There are many effective medications for panic disorder, including tricyclic antidepressants (TCAs), selective serotonin reuptake inhibitors (SSRIs), serotonin-norepinephrine reuptake inhibitors (SNRIs), and monoamine oxidase inhibitors (MAOIs). Researchers used a network meta-analysis to simulate head-to-head comparisons between them. Specifically, the probability that an agent had the best balance of efficacy and tolerability was measured using the surface under the cumulative ranking (SUCRA). The higher the SUCRA, the better the medication's rank.

The investigators searched three databases to identify RCTs that evaluated medications to treat panic disorder with or without agoraphobia. A total of 87 studies involving 12,800 participants and 12 drug classes met the inclusion criteria. Their mean patient age was 35, 64% were female, and the average duration of panic disorder was 6.9 years. The two primary outcomes were remission, defined as no panic attacks for at least one week at the end of a study, and dropouts. Secondary outcomes were adverse effects and symptom scores for depression and anxiety as measured by the Hamilton Depression Rating Scale, Montgomery-Åsberg Depression Rating Scale, and Hamilton Rating Scale for Anxiety.

Results

Compared to placebo, remission rates were significantly higher with benzodiazepines (relative risk [RR]=1.47), TCAs (RR=1.39), SSRIs (RR=1.38), MAOIs (RR=1.30), and SNRIs (RR=1.27). SUCRA rankings echoed these findings, placing benzodiazepines, TCAs, and SSRIs at the top. Buspirone and beta blockers ranked lowest.

Dropout rates were lowest for benzodiazepines and TCA-benzodiazepine combinations, and highest for buspirone and MAOIs. However, adverse events were also most frequent with TCAs and benzodiazepines, and least common with buspirone and SNRIs. Overall, balancing efficacy and tolerability favored SSRIs, especially sertraline and escitalopram. Fluvoxamine, paroxetine, and fluoxetine showed strong efficacy, but also higher rates of side effects. Citalopram had poorer efficacy and greater adverse effects.

Network meta-analyses assume that included studies of varied design are comparable. This analysis was further limited by a high risk of within-study bias, inconsistency and imprecision of findings, and short trial durations.