Introduction

Rheumatoid arthritis (RA) is a chronic inflammatory rheumatic disease leading to joint destruction, extra-articular manifestations, and systematic comorbidities. It is associated with progressive disability, increased mortality, and socioeconomic costs [1]. According to the treat-to-target (T2T) strategy, which was recommended by international task force American College of Rheumatology (ACR) and European League Against Rheumatism (EULAR), the primary target for RA patients is clinical remission or low disease activity (LDA). Remission is defined as the absence of signs and symptoms of significant inflammatory disease in patients with RA [2], which can be measured by criteria like DAS28, CDAI, SDAI, and 2011 ACR/EULAR criteria.

Real-world evidence (RWE) generally refers to information coming from electronic health records, billing data, registries, but not traditional tightly controlled trials [3]. Selected patients with higher adherence, more severe RA activity, and shorter study durations make it difficult for trials to generalize findings in larger and “random” population of patients in clinical practice. There have been many well-designed registries worldwide studying remission status of patients with RA for more than 30 years [4,5,6,7,8,9,10,11,12,13,14]. But we have insufficient knowledge of how frequently patients achieve remission and whether this “target” is practical or achievable in real-world circumstances as no meta-analysis of remission rate in real-world studies have been carried out. Herein, we conducted a meta-analysis and systematic review of published cohort studies to investigate remission rate and substantial predictors of remission in RA patients in real-world studies and to enhance the T2T strategy into daily practice.

Materials and methods

This meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement protocol [15].

Inclusion and exclusion criteria

Studies that met the following criteria were included:

  1. 1)

    Conducted in adult patients (≥ 18 years) with a diagnosis of RA according to the American College of Rheumatology (ACR) RA classification criteria;

  2. 2)

    Reported remission rates according to different sets of indexes;

  3. 3)

    Prospective cohort studies or consecutive cross-sectional studies. Consecutive cross-sectional studies were defined as an observational study with consecutive patients and reported remission rate of the whole cohort each year. Cross-sectional studies were included in systematic review but not in meta-analysis;

  4. 4)

    Carried out longitudinally and followed up patients for at least 3 months.

Studies that met any of the following criteria were excluded: (1) carried out in patients with any combined rheumatic diseases (like systematic erythematous lupus, Sjogren’s syndrome, dermatomyositis); (2) point remission rate studies where patients had only one visit; (3) Randomized controlled trials, review articles; (4) conference abstracts, as the limited data available in abstracts may hinder ability to assess the quality of study and data.

Literature search and study selection

We searched MEDLINE, EMBASE, and Scopus to identify all potential articles reporting remission rates in patients with RA. Details of search terms and strategies can be found in Supplementary Materials. The eligibility of the reference list retrieved by literature search was assessed independently by two reviewers and disagreements were resolved at each step by consensus. Selected articles were further examined to determine if they contained relevant information.

Data extraction

Data were extracted from the selected studies independently by two authors, using a predefined standardized form according to the Cochrane Handbook. Discrepancies were discussed by the two authors and arbitrated by a third independent reviewer. For each study, the following information were identified by two of the authors: study type, time span, mean follow-up time, baseline data of patients including age, gender, country, clinical characteristics (including disease duration, tender joint counts, swelling joint counts, rheumatoid factor positivity, mean DAS at baseline, remission rate), treatment regime, and predictors of remission. When the available information from the selected studies was incomplete, attempts were made to contact the corresponding authors for additional information. Data given in subgroups from one study were recalculated into one group before meta-analysis to balance their weights in meta-analysis. Sustained remission was defined as being in remission for at least twice in the observation period.

Quality assessment

Quality assessment (Table S1) of each study was performed independently by the two authors, using the Newcastle-Ottawa Scale (http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp), which graded studies according to the quality of selection, comparability, and outcome of study participants. Discrepancies were addressed by re-evaluation of the original article and discussion with a third independent reviewer.

Statistical analysis

We assessed the pooled remission rates of selected cohort studies at 3, 6, 12, and 24 months. Remission rates using different criteria were also listed. We employed a random-effects model to account for heterogeneity between studies in our sample when estimating the common effect size. Statistical evidence of heterogeneity across studies was examined using the Cochrane Q test and the Ι2 statistic. To identify potential sources of heterogeneity in remission rates reported, pre-specified subgroup analyses were conducted according to disease activity criteria, follow-up duration, geographical region, disease duration, sample size, use of biologic DMARDs (bDMARDs), and type of bDMARDs used. Sensitivity analyses were performed by serially excluding each study to evaluate its influence on the overall results. Publication bias was assessed using funnel plots and the Egger’s test. All analyses were performed using Stata Version 14.0 (StataCorp, Texas, USA), and a p value of less than 0.05 was considered statistically significant. In the summary of potential predictive factors of remission, only results from multivariate analysis (multiple linear logistic regression or Cox proportional hazards model) were included.

