Transitions over time

We went beyond investigating solely the state of remission and evaluated transitions between different disease activity states defined by the DAS28-ESR.

Disability over time

We described and characterised functional disability over time (using the HAQ) with a specific focus on the impact of erosive disease at baseline and the impact of time-varying disease activity on transitions between HAQ states.

Erosive damage

We investigated erosive disease, focusing on the impact of disease activity and disability at baseline on 1-year damage progression.

Patients

The REMIRA study recruited 152 adults with RA undergoing a treat-to-target management strategy for 12 months. Inclusion criteria comprised disease duration of ≤ 10 years, receiving stable doses of conventional DMARDs or biologics for > 6 months and DAS28-ESR scores of ≤ 3.2 for 1 month or longer before recruitment. Details of these patients are shown in Appendix 1, Table 30.

Assessments

Baseline data were collected on demographics, disease duration and current treatment. Three-monthly assessments comprised DAS28-ESR and its components (i.e. 28 tender joint count, 28 swollen joint count, patient global assessment and ESR), CRP levels, HAQ, EQ-5D, Functional Assessment of Chronic Illness Therapy – Fatigue (FACIT-F) for recording self-reported fatigue, and the Medical Outcomes Study SF-36 and its physical and mental subscales (PCS and MCS).^358–362

Posterioanterior radiographs of hands and feet at baseline and 12 months were used to define damage by radiographic erosions and damage progression by new erosions or worsening of existing erosions over 12 months.

Definition of disease states

Disease activity states were defined based on DAS28-ESR using the internationally agreed definition³⁶³ of a DAS28-ESR score of < 2.6 to indicate clinical remission, a DAS28-ESR score of 2.6–3.1 to indicate low disease activity, a DAS28-ESR score of 3.2–5.1 to indicate moderate disease activity and a DAS28-ESR score > 5.1 to indicate high disease activity.

Functional disability states were defined using the HAQ disability index categorised into four categories, (1) HAQ = 0, (2) HAQ = 0.1–0.49, (3) HAQ = 0.50–1.49 and (4) HAQ = 1.50–3.00, representing no functional disability, mild disability, moderate disability and severe disability, respectively. HAQ scores < 0.50 represent few difficulties (if any) in performing daily activities and scores > 1.50 reflect considerable difficulties or assistance required in performing daily activities.^364–367

Statistical methods

For modelling disease activity and functional disability over time, a multistate modelling approach³⁶⁸ based on Markov processes was adopted, as our interest lay in characterising the evolution of these disease processes as they transition between clinically meaningful disease states. In addition, multistate models naturally handle staged data where patients are under only intermittent observations. They allow the estimation of rates of transitions between the various states of disease activity or functional disability and easily incorporate the effects of covariates (both time-independent and time-dependent) on transition rates. Here, correlation among states of a patient at different assessments are directly modelled through the Markov assumption that the future evolution of the patient’s disease process depends only on his/her current disease state and not on his/her previous disease history.

For the disease activity process, we consider the three-state multistate model shown in Figure 21a, where direct transitions (forward and backwards) are allowed between adjacent states (i.e. between remission and low disease activity or between low disease activity and moderate to high disease activity). Direct transitions between remission and moderate to high disease activity are not allowed, although, as we model the disease activity process in continuous time, this simplifying assumption is not restrictive.

FIGURE 21

Multistate diagrams for (a) disease activity states and (b) disability states.

For modelling physical functional disability, we adopt a four-state multistate model, as shown in Figure 21b, where, again, direct transitions are allowed only between adjacent states (i.e. between no disability and low disability, between low disability and moderate levels of disability, and between moderate and severe disability).

We examined the separate (univariate) and joint (multivariate) effects of select demographic and clinical variables on the transition rates in the models for both disease activity and disability. These covariates were incorporated into the models through the proportional hazards assumption. The variables considered were sex, age, ethnicity, erosive disease and treatment all at baseline, disease duration and either the HAQ score or DAS28-ESR score updated at each visit when considering either the disease activity multistate models or the disability multistate models, respectively. Damage progression was investigated through logistic regression models where only covariates measured at baseline were considered as predictors.

