Sustainability of Treatment Success 5 Years after Exposure-Based Cognitive-Behavioral Therapy for Anxiety Disorders

Abstract

Introduction

Achieving sustainable success in the treatment of anxiety disorders remains a central objective in mental health care. Although research has demonstrated the short-term efficacy of psychotherapy, evidence regarding long-term sustainability is limited. This study examined treatment outcomes 5 years after prediction-error-based exposure therapy.

Methods

For 355 patients (616 eligible; 58% follow-up rate), newly collected follow-up data on anxiety symptoms and psychosocial functioning were compared to pretreatment, posttreatment, and 6-month follow-up data from a multicenter clinical trial characterized by high treatment fidelity.

Results

Improvements in anxiety symptoms and psychosocial functioning that were evident at posttreatment and 6-month follow-up were largely preserved after 5 years. No significant differences emerged between randomized groups of temporally intensified and non-intensified exposure. From 6 months to 5 years, overall remission rates remained stable, with the majority of patients exhibiting no reliable change in symptom severity. Reliable relapse occurred in 4.9% and reliable new remission in 6.5%. Most patients (63.4%) did not seek additional treatment. Among those who did, depression (64.2%) and anxiety (60.5%) were the most frequently cited reasons, although only a minority (6.0%) sought further treatment exclusively for anxiety. Additional treatment during the follow-up period was associated with higher symptom severity throughout assessments.

Conclusion

These findings highlight the sustainability and long-term public health benefits of exposure-based CBT for anxiety disorders. Most patients do not need additional treatments for mental disorders even 5 years after treatment. Nevertheless, further efforts are needed to optimize interventions for those patients who do not achieve remission or experience relapse.

Introduction

Anxiety disorders are among the most prevalent mental health conditions and impose a substantial individual and socioeconomic burden [1]. Without adequate treatment, anxiety disorders tend to persist with a waxing and waning pattern, frequently leading to the development of secondary conditions [1]. Consequently, providing effective and sustainable treatment remains a central goal in mental health care.

For psychotherapy, exposure-based cognitive-behavioral therapy (CBT) has consistently been shown to effectively improve anxiety symptoms and psychosocial functioning. Meta-analytic evidence indicates that CBT is superior to waitlist, placebo control, and other forms of psychotherapy [2–4]. Accordingly, international guidelines recognize exposure-based CBT as the first-line psychotherapeutic treatment for anxiety disorders [5].

In addition to the consistent evidence for posttreatment efficacy, recent meta-analytic evidence suggests that efficacy and remission rates are maintained at follow-up (FU) [2, 6]. However, most controlled trials typically include short FU assessments between six and 12 months [7]. Some evidence suggests that treatment benefits might mitigate after 12 months [8]. Naturalistic studies with longer FU periods up to 14 years yielded mixed results, for example, with remission rates between 18% and 77% in panic disorder (PD) and agoraphobia (AG) after 5 years [9–11]. Most of these studies are limited by a lack of initial control for treatment fidelity, inclusion of different or unspecified treatments alongside CBT, small sample sizes, and/or substantial within-study variability in FU duration. Importantly, most studies do not report additional treatment seeking during the FU period, which may bias estimation of long-term effects. Thus, further research with longer FU periods and larger samples receiving exposure-based CBT with high treatment fidelity is needed.

Besides additional treatments, several factors may impact the sustainability of treatment outcomes. One important factor is the treatment rationale, reflecting the assumed mechanism of change [12]. Traditionally, exposure-based CBT followed a within-session fear reduction rationale, in which exposure is considered successful if fear responses declined during confrontation with a feared situation in the absence of dysfunctional anxiety control behavior (such as avoidance) [13, 14]. Most previous research on sustained treatment success has focused on this rationale. More recently, clinical translation of inhibitory extinction learning models emphasized prediction-error-based learning as a rationale for exposure-based CBT [15, 16]. This rationale highlights the importance of violating and modifying individual core threat beliefs. Specifically, exposure is deemed successful if patients experience that their expected threat does not occur during exposure (“I did not get a heart attack.”) and subsequently change their threat beliefs. We recently demonstrated that change in threat beliefs after exposure predicts short-term treatment success [17]. Importantly, this rationale is hypothesized to enhance long-term treatment outcomes and reduce relapse [15]. In our randomized controlled trial (RCT), prediction-error-based exposure was highly effective and treatment success increased from posttreatment to 6-month FU [18]. However, evidence for long-term sustainability remains lacking. To this end, the present study collected novel FU data from the sample of the original RCT to examine treatment success after 5 years.

