Prevalence of and predictive factors for sustained disease-modifying antirheumatic drug–free remission in rheumatoid arthritis: Results from two large early arthritis cohorts

Authors


Abstract

Objective

Remission has become an attainable goal of rheumatoid arthritis (RA) treatment, especially since the advent of biologic antirheumatic therapy. Because little is known about patients who achieve disease remission with conventional treatment, we used 2 large independent inception cohorts to study the prevalence of and predictive factors for disease-modifying antirheumatic drug (DMARD)–free sustained remission after treatment with conventional therapy.

Methods

Remission of disease was assessed in 454 patients from the Leiden Early Arthritis Clinic (EAC) and in 895 patients from the British Early Rheumatoid Arthritis Study (ERAS) who fulfilled the American College of Rheumatology 1987 revised criteria for the classification of RA and were treated with conventional therapy. Sustained DMARD-free remission was defined as fulfilling the following criteria for at least 1 year: 1) no current DMARD use, 2) no swollen joints, and 3) classification as DMARD-free remission by the patient's rheumatologist. Predictive factors were identified by Cox regression analysis.

Results

Sustained DMARD-free remission was achieved by 68 of 454 patients (15.0%) in the Leiden EAC and by 84 of 895 patients (9.4%) in the ERAS. Six factors were associated with sustained DMARD-free remission in both cohorts: acute onset, short symptom duration before inclusion, not smoking, little radiographic damage at baseline, absence of IgM rheumatoid factor (IgM-RF), and absence of HLA shared epitope alleles. In the ERAS, low disease activity at baseline was also predictive of remission. Multivariate analyses revealed symptom duration and the absence of autoantibodies (anti–cyclic citrullinated peptide 2 and IgM-RF) as independent predictors.

Conclusion

Sustained DMARD-free remission in RA patients treated with conventional therapy is not uncommon. Symptom duration at presentation and the absence of autoantibodies are associated with sustained DMARD-free remission.

Patients with rheumatoid arthritis (RA) vary considerably in terms of their disease course and outcome. The spectrum of possible disease outcomes extends from debilitating, destructive joint disease to remission, the most favorable outcome (1). Since the introduction of biologic antirheumatic treatment, remission is increasingly being reported as a primary disease outcome of new therapeutic trials (2). In addition to the potent suppression of disease activity that can be achieved with these novel treatment agents, the recent insight that early initiation of antirheumatic therapy leads to better long-term outcomes (3, 4) has led to a shift in the current goals of RA treatment toward aiming for remission.

When studying remission in RA, there are several aspects of remission that need to be taken into account and that are inconsistently defined and used in the literature. The first inconsistency involves the duration of remission. Clinical trials frequently use a Disease Activity Score (DAS)–based definition of remission that does not require a specified followup period but receiving can be fulfilled at a single point in time.

The second factor that needs to be taken into account is the use of antirheumatic treatment. Remission can be interpreted as either a state of minimal disease activity while receiving antirheumatic treatment or as persistent resolution of the disease after discontinuation of therapy (5, 6). The DAS-based remission criteria do not require the discontinuation of drugs because they were intended to monitor treatment response in clinical trials (7, 8). Remission rates as reported in clinical trials, therefore, often reflect the number of patients who achieve minimal disease activity while receiving novel therapeutic agents. Data are scarce regarding how often remission persists after discontinuation of treatment (9), although this may better reflect disease resolution.

A third aspect of remission studies is the type of treatment that was administered. Despite the fact that remission rates after treatment with new therapeutic agents are now often reported, there are few data on remission after treatment with conventional therapy, such as nonsteroidal antiinflammatory drugs (NSAIDs) and nonbiologic disease-modifying antirheumatic drugs (DMARDs). However, such data would allow a better interpretation of the remission rates reported by investigators in clinical trials of novel agents.

In this study, we investigated the prevalence of and prognostic factors for sustained DMARD-free remission in 2 large, independent inception cohorts of RA patients treated with conventional therapy. In order to investigate remission as a definitive disease outcome, we used a stringent definition of remission: the sustained absence of synovitis for at least 1 year after the discontinuation of therapy with DMARDs. This definition approximates a definitive cure of the disease and, as such, is close to the meaning of remission as used for other diseases such as malignancies.

Investigating remission as a definitive disease outcome, resembling cure, is very important from a pathophysiologic point of view. Knowledge of which clinical or immunologic characteristics of the patient are associated with remission could fuel new hypotheses about the biologic pathways involved in disease persistence and resolution and would increase our understanding of the disease course of RA.

