Effectiveness of a clinical practice intervention in early rheumatoid arthritis

Authors


Abstract

Objective

To compare the outcome of early rheumatoid arthritis (RA) patients in a country where early clinics were established versus the outcome of patients in nonprotocolized clinics.

Methods

We compared 2 multicenter cohorts: an RA cohort derived from an early arthritis registry set in 36 reference hospitals in which a specific intervention was established (Evaluation of a Model for Arthritis Care in Spain [SERAP]), and a historical control cohort of patients with early RA attending 34 rheumatology departments (Prognosis in Rheumatoid Arthritis [PROAR] cohort). Effectiveness was tested by comparing the change in the Disease Activity Score in 28 joints (DAS28), the change in the Health Assessment Questionnaire (HAQ), and the change in the Sharp/van der Heijde radiologic score using marginal structural models.

Results

A total of 161 early RA patients were recruited in the PROAR cohort and 447 in the SERAP cohort. Being a SERAP patient was inversely correlated with activity, resulting in a decrease of −0.24 (95% confidence interval [95% CI] −0.39, −0.08) units in the population average of the DAS28 after adjustment was made. Moreover, intervention may be seen as a protective factor of radiologic damage, with a decrease of −0.05 (95% CI −0.09, −0.01) units in the logarithm of the total Sharp/van der Heijde score. On the other hand, a decrease in functional impairment was detected, but intervention was not statistically associated with HAQ changes.

Conclusion

Preventing major radiographic progression in a 2-year term inside structured and organized special programs for the management of disease, such as early arthritis clinics, are effective compared to nonprotocolized referrals, treatment, and followup.

INTRODUCTION

Rheumatoid arthritis (RA) is a chronic polyarthritis whose course is progressive in most individuals and leads toward joint structural damage, functional impairment, decreased quality of life, and increased morbidity and mortality (1–4). An early diagnosis and aggressive management of RA is deemed associated with a great likelihood of a favorable outcome (5, 6), is recommended by most guidelines (7, 8), and is widely accepted by rheumatologists (9).

Most RA patients are initially seen by nonspecialists who may not be aware of the importance of a very early diagnosis, a proper referral, and early therapy. The consequence is that patients are first attended in rheumatology clinics more than 6 or even 12 months after the onset of disease (9, 10). A study in 2000 showed that the median lag time from first symptom to rheumatology referral was 17 months, and the time to initiation of disease-modifying antirheumatic drugs (DMARDs) was 19 months (11).

Acknowledging this gap between primary and specialized care, many countries (UK, The Netherlands, Austria, Norway, France, Italy, Germany, Finland, US), including Spain, started early arthritis (EA) clinics specifically designed to facilitate referral from primary care in order to evaluate and to treat arthritis promptly (12–14).

Accordingly, in November 2004, the Spanish Society of Rheumatology promoted a nationwide project called the Evaluation of a Model for Arthritis Care in Spain (SERAP) project. The project aimed to establish Early Arthritis Units (EAUs) in rheumatology clinics throughout the country, with the intention to provide early detection and appropriate clinical intervention in patients with EA.

Our hypothesis was that the establishment of EAUs would improve the prognosis of EA in terms of both radiologic progression and function at 2 years compared to EA that had not been attended in EAUs. This hypothesis, although basic for justifying the setup of any clinic, was perhaps not addressed properly in the literature, since most evidence supporting the establishment of EAUs was based on studies comparing aggressive and early treatment versus others, not the setting of the clinic as a whole versus no program at all. Our study aimed to compare the outcome of RA patients in an area where early clinics were established versus the outcome of patients in other areas where referral from primary care was not based on specific procedures nor were there specific protocols for following or treating patients.

Significance & Innovations

  • The study of rheumatoid arthritis in its early stages should be a research priority and its management a priority in health care delivery.

  • The benefits of widespread establishment of early arthritis clinics in daily clinical practice are significant.

MATERIALS AND METHODS

We compared the following cohorts: 1) an RA cohort derived from an EA registry in which a specific intervention was established (SERAP cohort) and 2) a historical control cohort of patients with early RA attending rheumatology departments established to identify prognostic factors for severe disease, i.e., the Prognosis in Rheumatoid Arthritis (PROAR) cohort (members of the SERAP and PROAR Study Groups are shown in Appendix A).

PROAR cohort.

