Anxiety and depressive disorders, also known as emotional disorders (EDs) [1], have the highest prevalence rates among psychological disorders [2,3], and are associated with substantial costs [4,5] and disability [6,7]. In the past three decades, research efforts to decrease the burden of these disorders have led to the development and evaluation of cognitive behavioral treatments in randomized controlled trials (RCTs) for each ED (ie, disorder-specific protocols), such as depression [8,9] and several anxiety disorders, including generalized anxiety disorder (GAD) [10,11], panic disorder (PD) and agoraphobia (AG) [12], social anxiety disorder (SAD) [13], and obsessive compulsive disorder (OCD) [14].

Although there is a large body of evidence showing the efficacy and effectiveness of disorder-specific cognitive behavioral therapy (CBT), in the past 15 years, an increasing number of researchers have agreed that there are some problems that hinder the optimal deployment of these treatments. The main drawback of disorder-specific treatments stems from the high comorbidity rates observed among anxiety disorders and between anxiety disorders and depressive disorders, with comorbidity estimates for these disorders ranging between 40% and 80% [15,16]. Thus, because disorder-specific treatment protocols focus on treating a specific diagnosis, the accompanying comorbid disorders do not receive therapeutic attention [17,18]. This problem becomes clearer when we take into account research linking comorbidity to aspects such as greater severity [3], increased chronicity rates [19], and a worse clinical course [20]. Another problem with disorder-specific treatments is that subthreshold symptoms that do not meet diagnostic thresholds for a particular disorder, but might be important to treat, normally go untreated [17]. Similarly, these protocols do not address diagnoses that do not fit any specific category, despite their clinical relevance, that is, not otherwise specified (NOS) anxiety and depressive disorders [15]. Finally, each disorder-specific treatment requires the use of different handbooks and protocols, which increases the economic costs and the amount of training needed to gain the knowledge and skills necessary to cover the array of anxiety and depressive disorders [17].

Transdiagnostic treatments have emerged as an alternative to the traditional disorder-specific approach that has dominated CBT research for the past 30 years. Transdiagnostic treatments have been developed and tested in several RCTs for anxiety disorders [21-24] and anxiety and depressive disorders [25-27], and their number continues to grow. Moreover, the efficacy and effectiveness of transdiagnostic treatments have been shown in different meta-analytic reviews, comparing them with different control groups, such as waiting list, attention control, and treatment as usual (TAU) [28-31], with pooled effect sizes (Hedges g) in the medium to large range for overall measures of anxiety (0.65-0.82) and depression (0.79-0.84). Moreover, an additional meta-analysis reported equivalent effects of transdiagnostic treatments (g=1.06) and disorder-specific treatments (d=0.95) on anxiety outcomes [32]. However, these meta-analyses include a number of mixed studies of different orientations, such as theory-based transdiagnostic treatments and tailored CBT. Therefore, these studies make it difficult to determine the effectiveness of each orientation (eg, the effectiveness of theory-based transdiagnostic treatments). The effectiveness of transdiagnostic treatments has also been shown in a meta-analysis by García-Escalera et al [33], with pooled effect sizes of g=0.80 for anxiety and g=0.72 for depression. Unlike the abovementioned meta-analyses, this study has the particularity that it only included theory-based transdiagnostic treatments. The main characteristic of theory-based transdiagnostic treatments, also known as mechanistically transdiagnostic treatments or transdiagnostic treatments based on shared mechanisms [34], is that they are designed to address the common psychopathological processes underlying anxiety and depression. Among the mechanistically transdiagnostic treatments for EDs, the Unified Protocol (UP) [35,36] stands out as one of the most empirically supported transdiagnostic protocols for anxiety and depression [23,37,38]. The UP is a CBT transdiagnostic protocol developed to address the underlying psychopathological processes that are common to anxiety and depressive disorders, with a particular focus on neuroticism, (low) extraversion, and emotion dysregulation, which have been shown to play a key role in the onset and maintenance of these disorders [39,40]. Thus, the main goal of the UP is to teach patients strategies to regulate their emotions in a more adaptive way through the following core treatment modules: (1) present-focused emotional awareness, (2) cognitive flexibility, (3) identification and prevention of emotional avoidance patterns, (4) increasing awareness and tolerance to emotion-elicited physical sensations, and (5) graded (interoceptive and situational) exposure procedures. The overall efficacy of the UP was first shown in an RCT where it was compared with a waitlist control group [23] and, more recently, in a larger RCT where it was compared with well-established disorder-specific CBT protocols for anxiety disorders [37]. In addition, some research shows the long-term effects of the UP [41] and its ability to produce changes in the temperament dimensions of behavioral inhibition (BI) and behavioral activation (BA) [42]. BI and BA have been conceptualized as 2 neurological systems representing motivational tendencies that are sensitive to threat and reward environmental cues, respectively [43]. These aspects have been intimately linked to neuroticism and negative affect and extraversion and positive affect [42]. Another treatment approach that includes principles or components that could be useful to target EDs is dialectical behavioral therapy (DBT) [44,45]. DBT initially emerged as a theoretical model and a treatment approach for the treatment of suicidal behaviors and borderline personality disorder, with the general aim of teaching strategies to change patterns of emotion dysregulation [46]. Of the range of strategies, DBT places a special emphasis on increasing experiential awareness and acceptance (mindfulness what and how techniques), but it also includes behavioral strategies such as the opposite action, designed to diminish distress by engaging in behaviors or actions opposite to those associated with negative emotions [44]. More recently, DBT skills have been adapted and successfully applied to several anxiety and depressive disorders [47,48], which suggests that they could be used in a transdiagnostic manner to improve the symptomatology of these disorders. For example, DBT principles may be combined with other evidence-based components (eg, components of the UP) to strengthen their effectiveness. The strategy of integrating CBT principles and components from different evidence-based therapies and orientations is consistent with the notion of process-based CBT, defined by Hofmann and Hayes [49]. As the authors state, “modern CBT places much less focus on protocols for syndromes and more focus on evidence-based processes linked to evidence-based procedures” [49].

