Aim: There have been many changes in treatment for patients with inflammatory polyarthritis (IP) / rheumatoid arthritis (RA) over the past three decades, therefore collecting new data on the long-term outcome of these patients is important. This thesis aimed to review literature on baseline predictors of long-term disability; to describe the natural history of IP/RA over 20 years; to investigate associations between early treatment and mortality and disability; to investigate predictors of and outcomes following orthopaedic surgery, and to compare the 10 year outcome of two cohorts recruited 10 years apart. Methods: A systematic review was performed by implementing a pre-specified search strategy into the Medline and EMBASE databases. Data on baseline predictors of long-term disability (>=5 years) were extracted from identified studies. The analysis sections of this thesis used data from the Norfolk Arthritis Register, a primary-care based, prospective inception cohort. Analyses one and two used data from patients recruited from 1990-94 and followed for 20 years. The disease activity and disability over twenty years of patients were described using descriptive statistics. The disability of patients who received early treatment (ET; treatment 6 months after symptom onset) were compared to patients never receiving treatment (NT) using weighted mixed-effects linear regression models. The risk of mortality was compared between treatment groups using weighted pooled logistic regression. Predictors of joint surgery were assessed using a conditional risk set model. Disability of those who had surgery vs. those who did not was compared using a weighted mixed-effects linear regression model. The final analysis compared the outcome of patients recruited from 1990-94 (C1) with those recruited from 2000-04 (C2) using population average models and Cox regression. Results: Based on the current literature, female gender as well as higher baseline age, Health Assessment Questionnaire (HAQ), pain and Disease Activity Score (28) were associated with long-term HAQ scores. Over 20 years, the median disease activity of patients fell from high baseline levels and remained low over follow-up, whereas HAQ scores followed the characteristic J-shaped trajectory. There was reduced mortality risk in the ET and LT groups compared to the NT group. HAQ scores were comparable between the ET and NT groups (mean difference 0.03, 95% CI -0.06 to 0.12); the LT group had increased disability compared to the NT group (mean difference 0.10, 95% CI 0.02 to 0.17). The strongest predictor of surgery was HAQ score at the closest assessment to surgery (HR 2.11 per unit increase in HAQ, 95% CI 1.64 to 2.71). Patients had worse HAQ scores following surgery than patients not undergoing surgery (mean difference 0.17, 95% CI 0.03 to 0.32). Patients in C2 had reduced risk of all cause and cardiovascular disease (CVD) mortality compared to C1 (all cause: HR 0.72, 95% CI 0.56 to 0.95; CVD: SHR 0.58, 95% CI 0.37 to 0.93), and 17% lower swollen joint count (95% CI -23% to -10%). Tender joint count and HAQ scores were similar between the cohorts. Conclusion: Patients had low disease activity and moderate disability over 20 years. Early treatment is associated with improved long-term outcome. The HAQ is a strong predictor of outcome (later HAQ/surgery) and potentially should be used more in clinical practice. Patients in the new millennium have improved long-outcomes on some measures (mortality, swollen joints) compared to a decade earlier.