The PROAR cohort has been described in detail elsewhere (15). It is a nationwide prospective cohort followed for 5 years (2001–2005) and was assembled to identify the effect of independent variables on poor outcome in EA. Consecutive patients with EA (disease duration <1 year) with no previous use of any type of DMARDs were recruited among 34 rheumatology clinics in Spain. For comparison purposes, only the patients who fulfilled criteria for RA (16) and only their first 2 years of followup were considered.

Intervention: EAUs and SERAP cohort.

The SERAP registry has been described in detail elsewhere (12). In brief, the registry was set in 36 reference hospitals with rheumatology departments. Participating EAUs in reference hospitals were selected based upon the following 2 premises: 1) to ensure enough geographic coverage of the country and 2) the centers should guarantee any system in which patients were seen by a rheumatologist trained in RA within 15 days of the first visit to primary care. A primary care–based registry was established in the participating EAUs.

Each participating EAU selected primary care centers by convenience sampling. In each center, primary care physicians were trained in how to suspect EA and were requested to refer all patients with suspected EA to the EAU within a maximum of 15 days, according to preestablished definitions and as supported by specific referral protocols and materials. The referral procedures were fully flexible, and the units were encouraged to adapt the whole system to their own area, i.e., hiring staff or rearranging spaces as needed as long as the guaranteed 15 days and a trained rheumatologist were in place. A protocol on treatment was not mandatory, but the Spanish clinical practice guideline of RA (8), as well as the recommendations on the use of biologic agents of the Spanish Society of Rheumatology (17), were provided and their application encouraged. Both guidelines recommend immediate treatment at the time of diagnosis.

All patients with an RA diagnosis according to the 1987 American College of Rheumatology classification criteria (16) and who were registered during the first year after the registry was established became part of the longitudinal cohort that was followed for 2 years (2004–2006).

Baseline and followup assessment.

At baseline, and later at each half-yearly visit, the rheumatologists recorded sociodemographic and clinical characteristics of all the patients included in both cohorts. The following parameters were collected: sex, age, level of education, social class, employment status, smoking, history of transfusions, duration from onset of symptoms, prescribed treatment of EA, pain, patient global assessment of activity on a 0–100 visual analog scale, early morning stiffness, tender and swollen joint counts, erythrocyte sedimentation rate, and rheumatoid factor. Inflammatory activity was measured by the Disease Activity Score in 28 joints (DAS28) (18), and functional capacity was obtained from the Health Assessment Questionnaire (HAQ).

Primary outcome variables.

Effectiveness was tested by comparing the change in the DAS28 score, the change in the HAQ score, and the change in the total Sharp/van der Heijde radiologic score (TSS) from baseline to the end of the second year of followup between the 2 cohorts.

Radiographic assessment.

Conventional radiographs of the hands and feet were obtained at baseline and at 1 and 2 years. In the SERAP cohort, a random sample of radiographs was selected between all RA patients (Figure 1). The same trained observer (JI-C) scored all radiographs in both cohorts. Radiographic damage was assessed according to the modified Sharp/van der Heijde method (19), blinded to the identity and treatment status of the patients but not to the chronological order of the images.

Figure 1.

Flow chart of rheumatoid arthritis (RA) patients and radiographs in the Evaluation of a Model for Arthritis Care in Spain (SERAP) cohort (left panel) and the Prognosis in Rheumatoid Arthritis (PROAR) cohort (right panel).

Statistical analysis.

Continuous data in demographics, clinical characteristics, and radiologic damage were presented as means with SD or as medians with quartile ranges. Categorical variables were represented by percentages. Baseline characteristics between groups were compared using the chi-square test, Student's 2-tailed t-test, or the Mann-Whitney U test.

The radiographic end point, i.e., the change in TSS, was compared between the 2 cohorts using generalized estimating equations (GEEs) after logarithmic transformation. To further evaluate the change in the DAS28 and HAQ scores between baseline and the 2-year followup in both cohorts, a GEE model was also performed for each outcome. All models were fit assuming multivariate normal data based on histograms and literature. Results provided β coefficients, which are interpreted as unit changes.