In the past two decades, one of the most evident efforts made by researchers has been to take advantage of the possibilities offered by information and communication technologies to improve the assessment and treatment of psychological disorders. A clear example would be the use of the internet to increase the dissemination of empirically supported psychological treatments to anyone in need [50]. Research has shown promising and compelling evidence that internet-delivered psychological interventions are effective for a variety of psychosocial problems, including anxiety and depressive disorders [51,52]. The main advantages of internet-delivered treatments over traditional delivery methods (eg, face-to-face therapy) include widespread access and dissemination [53], a nonstigmatizing way of receiving psychological treatment [54], and increased cost-effectiveness [55].

Research has shown that transdiagnostic internet-delivered treatments are more effective than control groups [29] and that these treatments are at least as effective as individual and group face-to-face transdiagnostic treatments [33]. However, most of the existing literature on transdiagnostic treatments is limited to studies conducted in community settings, with few studies carried out in public contexts such as primary or specialized care [21]. Indeed, to our knowledge, no transdiagnostic internet-delivered treatments for anxiety and depression have been conducted in specialized public mental health care. This is somewhat surprising because transdiagnostic internet-delivered treatments in this particular setting could have several advantages for both clinicians and the patients attending these centers. First, anxiety and depressive disorders are disorders with the highest prevalence rates [2,3]. Second, resources in these settings are usually scarce, which affects both the quantity and the quality of the mental health care provided [56]. Third, a large percentage of patients with anxiety and depressive disorders do not receive treatment in mental health care centers [57]. In the specific case of Spain, most patients attending public mental health units suffer from anxiety and depressive disorders [58], the ratio of clinical psychologists to patients is one of the lowest in Europe [59], and patients have to endure long waitlists to receive treatment [60]. Fourth, transdiagnostic treatments can be provided at a lower cost (eg, in terms of training) [61]. Finally, the use of the internet can help improve mental health services, for instance, by reducing the waiting period to receive face-to-face treatment [62] or by implementing these treatments as part of a stepped-care model that takes patients’ profiles and needs into account. Thus, each patient can be assigned the most appropriate treatment [63], leaving face-to-face therapy for those patients who are less likely to benefit from internet-delivered interventions.

