Aim: The aim of this thesis was to validate the 2010 American College of Rheumatology/European League Against Rheumatism classification criteria for rheumatoid arthritis (RA), in particular with respect to its construct validity and the role of autoantibodies within the criteria. Methods: This thesis used data from the Norfolk Arthritis Register, a longitudinal inception cohort of adults (16 years and over) with inflammatory polyarthritis (IP), defined as at least 2 swollen joint for at least 4 weeks. The 2010 criteria were used to define RA, firstly in a re-estimation of the incidence rates (IR) with comparisons made to incidence defined by the previous criteria set; and secondly in a study comparing mortality rates in patients with RA to those of the general population, and how these rates changed over time. Analyses were performed testing the ability of the 2010 criteria to identify those patients with IP at increased risk of mortality, disability, disease severity and radiographic damage. The levels and number of autoantibodies present were investigated as predictors of mortality in patients with IP. The association between anti-carbamylated protein (anti-CarP) antibodies and long term disease outcomes were investigated. Results: The incidence of RA was 40 per 100 000 population; baseline IRs were similar to the cumulative IRs using the previous criteria set over 5 years. Patients who were seronegative were less likely to be classified as RA by the 2010 criteria. Mortality rates in patients with RA were higher compared to the general population (standardised mortality ratio 1.16, 95 percent confidence interval (CI) 1.04-1.29) and declined over the study period at the same rate as the general population. Patients with IP who fulfilled the 2010 criteria had increased risk of early death (hazard ratio (HR) 1.35, 95 percent CI 1.13-1.64), as well as increased levels of disability (beta 0.38, 95 percent CI 0.33-0.43), disease severity (beta 1.63, 95 percent CI 1.54-1.73) and radiographic damage (beta 0.33, 95 percent CI 0.20-0.47) throughout follow up. Patients with two autoantibodies had an increased risk of early death (HR 1.35, 95 percent CI 1.09-1.68), but there was no association with early death and the levels of these antibodies. Anti-CarP antibody positivity was independently associated with worse disability (beta 0.12, 95 percent CI 0.02-0.21) and disease severity (beta 0.23, 95 percent CI 0.07-0.39) throughout follow up. Conclusions: The 2010 ACR/EULAR classification criteria for RA identify patients with IP early in their disease course and recognise those at increased risk of mortality and poor outcomes. The 2010 criteria may miss a subgroup of seronegative patients who nevertheless have a poor prognosis. Novel autoantibodies may be useful to identify this subgroup.

Background: The ability to measure health and the value of improving or declining health is crucial to the evaluation of health care interventions. Many generic and disease specific health status measures exist for use in patients with rheumatoid arthritis (RA). The Overall Status in Rheumatoid Arthritis (OSRA) measure is a new and simple measure with early evidence of construct validity. Generic health profiles with attached utility weights such as the EuroQol EQ-5D and the SF-6D (calculated from the Medical Outcome Study 36-item short-form health survey) allow the quantification of a patient's health relative to perfect health and death, and can be used to estimate quality adjusted life years (QALYs). The EQ-5D is extensively used in RA, but has potential limitations. The SF-6D appears to have potential, but needs further evaluation. The aim of this thesis was to assess the validity and responsiveness of the EQ5D, SF-6D and OSRA in UK RA patients, and compare the performance and implications of the use of the EQ-5D and SF-6D.Methods and subjects: Patient data were obtained from three sources; the Steroids in Very Early Arthritis (STIVEA) (n=256) and British Rheumatoid arthritis Outcome Study Group (BROSG) (n=466) randomised controlled trials, and the British Society for Rheumatology Biologics Register (BSRBR) (n=129). The data used included lifestyle and demographic factors, disease activity (DAS28), functional disability (HAQ), X-rays to assess erosive damage, the EQ-5D and the SF-6D. The OSRA was collected only in the BROSG trial. Visual analogue scales (VAS) of pain and fatigue were collected in BROSG and STIVEA. Construct validity was tested by correlating the EQ-5D, SF-6D and OSRA with a range of outcome measures for RA. Responsiveness to change was assessed using minimum important differences (MID), effect size (ES) and standardised response means (SRM), and compared using ratios. EQ-5D profiles placing arthritis patients in utility states 'worse than death' (negative scores) were described and assessed using linear and logistic regression. The implications of using the EQ-5D and SF-6D in economic evaluation were compared by cost-effectiveness analyses of the BROSG trial. Results: The correlation of the EQ-5D and SF-6D was moderate to high (0.67). Both measures had moderate to high correlations with disease activity, physical function, joint damage and fatigue. The OSRA Activity (OSRA-A) and Damage (OSRA-D) correlated strongly with measures of related aspects of disease. The EQ-5D, SF-6D and OSRA discriminated between known differences in health status across groups defined by social deprivation and disease activity. The EQ-5D MID was 0.04 for improvement and 0.10 for deterioration. The SF-6D MID was 0.04 in both directions. The SF-6D was more responsive to improvement (EQ-5D: SF-6D ES ratio 0.78-0.88) and the EQ-5D more responsive to deterioration (ES ratio 1.14) in health. The OSRA-A was the most sensitive disease specific measure in the BROSG trial, and the OSRA-D was more responsive than the HAQ. The factors associated with being in a 'worse than death' health state were male gender, the HAQ, SF-36 mental composite scale, pain VAS, and erythrocyte sedimentation rate (a marker of inflammation). Pain was the predominant factor and was scored at the most extreme level in every worse than death profile. The cost-effectiveness analyses (BROSG trial), found net quality adjusted life years (QALYs) were greater for the EQ-5D (0.07) than the SF-6D (0.05), but had higher variance than the SF-6D. Conclusions: The EQ-5D and SF-6D appear valid and responsive to changes in health in RA, but measure subtly different aspects of health. There are issues with both measures, and cost-effectiveness conclusions of a study could differ according to which measure was used. The EQ-5D may be more likely to demonstrate that an intervention is cost effective than the SF-6D, due to its larger mean change in response to change in health status. The OSRA is valid for use in RA and its responsiveness suggests potential for inclusion in clinical trials.