Predictors and moderators of treatment dropout in cognitive-behavioral and psychodynamic therapies for panic disorder

Abstract

Introduction: Panic disorder patients who drop out of treatment typically do not remit from their disorder. How patient-level moderators influence dropping out of one panic-focused treatment over another has never been examined, nor in non-CBT treatments. Method: 200 patients with panic disorder with or without agoraphobia were randomized to receive cognitive–behavioral therapy (CBT), panic-focused psychodynamic psychotherapy (PFPP), or applied relaxation training (ART) across two sites. Therapy was twice a week for 12 weeks. A two-step variable search method was applied to identify potential prognostic predictors and moderators of patient dropout. Survival models predicting hazard of session-by-session dropout tested the resulting variables. Results: Across treatments, unemployment and higher psychosocial disability on the Sheehan Disability Scale predicted increased risk of dropout, while patients with higher anxiety sensitivity were more likely to complete treatment. Patients who reported experiencing childhood abuse had heightened dropout in ART, but not CBT or PFPP. Men were especially likely to complete PFPP. Session 2 expectancies and patient-rated alliance predicted lower dropout only in CBT. Conclusions: Patient-level factors may influence both whether patients will complete any treatment, and whether they continue in a particular panic-focused therapy. Moderators of dropout (e.g., abuse history) may inform treatment decisions for specific patients.

Trial registration: ClinicalTrials.gov identifier: NCT00353470.

Keywords:

• cognitive behavior therapy
• outcome research
• psychoanalytic/psychodynamic therapy
• anxiety

Supplemental Data

Supplemental data for this article can be accessed at https://doi.org/10.1080/10503307.2020.1784487.

Notes

1 Generally, findings were similar when excluding the very early dropouts randomized to treatment and completing baseline assessments but not attending their first therapy session. Both the ASI and SDS scores remained significant prognostic predictors (HR = 0.96, p = 0.016 & HR = 1.09, p = 0.002, respectively), while being unemployed or on disability dropped to a trend level but with a similar effect size (HR = 1.92, p = 0.083). For moderators, both gender (χ² = 6.97, p = 0.031) and childhood abuse history (χ² = 9.88, p = 0.007) remained significant. We did not include in our variable search two variables identified in previous manuscripts: intake PDSS severity, which indicated greater dropout in ART relative to CBT or PFPP (Milrod et al., 2016), and comorbidity of ADIS-diagnosed MDD, which was a general predictor of dropout (Keefe et al., 2019). As a sensitivity check, we examined including these variables in our models. In the moderation model, the PDSS interaction with condition dropped to a trend level (χ² = 5.22, p = 0.073), while both gender (χ² = 7.70, p = 0.021) and childhood abuse history (χ² = 7.77, p = 0.021) remained significant moderators. In the prediction model, comorbid MDD remained a significant predictor of dropout (HR = 1.88, p = 0.043), as did ASI (HR = 0.97, p = 0.019), SDS (HR = 1.09, p < .001), and being unemployed or on disability (HR = 1.88, p = 0.043).

Additional information

Funding

This work was supported by National Institute of Mental Health [grant numbers R01-MH070918, R01-MH070664]. Dr. John Keefe is supported by NIH/NCATS Grant # TL1-TR-002386 through the Clinical & Translational Science Center at Weill Medical College of Cornell University.