Taking all of this into consideration, in this study, an RCT was conducted to test the effectiveness of a transdiagnostic internet-delivered protocol for ED (EmotionRegulation), compared with TAU provided in Spanish public specialized mental health care. EmotionRegulation includes components of the UP and the skills from DBT (eg, mindfulness what and how techniques), and it was designed to target a wide range of EDs, including major depressive disorder (MDD), dysthymic disorder (DD), PD, AG, SAD, GAD, OCD, anxiety NOS, and depression NOS. It was hypothesized that (1) the EmotionRegulation group would outperform the TAU group on measures of overall anxiety and depression, temperament (ie, BI and BA), and health-related quality of life (QoL) at posttreatment; (2) these posttreatment changes would be maintained at follow-up; (3) a significantly greater clinical change would be observed in EmotionRegulation compared with TAU; and (4) participants in the EmotionRegulation group would rate the treatment as acceptable (scores on expectations and opinion ≥7/10).

A 2-armed RCT was conducted. Participants were randomly assigned in a 1:1 ratio to one of the following 2 conditions: (1) EmotionRegulation and (2) TAU. Participants were stratified by principal diagnosis, performing block randomization in blocks of 4 to ensure that all the principal diagnoses were equally represented across conditions. Computer-generated random number sequences were obtained using statistical software (Epidat 4.1, SourceForge). This task was performed by an independent researcher who was unaware of the characteristics of the study.

The study was conducted in compliance with the study protocol, the Consolidated Standards of Reporting Trials (CONSORT) statement [64,65], the CONSORT of Electronic and Mobile HEalth Applications and onLine TeleHealth guidelines [66], and the Declaration of Helsinki and good clinical practice. A full description of the study protocol has been reported elsewhere [67]. The RCT obtained ethical approval from the Ethics Committee of Universitat Jaume I (Castellón, Spain) and the Clinical Research Ethics Committees of 3 hospitals (Consorcio Hospitalario Provincial de Castellón, Hospital Universitario de la Ribera, and Hospital Universitario Vall d’Hebron). The study protocol was registered at ClinicalTrials.gov (NCT02345668) on July 27, 2015. The duration of the intervention period was 18 weeks for participants in both conditions, and participants’ assessments were conducted at pre- and posttreatment and at 3- and 12-month follow-ups. Both the intervention (EmotionRegulation) and the assessment instruments, except the diagnostic interview, were delivered through a web platform designed by our research group [68]. All transferred data were secured via Advanced Encryption Standard-256 encryption. Study researchers conducting posttreatment and follow-up assessments (ie, diagnostic interviews) were blinded to the participants’ treatment conditions. To ensure blinding of the evaluators, participants were informed that an independent researcher would contact them to conduct follow-up assessments, and they were asked not to disclose the treatment condition to which they had been allocated. This study reports pre- to posttreatment data and data at 3-month follow-up.

Several studies were considered for the calculation of the expected sample size [23,69,70]. On the basis of a minimum power of 0.80 in a 1-tailed test (ie, t test for differences between 2 independent means), an α of .05, and an estimated dropout rate of approximately 30%, a sample size of 78 participants per condition was determined to detect a posttreatment effect size of 0.40 (Cohen d) between the 2 conditions. In addition, based on the literature [71,72], an estimated dropout rate of approximately 30% was expected. Thus, the final sample size was set at 100 participants per condition (total of 200 participants). The G*Power software (version 3.1.9.4, Heinrich-Heine-Universität) was used to calculate the sample size [73].

