Treatment targets

Targets must balance the ideal with the practical. Stringent targets may deliver optimal outcomes in some individuals, but achieve fewer overall benefits than more readily achievable targets. The TITRATE programme adopted DAS28-ESR remission as its primary target because it is the most widely used composite remission assessment. There is also extensive evidence that achieving DAS28-ESR remission optimises health-related quality of life and function and minimises radiological damage.^211–215 Sustained remission over time is particularly important because it is associated with better outcomes than remission at a single time point.^216,217 We collected components of other composite measures to compare their value as targets during intensive management.

In this section, we evaluated four aspects of treatment targets: (1) comparisons of sustained remission (persistent remission after 6 months’ treatment) and point DAS28-ESR remission and low disease activity, (2) the impact of lesser improvements in DAS28-ESR, (3) limitations of DAS28-ESR in comparison to other composite assessments and (4) associations of DAS28-ESR components with health-related quality of life.

We focused on these aspects of treatment targets, as they are important and we had access to relevant observational and trial data. We could not examine all indices, as some of the necessary data are not collected in routine care settings or our published trials. As C-reactive protein (CRP) levels and physicians’ global assessments are not usually measured in routine practice in England, we could not study the SDAI.²¹⁸

Simple baseline outcome predictors

Using baseline measures to predict clinical outcomes may help routine practice. We therefore assessed the value of simple four-point scores, HAQ scores alone and mental health status.

We selected these areas on the basis of what was practical and likely to be used in clinical practice. There was a case to assess rheumatoid factor and subtypes and other autoantibodies in predicting RA outcomes,^81,219,220 but these were not used in the observational studies and trials that we could access.

Observational studies

We further studied the observational longitudinal cohort study established in 2005 at Guy’s Hospital. The study focused on 752 patients followed over ≥ 3 years. Details of these patients are given in Appendix 1, Table 28. We also studied 155 early RA patients who completed 12 months’ follow-up with clinical data at 0, 6 and 12 months in an observational study [Early Rheumatoid Arthritis Network (ERAN)]. Details of these patients are given in Appendix 1. We selected these studies because they involved patients treated in recent years. More historical studies were not used because patients received far less intensive management.

We also evaluated a compilation of single time point observational studies of outpatients with RA.^72,221–223 The patients included 747 European white RA patients, 197 black African/Caribbean British patients and 430 Arab patients seen in rheumatology clinics in Saudi Arabia. No specific treatment policies were followed in these patients. Details of these patients are given in Appendix 1, Table 29.

Clinical trials

We further evaluated the 379 patients completing the 24-month CARDERA trial⁷³ and the 179 patients completing the TACIT trial.⁷⁴ We also evaluated patients with established RA in the OPTTIRA (Optimizing Treatment with Tumour Necrosis Factor Inhibitors In Rheumatoid Arthritis) trial.²²⁴ The OPTTIRA trial, which lasted 6 months, enrolled 103 patients, and complete end-point data were available in 97 patients. Details of these patients are given in Appendix 1, Table 31.

Clinical assessments

We assessed disease activity using DAS28-ESR, CDAI and Routine Assessment of Patient Index Data 3 (RAPID3), disability using the HAQ, and health-related quality of life using EuroQol-5 Dimensions (EQ-5D) and the Short Form questionnaire-36 items (SF-36).^225–230

Further details are given in Appendix 2.

Statistical analyses

Data were analysed descriptively using means, SDs and 95% CIs or medians and IQRs for non-normally distributed data. Other tests included chi-squared tests, assessments of sensitivity and specificity, t-tests, regression analyses, Spearman’s correlations and multiple linear regression methods. Further details are given in Appendix 3.

Treatment targets: DAS28-ESR and disability

The relationships between remission and low disease activity with disability and quality of life were studied in early and established RA trials and an observational cohort. In these patients, both sustained remission and sustained low disease activity were relatively uncommon. Between 5% and 9% of patients had sustained remissions and 9–16% of patients had sustained low disease activity. More patients had remission and low disease activity at single time points. Between 35% and 58% of patients had an episode of remission and between 49% and 74% of patients had an episode of low disease activity.

Disease Activity Score for 28 joints based on ESR scores varied substantially after patients had achieved an episode of remission. End-point DAS28-ESR scores (at 12 and 24 months in TACIT and CARDERA trials,^73,74 and at final assessment in the observational study) in patients achieving an episode of remission showed that 53–61% of patients were still in remission, 9–18% of patients had low disease activity, 21–22% of patients had moderate disease activity and 4–8% of patients had high disease activity. These findings are shown in Figure 8.