Another important factor for the sustainability of treatment outcomes may be the temporal spacing of exposure. In our original RCT, two randomized groups were compared to examine the differences between temporally intensified (PeEx-I) and non-intensified prediction-error-based exposure (PeEx-S). Keeping a consistent dosage of exposure, temporally intensified exposure was delivered with multiple weekly and non-intensified exposure with weekly exposure sessions. Both variants showed equivalent symptom reduction. Temporally intensified exposure was, however, associated with shorter treatment duration, lower drop-out rates during exposure, and higher quality of life at 6-month FU [18]. This agrees with experimental research suggesting that shorter exposure intervals may accelerate symptom reduction [19, 20], though there is a risk of reduced long-term effects [21].

In summary, the present study aimed to examine the long-term sustainability of treatment success following exposure-based CBT for anxiety disorders. Building on a large multicenter RCT with rigorous control of treatment integrity, eligible patients were invited to complete a new FU assessment after an average of 5 years. Our objectives were to examine (1) the average change in anxiety symptoms and psychosocial functioning after 5 years, (2) the frequency and impact of additional treatments, and (3) individual reliable change and remission rates from 6-month to 5-year FU.

Methods

Procedures of the Original Trial

The full study protocol and main outcomes of the original trial are described elsewhere [18, 22]. Treatment comprised 12 sessions plus 2 booster sessions (each 100 min) of structured exposure-based CBT [23] (see online suppl. material at https://doi.org/10.1159/000549073). Treatment groups received the same treatment content, which differed in the temporal spacing of exposure sessions (5–10). Non-intensified standard exposure comprised weekly sessions (PeEx-S; 10 weeks from pre to post), while temporally intensified exposure comprised three sessions per week (PeEx-I; 6 weeks from pre to post). Treatment was delivered by trained study therapists who had to treat one certification case with high competence and adherence before being certified as study therapist and received supervision throughout their treatments. Overall, treatment integrity and therapist competence as well as patient satisfaction and engagement were high [18, 24].

Patients

In the original trial, patients with DSM-5 primary diagnosis of PD, AG, social anxiety disorder (SAD), and multiple specific phobias were included (see online suppl. material for inclusion/exclusion criteria). In total, 726 patients were included in the original trial [18]. For the 5-year FU, all patients who completed at least the posttreatment or the 6-month FU and provided consent for further FU assessments were eligible and recontacted (N = 616). Of these, 355 (57.6%) completed the new FU (FU completers); 61 (9.9%) did not participate despite providing informed consent, 18 (2.9%) refused, and 2 (0.3%) were deceased. The remaining 180 (29.2% of the total sample) had outdated contact information; i.e., the majority of 69% of the FU non-completers could not be invited to participate. Approximately half of the FU completers (52.1%) had received PeEx-I, and the other half (47.9%) had received PeEx-S in the original trial. Clinical and sociodemographic characteristics of FU completers at pre and at the 5-year FU are summarized in online suppl. Table S1. FU completers and non-completers did not differ regarding sociodemographic variables, primary diagnosis, prior treatment, or symptom reduction from pre to 6-month FU, but non-completers displayed more comorbidities at pre-treatment and higher overall anxiety symptoms (online suppl. Fig. S1, S2; online suppl. Table S2).

Assessments and Measures

In the original study, sociodemographic and clinical characteristics were collected before treatment. Outcome measures were obtained at pre, post, and 6-month FU. For the 5-year FU, a subset of outcome and service use measures was administered via a secure online system (REDCap), which had already been used in the original trial, ensuring patient familiarity (see online suppl. details). Data management was supervised by the Coordination Centre for Clinical Trials Dresden (KKS) according to GCP Guidelines. Selection of measures prioritized feasibility and minimizing patient burden. Consequently, only self-report measures were included, excluding the primary outcome of the original study, the clinician-rated HAM-A. Specifically, transdiagnostic anxiety symptoms were assessed using the DSM-5 Cross-Cutting Dimensional Scale for Anxiety Disorders (DSM-5 Cross-D) [25]. Disorder-specific measures included the Panic and Agoraphobia Scale for PD and AG (PAS) [26], the Liebowitz Social Anxiety Scale for SAD (LSAS) [27, 28], and the DSM-5 Dimensional Specific Phobia Scale for multiple specific phobias (DSM-5 SP) [25], as prespecified in Heinig et al. [22].