Participants were recruited from adult outpatients attending Spanish public specialized mental health care services (mental health units) to seek psychological and/or psychiatric treatment between July 2015 and June 2019. Initial recruitment was performed by clinical psychologists and psychiatrists working in these centers, and it took place in 3 different hospitals: Consorcio Hospitalario Provincial de Castellón (Castellón de la Plana), Hospital Universitario de la Ribera (Valencia), and Hospital Universitario Vall d’Hebrón (Barcelona). Recruitment was performed as follows: (1) once psychiatrists and clinical psychologists had identified a potential candidate, they offered the patient the possibility of participating in the study and described the study characteristics to him or her; (2) patients who were interested in participating gave their informed written consent, and the clinician filled out a document with the participants’ sociodemographic and clinical characteristics (moreover, in this stage, participants were provided with a document containing information about the study); (3) one of the researchers involved in the study contacted the participants by phone to schedule a face-to-face appointment to evaluate eligibility criteria using a structured diagnostic interview; and (4) whenever a participant met the eligibility criteria, an independent researcher (unaware of the study characteristics) was contacted to implement randomization, and participants completed the remaining assessment instruments (self-reported questionnaires) through web-based surveys.

Participants were selected based on the following inclusion criteria: (1) aged 18 years or older; (2) ability to understand and read Spanish; (3) having access to the internet at home and an email address; (4) meeting Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) diagnostic criteria [74] for ED (ie, MDD, DD, depression NOS, PD, AG, SAD, GAD, anxiety NOS, and OCD); (5) providing written informed consent; (6) not suffering from a severe mental disorder (schizophrenia, bipolar disorder, and alcohol and/or substance dependence disorder); (7) not presenting a high risk of suicide; (8) not suffering from a disabling medical disease that prevented the participant from carrying out the psychological treatment; and (9) not receiving another psychological treatment during the study (in the experimental group). Pharmacological treatment was allowed, but participants had to be taking the same dose during the 2 months before enrolling in the study. In addition, participants in the experimental group whose medication was increased or changed during the study period were excluded from the trial (decreases in pharmacological treatment were accepted). There was no monetary compensation for participation in the study under any of the treatment conditions.

The treatment and support protocols were administered by doctoral students with at least two years of experience in the diagnosis, psychological assessment, and application of CBT for different ED. Several steps were taken to ensure treatment fidelity. First, therapists had previously been trained in the application of the treatment modules. Second, a support protocol (ie, weekly phone calls and automated text messages) was developed to be applied to all the participants in the EmotionRegulation condition. This support protocol has been briefly described earlier, but more details can be found in the study by González-Robles et al [67]. Third, to increase diagnosis reliability, all therapists involved in the participants’ assessment were trained in the application of the diagnostic interview (MINI).

All analyses were performed using the Statistical Package for Social Sciences version 25.

First, chi-square tests for categorical data and independent samples t tests for continuous data were performed to confirm that there were no significant differences between the groups at baseline on any of the sociodemographic and clinical variables.

Intention-to-treat (ITT) analyses were performed following Newman’s guidelines [97], using maximum likelihood (ML) estimation through the expectation maximization imputation method. To handle missing data, we followed the procedure suggested by Hair et al [98]. First, we explored the types of missing data and determined that data were missing at the construct level. On the basis of this, we concluded that the data were susceptible to imputation. Second, the quantity of missing data was analyzed to ensure that none of the measures exceeded the recommended limits to implement this method [99]. Third, Little’s missing completely at random (MCAR) tests were carried out to analyze the pattern of missing data, concluding that all missing data were MCAR (X²₂₅=28.7; P=.28). Finally, a sensitivity analysis was performed on the main outcomes to compare the results of the per-protocol sample (ie, completers) with the imputed values. This analysis revealed that the ML estimation was not likely to produce biased estimations in the main analyses, reaching the same conclusions in both the completers and the imputed dataset (per-protocol: F_{BAI (1,127)}=6.1; P=.02; and F_{BDI (1,127)}=15.54; P<.001; ITT: F_{BAI (1,197)}=4.79; P=.03; and F_{BDI (1,197)}=12.97; P<.001).

