Randomized controlled trial of the effects of high-dose ondansetron on clinical symptoms and brain connectivity in obsessive-compulsive and tic disorders

Abstract

Objective:

Sensory phenomena (SP) are aversive sensations driving repetitive behaviors in obsessive-compulsive disorder (OCD) and Tourette’s disorder (TD) that are not well addressed by standard treatments. SP are related to functioning of an interoceptive-sensorimotor circuit that may be modulated by the 5-HT3 receptor antagonist ondansetron. The present study employed an experimental medicine approach testing the effects of 4 weeks of high-dose ondansetron (24 mg, n=27) compared to placebo (n=24) on SP severity and brain connectivity in a cohort of individuals with OCD and/or TD.

Methods:

Analyses examined changes in SP severity and, for OCD participants, overall OCD severity from Baseline to Final visits. Functional magnetic resonance imaging (fMRI) data were collected at both visits for analysis of intrinsic functional connectivity metrics characterizing global correlation (reflecting area “hubness”) and local correlation (reflecting near-neighbor coherence).

Results:

There were no significant differences between ondansetron and placebo in the reduction of SP or overall OCD severity in the full sample. In a subsample of OCD participants taking concomitant serotonin-reuptake inhibitors (SRIs), ondansetron significantly decreased overall OCD severity and global connectivity of medial sensorimotor cortex more than placebo. Longitudinal reductions in SP severity were related to decreases in right sensorimotor hubness in both groups and brainstem local coherence only in participants taking ondansetron.

Conclusions:

There was no effect of high-dose ondansetron on SP. However, when used as an augmentation to SRIs, ondansetron reduced overall OCD severity, which may be related to changes in the “hubness” of sensorimotor cortex. Ondansetron’s ability to modulate brainstem connectivity may underlie its variable effectiveness in reducing SP.

Introduction

Sensory phenomena (SP) are aversive or uncomfortable sensations driving repetitive behaviors in Obsessive-compulsive disorder (OCD) and Tourette’s Disorder (TD). Although cognitive-behavioral models of OCD propose that compulsions are driven by the motivation to prevent harm and avoid feared outcomes(1, 2), many individuals with OCD experience SP including both physical urges and “not-just-right experiences” that do not involve a concrete fear(3–9). In TD, SP manifest primarily as premonitory urges or “sensory tics”(10, 11) involving the build-up of somatosensory discomfort in the body until a tic is performed(10–16). SP are highly distressing(10, 12, 16, 17) and associated with reduced quality of life(13, 18), yet are not well addressed by standard treatments in OCD or TD(8, 9, 13, 19, 20).

Research into the neural underpinnings of SP is limited but points to an insular-sensorimotor network involved in processing sensation (interoception and somatosensation)(21–28) and motor responses. Prior work has found that gray matter volume in sensorimotor cortex (including postcentral gyrus [PostCG] and paracentral lobule) was higher in a group of OCD patients with SP compared to without SP(29), and we previously found that greater severity of SP within a small OCD cohort was associated with increased activity in mid-insula and sensorimotor regions including PostCG(30).

One potential agent that could be used to modulate interoceptive-sensorimotor circuitry is the serotonin type 3 (5-HT3) receptor antagonist ondansetron, FDA-approved for nausea and vomiting in dosages ranging from 4–24 mg(31, 32). Although the neurobiological effects of ondansetron are unclear, a high density of 5-HT3 receptors are located in spinal and brainstem neurons that relay ascending interoceptive and somatosensory signals to higher-order cortical regions including the insula, PostCG, orbitofrontal cortex, and anterior cingulate cortex [ACC] (33–35). In addition to its anti-emetic effects, ondansetron has been used to treat other sensory-based symptoms including pruritus(36), neuropathic pain(37), and visceral hypersensitivity(38), suggesting a more general role for the agent in modulating signals from the body. In a double-blinded placebo-controlled crossover study comparing activation in a sensorimotor task following single doses of 8, 16, and 24 mg of ondansetron, we found a dose-dependent decrease of activity in the insula, sensorimotor cortex (including PostCG, precentral gyrus [PreCG], supplementary motor area [SMA]), temporal cortex, and ACC(39). Although that study was conducted in control participants, the findings support the use of higher ondansetron dosages to modulate interoceptive-sensorimotor circuits relevant for SP.

