Background/Purpose: Disease progression in
longitudinal studies of rheumatoid arthritis (RA) is usually assessed by examining
a measure of disease over time fixed time intervals. We sought to use a
multi-state model: (1) to provide a descriptive
analysis of patients’ disease states defined by the DAS-28 disease index throughout
the first three years of treatment; (2) to determine the time spent in each
disease state; and (3) to estimate the probabilities of changing between disease
states.

Methods: Patients were selected from an ongoing
prospective study of RA  if they had active disease (>=
1 swollen or tender joint) and at least 2 follow-up visits to their
rheumatologist. The DAS-28 score was collected at each visit and assigned to
one of four category scores, from remission to high disease activity, using the
ACR criteria. A multi-state (4-state) Markov model was fitted to describe patient
progression through disease states over time to account for the irregular time
intervals in our data considering time as a continuous variable allowing us to
examine real-world disease course over time. Multiple piecewise transition
parameters for the Markov model were fit to determine if patients have
differential transitions at different time periods.

Results: There were 3014 visits in 586 patients
with visits distributed arbitrarily over the 3-year follow-up window. At
baseline, about 50% of patients were in high disease activity, but patients respond
rapidly, moving out of this health state spending on average 0.17 years, 95% CI
(0.19, 0.23) there. At baseline about 8% of patients were in DAS-28 remission,
which rose rapidly at 0.5 years 25% probability and 1 year with about 40% of
patients in remission. Once a patient achieved remission, the mean duration
before moving to another disease state was 0.81 years, 95% CI (0.67,
0.97). By 1.5 years after initiation of treatment, patients in each disease
state remained relatively constant indicating no net movement between health
states.  The piecewise model
indicated that patients at 6-months had a differential movements through
disease states then at any other time point. Patients starting in the highest
disease state took the longest about 1.1 years to reach their first remission
compared to 0.9 and 0.8 years for the moderate and low groups.

Conclusion: Individual patients transition between disease states in the
first 3 years of treatment at differential times; the first 6-months are
critical with the first year seeing rapid movement between states. Within the
first 6-months, dramatic reductions are realized in patients with high disease
activity in parallel with an observed increase in the proportion of patients
achieving remission. Our analysis indicates the critical first year of
treatment before a steady disease state with no net movement will be reached. Major
changes in the first year of treatment could be a result of treat-to target
strategy.