Level of functioning was assessed with the self-reported Global Assessment of Functioning (GAF; 0–100), quality of life with the EuroQOL five-dimensional health status measure (EQ-5D) [29], and disability with the World Health Organization Disability Assessment Schedule (WHODAS 2.0) [30]. The use of additional treatment was assessed by asking participants whether they had received psychotherapy or medication for any mental health condition between the 6-month and 5-year FU and the reason for this additional treatment (anxiety, depression, and/or other).

Data Reduction and Statistical Analyses

Main analyses focused on the sustainability of treatment outcomes 5 years after completion of exposure-based CBT. Specifically, we compared the severity of anxiety symptoms and the level of psychosocial functioning at 5-year FU with previous assessments (pre-, post-, and 6-month FU). To obtain more robust and comprehensive measures, less influenced by a single scale or by comorbid anxiety diagnoses, we calculated standardized composite scores for all four assessment points (as in Pittig et al. [17]) (see online suppl. Table S3). The composite anxiety score included all symptom measures, reflecting overall anxiety severity, with lower values indicating lower symptoms. The composite functioning score represents overall psychosocial functioning, where higher values indicate higher functioning.

To evaluate change in symptoms and psychosocial functioning, we compared composite anxiety and functioning scores at the 5-year FU to the previous three assessments (planned contrasts) using linear mixed models (see online suppl. material). To examine frequency and impact of additional treatment, we first calculated the proportion of additional psychotherapy and/or pharmacotherapy use as well as the self-reported reason (subdivided into anxiety, depression, other). Next, LMMs with composite anxiety or functioning scores as dependent variable accounting for additional treatment were calculated. Because we aimed to analyze the influence of additional treatments between 6-month and 5-year FU, only these two time points were included (for transparency, descriptive data for all assessments are presented in Fig. 1).

Fig. 1.

a–d Course of anxiety symptoms and psychosocial functioning from pretreatment to 5-year FU in 355 patients with primary PD, AG, SAD, and/or mSP. Composite scores represent a combination of multiple self-report measures. mSP, multiple specific phobias.

For remission status, remission was defined using established cut-offs for the diagnosis-specific measures corresponding to each patient’s primary disorder. Sustainability of remission rates from 6-month to 5-year FU was tested with generalized linear mixed models for binary dependent variables. Because remission status can change with small fluctuations on the relevant scales, we also calculated the individual Reliable Change Index (RCI) [31] and provided results for change in remission status in combination with RCI (see online suppl. material).

Sensitivity analyses using multiple imputation with chained equations for FU non-completers were conducted based on available previous measurements and a set of pre-treatment characteristics to account for a potential bias due to pre-treatment differences. All analyses were robust to sensitivity analyses with the exception that the increase in remission rates from 6-month to 5-year FU reached significance in the sensitivity analyses (see online suppl. material).

Results

Change in Symptoms and Psychosocial Functioning after 5 Years

For composite anxiety symptoms (Fig. 1a), we found no significant main or interaction effects involving treatment group, ps > 0.227. Importantly, there was a significant main effect of time, F(3, 1,024.68) = 563.09, p < 0.001. Planned contrasts showed significantly lower symptoms at 5-year FU compared to pre-treatment, t(1,020) = 34.16, p < 0.001, d = 1.92, 95% CI = 1.81–2.03, and to posttreatment, t(1,020) = 3.53, p < 0.001, d = 0.20, 95% CI = 0.09–0.31. No significant differences were found compared to 6-month FU, t(1,028) = −1.53, p = 0.126, d = −0.09, 95% CI = −0.20 to 0.02. These results highlight significantly lower symptoms at 5-year FU compared to pre-treatment and even posttreatment that sustained from 6 months with no differences between treatment groups.

For composite functioning (Fig. 1b), results were similar with no significant main or interaction effects involving treatment group, ps > 0.14. Importantly, there was a significant main effect of time, F(3, 1,021.29) = 176.82, p < 0.001. Planned contrasts showed significantly higher functioning at 5-year FU compared to pre-treatment, t(1,019) = 17.18, p < 0.001, d = 0.95, 95% CI = 0.84–1.06, but not to posttreatment, t(1,019) = 0.01, p = 0.999, d = 0.01, 95% CI = −0.11 to –0.11. There was a slight, but significant decrease in functioning from 6-month to 5-year FU, t(1,026) = 3.99, p < 0.001, d = −0.22, 95% CI = 0.11–0.34. These results highlight sustained improvement in functioning at 5-year FU compared to pre, which did not differ from posttreatment. There was a small decrease in functioning from 6-month to 5-year FU. Again, there were no significant differences between treatment groups.