Baseline features

Over 85% of the cohort had low levels of disease activity or were in clinical remission at entry. None of the patients had high baseline levels of disease activity. Approximately 50% of the cohort had low levels or no disability at baseline. Erosive disease was observed to be present in 40% of the study sample (n = 67). Most patients were receiving stable doses of methotrexate for over 6 months prior to entry. Fifty-three per cent of patients were on two or more RA medications at baseline.

Characterisation of disease activity states over time

Figure 22a displays the longitudinal profiles of disease activity states for the 150 patients with at least two visits in which DAS28-ESR scores were recorded. Of these patients, 95 had DAS28-ESR scores recorded at all five visits and 44 (46%) of those were observed to have sustained remission (i.e. observed in remission at all five visits). Table 19 summarises the numbers and types of transitions in disease activity states that were observed over the 12-month follow-up period. Seventy-two patients were not observed to have made any transitions out of their initial state. Twenty-eight patients were observed to have made one transition (16 patients were observed to have deteriorated and 12 patients to have improved). Fifty patients were observed to have made two or more transitions, with the majority (92%) of patients observed to have a fluctuating course of both deteriorations and improvements.

FIGURE 22

Longitudinal profiles of disease activity and disability over five visits. (a) Disease activity states; and (b) disability states. NA, not applicable.

TABLE 19

Frequencies and types of disease activity transitions by initial disease activity state

On fitting the three-state multistate model (see Figure 21a) with no covariates to the disease activity states, we estimated that there would be 7.28 (95% CI 5.22 to 10.16) transitions per 100 person-months from the remission state to the low disease activity state, 36.81 (95% CI 26.30 to 51.51) transitions per 100 person-months from low disease activity back to remission, and 32.21 (95% CI 20.39 to 50.86) and 24.22 (95% CI 15.26 to 38.51) transitions per 100 person-months from low disease activity to moderate/high disease activity and moderate/high back to low disease activity, respectively. Based on these transition rates, we estimate that the mean time spent in the three states of remission, low disease activity and moderate to high disease activity before exiting are 13.74 (95% CI 9.85 to 19.17) months, 1.45 (95% CI 1.08 to 1.94) months and 4.12 (95% CI 2.60 to 6.55) months, respectively. Therefore, on average, once a person enters the remission states they spend over 12 months in this state before transitioning out under a treat-to-target management strategy. In contrast, a person spends very little time continuously in either the low or moderate/high disease activity states before exiting.

The results in Table 20 show the final multivariate multistate model when covariates are explored on various transition rates. We find that males have a higher transition rate out of low disease activity to moderate/high disease activity states than females. The rate of transition to remission (from low disease activity) is 3.2 (95% CI 1.2 to 6.6) times faster for males than for females, and a 1-unit increase in HAQ increases by 2.9-fold (95% CI 1.8- to 4.7-fold) the rate of transitioning out of remission.

TABLE 20

Multivariate three-state multistate model for disease activity

Characterisation of functional disability over time

Figure 22b displays the longitudinal paths of disability states of the same 150 patients. The numbers of patients observed in the no disability, mild, moderate and severe disability states at first observed HAQ assessment were 64, 26, 49 and 11, respectively. Ninety-eight patients were observed to have HAQ measured at all five time points. Of these 98 patients, 23 (23.5%) remained disability free over all their visits (compared with 46.3% of patients who had five visits with DAS28-ESR measured being in sustained remission). Table 21 summarises the transition patterns in disability states that were observed over the 12-month follow-up period. Just under half of the 150 patients (n = 73) were observed to have not made any transitions out of their initial state. Twenty-five patients were observed to have made one transition (14 patients were observed to have deteriorated and 11 patients to have improved). Fifty-two patients were observed to have made two or more transitions and all but one patient were observed to have a fluctuating course of both deteriorations and improvements.

