Research Frontier | Retrospective Comparison of Transcutaneous Auricular Vagus Nerve Stimulation and Medication in Major Depressive Disorder
Introduction
For major depressive disorder (MDD), medication is not an option that every patient can tolerate or benefit from adequately. Transcutaneous auricular vagus nerve stimulation (taVNS) is being investigated as a nonpharmacological intervention, but whether it can improve core depressive symptoms when compared directly with a commonly used antidepressant requires cautious interpretation.
On October 22, 2026, Yu-Lin Wu and colleagues published an article online in BMC Psychiatry that retrospectively compared 8-week outcomes of taVNS and escitalopram, a selective serotonin reuptake inhibitor, for the treatment of MDD. After propensity score matching, each group included 120 patients.
The study's central finding was not that taVNS outperformed medication. The prespecified primary endpoint, change in Montgomery-Åsberg Depression Rating Scale (MADRS) score, did not differ significantly between groups. Between-group differences did emerge in sleep, cognition, anhedonia, and heart rate variability. These findings are better viewed as signals for subsequent trials than as an established efficacy advantage.
Study Background :Evidence Gap
In addition to persistent low mood, MDD is often accompanied by sleep disturbance, cognitive impairment, and anhedonia. Escitalopram primarily acts on the serotonergic system and is commonly used in clinical practice, but its onset of action takes time, and some patients experience adverse effects such as gastrointestinal discomfort, somnolence, and dry mouth. Cognitive impairment and anhedonia may not improve in parallel with core depressive symptoms.
taVNS uses an auricular electrode to stimulate the auricular branch of the vagus nerve and differs from both implanted vagus nerve stimulation and transcutaneous cervical vagus nerve stimulation. Its afferent signals may influence brain regions involved in mood, autonomic regulation, and arousal through the nucleus tractus solitarius (NTS). However, most previous evidence has come from sham-controlled trials or small studies. Direct comparisons of taVNS with standard medication in clinical populations with similar baseline characteristics remain limited.
Study Methods :Retrospective Cohort
The investigators reviewed patients treated at a single center in outpatient or inpatient settings from September 2021 to September 2025. Eligible patients were 18-65 years old, met the diagnostic criteria for MDD, had a baseline total score of at least 24 on the 17-item Hamilton Depression Rating Scale (HAMD-17), and had a MADRS total score greater than 35. Treatment was selected by the clinician based on patient preference, previous efficacy and tolerability, device availability, cost, and other factors rather than being randomly assigned.
The original cohort comprised 289 eligible patients: 152 in the escitalopram group and 137 in the taVNS group. The study used 1:1 nearest-neighbor propensity score matching with a caliper of 0.02, retaining 120 patients in each group. After matching, no significant differences were observed in measured variables such as age, sex, body mass index, educational attainment, duration of the current episode, recurrence status, and baseline symptom scores. Propensity scores can balance only variables that were recorded and entered into the model; they cannot eliminate unmeasured confounding or confounding by indication.
Escitalopram was started at 10 mg/day, adjusted within 1-2 weeks according to tolerability and clinical response, and maintained at 10-20 mg/day. taVNS was applied to the left cymba conchae at a frequency of 20 Hz and a pulse width of 200 μs. Current was adjusted to 1-5 mA according to each patient's sensory threshold. Sessions lasted 30 minutes and were given on 5 consecutive days followed by 2 rest days each week for 8 weeks. Neither group received other antidepressants, antipsychotics, or mood stabilizers during treatment.
The primary endpoint was the change in MADRS score from baseline to week 8. Other depression measures, the Hamilton Anxiety Rating Scale (HAMA), sleep, cognition, anhedonia, heart rate variability (HRV), and safety measures were designated exploratory outcomes. Continuous outcomes were analyzed primarily by analysis of covariance with the baseline value as a covariate, and adjusted between-group differences with 95% confidence intervals were reported. The P values for exploratory analyses were not adjusted for multiple comparisons. Neither patients nor outcome assessors were blinded, and the analysis included only patients who completed treatment and follow-up.
Study Results:Primary Outcome
At week 8, the changes in MADRS score from baseline were −22.72 ± 6.04 points in the escitalopram group and −22.26 ± 6.08 points in the taVNS group; the between-group comparison yielded P = 0.530. After adjustment for baseline, the between-group difference in week 8 MADRS score was 0.30 points (95% confidence interval, −1.07 to 1.67; P = 0.667). Clinical response rates were 65.83% and 63.33%, respectively (P = 0.686), and remission rates were 39.17% and 36.67%, respectively (P = 0.690).
The HAMD-17 findings were similar: changes were −18.64 ± 5.75 points and −18.51 ± 5.30 points in the two groups, respectively (P = 0.852), and the adjusted between-group difference at week 8 was 0.28 points (95% confidence interval, −0.91 to 1.47; P = 0.641). Anxiety scores decreased in both groups, but the between-group difference was not significant. Thus, the study did not find taVNS to be superior to escitalopram for core depressive or anxiety symptoms. Because this was not a noninferiority trial, the absence of a significant difference cannot be interpreted as equivalence or as evidence that taVNS is noninferior to medication.

Figure 1 Outcomes for Core Depressive Symptoms
The two tables show week 8 MADRS and HAMD-17 scores, changes from baseline, and MADRS-defined clinical response and remission rates. None of the key between-group comparisons reached statistical significance.
Exploratory Outcomes
For sleep, the adjusted between-group difference on the Pittsburgh Sleep Quality Index (PSQI) was −0.75 points (95% confidence interval, −0.83 to −0.66; P < 0.001), favoring taVNS. At week 8, 31.67% of patients in the taVNS group and 18.33% in the escitalopram group achieved good sleep (PSQI < 5; P = 0.017).