Frequency and Impact of Additional Treatment

Overall, two-thirds of the patients did not seek additional treatment (Fig. 2a and online suppl. Table S1). For the remaining third, the most frequent reason, as indicated in Figure 2b, was depression (64.2%), followed by anxiety (60.5%), and others (29.4%). Only 16.5% of those reporting additional treatment, i.e., 6.0% of the FU completers, reported that anxiety was the sole reason for additional treatment.

Fig. 2.

a, b Self-reported use of additional treatment and reasons between 6-month and 5-year FU.

For composite anxiety symptoms separated by additional treatment (Fig. 1c), analyses yielded a significant main effect of additional treatment, main effect: F(1, 345.98) = 37.46, p < 0.001, indicating overall higher anxiety symptoms in patients seeking additional treatment, d = 0.61, 95% CI = 0.42–0.81. No significant interaction involving additional treatment was found, F(1, 331.80) = 3.62, p = 0.058.

For composite functioning separated by additional treatment (Fig. 1d), an identical pattern was found. Analyses yielded a significant main effect of additional treatment, F(1, 339.30) = 39.00, p < 0.001, indicating overall lower functioning in patients seeking additional treatment, d = 0.60, 95% CI = 0.41–0.79. No significant interaction involving additional treatment was found, F(1, 327.45) = 1.20, p = 0.275.

Remission Rates and Individual Reliable Change

Descriptively, overall remission rates increased from post to 6 months and remained stable to 5-year FU (see Fig. 3a). For change in remission rates from 6-month to 5-year FU for patients without (Fig. 3b) or with additional treatment (Fig. 3c), results showed no significant difference between 6-month and 5-year FU, main effect: χ² (1) = 0.20, p = 0.656, OR = 0.96. Remission rates were lower in patients with additional treatment, main effect: χ² (1) = 35.10, p < 0.001, OR = 0.47. There was no significant interaction, χ² (1) = 0.19, p = 0.659, OR = 0.96. Results were replicated for the transdiagnostic symptom analyses (see online suppl. Table S4). Thus, overall remission rates remained stable, also for patients seeking additional treatment, for which remission rates were lower compared to patients without additional treatment.

Fig. 3.

Transition of remission status from post- to 6-month to 5-year FU for all patients (a) and subsamples of patient without (b) and with additional treatment (c) between 6-month and 5-year FU. Note that patients continued to receive two booster sessions to support self-management between posttreatment and 6-month FU.

While overall remission rates remained stable, some individual patients improved from no remission to new remission or relapsed from remission to no remission (Fig. 3). Table 1 therefore shows the relative frequencies of “sustained remission,” “new remission,” “relapse,” and “never remission” (for change from post to 5 years, see online suppl. Tables S5, S9).

Table 1.

Change in remission status and reliable change status

​	Total sample	Additional treatment between 6-month and 5-year FU
		no	yes
A: Change in remission status from 6-month to 5-year FU 37
Sustained remission	56.0% (178)	67.5% (135)	34.7% (17)
New remission	14.2% (45)	10.0% (20)	26.5% (13)
Relapse	13.5% (43)	11.5% (23)	12.2% (6)
Never remission	16.4% (52)	11.0% (22)	26.5% (13)
B: Reliable change from 6-month to 5-year FU 37
Reliable improved	10.4% (33)	8.3% (17)	14.2% (16)
No reliable change	76.7% (244)	82.0% (168)	67.3% (76)
Reliable deterioration	12.9% (41)	9.8% (20)	18.6% (21)

Relative (and absolute) frequencies in patients with and without additional treatment between 6-month and 5-year FU. A: Changes in remission status between 6-month and 5-year FU. Sustained remission = remission at both assessments. New submission = no remission at 6-month, but at 5-year FU. Relapse = remission at 6-month, but not 5-year FU. Never remission = no remission at both assessments. B: Reliable change based on RCI calculation with the DSM-5 Cross-D. n = 37 missings. Percentage is based on columns. For a combination of change in remission status and reliable change, see online supplementary material.