We used inverse probability of treatment weights (IPTW) and inverse probability of censoring weights (IPCW) to adjust for “confounder by indication” and bias due to loss to followup (20). At each followup visit, a logistic regression model was used to estimate the probability that the patient was receiving the observed treatment. Baseline-, previous-, and assessment-specific variables (time-dependent confounders such as the HAQ or the DAS28) were used in the model to predict the treatment decisions. The product of all the probabilities for each patient gave the overall probability of the treatment decisions that occurred over the length of followup. Weighted estimating equations (WEEs) are an extension of GEE and consist in weighting the GEE model with the product of IPTW and IPCW, so the model produces an unbiased estimate of the effect of the intervention on the progression of the DAS28, the HAQ, and the TSS (20). We also used multiple-imputation methods for the analysis of missing data with the mi command in Stata software. The imputed data and clinics were used in the estimation of weights but not in the WEE models.

Only demographic data and clinical variables producing significant associations (P < 0.10) with the outcome variables in bivariate analyses were entered into multivariate models, and then backward stepwise selection was applied in WEE models. The final models were reached by means of quasi-likelihood under the independence model information criterion (QIC) values (21). We used the smallest QIC value to choose among the following competing correlation structures: independent, autoregressive (order 1), unstructured, or exchangeable.

A sensitivity analysis was performed with data from 14 hospitals that contributed to both the PROAR and SERAP cohorts. All analyses were conducted using Stata Statistical Software (StataCorp), release 11.0, except for WEE models, which were calculated by using the command GENMOD of SAS/STAT, version 8.2 for Windows. P values less than 0.05 were considered significant.

Ethical issues.

Patients were required to sign a written consent form after being informed about the details of the study in both cohorts. Both studies were performed according to the principles of the latest Helsinki recommendations, the International Guidelines for Ethical Review of Epidemiological Studies (Council for the International Organizations of Medical Sciences, Geneva, 1991), and the recommendations of the Spanish Society of Epidemiology. The SERAP protocol and materials were approved by the Institutional Review Board (IRB) of the Hospital Institut Municipal d'Assistència Sanitària (Barcelona), and those of the PROAR study were approved by the IRB of the Hospital La Paz (Madrid).

RESULTS

A total of 161 early RA patients were recruited in the PROAR cohort and 447 in the SERAP cohort; the flow chart of the study is shown in Figure 1. Within these groups of patients, 107 and 97 radiographs of hands and feet were received from the PROAR cohort and the SERAP cohort, respectively.

Baseline characteristics of the included patients and the patients with radiographs are shown in Table 1. No statistical differences were observed in most clinical manifestation of the disease, and no statistical differences were observed between a random sample of patients with radiographs and patients without radiographs. At the end of the second year, the percentage of patients with methotrexate was superior in the SERAP cohort than in the PROAR cohort (70% versus 64%), although the difference did not reach statistical significance (P = 0.270).

Table 1. Baseline characteristics of both early RA cohorts*
 PROARSERAPPPROAR radiographsSERAP radiographsP
  • *

    RA = rheumatoid arthritis; PROAR = Prognosis in Rheumatoid Arthritis cohort; SERAP = Evaluation of a Model for Arthritis Care in Spain cohort; HAQ = Health Assessment Questionnaire; DAS28 = Disease Activity Score in 28 joints.

Patients, no.161447 10797 
Women, no. (%)113 (70)318 (71)0.819082 (77)66 (68)0.1700
Age, mean ± SD years54 ± 1555 ± 170.488854 ± 1553 ± 170.6687
Educational level, no. (%)  0.0680  0.2950
 Uneducated or elementary38 (24)69 (15) 19 (18)17 (18) 
 Primary58 (36)197 (44) 45 (42)45 (46) 
 Secondary51 (32)136 (31) 37 (35)25 (28) 
 University12 (8)44 (10) 5 (5)10 (10) 
Employment status, no. (%)  0.6660  0.7540
 Employed61 (38)148 (33) 41 (39)31 (32) 
 Retired34 (21)113 (25) 19 (18)20 (21) 
 Housewife39 (25)118 (26) 28 (27)28 (29) 
 Unemployed7 (4)24 (5) 6 (6)9 (9) 
 Work incapacity18 (11)43 (10) 11 (10)9 (9) 
Polyarticular extension of disease, no. (%)111 (70)343 (77)0.076078 (74)75 (77)0.5370
Course onset, no. (%)  0.3170  0.9550
 Acute50 (31)158 (35) 36 (34)33 (34) 
 Subacute111 (69)288 (65) 71 (66)64 (66) 
Disease duration, median (25th percentile, 75th percentile) months4.5 (2.5, 7.3)4.3 (3, 6.7)0.86574.3 (2.5, 7.5)4.3 (3.1, 6.3)0.7906
Rheumatoid factor positive, no. (%)81 (52)247 (56)0.354052 (50)54 (57)0.3000
HAQ score, mean ± SD1.4 ± 0.71.4 ± 0.80.41181.5 ± 0.61.3 ± 0.70.1220
DAS28 score, mean ± SD5.9 ± 1.25.5 ± 1.30.00566.0 ± 1.15.6 ± 1.20.0194
Smoking ever, no. (%)62 (39)183 (41)0.549031 (29)38 (39)0.1240
Total Sharp/van der Heijde score, median (25th percentile, 75th percentile)   12 (6, 20)7 (3, 18)0.0147
Radiographic progression rate, median (25th percentile, 75th percentile)   34.5 (14.9, 53.6)17.2 (5.8, 51.7)0.0136