FIGURE 8

End-point DAS28-ESR category after attaining point remission.

Individual patients showed marked levels of variation in their subsequent DAS28-ESR scores following attaining remission. The extent of this within-individual variability was similar across all three cohorts. Figure 9 shows DAS28-ESR scores for all patients following attaining remission in the longitudinal observations study in patients with at least five subsequent DAS28-ESR measures.

FIGURE 9

Within-individual variability in DAS28-ESR scores after remission for patients in the longitudinal observational study. DAS28-ESR scores for each individual patient following an episode of remission are plotted, provided there were at least five subsequent (more...)

Sustained and point remissions had varying impacts on end-point low disability and normal EQ-5D scores (Table 3). Sustained remissions were most specific for low disability (97–98%) and normal EQ-5D (93–97%), but lacked sensitivity (low disability: 19–29%; normal EQ-5D: 19–36%). Point remission gave a better balance between sensitivity and specificity (low disability: specificity 50–78% and sensitivity 68–89%; normal EQ-5D: specificity 42–72% and sensitivity 70–93%).

TABLE 3

Relationship between remission states and end-point low HAQ scores and normal EQ-5D scores

Attaining sustained low disease activity was also highly specific for low disability (92–96%) and normal EQ-5D (86–94%), but lacked sensitivity (low disability: 26–41%; normal EQ-5D: 33–41%). Low disease activity at any point was highly sensitive (low disability: sensitivity 87–97%; normal EQ-5D: sensitivity 87–100%), but had only moderate specificity (low disability: specificity 31–63%; normal EQ-5D: specificity 25–58%).

Treatment targets: optimal responses in DAS28-ESR scores

An alternative way of assessing the inter-relationship between DAS28-ESR scores, disability and quality of life was examining the impact of EULAR responses in clinical trial settings. This approach was taken in another study²⁰⁸ that evaluated the impacts of moderate and good EULAR responses on changes in HAQ scores at the end points of early and established RA trials.

Moderate EULAR responders’ mean HAQ scores decreased by 0.39 and 0.33 in the CARDERA and TACIT trials, respectively.^73,74 In contrast, EULAR good responders had reductions of 0.88 and 0.64, respectively. In both trials, the difference between moderate and good responders exceeded the minimum clinically important difference for HAQ scores (0.22). The differences in mean reductions of 0.49 and 0.30 between moderate and good responders were significant (p < 0.01, unpaired t-test).

There were similar findings for EQ-5D scores. In moderate EULAR responders, EQ-5D scores increased by 0.18 and 0.15. In good EULAR responders, EQ-5D scores increased by 0.30 in both trials. The differences (0.12 and 0.15) between moderate and good responders also exceeded the minimum clinically important difference, which is generally considered to be 0.07, and were significant (p < 0.01, unpaired t test).

In addition, the frequencies of large and minimal improvements in disability and quality of life were assessed in these patients (Figure 10). With HAQ scores between 41% and 18%, good EULAR responders had large decreases in HAQ score (> 1.00) in early and established RA. However, only 13% and 9% of moderate EULAR responders had such reductions. By contrast, only 21% and 20% of good EULAR responders had minimal changes in HAQ score (> 0.22), compared with 37% and 43% of moderate EULAR responders.

FIGURE 10

Changes in HAQ score in moderate and good EULAR responders. In early and established RA trials (CARDERA trial and TACIT trial, respectively). Shows per cent of patients with substantial (> 1.00) and minimal changes (< 0.22). (more...)

Treatment targets: DAS28-ESR and alternative assessments

The inter-relationships between the four components of DAS28-ESR scores were assessed over the four disease activity levels: (1) remission, (2) low disease activity, (3) moderate disease activity and (4) high disease activity. Initially, these inter-relationships were assessed in an observational study of 747 European white patients. This analysis showed that ESRs contributed most to mean DAS28-ESR scores at all activity levels. However, their contribution was greatest in remission, when ESRs accounted for 70% of DAS28-ESR scores. The contribution of ESR decreased to 40% of DAS28-ESR scores in active disease. In contrast, the contributions of tender joint count to overall DAS28-ESR scores declined as DAS28-ESR fell. Swollen joint counts and patient global assessments showed small stable contributions to DAS28-ESR scores over all disease activity levels.