Prior clinical trials using ondansetron in OCD have yielded conflicting results. The largest study-to-date in 168 OCD participants failed to identify a reduction in Yale-Brown Obsessive-Compulsive Scale (Y-BOCS, 40) score following 12 weeks of adjunctive low-dose ondansetron (1–1.5mg)(41). Smaller trials utilizing higher dosages have been more encouraging, with recent meta-analyses in OCD finding that ondansetron in total daily dosages of 4 mg or higher outperformed placebo (data from the present trial was provided to these authors for the limited purpose of their meta-analysis(42) and other adjunctive pharmacotherapies including antipsychotics and glutamatergic agents(43) in reducing Y-BOCS score. Only one published study has compared ondansetron to placebo in TD, with results reporting a significant reduction in global severity scores with ondansetron monotherapy for 3 weeks using dosages of up to 24 mg(44). Despite these promising findings, individual treatment response varies considerably within cohorts(42, 43) and no study has tested ondansetron’s effects on SP specifically. The present investigation employed an experimental medicine, neuroscience-informed approach to test the effects of high-dose ondansetron (24 mg) compared to placebo for 4 weeks on SP severity (measured with the University of Sao Paulo-Sensory Phenomena Scale [USP-SPS, 6]) and resting-state (or intrinsic) functional connectivity in a cohort of individuals with OCD and/or TD. Increasing research highlights the benefits of investigating neurocircuit mechanisms of psychiatric disorders using measures of intrinsic connectivity, which have higher signal-to-noise and are less affected by performance-related confounds than task-activation approaches(45, 46). We predicted that participants randomized to ondansetron would exhibit a greater reduction in SP severity compared to placebo and that this effect would be associated with decreased connectivity in interoceptive and sensorimotor circuitry. The inclusion of longitudinal connectivity measures enables both the investigation of neural effects of ondansetron as well as an examination of how neurocircuit connectivity changes in association with changes in SP, information that could be used to develop additional treatments for these impairing symptoms.

Materials and Methods

Subjects and Procedure

Participants were randomized to 24 mg ondansetron or placebo in a double blinded fashion. Prior to randomization, eligible participants completed a Baseline visit where clinical symptoms were assessed and MRI scanning took place. Participants were instructed to take the initial dose of study drug upon waking on the first day following the Baseline visit and continue to do so at approximately the same time each day until the day of the Final visit, which occurred approximately 4 weeks later (target duration was 28 days with flexibility up to 40 days, see supplement for dosing and trial duration rational). On the day of the Final visit, participants took their final dose of study drug in the laboratory 90 minutes prior to MRI scanning and clinical symptoms were re-assessed. Table S1 shows a detailed schedule of events.

A target sample size of 60 was set to yield >80% power to detect differences in symptom reduction between ondansetron and placebo arms based on effect sizes from randomized placebo-controlled studies in OCD(47) and TD(44) available at the time of study design (see supplement). One-hundred-ten participants were enrolled, 62 were randomized, and 51 contributed complete clinical and MRI data from both the Baseline and Final visits (Figure 1; supplement describes withdrawals). The final sample of 51 completers (asterisk, Figure 1) included 45 patients with OCD (including 5 with comorbid TD) and 6 patients with TD (without comorbid OCD). Twenty-seven participants were randomized to ondansetron (OND) and 24 to matched placebo (PL). Participants were enrolled across three sites (Icahn School of Medicine at Mount Sinai, New York University Grossman School of Medicine, and Nathan Kline Institute for Psychiatric Research) between June of 2017 and May of 2022. The study was approved by Institutional Review Boards at all sites and participants provided written informed consent. The study is registered on clinicialtrials.gov (NCT03239210) and results from a priori-registered analyses of clinical outcomes (USP-SPS and Y-BOCS scores) are reported in the main text. A priori-registered neural analyses compared the effects of ondansetron vs. placebo on insula and somatosensory cortex activation during a sensorimotor fMRI task. Results from these task comparisons are reported online at clinicaltrials.gov and in the supplement.

Figure 1. Consort diagram of study enrollment.