To analyze the effect of a protocolized intervention in early RA patients (SERAP cohort) on the outcomes over time, we first analyzed other variables to adjust for as covariates in the WEE models. Three different models were created, 1 for each outcome as the dependent variable: DAS28 score, HAQ score, or logarithm of TSS. Tables 2, 3, and 4 provide the results (β coefficients and 95% confidence intervals [95% CIs]) from each model, respectively, with separate regression coefficients for time, the covariates, and the cohort (with the SERAP cohort being the reference). DMARDs were introduced in the models individually.

Table 2. Variables associated with the progression of the DAS28*
DAS28Bivariate model, coefficient (95% CI)Multivariate model, coefficient (95% CI)
  • *

    DAS28 = Disease Activity Score in 28 joints; 95% CI = 95% confidence interval; anti-TNF = anti–tumor necrosis factor; PROAR = Prognosis in Rheumatoid Arthritis cohort; SERAP = Evaluation of a Model for Arthritis Care in Spain cohort.

  • †, §

    P < 0.001.

  • P < 0.01.

  • §

    P < 0.05.

Intercept4.09 (4.01, 4.18)2.38 (2.03, 2.73)
Time  
 Baseline visit1 (reference)1 (reference)
 6-month visit−2 (−2.14, −1.86)−2.16 (−2.35, −1.97)
 12-month visit−2.25 (−2.39, −2.1)−2.42 (−2.62, −2.22)
 18-month visit−2.41 (−2.57, −2.26)−2.6 (−2.8, −2.39)
 24-month visit−2.58 (−2.74, −2.42)−2.75 (−2.96, −2.54)
Women0.34 (0.15, 0.52)0.31 (0.17, 0.46)
Disease duration (months)0.02 (0, 0.05)0.02 (0.01, 0.04)
Rheumatoid factor positive0.25 (0.07, 0.42)0.32 (0.19, 0.45)
DAS28 (baseline)0.45 (0.4, 0.51)0.51 (0.46, 0.57)
Methotrexate use−1.37 (−1.51, −1.23)0.21 (0.03, 0.4)§
Anti-TNF use−0.86 (−1.25, −0.48)−0.36 (−0.78, 0.06)
Intervention  
 PROAR patient1 (reference)1 (reference)
 SERAP patient−0.26 (−0.46, −0.05)§−0.24 (−0.39, −0.08)
Table 3. Variables associated with the progression of the HAQ*
HAQBivariate model, coefficient (95% CI)Multivariate model, coefficient (95% CI)
  • *

    HAQ = Health Assessment Questionnaire; 95% CI = 95% confidence interval; DAS28 = Disease Activity Score in 28 joints; PROAR = Prognosis in Rheumatoid Arthritis cohort; SERAP = Evaluation of a Model for Arthritis Care in Spain cohort.

  • P < 0.001.

  • P < 0.05.

  • §

    P < 0.01.

Intercept0.87 (0.83, 0.92)−0.16 (−0.29, −0.03)
Time  
 Baseline visit1 (reference)1 (reference)
 6-month visit−0.73 (−0.8, −0.66)−0.17 (−0.25, −0.09)
 12-month visit−0.74 (−0.81, −0.66)−0.13 (−0.21, −0.04)§
 18-month visit−0.83 (−0.9, −0.75)−0.17 (−0.26, −0.08)
 24-month visit−0.83 (−0.9, −0.75)−0.14 (−0.23, −0.05)§
Age (10-years group)0.08 (0.06, 0.11)0.02 (0.01, −0.03)§
DAS280.3 (0.29, 0.32)0.26 (0.23, −0.28)
DAS28 (baseline)0.15 (0.12, 0.19)−0.13 (−0.16, −0.11)
HAQ (baseline)0.54 (0.49, 0.59)0.52 (0.48, 0.56)
Methotrexate use−0.51 (−0.59, −0.44)−0.07 (−0.14, 0)
Intervention  
 PROAR patient1 (reference)1 (reference)
 SERAP patient−0.04 (−0.14, 0.06)0.01 (−0.03, 0.06)
Table 4. Variables associated with the progression of the TSS*
TSS (in log scale)Bivariate model, coefficient (95% CI)Multivariate model, coefficient (95% CI)
  • *