These findings were replicated in two further observational studies of 197 black African/Caribbean British patients²²³ and 430 Arab patients.²²² They were also replicated in three clinical trials (i.e. CARDERA,⁷³ TACIT⁷⁴ and OPTTIRA²²⁴) that involved 97–369 patients. Figure 11 shows the contributions of ESR and tender joint counts to DAS28-ESR disease activity levels in all six of these observational studies and trials.

FIGURE 11

Contributions of different components to activity scores in different states of disease activity. Tender joints and ESR in DAS28 in data from six studies–,– and components CDAI and RAPID3 in data from the CARDERA trial. (a) DAS28 tender (more...)

Two other composite scores – CDAI and RAPID3 – were studied in a comparative manner in one early RA trial (i.e. the CARDERA trial⁷³). These alternative composite scores showed different patterns of variation across disease activity levels, which is also shown in Figure 11. With the CDAI, patient and assessor global assessments dominated in remission and swollen and tender joint counts dominated in active disease. With RAPID3, patient global score, pain score and HAQ score made relatively stable contributions to the overall score across all activity levels.

Treatment targets: DAS28-ESR and health-related quality of life

The association between different components of the DAS28-ESR and health-related quality of life was assessed using the SF-36. The inter-relationships were evaluated in clinical trials of 672 patients with early and established RA.

Linear regression models, which included all four DAS28-ESR components, examined the relationships to SF-36 physical component score (PCS) and mental component score (MCS). The regression models were adjusted for treatment, age, sex and disease duration. The regression models found significant correlations between patient global scores and both SF-36 summary scores in early and established RA (Table 4). Other components of DAS28-ESR had significant correlations in early RA patients, but did not have significant correlations in established RA patients.

TABLE 4

DAS28-ESR components and SF-36 PCS and MCS at final trial time points (assessed in multiple linear regression models)

Predictive factors: simple four-point scores

The first approach to predicting RA outcomes involved developing and testing a simple predictive score using data that are regularly collected in routine care settings. It focused on predicting persisting active RA. The score was developed in an observational study (ERAN),²⁰⁶ using 155 early RA patients who completed 12 months’ follow-up and had clinical data at 0, 6 and 12 months. Regression modelling identified three main predictors for persisting active disease: (1) tender joint counts, (2) HAQ scores and (3) ESR. Each of these predictors was then dichotomised (six or more tender joint counts, HAQ score of ≥ 1.0 and an ESR of ≥ 20 mm/hour) to give a four-point score. This index predicted persisting active disease (i.e. a DAS28-ESR score of > 3.2) at 6 and 12 months during follow-up in ERAN.

The value of this four-point score was then assessed in clinical trials in 558 patients with early and established RA. In the early RA trial, only 20% of patients with no predictors had persistent active disease, whereas 80% of patients with all three predictors had persistent active disease (Figure 12). This relationship was significant (p < 0.01). There was a similar relationship in the established RA trial, although this was weaker because none of the patients in the established RA trial had no initial predictive factors. In these patients, 20% of patients with one predictive factor had persistent active disease, compared with 60% of patients with all three predictors (p = 0.05).

FIGURE 12

Simplified predictors of persistently active disease in trial patients.

Predictive factors: high baseline Health Assessment Questionnaire score as an outcome predictor

A second approach to predicting outcomes used initial HAQ scores alone. The value of baseline HAQ scores in predicting outcomes was assessed in 558 patients in early and established RA trials^73,74 and 752 patients followed over ≥ 3 years in the observational study.

In both trials,^73,74 patients with low baseline HAQ scores (≤ 1.50) had significantly more good EULAR responses (both p = 0.013) and significantly lower final mean DAS28-ESR scores (both p > 0.001) than patients with high baseline HAQ scores (> 1.50). These differences are shown in Figure 13. In the observational study, patients with low baseline HAQ scores had significantly lower overall mean DAS28-ESR scores (p < 0.001). In both trials, patients with high initial HAQ scores (> 1.50) had the largest end-point decreases in HAQ score if they achieved good EULAR responses. The observational study showed the same pattern.

FIGURE 13

Initial high and low HAQ scores and changes in HAQ score in EULAR good responders. Mean and standard errors are shown.

Sequential changes in DAS28-ESR scores in both trials^73,74 and the observational study showed that patients with low baseline HAQ scores had lower mean DAS28-ESR scores at all subsequent follow-up time points (Figure 14). The pattern of response was similar in all three groups, although the difference in DAS28-ESR scores attributed to baseline HAQ was larger in the early RA trial⁷³ than in the established RA trial.⁷⁴

FIGURE 14

Sequential changes in mean DAS28-ESR scores in patients with high (> 1.50) and low (≤ 1.50) baseline HAQ scores. Standard errors are shown. (a) Early RA trial; (b) established RA trial; and (c) observational study.

