The goal of treatment of rheumatoid arthritis (RA) is to prevent pain and disability. A surrogate measure for disability is radiographic joint damage. Prevention of radiographic progression of joint damage has thus become a goal of treatment, and an outcome for many clinical trials. There are several issues that have not been clarified, however. First, is the rate of radiographic damage linear over time, or more rapid in early disease? There are data to suggest both. Second, does clinically assessed disease activity correlate with radiographic damage? Intuitively, the answer to this would seem to be yes but data from the ATTRACT trial and studies on bone erosion in animal models of RA have suggested that there may be a disconnect between the two.
These questions are not readily answered by conventional statistical analysis methods which use time-averaged estimates of radiographic progression and disease activity over a study interval. Time-averaged estimates do not take into account the variability of disease within an individual patient, and also assume a linear rate of radiographic progression. The use of GEE (generalized estimating equations) is a regression technique that eliminates these problems and is particularly useful in the longitudinal analysis of time-dependent relationships. Here, Welsing et al (Arthritis Rheum 50(7):2082, 2004) use GEE to investigate the longitudinal relationship of RA disease activity to radiographic damage in two different cohorts of RA patients.
Methods:
Data from two separate cohorts of RA patients were utilized. The first, the University Medical Center Nijmegen (UMCN) cohort, is an inception cohort of early RA patients followed since 1985. Patients are treated according to the discretion of their treating rheumatologist, but have regular assessments of disease activity (reported as DAS (disease activity scores) collected at baseline and every 3 months) and radiographs of hands and feet obtained at baseline and every three years. The second, the Combination Therapy in RA (COBRA) trial cohort, were participants in a 56 week multicenter double-blind trial comparing combination/step-down therapy with monotherapy (see study results). COBRA enrollees had regular assessments of disease activity (DAS28 collected at baseline, and at weeks 16, 28, 40, 56, and annually after the double-blind portion of the study concluded) and radiographs of hands and feet at baseline and every 6 months during the double-blind portion of the study, and every year thereafter. Radiographs were read in sequential order and scored according to the Sharp/van der Heijde method.
Regression models using GEE were constructed to evaluate trends in radiographic progression. Time was entered into the model first. Then baseline variables thought to be predictive of radiographic progression (age at baseline, sex, rheumatoid factor, baseline Sharp score, baseline DAS score) were entered into the model. Finally, disease activity was entered into the model. Disease activity was measured as DAS28 at individual time points for COBRA, or as mean interval DAS or SD (standard deviation) of the mean interval DAS at each 3-month interval for the UMCN cohort.
Results:
The baseline characteristics of patients enrolled in each of the cohorts was remarkably similar except that the COBRA group had higher disease activity at baseline (as required by the study protocol) and was followed for a longer period of time. The average age in both cohorts was 50-55, approximately 75% were seropositive for RF, and the majority of patients were women.
UMCN Cohort. The time averaged radiographic progression in this cohort was 9.5 Sharp points per year. Using GEE, the rate of radiographic progression was found to slow slightly with increasing disease duration. Of the baseline variables considered, only RF positivity and Sharp score at baseline were positive correlated with radiographic progression. However, the mean interval DAS, and SD of the mean interval DAS, were also positively correlated with radiographic progression even after adjusting for RF positivity and baseline Sharp scores. Thus, periods of higher disease activity (mean interval DAS) or fluctuating disease activity (SD of mean interval DAS) were associated with more radiographic progression. These associations were stronger in RF-positive patients than RF negative patients.
COBRA cohort. Almost identical results were obtained. The time averaged radiographic progression in this cohort was 7.7 Sharp points per year, but the rate of radiographic progression tended to slightly slow with increasing disease duration. The factors significantly associated with radiographic progression over time were RF status, baseline Sharp score, and treatment allocation. Individual DAS scores were independently associated with subsequent radiographic progression, even after adjusting for RF status, baseline Sharp score and treatment allocation. This association was only present in RF positive patients.
Conclusions:
Elevated or fluctuating RA disease activity is positively associated with subsequent radiographic progression, particularly in RF-positive individuals. Radiologic progression is not linear in individual patients.
Editorial Comments:
This well designed longitudinal study strongly suggests that the natural history of radiographic progression in RA is not constant over time, and that variations in the level of disease activity over time are largely responsible for the speed-ups and slow-downs in the rate of radiographic damage. That sustained moderate to high levels of disease activity lead to progressive radiographical joint damage is logical and not entirely unexpected. What is surprising is that fluctuations in the level of disease activity independently contribute to radiographic progression. These results strongly support an aggressive therapeutic approach to attain and maintain zero or minimal disease activity rather than tolerating periodic flares. Such an approach was adopted in the TICORA study and was quite effective in minimizing radiographic damage (Jane, see if you can find TICORA in ACR or EULAR highlights. WE may not have discussed it by that name. The study aimed for a DAS28 of less than some defined number.) Importantly, the present study also reaffirms RF positivity as a potent risk factor for progressive joint damage, and suggests that rheumatologists should be more aggressive in the treatment of RF positive patients.