    TSS = total Sharp/van der Heijde score; 95% CI = 95% confidence interval; PROAR = Prognosis in Rheumatoid Arthritis cohort; SERAP = Evaluation of a Model for Arthritis Care in Spain cohort.

  • P < 0.001.

  • P < 0.05.

  • §

    P < 0.01.

Intercept2.87 (2.79, 2.95)0.07 (−0.02, 0.16)
Time  
Baseline visit1 (reference)1 (reference)
 12-month visit0.18 (0.15, 0.2)0.18 (0.15, 0.21)
 24-month visit0.3 (0.25, 0.34)0.3 (0.26, 0.35)
TSS (baseline)0.98 (0.94, 1.01)0.97 (0.94, 1.0)
Rheumatoid factor positive0.08 (0.08, −0.08)0.04 (0.01, 0.08)
Intervention  
 PROAR patient1 (reference)1 (reference)
 SERAP patient−0.24 (−0.4, −0.09)§−0.05 (−0.09, −0.01)

Baseline mean activity was statistically higher in the PROAR cohort (nonintervention). The progression in the DAS28 scores over time are shown in the left panel of Figure 2A. The mean ± SD DAS28 score was 5.5 ± 1.3 at baseline and gradually decreased to 2.9 ± 1.3 in the second year in the intervention group. This decrease was also present in PROAR cohort patients, whose mean ± SD DAS28 scores dropped from 5.9 ± 1.2 to 3.3 ± 1.6. At the end of the second year, the DAS28 score was statistically different (P = 0.0373) between both groups.

Figure 2.

A, Box plot of DAS28 over time (months) in each PROAR cohort (left panel) and SERAP cohort (right panel). B, Box plot of HAQ over time (months) in each PROAR cohort (left panel) and SERAP cohort (right panel). DAS28 = Disease Activity Score in 28 joints; HAQ = Health Assessment Questionnaire; PROAR = Prognosis in Rheumatoid Arthritis; SERAP = Evaluation of a Model for Arthritis Care in Spain.

Results showed that the DAS28 score was positively associated with the DAS28 score at baseline, age, sex, disease duration, and rheumatoid factor positivity, and negatively associated with methotrexate use (Table 2). Finally, in the multivariate model, intervention (the SERAP cohort) was inversely correlated with activity, resulting in a decrease of −0.24 (95% CI −0.39, −0.08) units in the population average of the DAS28 after adjustments were made.

With regard to functional capacity, the mean ± SD HAQ score improved considerably from baseline (1.4 ± 0.7) to 6 months (0.6 ± 0.6) in the SERAP cohort. After that, the HAQ improved slightly until it reached 0.5 ± 0.6. The results were similar in the PROAR cohort (Figure 3). At the end of the second year, the HAQ score was not statistically different between groups without adjustment, nor was it different at baseline. Having an older age and increased HAQ score at baseline were both associated with worse progression of HAQ scores (Table 3). Regarding the effect of the DAS28 on the HAQ, worse DAS28 values were associated with worse HAQ scores. The effect of time was significant; a decrease of −0.14 (95% CI −0.23, −0.08) units was observed when comparing the HAQ at the end of study to baseline values. However, intervention was not statistically associated to HAQ decrements after adjustment.

Figure 3.

Cumulative probability plots of joint erosion score (JES) in Prognosis in Rheumatoid Arthritis (PROAR) and Evaluation of a Model for Arthritis Care in Spain (SERAP) cohorts. JES was calculated at 12 months (left panel) and at 24 months (right panel).