PATIENTS AND METHODS

Patient population.

The study population consisted of 2 cohorts: the Leiden Early Arthritis Clinic (EAC) cohort and the British Early Rheumatoid Arthritis Study (ERAS) cohort. The Leiden EAC cohort is an inception cohort of patients with recent-onset arthritis (<2 years of symptoms) that was initiated at the Department of Rheumatology, Leiden University Medical Center, in 1993 (10). The present study included patients who fulfilled the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) 1987 revised criteria for the classification of RA (11) at baseline (n = 369) or within the first year of followup (n = 85) and who presented to the Leiden EAC between 1993 and 2002.

The treatment strategy in the Leiden EAC differed according to the inclusion period. Patients included between 1993 and 1995 were initially treated with analgesics and were subsequently treated with hydroxychloroquine or sulfasalazine if they had persistent active disease. Between 1996 and 1998, patients who were included were promptly treated with chloroquine or sulfasalazine, while after 1998, the initial treatment strategy consisted of either methotrexate or sulfasalazine. Patients included in the Leiden EAC after January 2003 were not part of the current study because, due to their limited duration of followup, they only had a short period of time to achieve sustained DMARD-free remission and therefore had a high risk of being misclassified. The patients who presented after January 2003 did not differ markedly from those who were included earlier with regard to the rate and the predictive factors of sustained DMARD-free remission (data not shown).

Followup visits with standard clinical assessments were performed 2 weeks after the first presentation and yearly thereafter. Demographic characteristics, a swollen joint count (SJC) of 44 joints, and the score on the modified Stanford Health Assessment Questionnaire (M-HAQ) (12) were recorded at each visit. Laboratory evaluation consisted of erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) measurements and determination of IgM rheumatoid factor (IgM-RF) and anti–cyclic citrullinated peptide 2 (anti–CCP-2) antibodies. Furthermore, genotyping of the HLA–DRB1 region was performed to determine the number of RA-associated shared epitope (SE) alleles (13). Radiographs of the hands and feet were scored according to the modified Sharp/van der Heijde scoring method (SHS) (14).

To determine whether the prevalence and predictive factors that were identified in the Leiden EAC cohort could be replicated, we used a second inception cohort, the ERAS cohort. This cohort consists of RA patients who were recruited from 9 rheumatology departments in the UK from 1986 until 1996. During this period, consecutive patients who presented with RA according to the ACR criteria were included in the ERAS cohort if their RA symptoms had lasted <2 years and if no second-line antirheumatic medication had been used. ERAS patients were treated according to the rheumatologist's preference, which generally entailed a short course of analgesics followed by sequential monotherapy or combination therapy for more severe RA, with methotrexate, sulfasalazine, and hydroxychloroquine as favored drugs.

Baseline and yearly assessments included the Ritchie Articular Index (RAI) (15), an SJC of 44 joints, the M-HAQ, and measurements of ESR and IgM-RF. Disease Activity Scores were calculated according to the original formula (16). In addition, HLA–DRB1 genotyping was performed. Radiographs of the hands and feet were scored according to the Larsen method (17). In contrast to the Leiden EAC cohort, measurements of CRP and anti–CCP-2 were not performed in the ERAS cohort.

A number of patients failed to complete followup both in the ERAS (n = 384 [30%]) and in the Leiden EAC (n = 136 [23%]) cohorts. The most common reasons in both cohorts were death, moving, and withdrawal of consent. Because the definitive disease outcome (remission versus nonremission) could not be accurately determined for these patients, we report the results of the analyses without this group of patients. Therefore, in total, 454 Leiden EAC patients and 895 ERAS patients were included in the present study. If the patients who did not complete followup were included in the analysis as being part of the nonremission group, the identified predictive factors did not change (data not shown).

Definition of remission.

Sustained DMARD-free remission was defined according to the following 3 criteria: 1) no current use of DMARDs, 2) no swollen joints, and 3) classification as DMARD-free remission by the patient's rheumatologist. Corticosteroids were considered to be equivalent to DMARDs for the present study, while NSAIDs did not qualify as DMARDs. Patients had to fulfill all 3 criteria in order to be diagnosed as having disease in remission. To ensure that remission was not temporary but rather sustained and long-lasting, the absence of swollen joints had to have been observed by a rheumatologist for at least 1 year after discontinuation of DMARD therapy.