To test the first hypothesis and control for baseline differences, analyses of covariance (ANCOVAs) were performed to compare the effects of the groups on measures of anxiety, depression, temperament, and health-related QoL, taking condition as the between-subject variable and the pretreatment scores as covariates. The use of ANCOVAs for the analysis has been recommended by several authors as a more powerful tool to analyze data in studies with randomized designs [100,101].

To test the second hypothesis, a 2 (condition: EmotionRegulation vs TAU) × 3 (time: pretreatment vs posttreatment vs 3-month follow-up) mixed analysis of variance (ANOVA) was performed to test whether the differences between EmotionRegulation and TAU (ie, between-subjects factors) were maintained at follow-up (ie, within-subject factor). The following assumptions for the mixed ANOVA were analyzed: normality (Shapiro-Wilk test), homoscedasticity (Levene test), independence (nonparametric Runs test), and sphericity (Mauchly test). The degrees of freedom were corrected using Greenhouse-Geisser whenever the sphericity assumption was violated. Moreover, pairwise Bonferroni-corrected tests were used for posthoc comparisons.

To compute the magnitude of both within-group and between-group changes, effect sizes (Cohen d) were calculated by dividing the differences between means by the pooled SD. Effect sizes were interpreted according to Cohen convention: effect sizes of 0.20 are considered low, effect sizes of 0.50 are considered medium, and effect sizes of 0.80 and above are considered large [102].

To test the third hypothesis, we explored the clinical significance of the changes achieved by the participants as well as potential deterioration rates using Jacobson and Truax’s reliable change index (RCI) [103] for the main outcome measures (BDI-II and BAI) in the completer sample for posttreatment and follow-up measurements. First, the cutoff points for the posttreatment and follow-up scores were determined to be within the range of a functional distribution. The RCI was then calculated to test the clinically significant change, with an RCI of |1.96| or greater (P<.05). Finally, both criteria were taken into account to classify participants into the following 4 categories: (1) recovered: when the change is significantly reliable (RCI≥|1.96|; P<.05) and the posttreatment score is located within the range of the functional distribution (mean [SD 2]), (2) improved: when the change is significantly reliable but the posttreatment score is below the functional level, (3) not changed: when the change is not significantly reliable and the posttreatment score does not reach the functional level, and (4) deteriorated: when the change is significantly reliable but the posttreatment score is worse than the pretreatment score.

Finally, to test the fourth hypothesis, the scores on expectations and opinions were analyzed by calculating means and SDs for each of the items on the expectation and opinion of treatment scales. In addition, 1-way ANOVAs were performed to analyze the significance of the differences between expectations and opinions.

A flowchart of the study participants is displayed in Figure 1. A total of 326 patients expressed interest in the study, 281 of whom were assessed for eligibility. Of these 281, 67 participants were excluded from the study. A total of 214 participants were randomized to either EmotionRegulation (n=106) or TAU (n=108). In addition, 7 patients in each condition withdrew from the study before the pretreatment assessment. Consequently, these participants were not included in any of the analyses.

Regarding attrition, 35 participants in the EmotionRegulation condition (35/106, 33.0%) and 34 in the TAU condition (34/108, 31.5%) dropped out of the study (reasons for dropout are shown in Figure 1). In addition, 3 participants in the EmotionRegulation condition had to be excluded from the trial because of a change in their pharmacological treatment during the treatment period. Posttreatment data were obtained from 63 participants (63/99, 64%) in the EmotionRegulation condition and from 67 participants (67/101, 66.3%) in the TAU condition. Follow-up data were collected from 51 participants (51/99, 52%) in the EmotionRegulation condition and 56 participants (56/101, 55.4%) in the TAU condition. Finally, 99 participants in the EmotionRegulation group and 101 participants in the TAU condition were included in the ITT analysis.

Participants (N=200) had a mean age of 38.44 years (SD 10.80; range 18-68), and they were mostly females (138/200, 69.0%). Table 2 provides the sociodemographic and clinical characteristics for both conditions at baseline. There were no significant differences between the EmotionRegulation and TAU groups at baseline on any of the sociodemographic and clinical characteristics. Moreover, no significant differences were found for medication, principal diagnosis, number of comorbid diagnoses, or clinical severity on any of the measures.