Reliable change from 6-month to 5-year FU is shown in Table 1 (for RCI from post to 5 years, see online suppl. Table S6). Overall, most patients did not show reliable change in anxiety symptoms (76.7%), highlighting sustained effects. Patients without additional treatment had the highest rate of no reliable change. A small proportion of patients showed reliable improvement (10.4%), and another small proportion showed deterioration (12.9%). RCI results generally corresponded with the results for change in remission status with reliable improvement/deterioration (10.4%/12.9%) being slightly less frequent than new remission/relapse (14.2%/13.5%) in the whole sample. Combining the “relapse” status with RCI analyses (see online suppl. Tables S7 and S8), 4.9% showed reliable relapse, i.e., were classified as relapse and showed reliable deterioration. In contrast, 8.1% were classified as relapse but showed no reliable change in anxiety symptoms (and 0.5% missing). Combining the “new remission” status with RCI analyses, 6.5% showed new remission with reliable improvement, while 7.1% were classified as new remission, however, without reliable change.

Exploratory analyses yielded no significant differences between intensified and standard exposure in remission rates at 5-year FU and change in remission status (see online suppl. Table S10). Exploratory comparison of primary disorders (PD/AG vs. SAD vs. SP) yielded lower remission rates and rates of no reliable change in SAD but also higher rates of reliable improvement and new remission, which may suggest delayed effects for a subgroup of SAD patients with overall lower remission rates in SAD. However, these differences were not supported by analyses of dimensional treatment outcome, which yielded no significant effect of primary disorder on change in anxiety symptoms and psychosocial functioning (see online suppl. Tables S11, S12).

Discussion

This study investigated the long-term sustainability of treatment outcomes 5 years after exposure therapy for anxiety disorders. Unlike previous research, it is the first to examine long-term outcomes of prediction-error-based exposure in a large sample. Treatments in the original trial were characterized by strict control of treatment integrity. Reductions in anxiety symptoms and improvements in psychosocial functioning observed at posttreatment and 6-month FU were largely maintained after 5 years. Specifically, anxiety symptoms were significantly lower after 5 years compared to pre- and posttreatment and did not differ from the 6-month FU. These findings support the long-term stability of improvements, which were typically reported for shorter FU periods [2, 6] and do not support the notion that treatment benefits mitigate after 12 months [8]. Psychosocial functioning also remained stably improved compared to pretreatment, with a small but significant decline between 6 months and 5 years. Importantly, these outcomes were consistent across both intensified and standard exposure protocols, suggesting flexibility in session spacing. Overall, these results demonstrate sustainable treatment success over 5 years for most patients.

Overall remission rates remained stable from 6 months to 5 years after treatment, with most patients showing no reliable change in symptom severity over time. Relapse based on cut-off scores occurred in 13.5%, consistent with recent meta-analyses [8, 32]. This finding is noteworthy, as it suggests that relapse occurs less frequently than is often claimed in basic and translational research [33, 34]. Interestingly, of those classified as relapse, only 34.9% showed reliable deterioration in RCI analyses (i.e., 4.9% reliable relapse), while 58.1% did not show reliable change in anxiety symptoms, highlighting that the majority of relapse cases were due to a small, unreliable increase above cut-off scores. Moreover, new remissions occurred 14.2%, reliable new remission in 6.5%. Our findings of approximately 70% remission rates exceed the typical range of 50–60% reported in earlier research [35, 36]. Despite substantial FU retention, around 40% of eligible patients were lost to 5-year FU, which may have introduced bias. These FU non-completers show more comorbidities before treatment and higher anxiety symptom severity at the three previous assessments, but not lower symptom reduction after treatment. Results were robust to sensitivity analyses accounting for these differences. Notably, remission rates at posttreatment were similar to prior studies (∼50%) but increased at FU. Moreover, remission was defined as scoring below a well-established clinical cut-off, which has been recommended as a conservative approach [36]. Apart from these methodological considerations, high remission rates may be attributable to the novel rationale of exposure therapy utilized in our trial as well as the rigorous control of treatment fidelity, highlighting the long-term benefits of bona fide exposure therapy.

A closer examination at the individual patient level and additional treatment use provided further insights. About two-thirds of the patients (63.4%) did not seek additional treatment during the FU period. Across all analyses, these patients were characterized by lower anxiety symptoms, higher psychosocial functioning, the highest remission rates (close to 80%), the lowest relapse rates, and the greatest stability in terms of reliable change, with only a few patients showing reliable improvement or deterioration over time. For this majority of patients, exposure-based CBT delivered over 6–10 weeks resulted in long-lasting improvement without the need for further treatment. These findings underscore the considerable long-term individual and public health benefits of psychotherapy for anxiety disorders [37].