Samples indicated with an asterisk and in bold font indicate participants used for the completer analysis. Results from sensitivity analyses conducted using last observation carried forward (LOCF) analyses in the intent-to-treat sample consisting of all randomized participants (samples indicated with +) and per-protocol (PP) analyses (excluding those with protocol deviations, samples indicated with ^) on clinical outcomes did not differ from completer sample findings and are presented in the supplement. Two participants did not take the medication at the Final visit: the participant in the placebo group missed only the last dose (left medication at home) and the participant in the ondansetron group missed the last dose as well as the preceding 3 doses (lost medication near end of trial). Results from sensitivity analyses without these and other individuals with protocol deviations are described in the supplement.

DSM-5 diagnoses were made using the Mini-International Neuropsychiatric Interview (MINI, 48, see supplement). As per protocol, participants taking medication and those with comorbidities were permitted to increase the population validity of the trial, with effects tested post-hoc(49). Lifetime history of schizophrenia spectrum, bipolar, alcohol or substance use (greater than mild), or major developmental disorders was exclusionary; the most frequent comorbidities were anxiety and attention-deficit-hyperactivity disorders (Table S4 list comorbidities). Half of completers (n=26, 51%) were free of concomitant psychotropic medications. The remaining (n=25) were taking psychotropic medication consisting primarily of serotonin-reuptake-inhibitors (SRIs; 88%, Tables S2, S3), on stable dosages for at least 6 weeks before enrollment. Ten participants were in concurrent psychotherapy that had been initiated prior to randomization and none were in cognitive behavioral therapy (CBT) for OCD that had been initiated within the past 12 months. Twenty-nine participants had a history of at least one past psychotropic medication trial. Importantly, none of these variables differed between ondansetron and placebo groups. USP-SPS score of moderate or higher (>6) was required for inclusion to enrich the sample with higher-SP participants. Baseline characteristics are in Table 1 and elaboration of inclusion/exclusion criteria, screening procedures, and baseline characteristics are in the supplement.

Table 1:

Demographic and clinical characteristics in full sample of completers and OCD cohort.

	Ondansetron		Placebo		Group comparisons
	mean/N	SD/percent	mean/N	SD/percent
All completers (n=51)	N=27		N=24
Baseline
Age (years)	31.0	10.0	29.1	12.5	ns
Education (years)	15.6	1.6	15.0	2.1	ns
Sex assigned at birth (M / F)	14 / 13		10 / 14		ns
Concomitant medication use (Y / N)	12 / 15		13 / 11		ns
DSM-5 comorbidity (Y / N)	21 / 6		16 / 8		ns
Doses received	28.9	5.2	30.7	3.3	ns
University of Sao Paulo-Sensory Phenomena Scale (USP-SPS) total	9.2	1.8	9.1	2.3	ns
Baseline vs. Final: USP-SPS
Change score (Baseline–Final)	1.2	2.1	1.6	2.1	ns
N decreased (from Baseline to Final)	16	59%	18	75%	ns
OCD (n=45)	N=23		N=22
Baseline
Age (years)	31.3	10.6	30.0	12.8	ns
Education (years)	15.4	1.7	15.1	2.2	ns
Sex assigned at birth (M / F)	11 / 12		8 / 14		ns
Concomitant medication use (Y / N)	10 / 13		12 / 10		ns
DSM-5 comorbidity (Y / N)	18 / 5		16 / 6		ns
Doses received	28.8	5.6	30.9	3.4	ns
Yale-Brown Obsessive Compulsive Scale (Y-BOCS) total	26.6	3.9	22.8	6.2	t43=2.4, p=0.02^
Baseline vs. Final: Y-BOCS
Change score (Baseline–Final)	4.0	4.9	2.9	5.9	ns
N decreased (from Baseline to Final)	20	87%	13	59%	χ2 =4.5, p=0.04+

For continuous variables (age, education, doses received, baseline USP-SPS, baseline Y-BOCS, and change scores), values in cells represent means and standard deviations (SD). For group variables (sex assigned at birth, concomitant medication use, and DSM-5 comorbidity), values in cells represent number (N) of participants in each group. “N decreased” rows show the number and percentage of participants who exhibited any decrease in clinical score from Baseline to Final visits.