Participants in the EmotionRegulation condition were provided a mean of 49.97 min (SD 41.20 min) of clinician support delivered through phone calls. In addition, an initial face-to-face session was scheduled with all the patients in both conditions (EmotionRegulation and TAU) to explain the study and perform the screening assessment, with an approximate duration of 60 min for each participant. Regarding automated support, that is, weekly text messages, participants in the EmotionRegulation group were sent a mean of 24.61 text messages (SD 8.80).

Table 3 provides the means and SDs for the 2 conditions at pre- and posttreatment and at 3-month follow-up on both primary and secondary outcome measures. Additional data on change scores and within-group effect sizes for diagnosis-specific measures are reported in Multimedia Appendix 1.

The results for the significance of the clinical changes on measures of overall depression and anxiety (BDI-II and BAI, respectively) are summarized below.

Table 5 provides the means and SDs for expectations and opinions about treatment in the completer sample of the EmotionRegulation condition (n=63). As hypothesized, the results indicate that participants reported high scores on all the items measuring treatment expectations (scores between 7.56 and 7.81): logic of the treatment, satisfaction with the treatment, recommending the treatment to other people with similar problems, usefulness of the treatment for other psychological problems, and usefulness of the treatment for one’s specific problem. After receiving the intervention, scores for treatment opinions were generally higher than scores for treatment expectations (scores between 7.67 and 8.24).

A 1-way repeated measures ANOVA showed that, compared with treatment expectations, the opinion of treatment was significantly better for item 1 (logic of the treatment; F_1,62=7.81; P=.007; η² partial=0.11), item 3 (recommending the treatment to other people with similar problems; F_1,62=4.80; P=.03; η² partial=0.07), and item 4 (usefulness of the treatment for other psychological problems; F_1,62=4.92; P=.30; η² partial=0.07). No significant differences were found for item 2 (satisfaction with the treatment; F_1,62=3.05; P=.09; η² partial=0.05) or item 5 (usefulness of the treatment for one’s specific problem; F_1,62=.21; P=.65; η²partial=0.003).

The objective of this RCT was to explore whether a transdiagnostic internet-delivered protocol (EmotionRegulation) could be effective in treating a wide range of anxiety and depressive disorders, compared with TAU as provided in Spanish public specialized mental health care services. The effectiveness of EmotionRegulation was evaluated on measures of overall anxiety and depression, temperament (ie, BI and BA), and health-related QoL. Finally, expectations and opinions were evaluated to examine the acceptability of EmotionRegulation for the patients. To our knowledge, this is the first RCT to report data on a transdiagnostic internet-delivered protocol for ED in public specialized mental health care.

Regarding our first hypothesis, the ANCOVAs revealed that participants in the intervention group (EmotionRegulation) improved their depression and anxiety symptoms at posttreatment to a greater degree than participants in the TAU group. Regarding the magnitude of the changes at posttreatment, the analyses showed small but significant between-group effect sizes for both depression and anxiety. With regard to measures of temperament, the findings were mixed. Overall, although patients in the EmotionRegulation group showed better scores on the 2 subscales of the BIS and BAS than the TAU group, only the differences on the BAS subscale were significant, favoring the EmotionRegulation group. These results suggest that the intervention tested in this study was able to modify temperament, in line with a mechanistically transdiagnostic approach that assumes the existence of underlying mechanisms that account for the occurrence of specific symptoms. To our knowledge, only 1 study has previously investigated the effect of a mechanistically transdiagnostic treatment on the dimensions of BI and BA [42]. Similar to the results obtained in our study, in the study by Carl et al [42], both BI and BA improved following treatment with the UP compared with a waitlist control group, with small between-group effect sizes found for both the BIS and BAS temperament dimensions. Moreover, these authors showed associations between the gains in temperament dimensions and symptoms of anxiety and depression, particularly for BI, because lower BI scores were associated with greater improvements in anxiety and depressive symptoms. However, a limitation of this study was that the sample size was small, and so the findings were preliminary. Thus, future RCTs of mechanistically transdiagnostic treatments should analyze the extent to which these interventions are able to modify BI and BA and other related temperament dimensions such as negative and positive affect to shed more light on this question [104,105]. Finally, as hypothesized, EmotionRegulation was superior to TAU in improving health-related QoL, with a small but significant between-group effect size observed for the EQ-5D-3L scale. In this regard, it is known that pharmacological treatment could be linked to some variables associated with health-related QoL, such as fatigue and sexual functioning [106]. Therefore, decreases in medication may have indirectly influenced the QoL in some patients. However, these decreases were not measured, and hence, no analyses could be performed to analyze this aspect.