A smaller proportion of patients (36.6%) engaged in additional psychotherapy and/or pharmacotherapy. While symptom reduction during treatment was comparable, these patients consistently exhibited greater symptom severity and functional impairment throughout the study period. Interestingly, about half of these patients achieved remission of their primary disorder by the 6-month FU, but still pursued additional treatment thereafter. These findings may indicate a resurgence of anxiety symptoms during the FU period, necessitating further treatment for some patients to maintain treatment gains. However, among those seeking additional treatment, relapse rates were lower than sustained remission rates and most showed no reliable symptom change from 6-month to 5-year FU. It might be possible that the symptom resurgence was not captured by one FU after 5 years; i.e., additional treatment reduced returning symptoms before the assessment. Alternatively, additional treatments may have targeted comorbid conditions rather than the primary disorder addressed in the trial, as most comorbidities were not excluded. Although anxiety was frequently cited as a reason for additional treatment (60.5%), depression was the most common reason (64.2% of patients seeking further treatment). Notably, only 16.5% of those reporting additional treatment – representing 6.0% of the total FU sample – identified anxiety as the sole reason for additional treatment. These results suggest that most additional treatments may have been related to comorbid problems. Future research with repeated FUs at shorter intervals is however needed for direct evidence on the trajectories of primary symptoms, comorbidities, and service use over time. Future research should also examine differential effects of additional psychotherapy and pharmacotherapy, which requires detailed analyses of dosage, duration, discontinuation, and changes in medication to better understand the impact of additional treatment.

Despite the overall positive outcomes, further optimization and personalization of anxiety disorder treatment are needed. In the overall sample, 16.4% of patients never achieved remission at any assessment, and 4.9% experienced reliable relapse between the 6-month and 5-year FU. These findings extend recent perspectives on “difficult-to-treat” anxiety disorders [38] to long-term outcomes. While the term “difficult-to-treat” does not imply that patients themselves are difficult or that effective treatments should be “easy,” it underscores the current limitations of first-line treatments for a subset of patients. This highlights the need for ongoing translational research on mechanisms of change in anxiety treatment to optimize existing approaches and develop novel treatment options [12]. Additionally, identifying reliable predictors of treatment response may facilitate the personalization of initial treatment decisions and inform the inclusion of more robust relapse prevention strategies for certain patients.

Some limitations should be considered when interpreting these results. First, results are observational due to the lack of an untreated control group, which inclusion would, however, raise ethical concerns. Additional influencing factors during the 5-year FU period might have contributed to long-term outcomes, complicating causal interpretation of treatment effects. More data are especially needed to understand the causes of relapses and new remissions. Notably, a major increase in remission rates occurred between posttreatment and 6-month FU, a period during which patients continued to receive booster sessions to support self-management, suggesting that new remissions were likely attributable to treatment. More data on new remissions until 5 years are needed. Second, the primary outcome of the original trial (HAM-A) was not assessed to enhance feasibility and response rates. Nevertheless, the used self-report measures are well established for assessing symptom change, and the current study also included measures of psychosocial functioning. Furthermore, remission rates were based on the primary disorders targeted in the original trial, potentially overlooking comorbid conditions. Remission analyses using transdiagnostic anxiety measures, however, yielded comparable results. A more detailed analysis of comorbid disorders beyond the anxiety spectrum, particularly depression, may be beneficial for evaluating the generalization of treatment effects. Finally, although the FU rate was considerably high after 5 years, approximately 42% of patients did not complete the FU, introducing a risk of bias. Although a potential bias cannot be ruled out, the majority of FU non-completers (69%) could not be invited due to a lack of valid contact information and thus did not refuse to participate. In addition, analyses were robust to sensitivity analyses accounting for pre-treatment differences between FU completers and non-completers.

Conclusion

In summary, this large-scale FU demonstrates that prediction-error-based CBT yields robust and lasting remission of anxiety symptoms and maintenance of functioning for most patients, with relatively low rates of relapse or need for additional anxiety-focused treatment over 5 years. These outcomes reinforce the clinical utility of exposure-based CBT and suggest priorities for ongoing monitoring and personalizing intervention for patients with persistent or comorbid conditions. Finally, our findings indicate that intensified and accelerated treatment is not associated with poorer long-term outcomes, thereby supporting flexibility in the spacing and scheduling of therapeutic sessions. This flexibility may improve speed, access, adherence, and personalization of care without compromising efficacy.

