Pain is a principal presenting symptom of rheumatoid arthritis, and persistent pain despite low inflammatory activity is now the subject of its own systematic review literature. In 2026 the British Society for Rheumatology issued a guideline dedicated to pain management in inflammatory arthritis. Together these publications establish that control of inflammation and control of pain are related but separable outcomes. Preclinical arthritis studies are therefore designed with the same separation, because a candidate scored only on joint swelling has been characterized on only one of two clinically relevant outcomes.
The collagen-induced arthritis (CIA) model produces disease chronically, through active immunization. Animals are immunized with collagen, the immune system mounts an autoimmune response, and the resulting inflammation progressively destroys joint cartilage and bone in a course that reproduces the pathophysiology of rheumatoid arthritis. In mice the study runs 42 days and in rats 18 to 21 days. Because the model includes the immunization and the autoimmune response it provokes, it accommodates candidates acting on the immune process that generates the arthritis as well as anti-inflammatory agents acting downstream.
Collagen antibody-induced arthritis (CAIA) omits the immunization step. Anti-collagen antibodies are administered directly, bind collagen in joint tissue, and trigger an immediate inflammatory response with rapid joint swelling and cartilage degradation over a 12 day course in mice. The model reproduces the acute phase of rheumatoid arthritis on a timeline suited to early screening, while bypassing the upstream immune biology. A candidate directed at antibody effector mechanisms or acute joint inflammation is suited to CAIA, whereas a candidate directed at the adaptive response that produces those antibodies requires CIA.
Both models include the conventional efficacy readouts. In mice, CIA responds to dexamethasone and to Enbrel, which lowered arthritis score relative to vehicle and reduced anti-collagen type II antibody levels after their peak. In CAIA, systemic dexamethasone reduced clinical score, held paw thickness constant while vehicle animals developed swelling, and lowered the histopathological score that quantifies joint damage. The same dose administered intra-articularly was only partially effective on paw thickness, a comparison relevant to locally delivered candidates.
The assessment panel in both models extends beyond swelling into evoked pain testing, with von Frey, Hargreaves, hot plate, and dynamic weight bearing available alongside body weight, disease score, and paw thickness. In the rat CIA model, arthritic rats showed reduced withdrawal force on von Frey by day 11 of the study and shorter withdrawal times on the Hargreaves test. These changes indicate mechanical allodynia and thermal hypersensitivity developing alongside the joint pathology. In CAIA mice, dexamethasone treatment increased hot plate withdrawal latency, indicating reduced thermal sensitivity concurrent with its anti-inflammatory effect.
Swelling and pain do not always improve together, and measuring both in the same animals demonstrates whether a candidate relieves one, the other, or both. A candidate that reduces paw thickness but leaves von Frey thresholds low is controlling swelling without controlling nociception, which corresponds to the residual pain phenotype described in the clinical literature. A candidate with the opposite profile, analgesic without being anti-inflammatory, would be undetected in a study that scores joints alone. Neither profile can be identified unless pain is measured directly and longitudinally, because a single terminal timepoint cannot establish whether analgesia preceded, followed, or tracked the swelling.
Biomarker, histological, and behavioral endpoints together link mechanism to symptom. Biomarker panels quantify the inflammatory drive, and in the CIA model dexamethasone reduced IL-1β and PGE₂ relative to the vehicle control. Histology grades the consequences of inflammation for the joint, from soft tissue fibrosis to cartilage damage and reactive bone formation. Behavioral pain testing quantifies nociceptive sensitivity. A data package containing all three layers can demonstrate whether a candidate lowered the inflammatory mediators, protected the joint, and relieved the pain, or identify which of those effects was absent.
Pain defines this patient population and does not always resolve when inflammation is controlled. Direct assessment of pain in the preclinical arthritis study, alongside swelling, is therefore a necessary part of characterizing a candidate for these patients. This applies to any candidate whose therapeutic value depends on pain relief, whether an anti-inflammatory agent, a non-opioid analgesic, or a biologic acting upstream of both. Both models include dexamethasone as a reference compound, and the pain endpoints are included in the same assessment panel as the swelling, biomarker, and histology endpoints. Study design discussions are welcome at neuro@mdbiosciences.com.
References:
British Society for Rheumatology guideline for pain management in people with inflammatory arthritis. Rheumatology, 2026.