At baseline, median Sharp/van der Heijde scores were already larger in PROAR cohort patients than in SERAP cohort patients, and the progression rate was double in the nonintervention group (Table 1). The median TSS (25th percentile, 75th percentile) of PROAR cohort patients worsened from 12 (6, 20) at baseline to 21 (11, 28) after 2 years (P < 0.001). The same happened in the SERAP cohort, with the median TSS (25th percentile, 75th percentile) worsening from 7 (3, 18) to 13 (5, 25) after 2 years (P = 0.033). Differences between the cohorts were also significant after 2 years (P = 0.020). The percentage of patients with no progression of erosions is clearly higher in the SERAP group (P < 0.001), as shown in the cumulative probability plots of the joint erosion score (Figure 3). Finally, time, rheumatoid factor positivity, TSS score at baseline, and intervention were also independently associated with TSS (Table 4). Nevertheless, while TSS at baseline, rheumatoid factor positivity, and time are associated with radiographic progression, intervention may be seen as a protective factor of radiologic damage, especially for erosions. The results for intervention in the sensitivity analysis (data not shown) were overlapping with the primary analysis but with wider CIs.

DISCUSSION

Acknowledging the difficulties of designing a clinical trial in health services, the Spanish Society of Rheumatology designed a comparative study to contrast the outcome of RA inside and outside of the EA clinic setting. We established 36 EA clinics with flexible written recommendations for referral, access, followup, and treatment as part of a new model of care, and we compared the outcome of early RA in such clinics with patients included in a historical early RA cohort in which patients had been treated in regular rheumatology clinics.

The findings of our work may be summarized as follows: patients may benefit from attending structured and organized programs for the management of disease, such as EA clinics. Disease activity, as assessed by the DAS28, improves in patients with early referral, early diagnosis, and early therapy despite the procedures used. This study also highlights the fact that radiologic progression is faster in patients outside protocolized consults than in EA clinics, at least during the first 2 years.

Our findings are in agreement with other results. Two meta-analyses of clinical trials and observational studies documented the benefits of early intervention compared with delayed treatment, especially in those patients with more severe disease (6, 22). As shown in the Helsinki Cohort (1986) (23), patients with less duration of symptoms have more active disease at presentation at the clinic. In an observational study, Bukhari et al (24) evaluated the role of early DMARD treatment on radiographic outcome at 5 years after symptom onset. The results showed that the progression in the Larsen score was greater in those patients who delayed their DMARD treatment more than 6 months after symptom onset. In another nonrandomized comparison at the Leiden EA clinic (25), RA patients treated early had a more favorable course than those treated later. The mean Sharp/van der Heijde score in the delayed group progressed statistically further, and there was a significantly greater improvement in the median DAS28 in the early treatment group compared with the delayed treatment group.

On the other hand, early treated patients with recent-onset inflammatory polyarthritis recruited in the Norfolk Arthritis Register (26) experienced a nonsignificant improvement in functional outcome compared with those never treated in the adjusted analysis. At the Leiden EA clinic (25), the HAQ score showed modest improvements in both groups, but differences were not statistically significant after 2 years of followup.

One could argue that the differences between groups in our study were not important despite high consistency among end points analyzed. For example, some authors have set the value of ∼1.3 units of the DAS28 as the minimum clinically important difference (MCID) (27), but the mean DAS28 is expected to improve in the first 2 years of treatment in any naive patient, and finding differences beyond this improvement is of merit. The MCID for the HAQ was ∼0.22 units (28). Nonetheless, physical function is slow enough not to really worsen in 2 years, as in this phase of the disease it relies mainly on activity, and HAQ responsiveness is inversely associated with mean disease duration in RA (29). Again, finding differences between the groups would have been very challenging. Regarding the radiographic scores, the MCID for TSS was ∼6 units (30); as in our study the difference was ∼1 unit after undoing the log transformation. Patients in the PROAR cohort had already had a significant progression when they entered the cohort despite no differences in disease duration. Therefore, they had less room for worsening than the SERAP cohort patients, since earlier disease, when treated appropriately, does not progress as rapidly as more advanced disease. Yet, the PROAR group still progressed further than the SERAP group (Figure 3). It is plausible that patients entering the PROAR cohort were more advanced than SERAP patients, as they had been referred based upon clinical judgment of the primary care physicians and not via a previous training and established criteria.