^{^}Participants taking ondansetron had higher baseline Y-BOCS scores than those taking placebo despite random assignment to group.

⁺Significantly higher proportion of OCD participants in the ondansetron group exhibited reductions in Y-BOCS scores from Baseline to Final visits than the placebo group.

Clinical assessments

The primary clinical outcome was change in SP severity from Baseline to Final visits as assessed with the USP-SPS(6) (see supplement). The secondary clinical outcome was overall OCD symptom severity in the cohort of OCD completers (n=45) using the Y-BOCS (40). The Yale Global Tic Severity Scale (YGTSS (50) was completed in participants with TD (n=11) but not analyzed further due to the small sample size (baseline scores are in supplement). Symptom dimensions in the OCD cohort are described in the supplement (Table S5). Adverse events (AEs) were assessed for the entire randomized sample using the patient-rated inventory of side effects (PRISE (51)) at the Baseline visit and at the end of each week throughout the trial. Further details on AE assessment are in the supplement.

Neuroimaging data acquisition and preprocessing

Participants underwent two eight-minute eyes-open resting-state scans (480 volumes, TR=1s) on Baseline and Final visits. The CONN-fMRI Functional Connectivity Toolbox(v21a)(52) was used for individual-level denoising, which included motion scrubbing and the removal of noise and other artifactual confounds using component-based correction. Data were band-pass filtered between 0.008 and 0.09 Hz. Acquisition and processing details are in the supplement.

Data Analysis

Clinical symptoms: USP-SPS and Y-BOCS

Separate mixed-model analyses-of-variance (ANOVAs) analyzed USP-SPS or Y-BOCS score as dependent variables with visit (Baseline, Final) and treatment condition (OND, PL) as independent variables. Main effects of visit and treatment condition and their interactions were interrogated. Additionally, separate 2 x 2 ANOVAs examined change in USP-SPS or Y-BOCS score from Baseline to Final visits (larger numbers reflecting bigger decreases) as dependent variables with treatment condition (OND vs. PL) and concomitant psychotropic medication (presence vs. absence) as independent variables. Different 2 x 2 ANOVAs that were identical to those described above except that the concomitant medication variable was replaced with a psychiatric comorbidity variable (also as presence vs. absence). Main effects of both variables and their interactions were interrogated in all ANOVAs. Given the original aim to investigate transdiagnostic mechanisms of SP and the small sample size of TD participants, analysis of USP-SPS scores were not stratified based on diagnosis.

There was considerable between-subject variability in change scores during the trial, with some individuals in both OND and PL groups showing decreases in USP-SPS and Y-BOCS scores and others not. Responders were formally defined as participants who had a decrease in score by at least 35%(53), the proportions of which were compared between OND and PL using chi-square analyses. Additional analyses compared OND and PL groups in the proportion of participants showing any decrease in clinical scores between Baseline and Final visits.

To characterize the robustness of clinical findings from the completer sample, multiple sensitivity analyses were conducted on clinical outcomes. Intent-to-treat (ITT) analysis in the full randomized sample used last observation carried forward (LOCF; +in Figure 1). Per-protocol sensitivity analysis (PP; în Figure 1)(54) excluded up to 7 individuals with protocol deviations. Additional sensitivity analyses tested the effects of excluding one OCD participant in the OND group who exhibited a much larger clinical response (20-point reduction in Y-BOCS score) than other participants in the OND group. Finally, due to the high overlap of symptoms driving scoring for the USP-SPS and Y-BOCS (r=0.7, see supplement), follow-up analyses of neural effects related to USP-SPS statistically adjusted for Y-BOCS score in order to determine specificity to SP. Overall, results from sensitivity analyses did not differ substantively from those obtained in the completer analyses and are presented in the supplement.

Although not a main aim, exploratory analysis probed for baseline demographic and clinical predictors of USP-SPS and connectivity changes. Results did not reveal any significant predictors and are presented in the supplement.

Neuroimaging data: functional connectivity

Connectivity metrics

Two whole-brain functional connectivity measures were examined. First, global correlation measured the average correlation between the timecourse of a given voxel and every other voxel in the brain(52), with greater global correlation thought to reflect greater “hubness” or interconnectedness of an area, akin to degree centrality(55). Second, integrated local correlation measured the average correlation between the timecourse of a given voxel and its nearest neighbors(52, 56), with greater local correlation thought to reflect increased local coherence of an area. Local correlation is similar to the measure of regional homogeneity(57, 58), with several noted advantages(56).