In both the Leiden EAC and the ERAS cohorts, all patients were seen approximately every 4 months by their rheumatologists even when they had very low disease activity. In addition, standardized followup visits were performed annually to collect data for the Leiden EAC and the ERAS. In the Leiden EAC, all patients with disease in remission were discharged from the outpatient clinic after at least 1 year of observation after discontinuation of DMARDs in the absence of joint swelling. Most patients in the Leiden EAC cohort who achieved remission of disease were followed up longer than the minimum requirement of 1 year; the median time of observation after discontinuation of DMARDs in the absence of swollen joints was 2.5 years. Patients who had a recurrence of their arthritis after discharge could easily return to the Leiden University Medical Center, which is the only referral center for rheumatology in a health care region of ∼400,000 inhabitants. The frequency of disease relapse was recorded, and patients with disease relapse (n = 6) were included in the nonremission group. In the ERAS cohort, the majority of patients who had achieved disease remission continued to undergo their yearly assessments, which provided an opportunity to determine whether there was continued absence of joint swelling.

Statistical analysis.

Summary statistics were generated to investigate the prevalence of sustained DMARD-free remission and the baseline characteristics of patients with or without disease in remission. Patients included in the ERAS had a longer average period of followup than did patients in the Leiden EAC, since the ERAS was initiated in 1986 whereas the Leiden EAC was started in 1993. To avoid skewing of the results due to the difference in followup time, the present analysis used data from the first 10 years of followup for all patients.

Baseline variables were assessed for their ability to predict remission by univariate and multivariate regression analysis. To take into account the difference in followup times among patients, we performed Cox regression analyses after verifying that the proportional hazards assumption was satisfied. In Cox regression models, the dependent variable is the “time-to-event,” which consisted of the time to remission for the patients with remission and the time to the last followup visit (with a maximum of 10 years) for the patients without remission. The time of remission was defined in the Leiden EAC as the date at which DMARDs were discontinued due to remission and in the ERAS as the date of the first annual study visit when patients had DMARD-free remission of disease. The analysis was also performed with a later date defined as the time of remission (the date described above plus 1 year), which led to similar results. This indicates that the predictive factors were stable regardless of the exact date used to define remission.

In order to investigate the predictive ability of baseline characteristics in univariate analysis, each variable was included as a covariate in a separate nonconditional Cox regression analysis. The results of the univariate analyses were subjected to correction for multiple testing using the Holm method (18). Subsequently, multivariate Cox regression analysis was performed to identify significant independent predictors for achieving remission. As possible explanatory variables, all baseline variables with a P value less than 0.10 in univariate analyses were included in the model. A 2-step modeling approach was performed that in the first step, identified independent predictive variables by a backward step selection procedure that removed variables with a P value greater than 0.10. To verify that the identified predictive variables were indeed independent predictors for the entire cohort, they were then entered as covariates into a second multivariate Cox regression analysis (without a stepwise selection procedure).

If there is a highly significant correlation between 2 variables, they cannot be entered into the same multivariate model, since this results in overcorrection and annulment of the effect. In both cohorts, symptom duration and radiographic damage (as measured by the SHS in the Leiden EAC cohort and by the Larsen score in the ERAS cohort) were highly correlated. Symptom duration was included in the multivariate models because a longer symptom duration most likely caused a higher radiographic score and not vice versa. In the ERAS cohort, the RAI, SJC, and DAS were highly correlated since the DAS is a component measure consisting of, among other measures, the RAI and the SJC; therefore, the DAS was omitted from the multivariate model.

In the Leiden EAC cohort, several patients (<50) had missing values for baseline continuous variables, which would have resulted in their exclusion from the multivariate model. For these patients with missing values, multiple imputation analysis was performed using Stata software, release 10, with the ice and mim packages (StataCorp, College Station, TX). Multiple imputation was applied to the final Cox regression model identified through the preliminary variable selection step (predictive variables were symptom duration, CRP, and anti–CCP-2). The full database was used to generate 25 imputations, each from 10 cycles and incorporating both survival and censoring information within the multiple imputation prediction equations. In the ERAS cohort, the number of patients with missing values was very limited, and for that reason, no imputation was performed.

All analyses were performed using SPSS version 14.0 (SPSS, Chicago, IL), except for the multiple imputation. P values less than or equal to 0.05 were considered significant, except for the results of univariate analysis, for which subsequent correction for multiple testing was performed.

RESULTS

Prevalence of sustained DMARD-free remission.