When the PROAR cohort was established in 2001, biologic therapies were available in Spain and there were national guidelines to treat RA (31, 32). Participating rheumatologists were instructed not to vary their practice, as the objective was to assess prognostic factors of EA in real-life conditions. Three years afterward, in the SERAP program, rheumatologists were instructed to specifically follow the Spanish national guidelines available at that time, one of which was an update of a previous guideline (31, 33), although the compliance with guidelines was not assessed. The difference in outcome between cohorts may be due partly to the prevailing guidelines at the time of the respective studies (34), although the baseline years were not so separated in time and the specific guidelines available were not that different. Yet, all these differences occurred despite a flexible intervention that was open to clinical variability.

The major strength of the comparison is that the health area, the variables, and the timing of evaluation (6 months) overlapped in both cohorts. Furthermore, the SERAP cohort consists of many EAUs spread all over the country with very different settings and background policies, which underscores the benefit of any form of EAU with a given protocol for referral and management. The widespread establishment of EA clinics (12–14) has been based on studies of the efficacy of partial strategies mainly related to early aggressive treatment (35) or a strategy of intensive outpatient management that has already shown a substantial improvement (36), whereas our results are a reflection of an intervention in daily clinical practice.

Still, some limitations are in the research agenda; the evaluation of the impact of the treatment delay on the long-term outcome of AR is difficult (37). On one hand, how early is early, or what is the best time to start treatment? The “window of opportunity” has no formal definition (38, 39). This arbitrariness is well known, albeit the concept is accepted by most rheumatologists and the “earlier the better” principle of treatment prevails (7). On the other hand, the extent to which improvement is maintained in the long term needs clarification. This was evaluated by Verstappen et al in a 5-year followup study (40). In this study, the results favoring early DMARD treatment after the first year were not as evident after 5 years. Unfortunately, with only 2 years of followup, our study does not help to unveil the effect.

Another limitation of our study was the presence of missing data and censoring during followup, as well as confounding by indication. The use of marginal structural models to adjust for such confounders should ameliorate these effects. In addition, patients included in either the SERAP or the PROAR cohort were not clearly subject to selection bias, as they were recruited from over the whole country. Baseline characteristics were representative and, like other cohorts of early RA (25, 41), were different from most patients with established RA observed in the clinical setting, of whom 70% or more had a positive rheumatoid factor test and most had progressive disease with radiographic progression, severe functional declines, and controlled activity (42). However, one may think that a historical cohort, despite clear similarities and adjusting for confusion, may not be the perfect comparator. Unfortunately at present, not offering patients fast, early access to a rheumatologist, tight control, and a treat-to-target approach, which is the intervention in the SERAP cohort, would not be justifiable.

In conclusion, the study of RA in its early stages should be a research priority, and its management a priority in health care delivery. EA clinics are effective in improving activity and in preventing major radiographic progression in a 2-year term, compared to nonprotocolized referral and followup. Ideally, the clinic should be able to provide easy access to patients for expert consultation within 2 weeks of referral to an EA clinic and offer early protocolized response to intervention.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Carmona had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Carbonell, González-Álvaro, Sanmarti, Balsa, Carmona.

Acquisition of data. Carbonell, González-Álvaro, Balsa, Román-Ivorra, Ivorra-Cortéz, Lisbona, Alperi.

Analysis and interpretation of data. Descalzo, González-Álvaro, Hernandez-Barrera, Román-Ivorra, Jiménez-Garcia, Carmona.

ROLE OF THE STUDY SPONSOR

Abbott Laboratories provided the financial support for the arthritis units and for the study, but did not participate in the study design, data collection, analysis, or writing of the manuscript, and publication was not contingent on the approval of Abbott Laboratories.

Acknowledgements

The authors gratefully acknowledge the invaluable aid of Jesús Maese, Paloma Sánchez, and Silvia Herrera from the Spanish Society of Rheumatology with data entry, monitoring, and logistics of the study, as well as Estíbaliz Loza for editing the final manuscript.