Group-level modelling

Four regression models (see supplement) examined changes in connectivity (Baseline-Final) for global and local correlation metrics with the following predictors: treatment condition (OND, PL); change in clinical score from Baseline to Final visits (given large variability of symptom changes within both treatment conditions, Figure 2); and the interaction between treatment condition and clinical change score (to identify whether neural correlates of symptom change were specific to treatment). Separate models were created using USP-SPS as the score in the full completer sample and Y-BOCS as the score in the OCD-only cohort.

Figure 2. Change in clinical symptoms from Baseline to Final visits.

From left to right for (A), (B), and (C): Clinical scores at Baseline and Final visits for each participant within the ondansetron (blue) and placebo (gray) groups (shades of colors are varied to assist in visually distinguishing participants and do not reflect direction of symptom change). Yellow markers are placed at the group means for each time point. Pie charts display the number of participants in each group who showed a decrease in clinical score from Baseline to Final visits (solid black slice) compared to those who did not (striped black slice). Bar graphs show mean difference score (Baseline minus Final) for ondansetron (blue) and placebo (gray) groups (positive difference reflects a decrease in score). Error bars represent standard error of the mean. A) University of Sao Paulo-Sensory Phenomena Scale (USP-SPS) scores for all completers (n=51); B) Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) scores displayed for all OCD completers (n=45); C) Y-BOCS scores displayed for OCD completer subsample taking concomitant psychotropic medications (n=22).

Due to symptom results showing greater reduction of Y-BOCS scores in OND than PL groups within the subsample of OCD patients taking concomitant psychotropic medications (see Results, Table S3), we also conducted a targeted fMRI analyses comparing OND and PL groups in change in connectivity from Baseline to Final visits specifically within this OCD subsample.

Following recommendations(59), effects of site-based clustering in connectivity were addressed by including site as a covariate (predictor) for all imaging analyses, as implemented in other multisite neuroimaging studies(60, 61). Neuroimaging analysis employed rigorous significance testing using non-parametric permutation tests (10,000 permutations) with whole-brain cluster-level family-wise-error-correction to p<0.05 with cluster-defining threshold of p<0.0025.

Results

Adverse events (AEs)

Overall, high-dose ondansetron was well-tolerated and safe with no serious AEs occuring. Non-serious AEs with an incidence of at least 5% in the randomized sample were constipation, headache, diarrhea, dizziness, fatigue, dry mouth, nausea, and restlessness (Table S6). The only AE that exhibited a significantly higher incidence in OND than PL was constipation (χ²=17.3, p<0.001), which occurred in 52% of participants taking ondansetron and 3% of participants taking placebo. Additional details on AEs as well as compliance are in the supplement.

Clinical symptoms

Average change in USP-SPS and Y-BOCS scores

In the analysis of USP-SPS scores in the full completer sample, both OND and PL groups showed small but significant decreases from Baseline to Final visits (main effect of visit, F_1,49=22.9, p<0.001), which did not differ between groups (no interaction between visit and treatment)(Figure 2A, Table 1). Similar effects were found within the cohort of OCD participants, where a main effect of visit (main effect of visit, F_1,43=18.8, p<0.001) on USP-SPS scores was found, with no other effects.

In an analysis of Y-BOCS scores within the OCD cohort, there were significant reductions based on visit (F_1,43=17.9, p<0.001), which also did not differ between OND and PL groups (no interaction)(Figure 2B, Table 1). There was also a main effect of treatment condition (F_1,43=4.8,p=0.034) due to higher Y-BOCS scores in the OND group at both visits despite random assignment to treatment.