Furthermore, to evaluate the significance of the clinical gains, Jacobson and Truax’s RCI was obtained for the principal outcome measures, that is, BDI-II and BAI.

Consistent with the aforementioned results, significant differences were observed in the proportion of patients who achieved a functional change in depression scores, with a significantly higher number of patients within the functional range in the EmotionRegulation condition (85/99, 84%) than in the TAU group (59/101, 58.4%). Regarding anxiety scores, although no significant differences were observed between the 2 groups in the proportion of patients reaching a functional change, a higher proportion of patients in the EmotionRegulation group (72/99, 73%) than those in the TAU group (65/101, 64.4%) reached a functional change. Furthermore, significant differences were found in the proportion of patients recovered, improved, did not change, and deteriorated, with better general results for the intervention group than for the TAU group. At follow-up, there were no differences in the proportion of patients reaching a functional change in either depression or anxiety. Therefore, it can be concluded that these differences tended to diminish between groups at follow-up.

Regarding the acceptability of EmotionRegulation, expectations about the treatment were high (scores ≥7). After receiving the treatment, compared with expectations, the participants rated the program as significantly more logical, more recommendable for other people with similar problems and more useful for the treatment of other psychological problems. The study of acceptability is important because expectations about treatment have been shown to affect treatment outcomes [107]. Furthermore, because most transdiagnostic internet-delivered protocols have been conducted in community samples [108], it is necessary to continue to explore the acceptability of these interventions in specialized care.

Overall, the findings showed that the transdiagnostic internet-delivered protocol tested in this RCT was more effective than TAU for the treatment of anxiety and depressive disorders in public specialized mental health care. On the one hand, the results show that EmotionRegulation led to greater improvements at posttreatment, and these gains were maintained at follow-up. Regarding TAU, the results reveal that patients undergoing TAU also experienced improvement over time, but it was slower and less pronounced than in patients receiving EmotionRegulation. The generally lower intensity of TAU (ie, lower frequency of therapy sessions) and the fact that for most patients TAU was limited to pharmacotherapy (with no access to psychological treatment) might partly account for these results. On the other hand, as anticipated, and in line with previous studies conducted by our research group using the same treatment platform [94,109], scores on expectations and opinions demonstrated EmotionRegulation’s acceptability for participants.

The results obtained in this RCT have implications for both research and clinical practice, especially in the context of public specialized mental health care. First, the findings obtained in this study support the effectiveness of a mechanistically transdiagnostic internet-delivered protocol for the treatment of EDs, and they contribute to the literature in this particular field [23,27,36,110-112]. Specifically, a combination of the components of the UP and DBT regulation skills was found to be more effective than TAU in treating ED. Second, as far as we know, this is the first study to explore a transdiagnostic internet-delivered treatment in public specialized mental health care. As mentioned earlier, most research on transdiagnostic web-based treatments has been conducted in community settings, with a few of these studies carried out in primary care [108]. The results showed that the intervention was found to be more effective than TAU on the measures of generic depression, anxiety, and QoL. These results are consistent with the literature showing the superiority of CBT over TAU. For instance, a meta-analysis showed that CBT outweighed TAU, with effect sizes in the medium range on measures of generic anxiety (Hedges g=0.70) and depression (Hedges g=0.69) [70]. As surprising as these data might seem, the truth is that current public mental health services still have to deal with a number of barriers that hinder appropriate care delivery, such as excessive waiting times to access mental health care [60], low frequency of sessions [113], or inadequate follow-up care [56]. Moreover, the lack of training in evidence-based treatments among professionals further adds to this problem [114]. Finally, in Spain, several RCTs have been conducted using the internet to provide evidence-based treatments, showing that they are effective for the treatment of ED and, in particular, depression, in community samples [94] and primary care [115,116], and others are underway [117]. This study demonstrated that an internet-delivered protocol for ED was effective in public specialized care, a setting with a high demand, but much less explored, thus adding to the literature on these treatments for ED. Furthermore, the use of TAU as the control condition may help to answer the question of “whether a new treatment or an evidence-based psychotherapy really surpasses in outcome effects what is ordinarily done at a given clinic” [118], thus helping to make clinicians, researchers, and policy makers aware of the limitations and aspects that should be improved in this specific setting.