The baseline characteristics of the RA patients in the Leiden EAC and ERAS cohorts are shown in Table 1. On average, ERAS patients had a longer disease duration and a higher number of swollen joints at baseline, and they were more often IgM-RF positive and HLA SE allele positive than Leiden EAC patients.

Table 1. Baseline characteristics of the patients in both cohorts*
 Leiden EAC (n = 454)ERAS (n = 895)
  • *

    EAC = Early Arthritis Clinic; ERAS = Early Rheumatoid Arthritis Study; SJC = swollen joint count; IQR = interquartile range; IgM-RF = IgM rheumatoid factor; SE = shared epitope; SHS = modified Sharp/van der Heijde score; ND = not done.

Period of inclusion1993–20031986–1996
Age, mean ± SD years56 ± 1652 ± 13
Women, %6969
Symptom duration, mean ± SD months6.4 ± 98.3 ± 6
SJC in 44 joints, median (IQR)8 (4–14)14 (7–25)
IgM-RF positive, %5863
HLA SE positive, %6672
SHS (0–448), mean ± SD8.0 ± 11ND
Larsen score (0–250), mean ± SDND3.7 ± 9

Sustained DMARD-free remission was achieved by 68 of the 454 patients (15.0%) in the Leiden EAC cohort and by 84 of the 895 patients (9.4%) in the ERAS cohort. The median time to remission was 43 months in both cohorts (interquartile range [IQR] 24–67 months for the Leiden EAC and 18–60 months for the ERAS) (Figure 1A).

Figure 1.

A, Kaplan-Meier curves of the percentage of patients with sustained disease-modifying antirheumatic drug (DMARD)–free remission in the Leiden Early Arthritis Clinic (EAC) cohort compared with the British Early Rheumatoid Arthritis Study (ERAS) cohort. B, Kaplan-Meier curves of the percentage of patients in the Leiden EAC cohort with sustained DMARD-free remission, stratified by period of inclusion. Differences between the different strata were not significant (P = 0.52 by log rank test across all strata). The line representing the patients included from 1999 until 2002 terminates at 8 years of followup because this was the maximum followup for patients in this stratum (data were collected in 2007).

A subgroup analysis was performed to determine whether the prevalence of remission in the Leiden EAC cohort would have been different if only the 369 patients who presented with RA at baseline had been included (leaving out the 85 patients who developed RA within the first year of followup). This revealed a prevalence of sustained DMARD-free remission similar to that in the entire group (14.9% [55 of 369]).

The prevalence of remission steadily increased in the first part of the followup period until ∼7 years after inclusion, after which it remained relatively stable. To determine whether patients who were included at a later time point and who had therefore undergone a different treatment strategy had a higher chance of achieving remission or achieved remission faster, analyses in both cohorts were stratified according to the inclusion period. This revealed that the rate of achieving remission did not differ between patients in different inclusion periods (P = 0.52 by log rank test), as shown for the Leiden EAC cohort in Figure 1B. The same was true for patients who were included in different time periods in the ERAS (data not shown).

Univariate analysis.

In order to investigate which baseline patient characteristics could predict sustained DMARD-free remission, univariate Cox regression analysis was performed. The following variables were associated with sustained DMARD-free remission in the Leiden EAC: symptom duration at baseline, smoking, CRP, IgM-RF and anti–CCP-2 positivity, radiographic damage as measured by the SHS, and HLA SE alleles (Table 2). A subgroup analysis of the 369 patients who presented with RA at baseline (without the 85 patients who developed RA within the first year of followup) again led to the same results. Of these identified factors, the presence of IgM-RF, the presence of HLA SE alleles, smoking, symptom duration at baseline, and acute start of symptoms could be replicated in the ERAS (Figure 2). The association between sustained DMARD-free remission and radiographic damage (as measured by the SHS in the Leiden EAC cohort and by the Larsen score in the ERAS cohort) could also be replicated.

Table 2. Leiden EAC univariate analysis: of baseline characteristics of the patients who did and those who did not achieve sustained DMARD-free remission*
 Remission (n = 68)No remission (n = 386)HR (95% CI)P
  • *

    For variables that were normally distributed, the mean ± SD is reported. For variables that were not normally distributed, the median and IQR are reported along with the mean ± SD, since Cox regression uses parametric analysis. For dichotomous variables, the number (%) of patients is listed relative to the total number of patients for whom information was available about the characteristic under investigation. DMARD = disease-modifying antirheumatic drug; RA = rheumatoid arthritis; M-HAQ = modified Health Assessment Questionnaire; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; anti–CCP-2 = anti–cyclic citrullinated peptide 2 (see Table 1 for other definitions).