There were no main effects of or interactions with concomitant psychotropic medication use on USP-SPS change score (Table S2). For the parallel analysis of Y-BOCS change scores in OCD participants (Table S3), there was a significant interaction between treatment condition and concomitant medication use (Cohen’s f=0.32, F_1,41=4.6, p=0.039). The OND group exhibited a larger change score (computed as Baseline minus Final) than the PL group within the subsample of OCD patients taking concomitant psychotropic medications (Figure 2C), but not for medication-free OCD participants. Of note, each concomitantly-medicated OCD participant in the OND group was taking an SRI (Table S3). Interestingly, this interaction effect was driven by a lack of placebo response in concomitantly-medicated participants (Figure 2C), which could be due to the fact this subsample was not only taking concurrent SRIs but also has a greater proportion of individuals with past SRI use (50%) than medication-free participants (30%) (see supplement).

Analyses examining the effects of psychiatric comorbidity on USP-SPS and Y-BOCS change scores revealed no main effects or interactions with treatment condition in the full sample or OCD cohort.

Proportions of participants showing decreases in clinical scores

When examining USP-SPS change scores, neither the proportion of individuals classified as responders (score reduction of at least 35%) nor the proportion of individuals exhibiting any decrease in USP-SPS score differed between OND and PL groups (Figure 2A, Table 1).

When examining Y-BOCS scores in the OCD cohort, the proportion of responders did not differ between OND and PL groups. However, a higher proportion of OCD participants taking OND had any decrease in Y-BOCS scores from Baseline to Final visits than PL (OND 87% vs PL 59%; χ²= 4.5, p=0.03; Odds Ratio [OR]=4.62, Figure 2B, Table 1). In OCD participants taking concomitant medications, the proportion showing a decrease in Y-BOCS scores from Baseline to Final visits was significantly higher for OND (90%) than PL (33%) (χ²= 7.2, p=0.01; OR=18, Figure 2C) groups. By contrast, there was no difference between OND and PL groups in the proportion showing a Y-BOCS decrease amongst medication-free OCD participants.

Functional connectivity

The neural findings reported here are derived from secondary analyses of study data. Results from the primary (registered) neural analyses failed to confirm a priori hypotheses of ondansetron-related reductions in insula and somatosensory cortex activation during an fMRI task previously shown to be sensitive to single doses of the drug (39). In order to strike a balance between transparency and the dissemination of findings to advance understanding of OCD and SP, the description and discussion of results from task-based analyses are detailed in the supplement.

Changes in global correlation from Baseline to Final visits

For the model in all completers, there was a significant main effect of USP-SPS change score such that participants with larger reductions in USP-SPS scores from Baseline to Final visits also exhibited greater decreases in the global connectivity of right sensorimotor cortex (including PostCG and PreCG, k=652, 50,−6,44, BAs 3/4/6, FWE-p<0.037), irrespective of treatment condition (Figure 3, top). There was no main effect of Y-BOCS change score that survived whole-brain correction in the OCD cohort model (supplement describes subthreshold findings) and no other effects in the full completer or OCD cohort models.

Figure 3. Change in functional connectivity associated with reduction of symptoms from Baseline to Final visits.

Top panel: Greater decrease of global correlation in right sensorimotor cortical regions including postcentral and precentral gyri was associated with larger reduction in sensory phenomena (USP-SPS score) in both ondansetron (blue circles) and placebo (gray squares) groups. Bottom panel: Greater decrease in local correlation in the brainstem was associated with a larger reduction in USP-SPS score in the ondansetron (blue) but not placebo (gray) groups. For both panels, y-axes plot the difference in connectivity values (z-transformed correlation coefficients, calculated for each metric as described in the Methods) between Baseline and Final visits extracted from region shown after regressing out effects of site. Higher values for both axes represent larger decreases (in USP-SPS score and connectivity) from Baseline to Final visits. Colorbar represents t-score.

Changes in local correlation from Baseline to Final visits

For the model in all completers, there was a significant interaction between treatment condition and USP-SPS change score in a large area of the brainstem (k=743, −4, 46, 62, FWE-p<0.031, Figure 3 bottom) such that greater decreases in local correlation were associated with larger reductions in USP-SPS score in the OND but not PL group. There were no interactions between treatment condition and Y-BOCS change score that survived whole-brain correction in the OCD cohort model (supplement describes subthreshold findings) and no other significant effects in the full completer or OCD cohort models.

