Trajectory and magnitude of response in adults with anxiety disorders: a Bayesian hierarchical modeling meta-analysis of selective serotonin reuptake inhibitors, serotonin norepinephrine reuptake inhibitors, and benzodiazepines.

BACKGROUND

How the trajectory of response to medication (and placebo response) varies among selective serotonin reuptake inhibitors (SSRIs), serotonin norepinephrine reuptake inhibitors (SNRIs), benzodiazepines and across anxiety disorders is unknown.

METHODS

We performed a meta-analysis using weekly symptom severity data from randomized, parallel-group, placebo-controlled trials of SSRIs, SNRIs, and benzodiazepines in adults with anxiety disorders. Response was modeled for the standardized change in anxiety using Bayesian hierarchical models.

RESULTS

Across 122 trials (N=15,760), SSRIs, SNRIs, and benzodiazepines produced significant improvement in anxiety compared to placebo. Benzodiazepines produced faster improvement by the first week of treatment ( < 0.001). By week 8, the response for benzodiazepines and SSRIs ( = 0.103) and SNRIs ( = 0.911) did not differ nor did SSRIs and SNRIs differ ( = 0.057), although for patients with generalized anxiety disorder (GAD), the benzodiazepines produced greater improvement than SNRIs at week 8 (difference - 12.42, CrI: -25.05 to -0.78,  = 0.037). Medication response was similar across anxiety disorders except for benzodiazepines, which produced greater improvement over the first 4 weeks compared to SSRIs and SNRIs in panic disorder. For SSRIs and SNRIs, women improved more than men, and for benzodiazepines, older patients improved more compared to younger patients. Finally, placebo response plateaued by week 4 of treatment, and, at week 8, social anxiety disorder trials had lower placebo response compared to other anxiety disorders.

CONCLUSIONS

Benzodiazepines show early improvement compared to SSRIs and SNRIs. However, by week 8, all treatments yield similar results. Patient characteristics influence the improvement trajectory and magnitude, suggesting potential for personalized medication selection.