TABLE 21

Frequencies and types of disability transitions stratified by initial disability state

The results in Table 22 of fitting the final multivariate four-state multistate model provided evidence for the effects of disease duration, ethnicity and disease activity on transitions between various disability states. More precisely, we found that a 1-year increase in disease duration reduces the rates of transitioning both in and out of the no disability state by 0.8. The rate of transition to moderate disability from mild disability state increased 1.17-fold (95% CI 1.03 to 1.33) for every additional year of disease. Patients who were white had a slower rate of transitioning in and out of the no-disability state than non-white patients, and a 1-unit increase in DAS28-ESR score increased the transition rate from moderate to severe levels by around twofold (95% CI 1.24-fold to 3.92-fold). No evidence of effects of erosive disease and exposure to treatments at entry on the various transitions between disability states was found.

TABLE 22

Multivariate four-state multistate model for disability

Damage progression

At 1-year follow-up, 71 of the 82 (86.6%) patients observed not to have erosive disease at baseline remained damage free and four patients were observed to have progressed. For the 67 patients with baseline erosive disease, 46 (68.7%) were observed not to have progressed further and 16 were observed to have further progression. There was clear statistical evidence, as expected, that the 1-year damage progression rate was higher in those patients with baseline erosive disease than in those without (p = 0.002).

The final multivariate logistic regression model investigating baseline predictors of 1-year damage progression identified ethnicity, disease activity and erosive disease at baseline as potentially important factors (Table 23). Patients who were white were found to have higher odds of damage progression (OR 9.47, 95% CI 1.11 to 80.47; p = 0.04) than non-white patients, patients with a higher DAS28-ESR score at baseline were more likely to progress than those with lower scores (OR 2.1, 95% CI 1.04 to 2.12 for a 1-unit higher DAS28-ESR score) and RA patients with erosive disease present at baseline were also at greater risk of progression (OR 7.3, 95% 2.08 to 25.59). No statistical evidence was found for an effect of physical functional disability on damage progression (p = 0.225) after accounting for ethnicity, disease activity and erosive disease at baseline.

TABLE 23

Multivariate logistic regression model for 1-year damage progression

Defining remission

There is no ideal definition of clinical remission. Even sustained clinical remission may not represent an underlying state of true biological remission. Although 44 patients in the REMIRA cohort had sustained remission over the 12 months, 5 of the 44 (11%) patients had damage progression after 1 year. Clinical remission may overlook subclinical synovitis and, consequently, true remission might be better characterised molecularly using laboratory markers.

Refractory disease

We found that 18 patients in the REMIRA cohort had moderate to high disease activity states at their last visit without ever showing any clinically meaningful improvement in their disease activity. These patients may represent a refractory group of patients who do not respond well to current therapies. However, our data also suggest that, once remission is achieved, patients under a treat-to-target strategy will tend to stay in remission for > 1 year, on average, before exiting this state of clinical disease quiescence.

Variability of remission

In the REMIRA cohort of 152 RA patients, we observed various patterns of disease activity and physical functional disability over 12 months. Forty-four patients were observed in sustained remission over all visits, whereas 23 patients were observed to be disability free at all visits and 14 patients were observed to have been both in sustained remission and disability free. Sizeable proportions of patients were observed to have a fluctuating disease course (30.7%, n = 46) and to have fluctuating levels of functional impairment (34%, n = 51). These fluctuating patterns may reflect the difficulty in controlling patients’ disease, even under a treat-to-target strategy.

Factors influencing transitions between disease states

We found that sex and levels of functional disability had an impact on the rate of transitions between disease activity states. In particular, males tended to have a higher rate of transitions out of low disease activity to moderate disease activity than females, although higher levels of disability increased the rate of transitioning out of the remission state. We also found that higher levels of disease activity increased the rate of transitioning from moderate to severe levels of disability and that ethnicity and disease duration also influenced the transition rates between disability states, in particular transitions in and out of the mild disability state.

Factors influencing erosive progression

As expected, the 1-year damage progression rate was higher in those patients with erosions at baseline. Ethnicity and disease activity at baseline were identified as possible predictors of damage progression. No evidence was found for functional disability as a predictor of 1-year damage progression in RA patients undergoing a treat-to-target strategy.