Although the results of this RCT are promising, they should be interpreted in light of the following limitations. First, although several measures were taken to minimize attrition (eg, guidance was provided to all the patients participating in EmotionRegulation), the number of patients who dropped out of the study was high (around 35%). However, this proportion was close to what is typically observed in the literature on internet-delivered psychological treatments (ie, approximately 30%-35%) [71]. Moreover, attrition in the TAU condition was similar (34/108, 31.5%). Second, this RCT was not powered to detect differences in disorder-specific measures (ie, for GAD, PD or AG, SAD, and OCD). Therefore, future studies that meet the minimum levels of statistical power to detect differences in these measures are warranted. Third, although the frequency of the session in TAU was low and most patients in this condition were receiving only pharmacotherapy (as observed by the researchers), these data were not monitored during the trial. Finally, the results for acceptability (ie, expectations and opinions) might not be entirely representative because data from patients who dropped out of the intervention were not included in these analyses.

The effectiveness of a mechanistically transdiagnostic internet-delivered protocol for ED was compared with TAU in public specialized mental health care. Although the results are promising, more research in specialized care should be conducted to extend the findings obtained in this study. First, research on predictors and moderators of treatment outcomes and dropout in this specific setting can help to delineate the profiles of participants who are more likely to benefit from these treatments and help to answer the classic question, “what treatment, by whom, is most effective for this individual with that specific problem, under which set of circumstances?” [119]. Moreover, we believe that the integration of principles from different evidence-based protocols (eg, components of the UP and DBT) can be a powerful strategy that should guide future research on evidence-based psychotherapy. This strategy is in line with process-based CBT and the new foundational question proposed by Hofmann and Hayes [49] ( ie, “What core biopsychosocial processes should be targeted with this client given this goal in this situation, and how can they most efficiently and effectively be changed?”) . Second, despite the huge advances experienced by the field of internet-delivered treatments in the past two decades, high dropout rates remain a major challenge in the field. To continue to improve current and future internet-delivered interventions, future studies should strive to include dropouts in the analysis of acceptability using both qualitative and quantitative approaches. Moreover, to ensure the integrity, quality, and replicability of these studies, adherence to existing research guidelines is of paramount importance in this endeavor. Third, although we did not assess the acceptability of clinicians involved in the RCT (ie, psychiatrists, clinical psychologists, and nurses), it is worth mentioning that some of them refused to participate in the recruitment process, which might reflect negative attitudes toward internet-delivered interventions among these professionals. In this scenario, research efforts should be made to inform clinicians and state holders about the benefits of internet-delivered treatments, especially because they are seen as authority figures and, therefore, their attitudes can have a major impact on patients’ perceptions. Finally, although the need for efficacy studies is out of doubt, we believe that it is of paramount importance to conduct more implementation research [120]. These studies may provide a much deeper understanding of implementation variables that can either facilitate or hamper the effective uptake of evidence-based protocols in real clinical practice, such as the attitudes of clinicians and other professionals toward internet-delivered treatments, or economical and logistic aspects that are difficult to implement.