Figure 2 Anxiety and Sleep Outcomes
In the upper panel, green represents the escitalopram group and magenta represents the taVNS group. Anxiety scores decreased in both groups, but the between-group difference was not significant. The lower table shows an exploratory difference in sleep outcomes favoring taVNS.
The mean change from baseline on the Montreal Cognitive Assessment (MoCA) was −0.02 ± 3.46 points in the escitalopram group and 0.96 ± 3.31 points in the taVNS group. The adjusted between-group difference was 0.67 points (95% confidence interval, 0.09 to 1.26; P = 0.024). Changes on the Snaith-Hamilton Pleasure Scale (SHAPS) were −11.43 ± 7.97 and −13.44 ± 7.19 points, respectively, with an adjusted between-group difference of −1.37 points (95% confidence interval, −2.68 to −0.07; P = 0.040). Both results favored taVNS, but they were exploratory findings without adjustment for multiple comparisons. Repeated MoCA testing may also be affected by practice effects.
Unadjusted HRV changes also favored taVNS. The median change in the standard deviation of normal-to-normal intervals (SDNN) was 6.11 ms versus 2.33 ms (P = 0.006); the root mean square of successive differences between normal-to-normal intervals (RMSSD) was 4.99 ms versus 2.06 ms (P = 0.016); and the low-frequency/high-frequency power ratio (LF/HF) was −0.43 versus −0.16 (P < 0.001). However, in Table 6 of the original article, the direction of the adjusted RMSSD difference and the magnitude of the adjusted LF/HF difference are inconsistent with the within-group changes reported in the same table.

Figure 3 Cognitive Anhedonia and Heart Rate Variability Outcomes
In the upper panel, blue represents the escitalopram group and green represents the taVNS group. The simple between-group P values at week 8 shown in the figure are not the same statistics as the baseline-adjusted analyses of covariance reported in the text. The lower table is reproduced unchanged; as noted above, the adjusted between-group differences for RMSSD and LF/HF are internally inconsistent, so statistical significance alone should not be used to infer a mechanism.
During the 8-week period, nausea or gastrointestinal discomfort, somnolence or daytime fatigue, and dry mouth occurred in 18.33%, 14.17%, and 11.67% of the escitalopram group, respectively, compared with 3.33%, 2.50%, and 1.67% of the taVNS group. In contrast, local ear-skin redness or stinging occurred in 13.33% of the taVNS group and in none of the escitalopram group. All of these differences were statistically significant.
Serious adverse events, suicidal behavior or attempts, and treatment-emergent mania did not differ significantly between groups. However, the event counts were small and do not establish that the risks were equivalent. The electronic medical records did not systematically grade adverse-event severity. In addition, the ear was inspected after every taVNS session, so local reactions may have been more readily detected than adverse effects in the medication group.

Figure 4 Adverse Events
Local ear-skin redness or stinging predominated in the taVNS group, whereas the escitalopram group reported more nausea or gastrointestinal discomfort, somnolence or daytime fatigue, and dry mouth. The table describes records from patients who completed 8 weeks of treatment and does not represent discontinuation or tolerability among all patients who started treatment.
Mechanistic Discussion:Observations and Inferences
The authors proposed that taVNS may regulate sleep-wake networks through the vagal afferent-NTS pathway, influence attention and working memory through the locus coeruleus-norepinephrine system, and contribute to improvement in anhedonia through midbrain reward pathways. Although previous research supports these explanations, the present study did not collect neuroimaging, neurotransmitter, or inflammatory measures and did not manipulate these pathways. It therefore cannot establish a specific mechanism.
HRV provides a physiological window into autonomic change, but it is not direct evidence that a vagal mechanism has been established. The study used only a 5-minute resting measurement at a fixed time of day. Rhythm classification relied on automated device detection rather than independent electrocardiographic interpretation, and all patients were included in the HRV analysis without additional rhythm-based exclusions. Together with the internal numerical issues in Table 6, the HRV findings should be treated as exploratory signals requiring confirmation.
The study's principal limitations arise from its design: it was single-center, retrospective, nonrandomized, and unblinded. Treatment selection was influenced by patient preference, previous treatment experience, device availability, and cost. The analysis included only patients who completed treatment and follow-up and did not report all interim discontinuations. Residual confounding may therefore explain some differences despite propensity score matching. Eight weeks of observation is also insufficient to determine whether benefits persist, whether relapse occurs after treatment stops, or how safe the interventions are over the long term.
Future Directions:Randomized Validation
The next step should be a multicenter, preregistered randomized trial with a sham-stimulation or active-medication control. Blinded assessment and intention-to-treat analysis should be used to test the main findings, and the protocol should prespecify the priority of sleep, cognition, anhedonia, and HRV outcomes together with the approach to multiplicity adjustment. Discontinuation, adherence, and adverse-event severity should be documented, and follow-up should be extended to determine whether benefits persist and whether effect sizes reach a clinically important difference that patients can perceive. Objective sleep monitoring, standardized electrocardiographic interpretation, and neuroimaging or biomarker measures could also be incorporated to test whether physiological changes truly mediate clinical improvement and to compare stimulation doses and identify patients most likely to benefit.
References
Wu, Y.-L., Ding, L., Yu, C.-Y., et al. (2026). Comparative effectiveness of transcutaneous auricular vagus nerve stimulation and escitalopram for depressive symptoms in major depressive disorder. BMC Psychiatry. Article in Press. https://doi.org/10.1186/s12888-026-08785-5