarding the feasibility of the UP with TAU in a Japanese clinical setting. No serious and only a few (n = 13) non-serious study-related adverse events occurred for eight out of 52 patients in the UP-TAU group (15.4%) during the intervention period. This is within the range of expected undesired events of 5%–20% for psychotherapy patients (Linden & Schermuly-Haupt, Reference Linden and Schermuly-Haupt2014). The treatment fidelity score was good, which is consistent with Barlow's trial (4.44 v. 4.01 in our study). The rate of dropout in our study (5.8%) was low compared to previous trials (Barlow et al., Reference Barlow, Farchione, Bullis, Gallagher, Murray-Latin, Sauer-Zavala and Cassiello-Robbins2017; 12.5%, Marnoch, Reference Marnoch2014; 12.5%, Bameshgi et al., Reference Bameshgi, Kimiaei and Mashhadi2019; 11.8%). The mean expectancy scale score was consistent with the trial conducted by treatment developers (Sauer-Zavala et al., Reference Sauer-Zavala, Boswell, Bentley, Thompson-Hollands, Farchione and Barlow2018) [63.46 (our study) v. 65.71 (Barlow's study)]. The major approach for psychiatric treatment for depressive and anxiety disorders in Japan is prescribing psychotropic medications with very brief routine outpatient examinations. Unfortunately, the provision of CBT for depression did not increase in the 6 years after it was first covered by Japan's national insurance scheme in 2010 (Hayashi et al., Reference Hayashi, Yoshinaga, Sasaki, Tanoue, Yoshimura and Kadowaki2020). Several factors are thought to have contributed to the current situation. These include an insufficient number of trained CBT therapists. Transdiagnostic psychotherapies, such as UP, are expected to reduce the cost and burden of training practitioners in multiple evidence-based treatments (Martin et al., Reference Martin, Murray, Darnell and Dorsey2018; McHugh & Barlow, Reference McHugh and Barlow2010; Murray et al., Reference Murray, Metz, Callaway, Stein, Bass and Hofmann2019). Our results offer valuable insights on the possible applications of UP in global mental health, particularly its extended applicability to depressive disorders and the Asian population.
This study must be interpreted while considering the following limitations. First, although we intentionally included heterogeneous clinical populations to test the efficacy of UP for depressive and anxiety disorders, such broad inclusion of participants might hinder the interpretation of the results in comparison with previous clinical trials conducted under traditional diagnostic systems. Second, the statistical test did not indicate blinding success. The proportion of the evaluators' correct guesses was 69.8% at the primary time point. This is relatively higher than that of randomized controlled trials regarding schizophrenia and affective disorders (62.0%) (Baethge, Assall, & Baldessarini, Reference Baethge, Assall and Baldessarini2013), though 74% (71/96) of the assessment was conducted under the conditions in which the evaluator's guess was ‘not at all sure.’ Third, our comparison group consisted of the condition of waiting for the UP while receiving TAU. We selected this condition because it is the most clinically relevant comparison in our Japanese setting. However, the nocebo effect might have affected our results. Fourth, although our participants were clinically heterogeneous, they were homogenous regarding race, ethnicity, and nationality. We must be cautious regarding the generalizability across national or cultural backgrounds. Fifth, there are no long-term follow-up data. Sixth, this study was not powered to test multiple outcomes. Therefore, our results for secondary outcomes and our sensitivity analyses should be interpreted as explorative. Seventh, we only reported the results regarding the pre-determined primary and secondary outcomes; all these outcomes were assessed by blinded assessors.
Conclusion
This study provides evidence that the addition of UP on TAU can be efficacious for patients with depressive and/or anxiety disorders in reducing depression. Specifically, our results strengthen the evidence of UP for patients with a principal diagnosis of MDD. Because our results regarding treatment response, remission, and loss of diagnosis were weak, future studies need to investigate strategies to enhance these aspects, for example, by examining moderating and mediating factors.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291721005067
Acknowledgments
Authors would like to tell the appreciation to Ayako Kanie MD, PhD for general administrative support and recruiting participants, Yukiko Hanada MA and Eriko Mizokawa MA for coordinating the trial, Asami Komazawa PhD, Chiaki Nakayama PhD, Ryo Horita PhD for collecting data, Naotsugu Hirabayashi MD, PhD, Fumi Imamura MA, Haruka Matsuda MD, Nahoka Kobayashi MD, Aiichiro Nakajima MD, PhD, Yasue Mitamura MD for recruiting participants, Tsunakuni Ikka, PhD, Kazushi Maruo, PhD, and Norio Sugawara PhD for consulting the change of protocol regarding the re-estimation of needed sample size, Kyoko Akutsu for managing data set and conducting on-site monitoring, Koki Takagaki, PhD, for reviewing the minutes of blinded interpretation, and Ono Yutaka MD, PhD, and Atsuo Nakagawa MD, PhD, for serving as a member of data and safety monitoring boards. The authors respectfully would like to show appreciation to participants for their effort to participate in the trial.
Financial support
This study was supported by a Grant-in-Aid for Young Scientists (A) (25705018 and 17H04788) awarded to MI from the Japanese Society for the Promotion of Science, an NCNP Intramural Research Grant (24-4, 27-3, 30-2) for Neurological and Psychiatric Disorders, and by AMED under Grant Number JP20dk0307084. The sponsors had no role in the preparation of the data or manuscript.
Conflict of interest
Dr Masaya Ito received royalties from Shindan to Chiryo Sha for treatment manuals and related books for the UP, from Sogen sha, and from several publishing companies in Japan. He also received honorariums for workshops and supervisions regarding the UP and for workshops regarding CBT. He received consultant fees from TIS Inc. Dr Masaru Horikoshi received consultant fees from Mitsubishi Tanabe Pharma and TIS Inc. and a research grant from Mitsubishi Tanabe Pharma and Jolly Good Inc. Yasuyuki Okumura received personal fees from MSD K.K., Otsuka Pharmaceutical, Cando, Japan Medical Data Center, and Japan Medical Research Institute. Yasuyuki Okumura is an employee of the Real World Data, Co., Ltd. and the president of the Initiative for Clinical Epidemiological Research. The other authors declare no conflict of interest.
Ethical statement
The trial protocol has been reviewed and approved by the institutional review board of the National Center of Neurology and Psychiatry (accepted in October 2013; A2013-092). All participants provided written informed consent. The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
Open access