Appendix A

MEMBERS OF THE SERAP AND PROAR STUDY GROUPS

Members of the SERAP and PROAR Study Groups are J. Manero, A. Pecondón (Hospital Miguel Servet, Zaragoza); F. Navarro, B. E. Hernández, F. J. Toyos, S. Ricca (Hospital Universitario Virgen de la Macarena, Sevilla); S. Reneses, A. García (Hospital Virgen del Rocío Sevilla); J. L. Marenco, L. Mayordomo (Hospital Universitario Valme, Sevilla); A. Fernández-Nebro, M. V. Irigoyen, M. Pérez (Hospital Carlos Haya, Málaga); E. Raya, A. Rueda (Hospital Clínico, Universitario San Cecilio, Granada); F. J. Ballina, M. Alperi (Hospital Central de Asturias Oviedo); L. Espadaler, J. Fiter (Hospital Son Dureta, Palma de Mallorca); A. Naranjo (Hospital de Gran Canaria Dr. Negrín, Las Palmas de Gran Canaria); M. Brito, J. Sánchez (Hospital Nuestra Señora De Candelaria, Santa Cruz de Tenerife); V. Rodríguez (Hospital Universitario Marqués de Valdecilla, Santander); J. del Pino (Hospital Universitario de Salamanca, Salamanca); M. Fernández (Hospital Universitario de Guadalajara, Guadalajara); J. Maymó, J. Carbonell, P. Lisbona (Hospital Del Mar, Hospital Esperança, Barcelona); R. Sanmartí, J. Cañete, G. Salvador, M. V. Hernández (Hospital Clínic de Barcelona, Barcelona); P. Barceló, E. Moreno (Hospital Vall d′Hebrón, Barcelona); J. Valverde, F. J. Narvaez, X. Juanola (Hospital Bellvitge, Barcelona); M. Larrosa (Hospital de Sabadell, Sabadell); C. Díaz, B. Nishishinya, A. Laiz (Hospital Santa Creu I Sant Pau, Barcelona); X. Tena, L. Mateo (Hospital Germans Trias, Badalona); E. Pascual, T. Pedraz, A. Martínez (Hospital General Universitario Alicante, Alicante); J. J. García (Hospital Universitario La Fe, Valencia); J. Calvo (Hospital General de Valencia, Valencia); J. A. Román, J. Ivorra, R. Hortal (Hospital Universitario Dr. Peset, Valencia); J. A. González, F. J. Navarro, J. Tovar (Hospital General de Elche, Elche); J. M. Salazar, J. L. Álvarez (Hospital Regional Infanta Cristina, Badajoz); J. Graña, F. Galdo, M. Freire (Hospital Juan Canalejo, A. Coruña); J. J. Gómez-Reino, A. Mera (Hospital Clínico Universitario, Santiago); A. Laffon, I. González-Álvaro, M. R. García de Vicuña, A. Ortiz (Hospital Universitario la Princesa, Madrid); I. Mateo, M. R. González (Hospital Doce de Octubre, Madrid); A. Balsa, T. Cobo, A. Hernández (Hospital La Paz, Madrid); A. Zea, S. Rodríguez, A. Bardal (Hospital Ramón y Cajal, Madrid); L. Carreño, I. Monteagudo (Hospital Gregorio Marañón, Madrid); C. Marras, E. Soriano (Hospital Virgen de la Arrixaca, Murcia); M. Figueroa, O. Maiz (Hospital Donostia, San Sebastián); A. Alonso (Hospital de Cruces, Baracaldo); A. Corrales (Hospital Comarcal de Laredo, Laredo); A. Cruz (Hospital Severo Ochoa, Madrid); B. Ribas (Hospital San Juan de Dios, León); C. Mata (Hospital Comarcal Sierrallana, Torrelavega); E. Ciruelo (Hospital General de Segovia, Segovia); E. Chamizo (Hospital de Mérida, Mérida); F. Martínez (Hospital Universitario Reina Sofía, Córdoba); F. Pérez (Hospital General de Requena, Requena); G. Pérez (Hospital del Insalud-Ceuta, Ceuta); J. Quirós (Hospital Fundación Alcorcón, Alcorcón); J. Rivera, T. González (Instituto Provincial de Rehabilitación, Madrid); J. García-Arroba, M. Cantalejo (Hospital Universitario de Getafe, Getafe); J. A. Cabezas (Complejo Hospitalario San Millán-San Pedro, Logroño); J. Vázquez (Hospital Virgen de la Luz, Cuenca); M. C. García (Hospital de Malalties Reumatiques, Barcelona); M. Pujol (Hospital Mutua Terrassa, Barcelona); M. A Guzmán (Hospital Virgen de las Nieves, Granada); M. Riera (Hospital Creu Roja, Barcelona); M. Rodríguez (Complejo Hospitalario Cristal-Piñor, Orense); R. Roselló (Hospital General San Jorge, Huesca); S. M. Gelman (Hospital General de Manresa, Manresa); J. Maese (Unidad de Investigación de la SER, Madrid).

Ancillary