SSRI treatment of obsessive-compulsive disorder as motion across a bistable fold: a calibrated circuit plasticity model of response, remission, and augmentation

Technical Summary: SSRI Treatment of OCD as Motion Across a Bistable Fold

Problem Statement
Current quantitative accounts of psychiatric treatment often operate at isolated scales: pharmacokinetic models describe receptor occupancy, biophysical models describe neural circuits, and statistical models describe dose–response curves. These levels are rarely coupled, leaving a gap in explaining clinical phenomena such as why treatment response builds over weeks, why some patients plateau, and why severe Obsessive-Compulsive Disorder (OCD) often fails to remit with Selective Serotonin Reuptake Inhibitors (SSRIs) alone. Existing dynamical models of OCD circuitry establish bistability (the coexistence of healthy and pathological states) but treat bifurcation control as an abstract parameter, disconnected from drug exposure, transporter occupancy, or clinical rating scales. Consequently, no existing model explains how a specific SSRI dose moves the system's state landscape or predicts the limits of remission based on baseline severity.

Methodology
The authors present a calibrated, multi-scale computational model that couples four distinct layers into a single system:

1. Pharmacokinetics: An extracellular serotonin engine modeling transporter (SERT) occupancy, autoreceptor desensitization (5-HT1A), and clearance, determining the steady-state extracellular serotonin concentration ($e_C$).
2. Neural Circuit: A mean-field basal-ganglia–thalamocortical circuit (based on van Albada & Robinson) where the caudate D1 firing rate ($\phi_{d1}$) encodes symptom severity.
3. Plasticity: A slow, activity-dependent remodeling of corticostriatal synaptic weights ($G_{CS}$) governed by a Bienenstock–Cooper–Munro (BCM) rule referenced to a healthy baseline. This rule creates a bistable landscape with a healthy fixed point ($G_{CS}=0$) and an elevated OCD attractor, separated by an unstable saddle.
4. Observational Readout: An explicit law mapping the circuit's firing rate to the Yale–Brown Obsessive-Compulsive Scale (Y-BOCS) score.

The model is calibrated against net (drug minus placebo) Y-BOCS trajectories from 13 dose-arms of six SSRIs. The calibration fixes nearly all parameters from literature or circuit constraints, leaving only two shared coupling constants (the 5-HT1B plasticity coupling and the autoreceptor desensitization coupling) to be fitted. The patient-specific vulnerability is encoded in a plasticity gain parameter ($\alpha$), which is back-solved from each patient's baseline severity.

Key Contributions

• Unified Multi-Scale Framework: The paper constructs the first model linking receptor occupancy, circuit dynamics, weeks-long plasticity, and clinical scores into a single causal chain.
• Bistable Fold Mechanism: It demonstrates that the OCD state is a stable fixed point separated from health by a saddle. Effective treatment is defined not merely as damping activity, but as the annihilation of the OCD attractor and the saddle via a saddle-node bifurcation (a "fold").
• Severity-Gated Remission: The model predicts that remission is gated by baseline severity. Mild-to-moderate patients can cross the fold (annihilate the OCD state) at tolerable SSRI doses, whereas severe patients cannot reach the fold because the required serotonin elevation would exceed physiological/toxic limits.
• Augmentation Rationale: By identifying distinct "routes to the fold," the model provides a mechanistic rationale for augmentation. While SSRIs act on the "sink" (depression sink via serotonin), augmentation agents like memantine (acting on the plasticity source gain $\alpha$) and antipsychotics (acting on the circuit argument $\phi_{d1}$) can push the system toward the fold even when SSRIs alone cannot.

Results

• Calibration and Fit: Using two shared parameters, the model reproduces the dose-response and weeks-long onset trajectories of six SSRIs across 84 data points with a root-mean-square error of 0.83 Y-BOCS points. It correctly predicts the saturation of benefit at high doses (e.g., fluoxetine 60 mg) due to SERT occupancy limits.
• Remission vs. Response: The model distinguishes between "response" (a reduction in symptoms while the OCD attractor still exists) and "remission" (annihilation of the attractor). Patients with severe baseline scores ($Y_{model,0} \gtrsim 24$) are predicted to be "responders" who relapse upon drug withdrawal because they never crossed the fold. Only patients with milder severity can achieve durable, off-drug remission.
• Differentiation from Graded Models: A matched "graded-null" model (monotone response without a fold) fits population mean data equally well. However, the bistable fold model predicts unique individual-level signatures absent in the graded model:
  1. Discontinuity: A sudden drop in symptoms at the fold threshold.
  2. Bimodality: A bimodal distribution of outcomes in treated cohorts (remission cluster vs. residual OCD).
  3. Irreversibility: Durable off-drug remission in a subset of responders (hysteresis), whereas the graded model predicts full reversibility.
  4. Critical Slowing: Rising autocorrelation and variance near the fold, serving as an early-warning signal.
• Augmentation Predictions: The model quantifies the "distance to the fold" for severe patients. It predicts that memantine (reducing LTP gain) and antipsychotics (reducing caudate firing) can annihilate the OCD attractor in severe cases where SSRIs fail, provided the combined effect is sufficient to reach the fold.

Significance and Claims
The paper claims that the failure of SSRIs to cure severe OCD is not a failure of the drug's potency but a structural limitation of the system's geometry: the fold is unreachable via serotonin elevation alone without toxicity. The significance lies in shifting the therapeutic paradigm from "dosing up" to "crossing the fold" via complementary mechanisms.

The authors explicitly state that population-level mean data cannot distinguish between a bistable fold and a graded response; the distinction relies on individual-level dynamics (irreversibility, bimodality, critical slowing). The model serves as a proof of principle for combination therapy in treatment-resistant OCD, suggesting that stacking agents acting on different terms of the plasticity equation (sink, source gain, and circuit argument) can achieve remission where monotherapy fails. The authors remain modest, noting that the model is a calibrated proxy and that specific dosing rules, pharmacokinetic interactions, and side-effect profiles are outside its scope. The "severity-gated" prediction is presented as a falsifiable empirical content of the model, not a definitive clinical protocol.