Changes in connectivity based on concomitant medication use in OCD

Due to the above-described symptom results indicating that OND reduced Y-BOCS score more than PL in the subsample of OCD patients taking concomitant psychotropic medication but not medication-free participants, targeted analyses probed for changes in connectivity metrics that might mirror this clinical finding. In OCD patients taking concomitant medications, there was a significantly greater decrease for OND compared to PL in the global correlation of an area of medial and right sensorimotor cortex (PostCG and SMA, k=530, 24, −32, 60, BAs 3/4/6, FWE-p=0.042, Figure 4). By comparison, in concomitant medication-free OCD participants, there were no global correlation differences between OND and PL.

Figure 4. Ondansetron-related changes in functional connectivity within the OCD subsample taking concomitant psychotropic medications.

Significantly greater decrease in global correlation of a medial sensorimotor region including postcentral gyrus and supplementary motor area (SMA) for participants taking ondansetron (blue) compared to placebo (gray). Bar graph plots the difference in connectivity values (z-transformed correlation coefficients, calculated for global correlation metric as described in the Methods) between Baseline and Final visits extracted from region shown after regressing out effects of site. Higher values represent larger decreases in connectivity during the trial. Colorbar represents t-score.

Analyses of local correlation changes from Baseline to Final visits found no significant differences between OND and PL groups in OCD participants taking concomitant medications. However, in concomitant medication-free OCD participants, local correlation decreased from Baseline to Final visits significantly more in PL compared to OND in left occipital cortex (Figure S1).

Discussion

Contrary to hypotheses, four weeks of high-dose ondansetron did not reduce the severity of sensory phenomena or overall symptoms of OCD in a heterogenous but representative (see supplement) sample of individuals with OCD and/or TD. However, in the subsample of OCD participants who were taking concomitant psychotropic medications (Table S3), those receiving ondansetron had a significantly greater reduction in overall OCD severity (Y-BOCS score) and a decrease in the global connectivity of medial sensorimotor cortex than those taking placebo, effects that were not found in OCD participants who were not taking concomitant medications. In the group of concomitantly-medicated OCD patients taking OND, all were on at least one SRI. Although this result was not predicted a priori, the majority of prior studies of 5-HT3 receptor antagonists in OCD have utilized ondansetron as an SRI augmentation(42, 62), an approach supported by the current findings. These findings contrast with the negative results obtained in a much larger randomized trial in OCD testing the efficacy of SRI augmentation with very low-dose ondansetron (41), perhaps due to differing dosages. Although ondansetron is highly permeable to the blood brain barrier (63), it is a substrate for the efflux transporter P-glycoprotein (63) and is measured in cerebrospinal fluid (CSF) at approximately 15% of the concentration detected in plasma after oral dosing (64). It is thus possible that high doses are required for sufficient brain penetrance. The effect size of the difference between OND and PL in reduction of Y-BOCS score for concomitantly-medicated OCD participants was moderate-to-large, but the absolute magnitude of the change in Y-BOCS score from Baseline to Final visits in the ondansetron group was slightly over 4 points. While not a large change, it represents a 15–20% drop for individuals with moderate-to-severe OCD, consistent with prior studies using short-term ondansetron augmentation in OCD(47, 65). A trial duration longer than 4 weeks would be feasible given the safety and tolerability of high-dose ondansetron in this trial and might have yielded a larger clinical response. This is a particularly important consideration for future work given that the reduction in Y-BOCS score was less than the 25% decrease required to define partial treatment response (53) and it remains unclear whether this amount of symptom change translates into broader clinical and functional improvement in individuals with OCD.