Acknowledgments

We would like to thank the following individuals for their help: Jule Dehler, Dorte Westphal (Dresden), Verena Pflug, Dirk Adolph, Cornelia Mohr, Jan Cwik (Bochum), Anne Pietzner, Jörg Neubert (Greifswald), Carsten Konrad, Yunbo Yang, Isabelle Ridderbusch, Hanna Christiansen, Anne Maenz, Sophia Tennie, Jean Thierschmidt (Marburg), Kathrin Zierhut, Kristina Dickhöver, Markus Winkler, Maria Stefanescu, Christiane Ziegler, Heike Weber (Würzburg), Nathalia Weber, Sebastian Schauenberg, Sophia Wriedt, Carina Heitmann (Muenster), Annika Evers (Cologne), Isabel Alt, Sophie Bischoff, Jennifer Mumm, and Anne Schreiner (Berlin).

Statement of Ethics

The original RCT was preregistered (NIMH Protocol Registration System: 01EE1402A; German Register of Clinical Studies: DRKS00008743), approved by the TU Dresden Ethics Review Board (EK234062014, 11/14/2014), and conducted in accordance with the Declaration of Helsinki. All participants provided written and verbal informed consent. Procedures for the 5-year FU were approved by amendment to the TU Dresden Ethics Review Board, and participants provided renewed informed consent.

Conflict of Interest Statement

Volker Arolt was member of advisory boards and received honoraria for presentation from AstraZeneca, Gilead, Janssen-Cilag Lundbeck, and Sanofi. Katharina Domschke is a member of the Neurotorium Editorial Board, Lundbeck Foundation, and has received speaker’s honoraria from Janssen Cilag GmbH. Tilo Kircher received funding for education and symposia from Lundbeck, Lilly, Pfizer, and Aristo. Jens Plag received lecture honoraria from Aristo and consultancy fees from Pfizer. Jürgen Deckert is an investigator in the EU-funded PReDicT study with P1vital and co-applicant in the Bavarian state government-funded InDepth study with BioVariance. Silvia Schneider was a member of the journal’s Editorial Board at the time of submission. All other authors have no financial interests or conflicts of interest regarding the manuscript.

Funding Sources

This study was funded by the Federal Ministry of Education and Research of Germany (BMBF) (PROTECT-AD project P1: 01EE1402A). The funder had no role in the design, data collection, data analysis, and reporting of this study.

Author Contributions

Andre Pittig: conceptualization, methodology, formal analysis, investigation, resources, data curation, writing – original draft, –review, and editing, visualization, supervision, project administration, and funding acquisition. Ingmar Heinig: conceptualization, methodology, investigation, writing – review and editing, supervision, and project administration. Volker Arolt, Katharina Domschke, Thomas Fydrich, Andreas Ströhle, and Hans-Ulrich Wittchen: resources, writing – review and editing, and funding acquisition. Christina Bartnick, Maike Hollandt, Isabelle Ridderbusch, Yunbo Yang: investigation, writing – review and editing, and project administration. Udo Dannlowski and Benjamin Straube: resources, writing – review and editing, and supervision. Jürgen Deckert: conceptualization, resources, writing – review and editing, supervision, and funding acquisition. Stephan Goerigk: conceptualization, writing – review and editing, supervision, and formal analysis. Alfons O. Hamm: conceptualization, resources, writing – review and editing, and supervision. Tilo Kircher, Jürgen Margraf, Paul Pauli, Winfried Rief, and Silvia Schneider: resources and writing – review and editing. Katja Koelkebeck: investigation, resources, writing – review and editing, and supervision. Ulrike Lueken, Peter Neudeck, and Jan Richter: conceptualization, investigation, resources, writing – review and editing, and supervision. Jürgen Hoyer: conceptualization, resources, supervision, project administration, writing – review and editing, and funding acquisition.

Funding Statement

This study was funded by the Federal Ministry of Education and Research of Germany (BMBF) (PROTECT-AD project P1: 01EE1402A). The funder had no role in the design, data collection, data analysis, and reporting of this study.

Data Availability Statement

The data that support the findings of this study are not publicly available due to privacy or ethical restrictions. Further inquiries can be directed to the corresponding author.