The mechanistic gap in orofacial pain: A call for cross-disciplinary collaboration

In clinical practice, symptom-based phenotyping alone rarely enables reliable treatment protocols. Patients with identical diagnoses such as a painful temporomandibular disorder (TMD) per Diagnostic Criteria for TMD (e.g., myofascial pain, arthralgia) or neuropathic-like orofacial pains (OFPs) per International Classification of Orofacial Pain often follow markedly different trajectories . One patient may improve with conservative care and resume normal function while another cycles through splints, injections, medications, and referrals, experiencing persistent limitations, disruptive flares, and unpredictable relapse. Unfortunately, this variability is common in all OFP clinics and explains, at least in part, the current low rate of clinical usage for such purported standardized diagnostic tools based on symptom-driven diagnostic labels . Even refined phenotyping frequently leaves many patients refractory to care, as current criteria poorly predict treatment response, pain persistence, or relapse risk, not to say the absence of a connection with pathophysiology.

Meanwhile, basic pain research has advanced the understanding of craniofacial nociception, revealing heterogeneous trigeminal sensory neuron programs, glial and immune contributions, and circuit-level regulation in brainstem and higher centers. Human and mouse trigeminal ganglion atlases highlight this biological diversity, showing why one clinical label can reflect multiple underlying drivers. The biology of axis II modulates the clinical impact, albeit partially. The core challenge is not insufficient mechanistic detail and understanding, but the lack of a clear translational pathway from discovery to clinical relevance. Clinicians need better tools to classify patients, estimate prognosis, select treatments, and track meaningful outcomes. A minimal-consensus, circuit-informed framework is therefore required to make mechanistic insights testable and actionable in routine care.

The field of orofacial pain continues to grow in both clinical and basic research settings. More and more similarities with other fields of pain medicine are emerging from an etiological, diagnostic, and treatment perspective. However, without structured collaboration between clinicians and researchers, preclinical models and endpoints often prioritize experimental convenience over real-world clinical decisions. The currently available classification systems make no exception.

This “mechanistic gap” arises from several mismatches. First, clinical labels rely primarily on symptoms and exam findings, yet the same diagnosis can stem from distinct peripheral triggers, trigeminal pathways, or immune/glial states. The label “arthralgia” used as an umbrella to indicate conditions such as TMJ retrodiscal tissue inflammation and intracapsular effusion, which do not share many commonalities in clinical decision making, is one example. Second, research endpoints frequently focus on isolated pathway effects, while clinicians require decision-grade evidence predictive of response, persistence, and relapse. Third, methodological differences in definitions, instruments, sampling, and follow-up hinder reproducibility and data integration. Finally, common animal readouts – such as von Frey withdrawal thresholds or Mouse Grimace Scale – fail to capture patient-relevant features like chewing-evoked pain, flare patterns, sleep disruption, mood impact, or functional limitations.

These apparent disconnects slow translation. Preclinical studies can rapidly establish causality in inflammation or nerve injury models, but many still rely on evoked sensitivity measures that do not reflect the disabilities driving clinical care. The concept of treatment-demanding patients is not necessarily implying the same construct as symptom-driven categorization for research purposes. Clinically, heterogeneous cohorts mixing myogenous and arthrogenous TMDs, neuropathic-like pain, and headache, often with variable treatment histories, combined with limited deep biological sampling, make results difficult to synthesize into practical stratification rules. In addition, the frequent absence of psychosocial monitoring cripples the core concept of biopsychosocial model of pain. Broad symptom-based trial enrollment further dilutes effect sizes by blending distinct subtypes.

To bridge this gap, a decision-focused framework might be proposed by explicitly linking routine diagnosis to measurable biology. It organizes pain mechanisms into practical layers with corresponding clinical results:

• Peripheral driver level: Provocation and localization testing to identify input-driven pain and guide local therapies.

• Trigeminal level: Compact quantitative sensory testing (QST) with facilitation and inhibition measures, aligned with animal models that include function-relevant outcomes such as bite force, gnawing behavior, or meal-pattern disruption.

• Brainstem/modulation level: Tests of temporal summation and conditioned pain modulation to distinguish facilitation from impaired inhibition.

• Network level: Longitudinal patient-reported outcomes tracking flares, relapse, function, sleep, and affect.

Near-term implementation can use scalable, clinic-friendly tools, e.g., hypothesis-driven inflammatory/neuroimmune blood or saliva panels, salivary cortisol for stress-axis assessment, and brief sensory profiling. A re-discovery of basic medical semeiotics (e.g., orthopedic and neurological test) and terminology is also recommended, and should not be abandoned in favor of diagnostic strategies favoring the medium-average practitioner. Medium-term additions could include targeted metabolomics or microbiome signatures when relevant. All measures should be evaluated against standardized, decision-relevant endpoints (response, persistence, relapse) using fixed follow-up windows and prospective multi-site validation. In addition, work to develop better classification schemes for clinical usage needs to be a priority early on.

Successful translation requires infrastructure supporting harmonized phenotyping, biospecimen collection, and shared data standards. Within an OFP clinic, a practical first step is establishing a recurring “Mechanism/Terminology-to-Decision” conference to map real cases to testable hypotheses and paired clinical-preclinical plans. A minimal registry battery covering pain intensity and impact, jaw function, brief sleep/mood screens, oral behaviors, and a compact sensory module should be collected with consistent timing.

Biospecimens (blood, saliva; stool when indicated) could be routinely linked to phenotype, treatment history, and longitudinal outcomes. Translation must operate as a bidirectional loop, i.e., registry-derived subtypes guide model selection and endpoints, while preclinical findings report at least one patient-relevant functional or ongoing-pain measure alongside traditional evoked assays. Early human studies should prioritize prospective stratification by subtype and sensory phenotype when possible.

Finally, workforce development should align with this vision. Training programs can help channel interested residents and fellows into NIH-supported or other similar pathways (e.g., T90/R90) to equip future specialists with skills to generate decision-grade evidence, rather than only the delivery of care. By aligning language, measures, and endpoints around clinical decisions, the field can accelerate subtype discovery, biomarker validation, and personalized prevention and therapy.