The relationship between baseline HAQ scores, subsequent remissions during treatment and changes in HAQ scores with treatment are shown in Table 5. The analysis shows two main things. First, patients with high baseline HAQ scores had significantly fewer remissions. Second, when remissions occurred, especially several remissions, the changes in HAQ scores were greatest in patients with high baseline HAQ.

TABLE 5

Initial high and low HAQ scores, remissions and changes in HAQ score

Predictive factors: anxiety and depression

The final approach to predicting outcomes assessed the impact of anxiety and depression using EQ-5D, which were related to outcomes in 379 patients in an early RA trial⁷³ using linear regression models.

In unadjusted regression models, patients with moderate and high levels of depression and anxiety at baseline had higher HAQ and DAS28-ESR scores over time and at the trial end point. After adjusting for age, sex, disease duration, time, treatment type, baseline HAQ and DAS28-ESR scores and rheumatoid factor status, there were no longer between-group differences for HAQ score (Table 6). However, there continued to be a significant relationship between high levels of depression and anxiety at baseline and higher end-point DAS28-ESR scores.

TABLE 6

Post-treatment mean differences and standardised mean differences by baseline level of depression/anxiety symptoms

At the trial end point, 80 (21%) patients had remissions (i.e. DAS-28 scores < 2.6). Patients with moderate levels of depression and anxiety at baseline had fewer clinical remissions than patients with no depression and anxiety at baseline [odds ratio (OR) 0.50, 95% CI 0.29 to 0.88; p = 0.02]. Patients with high levels of depression and anxiety symptoms at baseline also had reduced odds of reaching remission; however, this comparison was not significant (OR 0.77, 95% CI 0.25 to 2.33; p = 0.64). This is likely to reflect the small number of patients with extreme symptoms at baseline (n = 24), reducing the power to find a significant effect because of an imprecise estimate.

Secondary analyses of existing data

Most studies in this section were post hoc analyses of existing data. They did not address prespecified hypotheses. As a consequence, caution is needed interpreting their significance.

Rheumatoid factor and other autoantibodies

The prognostic studies did not consider the impact of rheumatoid factor isotypes or anti-citrullinated peptide antibodies. These are recognised response predictors that, together with smoking status, are linked to rheumatoid factor positivity.^231–234 However, as autoantibodies are measured in many different ways across centres, it is impractical to use them in current clinical practice studies. Smoking status is also not usually recorded in routine clinical practice.

Diagnostic criteria and intensive management strategies

Rheumatoid arthritis patients in all studies assessed were enrolled before the introduction of new diagnostic criteria for the classification of RA. There is evidence that these new criteria change the patients classified as having RA, particularly patients with seronegative disease.^45,235–238 The TITRATE trial uses the most recent criteria and it would have been a mistake not to do so. This change makes it difficult to relate historical findings exactly to new data.

The trials studied did not involve the same intensive management strategies used in the TITRATE programme and the impact of changing treatment in patients who did not achieve sustained remissions was not examined. These are general limitations with all trials involving intensive management strategies.^239–244

Duration of DAS28-ESR remission

We found that achieving sustained DAS28-ESR remission gave the greatest chance of minimising disability and maximising health-related quality of life; however, this was uncommon, reflecting international experience with sustained remission.^{217,245–249} More patients benefited when the treatment target was to achieve DAS28-ESR remission at any time during follow-up.

End-point remission and low disease activity were both reasonable targets.

Other assessments of remission

Disease Activity Score for 28 joints based on ESR scores were dominated by the ESR at low levels and in remission. Other composite disease activity assessments, such as CDAI and RAPID3, show different patterns in their components as disease activity changes. However, there was no reason to favour one composite index over another. These findings reflect the ongoing debate about how best to use composite indices in assessing RA disease activity.^250–254 We also found that patient global assessment, a component of most composite measures, was most closely associated with patient-assessed health-related quality of life. Several other recent reports^255–257 have highlighted the importance of patient global assessments in RA.

Simple outcome predictors

Poor outcomes were predicted by several simple baseline measures, including a simple four-point predictive score, initial HAQ score and the presence of anxiety and depression. The situation with the HAQ was complex, as the largest improvements with treatment were seen in patients who had high initial HAQ scores and then showed substantial clinical improvements and achieved remission. Other research has highlighted the relevance of baseline HAQ score^258–261 and depression^262–265 in predicting RA outcomes.