  • The hazard ratio (HR), which is the effect measure generated by Cox regression analysis, can be interpreted similar to an odds ratio, meaning that a higher HR signifies a greater chance of remission. HRs are reported along with their 95% confidence intervals (95% CIs).

  • P < 0.05.

  • §

    The Ritchie Articular Index and the Disease Activity Score were not reported in the Leiden EAC.

  • Significant after correction for multiple testing by the Holm method.

Age, mean ± SD years59 ± 1756 ± 161.02 (0.99–1.03)0.067
Women, no. (%)50 (74)262 (68)1.28 (0.74–2.19)0.38
Symptom duration at baseline, months    
 Median (IQR)3 (1–6)5 (2–9)0.94 (0.88–0.99)0.021
 Mean ± SD4.6 ± 56.7 ± 9
Smoking, no. (%)22 (33)175 (47)0.56 (0.34–0.94)0.028
Family history of RA, no. (%)12 (19)111 (31)0.55 (0.30–1.04)0.064
Absence of comorbidities, no. (%)43 (64)233 (63)0.98 (0.59–1.61)0.92
Acute start of symptoms, no. (%)43 (64)192 (53)1.55 (0.94–2.56)0.084
Start of symptoms in small joints, no. (%)38 (57)175 (47)1.48 (0.91–2.40)0.11
Symmetric start of symptoms, no. (%)46 (72)239 (67)1.24 (0.72–2.14)0.44
SJC in 44 joints§    
 Median (IQR)8.0 (4–12.5)8.0 (4–14)1.00 (0.96–1.04)0.82
 Mean ± SD9.6 ± 79.8 ± 7
M-HAQ score (range 0–3)    
 Median (IQR)1.1 (0.6–1.6)1.0 (0.5–1.5)1.06 (0.74–1.52)0.76
 Mean ± SD1.1 ± 0.71.1 ± 0.7
ESR, mean ± SD mm/hour38 ± 2544 ± 280.99 (0.98–1.00)0.14
CRP, mg/liter    
 Median (IQR)17 (8–31)20 (9–45)0.99 (0.98–1.00)0.039
 Mean ± SD24 ± 2435 ± 40
IgM RF positive, no. (%)14 (21)247 (64)0.17 (0.10–0.31)<0.001
Anti–CCP-2 positive, no. (%)8 (13)236 (66)0.09 (0.04–0.18)<0.001
SHS (range 0–448)    
 Median (IQR)3 (1–8)5 (2–11)0.95 (0.90–0.99)0.026
 Mean ± SD5 ± 69 ± 12
HLA SE positive, no. (%)33 (50)247 (69)0.46 (0.29–0.75)0.002
Figure 2.

Predictors of sustained DMARD-free remission identified in both cohorts. Values are hazard ratios with 95% confidence intervals. RF = rheumatoid factor; SE = shared epitope (see Figure 1 for other definitions).

In addition, univariate analysis in the ERAS cohort revealed several other predictive variables (SJC in 44 joints, RAI, M-HAQ score, and the 4-variable DAS) (Table 3). A subanalysis in the ERAS cohort of the effect of current or past smoking compared with never smoking as the reference group revealed that current smoking was associated with the lowest chance of achieving remission, with a hazard ratio (HR) of 0.27 (95% confidence interval [95% CI] 0.08–0.87), followed by past smoking (HR 0.78 [95% CI 0.39–1.58]). In both the Leiden EAC and the ERAS cohorts, there was not a significant gene-dose effect of the HLA SE alleles.

Table 3. ERAS univariate analysis of baseline characteristics of the patients who did and those who did not achieve sustained DMARD-free remission*
 Remission (n = 84)No remission (n = 811)HR (95% CI)P
  • *

    For variables that were normally distributed, the mean ± SD is reported. For variables that were not normally distributed, the median and IQR are reported along with the mean ± SD, since Cox regression uses parametric analysis. For dichotomous variables, the number (%) of patients is listed relative to the total number of patients for whom information was available about the characteristic under investigation. DMARD = disease-modifying antirheumatic drug; RA = rheumatoid arthritis; BMI = body mass index; RAI = Ritchie Articular Index; M-HAQ = modified Health Assessment Questionnaire; ESR = erythrocyte sedimentation rate; DAS = Disease Activity Score (see Table 1 for other definitions).