The paradoxical effect whereby the blockade of serotonin at 5-HT3 receptors (via ondansetron) augments the effects of serotonin-reuptake inhibitors has been previously explained via reference to effects of 5-HT3 antagonism on neurotransmitter systems other than serotonin(66, 67). The majority of 5-HT3 receptors are co-localized on GABA interneurons(68, 69), where their blockade contributes to downstream effects on multiple neurotransmitter systems including the inhibition of cortico-mesolimbic dopamine(31, 62, 68). It is possible that synergistic effects on serotonin and dopamine systems may underlie ondansetron’s effectiveness as an augmentation for SRIs, similar to mechanisms proposed to explain the efficacy of antipsychotic augmentation in OCD(66, 67, 70). It has also been shown that 5-HT3 receptor blockade may actually lead to greater serotonin release through a multi-system feedback loop in which 5-HT3 blockade reduces GABA inhibition of pyramidal glutamate neurons, which in turn stimulate not only serotonin release but also acetylcholine and norepinephrine(71, 72). Finally, rather than being driven by neurotransmitter interactions between SRIs and ondansetron, the efficacy of ondansetron only in OCD participants taking SRIs could be related to the relative treatment-resistance of this group compared to medication-free participants. OCD participants taking SRIs exhibited clinically impairing symptoms despite current and past SRI utilization (see supplement), and it is possible that the biological mechanisms associated with treatment-resistant OCD are particularly sensitive to modulation by ondansetron. Although there are several potential reasons for the clinical effects found in this study, the localization of changes to global connectivity of medial sensorimotor cortex suggests that this brain area may be the target of these effects.

Greater reduction of USP-SPS scores in both treatment conditions was related to a decrease of global connectivity in right sensorimotor cortex, primarily in postcentral gyrus. This finding provides additional support for the link between SP and postcentral gyrus, which we previously identified using a cross-sectional analysis in a small sample of OCD patients(30). As with our prior study, SP in this study was related to postcentral gyrus connectivity even after accounting for variance explained by overall OCD severity (see analysis in supplement), indicating that hyperfunctioning of this area may be a biomarker of SP specifically. We also previously found that activity in the same right postcentral region was significantly reduced in a group of controls following single 24 mg doses of ondansetron compared to placebo in a sensorimotor fMRI task(39), yet an analysis of postcentral activation in the present trial using the same task did not replicate this finding (see supplement for discussion of potential reasons). Although it is unclear why functional connectivity analyses revealed sensorimotor changes that were not evident in task analyses (results in supplement), it is conceivable that the task’s ability to robustly elicit activation in the sensorimotor circuit overwhelmed ondansetron’s ability to modulate it. That is, the task effect may have been stronger than the ondansetron effect. Additional research will be needed to determine whether biomarkers of treatment response to ondansetron extend to any type of task-evoked activity or are limited to measures of functional connectivity during an unconstrained resting state.

Interestingly, local correlation in a large region of the brainstem decreased from Baseline to Final visits more in participants with greater reductions in sensory phenomena, but only in the ondansetron group. The brainstem is a known target of ondansetron, containing a high density of 5-HT3 receptors in several brainstem nuclei including the nucleus of the solitary tract(31, 68), an area that is contained within the identified brainstem cluster. These results suggest that ondansetron’s efficacy in reducing sensory phenomena may depend on its ability to alter the function of brainstem targets of the drug. It is unclear why ondansetron would modulate brainstem connectivity for some individuals but not others, and exploratory analyses did not find any baseline clinical predictors of ondansetron-related brainstem connectivity changes (see supplement). Future work will be needed to better understand individual factors affecting ondansetron’s effects on the brain, with the goal of facilitating its modulation of the brainstem in order to reduce SP.

There are several limitations to the current study. Although we used stringent multiple comparison correction in neuroimaging analyses, tests on clinical outcomes were not corrected and should be interpreted with caution. The higher baseline Y-BOCS score in participants taking ondansetron compared to placebo despite random assignment to group highlights the need for future trials to enroll larger samples and prioritize baseline matching. Despite the harmonization of scanner sequences and statistical adjustment for site, the use of two scanners for data acquisition in the relatively small sample may have introduced noise that negatively impacted power. Although we aimed for a cohort with a more even distribution of TD and OCD participants, the final sample of TD participants was too small to enable conclusions regarding ondansetron in that group. Indeed, only two individuals in the OCD subsample taking SRIs had comorbid TD, and future work with larger samples will be important to determine whether the presence of comorbid tics influences ondansetron’s efficacy. The relatively short-term 4-week duration of the trial was chosen based on a variety of factors (see supplement) but may not have been sufficiently long particularly when using ondansetron as an augmentation to SRIs(65). Despite these limitations, this is the first study to examine the effects of high-dose ondansetron on SP and brain connectivity, with results revealing the potential utility of ondansetron as an augmentation to SRIs in OCD as well as biomarkers of clinical change that could be the target of future treatment development.