  • The hazard ratio (HR), which is the effect measure generated by Cox regression analysis, can be interpreted similar to an odds ratio, meaning that a higher HR signifies a greater chance of remission. HRs are reported along with their 95% confidence intervals (95% CIs).

  • P ≤ 0.05.

  • §

    Significant after correction for multiple testing by the Holm method.

  • C-reactive protein and anti–cyclic citrullinated peptide 2 were not measured in the ERAS.

Age, mean ± SD years51 ± 1652 ± 131.00 (0.98–1.01)0.58
Women, no. (%)53 (63)560 (69)0.78 (0.50–1.22)0.28
Symptom duration at baseline, months    
 Median (IQR)6.0 (3–9)7.0 (4–12)0.96 (0.92–1.00)0.038
 Mean ± SD7.0 ± 58.5 ± 6
Smoking, no. (%)13 (28)268 (42)0.54 (0.29–1.02)0.059
Family history of RA, no. (%)20 (24)211 (26)0.87 (0.53–1.44)0.59
BMI (kg/m2), mean ± SD25 ± 426 ± 50.98 (0.93–1.04)0.54
Acute start of symptoms, no. (%)51 (61)383 (48)1.71 (1.10–2.67)0.017§
Start of symptoms in small joints, no. (%)25 (30)204 (25)1.27 (0.80–2.04)0.31
Symmetric start of symptoms, no. (%)63 (81)615 (78)1.18 (0.67–2.07)0.56
SJC in 44 joints    
 Median (IQR)9.5 (4–21)15 (7–26)0.97 (0.95–0.99)0.005§
 Mean ± SD13 ± 1217 ± 13
RAI (range 0–81)    
 Median (IQR)4.0 (2–12)10 (5–17)0.91 (0.88–0.95)<0.001§
 Mean ± SD6.5 ± 613 ± 11
M-HAQ score (range 0–3)    
 Median (IQR)0.6 (0.1–1.1)1.0 (0.5–1.6)0.51 (0.36–0.71)<0.001§
 Mean ± SD0.8 ± 0.71.1 ± 0.7
ESR, mean ± SD mm/hour36 ± 3040 ± 270.99 (0.99–1.0)0.21
IgM-RF positive, no. (%)29 (36)520 (66)0.31 (0.20–0.50)<0.001§
Four-variable DAS, mean ± SD3.3 ± 14.3 ± 20.65 (0.55–0.76)<0.001§
Larsen score (range 0–250)    
 Median (IQR)0.0 (0–2)0.0 (0–4)0.94 (0.88–1.00)0.050
 Mean ± SD1.7 ± 33.9 ± 9
HLA SE positive, no. (%)34 (56)440 (74)0.47 (0.28–0.78)0.003§

To correct for multiple testing, the Holm method was applied. This revealed that autoantibody status and the absence of the HLA SE alleles were significant predictors of remission in both cohorts. Furthermore, acute onset, the RAI, the SJC in 44 joints, the 4-variable DAS, and the M-HAQ score at baseline were significant predictors of remission in the ERAS cohort.

Multivariate analysis.

To investigate which of the patient characteristics were independent predictors of remission, multivariate Cox regression analysis was performed using a 2-step modeling approach. Table 4 shows the results of the multivariate analysis for both cohorts. Three variables were found to be independent predictors of sustained DMARD-free remission in the Leiden EAC cohort, of which 2 (baseline CRP and anti–CCP-2 status) were significant (P < 0.05). In the ERAS, these 2 characteristics were not available, but autoantibody status in the form of IgM-RF positivity was the most significant predictive variable. Symptom duration showed a trend toward significance in the Leiden EAC cohort (P = 0.07) and was significantly associated with remission in the ERAS cohort (P = 0.029).

Table 4. Multivariate analysis of independent predictors for achieving sustained DMARD-free remission*
Cohort, patient characteristicHR (95% CI)P
  • *

    DMARD = disease-modifying antirheumatic drug; CRP = C-reactive protein; anti–CCP-2 = anti–cyclic citrullinated peptide 2; RAI = Ritchie Articular Index; M-HAQ = modified Health Assessment Questionnaire (see Table 1 for other definitions).

  • The hazard ratio (HR), which is the effect measure generated by Cox regression analysis, can be interpreted similar to an odds ratio, meaning that a higher HR signifies a greater chance of remission. HRs are reported along with their 95% confidence intervals (95% CIs).

Leiden EAC  
 Symptom duration in months0.95 (0.90–1.00)0.07
 CRP in mg/liter0.99 (0.98–1.0)0.040
 Anti-CCP2 positive0.09 (0.04–0.20)<0.001
ERAS  
 Symptom duration in months0.94 (0.89–0.99)0.029
 Acute start of symptoms2.03 (1.15–3.59)0.015
 RAI0.92 (0.88–0.97)0.001
 M-HAQ score0.66 (0.44–0.99)0.044
 IgM-RF positive0.28 (0.16–0.49)<0.001
 HLA SE positive0.44 (0.26–0.73)0.002

The fact that CRP and anti–CCP-2 status were not available in the ERAS cohort made it difficult to assess the extent to which the results of the ERAS replicated the findings in the Leiden EAC cohort. The multivariate analysis of the Leiden EAC cohort was therefore also performed after exclusion of these variables. This analysis revealed the following independent predictive variables: symptom duration (HR 0.96 [95% CI 0.91–1.01]), IgM-RF positivity (HR 0.19 [95% CI 0.11–0.35]), and presence of HLA SE alleles (HR 0.59 [95% CI 0.36–0.96]). These 3 factors were also identified as independent predictors in the ERAS cohort, which therefore provides a complete replication of the data from the Leiden EAC cohort.

Radiographic progression.

Although the present study aimed to find baseline characteristics predictive of achieving sustained DMARD-free remission, the rate of joint destruction during followup was also compared between the remission and nonremission groups in the Leiden EAC cohort. Patients in the remission group had a lower level of joint damage after 1, 3, and 5 years of followup compared with patients in the nonremission group (median SHS [IQR] 4 [1–9], 6 [1–14], and 6 [3–15], respectively, versus 11 [5–24], 20 [9–39], and 26 [12–50], respectively).

Remission defined according to the ACR criteria.

To determine whether the results would have been different if remission had been defined according to the modified ACR criteria (19,20), this definition was also applied to the Leiden EAC cohort. Remission according to the ACR criteria was achieved by 97 of the 454 patients (21.4%), including all 68 patients who achieved sustained, DMARD-free remission. The remaining 29 patients were still taking DMARDs at the time they fulfilled the ACR remission criteria. Investigation of which baseline factors were predictive of achieving remission as defined by the ACR criteria revealed almost completely the same predictive factors (data not shown).

DISCUSSION

The present study investigated the prevalence of and predictive factors for sustained DMARD-free remission in RA patients treated with conventional therapy. The long-term followup data from the Leiden EAC and the British ERAS cohorts revealed that sustained DMARD-free remission occurs in 9–15% of RA patients. Furthermore, sustained DMARD-free remission can be predicted by several clinical variables at first presentation that are routinely assessed in the outpatient clinic. Most predictive characteristics that were identified in the Leiden EAC cohort could be replicated in the ERAS cohort. In this study, we used a stringent definition of remission: the sustained absence of synovitis for at least 1 year after the discontinuation of therapy with DMARDs. This definition approximates a definitive cure of the disease. The present findings therefore provide important new insights with regard to disease resolution and the factors underlying this process.

Notably, the frequencies of sustained DMARD-free remission were roughly comparable in the ERAS and Leiden EAC cohorts (9.4% and 15.0%, respectively). The finding that the prevalence was somewhat lower in the ERAS cohort may be due to the fact that patients in the ERAS had a longer symptom duration at presentation and a higher SJC in 44 joints and were more often IgM-RF positive and HLA SE positive than patients included in the Leiden EAC cohort. These factors were found to be independently associated with a smaller chance of achieving remission, and they are known to be associated with a severe disease course. This indicates that patients included in the ERAS cohort were less likely to achieve sustained DMARD-free remission than patients in the Leiden EAC cohort.

Investigators in previous studies of remission in patients treated with conventional therapy have also reported remission rates of 10–20% (1, 21–24). However, remission was not defined as a sustained disease outcome in these previous investigations, but rather as an episodic phenomenon. In the present study, the long-term followup data from the Leiden EAC and ERAS cohorts enabled us to investigate remission as a definite disease outcome that most closely resembles a cure of the disease.

We cannot entirely exclude the possibility that treatment differences between the Leiden EAC and the ERAS cohorts may have affected the prevalence of remission. However, in the Leiden EAC, stratification for inclusion period (reflecting different treatment strategies) showed comparable remission rates for the 3 inclusion periods. This indicates that changes in treatment strategy during the course of the study did not have a major effect on the prevalence of sustained DMARD-free remission.

Furthermore, in order to exclude the possibility that patients who achieved remission had been treated more stringently, a detailed investigation of medication use was performed in both cohorts. This revealed that the patients who achieved remission had not been treated more stringently, but rather, they tended to have been treated less aggressively (e.g., with NSAIDs only [in ∼25% of patients] or with less potent DMARDs [data not shown]). In both cohorts, patients who had been treated with DMARDs had a significantly lower chance of ever achieving remission (in the Leiden EAC cohort, HR 0.12 [95% CI 0.07–0.21]; in the ERAS cohort, HR 0.08 [95% CI 0.05–0.12]). This suggests that the rheumatologists deemed patients in the remission group to have less severe disease that did not need to be treated with DMARDs. This is consistent with the other results of the univariate analysis, which revealed that patients in the remission group had less radiographic damage (in the Leiden EAC and in the ERAS cohorts) and lower SJCs, RAIs, and M-HAQ scores (in the ERAS cohort) at baseline. The effect of ever using DMARDs was not included in the multivariate models because the aim of the analysis was to identify baseline characteristics of the patients that were predictive of remission.

One may consider the observed prevalence of remission (9–15%) to be surprisingly high, leading to the question of whether these patients “really” had RA. However, all patients fulfilled the ACR 1987 criteria for RA, a set of criteria that although not perfect, is nonetheless the most frequently applied. In addition, remission was achieved after a median of 3.5 years, indicating that patients had a chronic polyarthritis for this lengthy period. Finally, the majority of patients in the remission group had radiographic joint destruction (although this was less severe than in the nonremission group); this is considered a hallmark of RA.

The current study identified several factors predictive of sustained DMARD-free remission. Autoantibody status, HLA SE alleles, smoking, an acute onset of symptoms, and the symptom duration before inclusion were found to be predictive in both cohorts. The findings on the association between the absence of autoantibodies and remission are consistent with previous studies documenting a strong association between autoantibody status (IgM-RF and anti–CCP-2) and the rate of joint destruction in RA (25–27). The current observations illustrate that these autoantibodies are also important persistency factors.

The majority of predictive factors identified in the Leiden EAC cohort could be replicated in the ERAS cohort, particularly when the same baseline characteristics were included in the multivariate models. The ERAS then provided a perfect replication of the results of the Leiden EAC study. There were also some differences between the 2 cohorts with regard to the predictive factors that were identified. In the ERAS cohort, disease activity at baseline, as reflected by the RAI, SJC in 44 joints, 4-variable DAS, and M-HAQ score, was inversely associated with achieving remission, while in the Leiden EAC cohort, the SJC in 44 joints and the M-HAQ score were not associated with remission. Part of this difference may be attributable to the higher disease activity on average in the ERAS cohort at baseline (as apparent from the higher median SJC in 44 joints and the accompanying larger range of values), which lends more power to the ERAS to find an association between disease activity at baseline and definite disease outcome.

Interestingly, several factors that were recently described to predict remission after treatment with anti–tumor necrosis factor therapy, such as age, sex, and comorbidity, could not be replicated in the present study (28, 29). This could very well be the result of the different definitions of remission that were used. In those studies, remission was defined as a DAS <1.6 or a DAS in 28 joints <2.6 during treatment, and it is therefore conceivable that the factors that were identified predict treatment response rather than long-term disease resolution (30). The difference in predictive factors emphasizes that low disease activity during treatment is different from remission that persists after discontinuation of antirheumatic therapy.

In conclusion, 9–15% of RA patients who received conventional therapy achieved sustained DMARD-free remission. Several factors that can easily be determined in clinical practice, such as symptom duration before presentation and the absence of IgM-RF and anti–CCP-2 antibodies, are consistently associated with sustained DMARD-free remission. In light of our findings that RA can resolve in ∼10% of patients, additional studies are warranted to further elucidate the underlying mechanisms.

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 published. Dr. van der Woude 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. Van der Woude, Young, van der Heijde, Huizinga, van der Helm-van Mil.

Acquisition of data. Van der Woude, Young, Jayakumar, Toes, van der Heijde, van der Helm-van Mil.

Analysis and interpretation of data. Van der Woude, Young, Mertens, van der Heijde, Huizinga.

Acknowledgements

We gratefully acknowledge Michael van der Linden, MD, of the Department of Rheumatology at Leiden University Medical Center, for his help with the radiographic data of the Leiden EAC. We also gratefully acknowledge Ronald Brand, PhD, of the Department of Medical Statistics at Leiden University Medical Center, for his help with the statistical analysis.

Ancillary