Availability of data and material

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Results

Literature search

The search of MEDLINE, EMBASE, and Scopus databases up to November 17, 2017 resulted in a total of 8323 articles, of which 6782 remained after duplicates were removed. A total of 4596 articles that did not meet the inclusion criteria were excluded after title and abstract screening and the 441 remaining articles were reviewed for full texts. Finally, 31 studies including ~ 82,450 patients were selected. Twelve studies were included in subsequent meta-analysis, and all 31 studies were included in qualitative analysis. A flowchart of article screening for the systematic review has been illustrated in Fig. 1.

Fig. 1
figure 1

Flowchart of study selection

Full size image

Study characteristics

For included studies, 19 in Europe, 7 in North America, 1 in Asia, 2 in South America, 1 in Africa, and 1 registry in Oceania were included (Table 1). The METEOR registry was carried out in 32 countries, including countries in Europe and Asia (Japan and India). The 31 included studies were mostly carried out in the first 15 years of the twenty-first century and were published between 2006 and 2017.

Table 1 Characteristics of included studies
Full size table

Main characteristics of the 31 selected studies were summarized in Table 1. Mean age was between 50 to 60 in 29/31 studies. Gender distribution was similar across registries (ranged from 64 to 88.8%), except that 90% patients in the VARA registry were male. Differences in patients’ characteristics were noticed, mainly in disease duration (mean duration ranged from 5.9 months to 16.6 years), disease activity at baseline (mean DAS ranged from 3.15 to 6.7), and treatment strategies. General treatment strategies were similar among studies, though variance existed. Methotrexate were used in 28 cohorts, percent of patients on MTX ranged from 40.9 to 97.7%. Glucocorticoids (GCs) were used in 15 studies, and proportions of patients on GCs ranged from 31.0 to 84.1%. Though most studies contained patients in cs-/bDMARDs monotherapy or combination therapy, there were both studies where every patient was on biologics and studies where no patient was receiving TNFis or non-TNFi biologics.

Remission rate

Ranges of remission rates according to their remission criteria, grouped by corresponding follow-up duration, were summarized in Table S3, where a trend of remission rate increasing with follow-up duration was roughly noticed.

DAS28

Twenty-one registries reported remission rate according to DAS28 criteria, making it the most frequently used criteria (Table S3). Pooled remission rates (DAS28 < 2.6) were 17.2%, 16.3%, 21.5%, and 23.5% for 3-, 6-, 12-, and 24-month follow-ups, respectively (Fig. 2). Significant heterogeneity among studies was noted (I2 ranged from 83.5 to 96.8%). Further subgroup analyses were conducted for the 6- and 12-month remission rates. Regarding treatment strategies (Table S4), all patients from 6-month group were treated with bDMARDs. Subgroup analyses showed that patients who received TNFi (I2 97.1%, p 0.000) had higher remission rate than patients who received non-TNFi biologics (I2 0.0%, p 0.687). For the 12-month group, 26.1% of TNFi patients (I2 91.6%, p 0.000) reached remission while 13.8% of non-TNFi patients reached remission (I2 67.2%, p 0.081). Subgroup analyses by geographical region, disease duration, sample size, and biologics use showed no significant change in heterogeneity (results not shown). Rate of remission or LDA (DAS28 < 3.2) ranged from 35.0 to 61.7% for 6 months, 31.6 to 55.0% for 12 months, and stabled at around 64% for 24, 36, and 60 months, respectively.

Fig. 2
figure 2

Pooled remission rate of patients with RA. Grouped by follow-up duration: a) 3 month, b) 6 month, c) 12 month, d) 24 month. Remission: number of patients meet DAS remission; Total: number of patients in the study

Full size image

Sensitivity analysis and assessment of publication bias

Sensitivity analysis (Table S2) suggested that the 3-, 6-, 12-, and 24-month remission rate was stable and omitting a single study did not change the significance of the pooled survival rates, while removal of the study by Couderc et al. led to the largest increase in the 24-month remission rate. Visual inspection of the funnel plots revealed mild to moderate asymmetry (Fig. S1). According to the Egger’s test, there was no evidence of publication bias in studies reporting on remission rates of 3-, 6-, 12-, and 24-month.

Predictors of remission

We summarized significant predictors of remission reported by the included studies (Table 2). Though predictors varied across studies, some of them remained consistent.

  1. 1)

    Gender, five studies showed that male was an independent predictor of remission.

  2. 2)

    Education level, higher level of education is considered as an independent predictor of remission in two studies.

  3. 3)

    Baseline clinical characteristics, eight studies addressed clinical characteristics at baseline as independent predictors. Lower TJC28, lower disease activity, and lower HAQ at baseline were positively associated with remission, while higher baseline PhGA, PtGA, HAQ, SJC28, ESR predicted lower remission rate.

  4. 4)

    Glucocorticoids use at baseline, three studies found that GC use at baseline predicted a reduced likelihood of treatment response.

Table 2 Predictors of RA remission in included studies
Full size table

Sustained remission

Eight studies reported sustained remission rates (Table 3). Although no agreement has been reached, sustained remission is commonly defined in two ways: (1) remission at two consecutive visits (with certain period apart), or (2) remission lasted a certain of time after the first one. Generally, sustained remission for 1 year occurred in 7.7–37.0%, 19.6–31.3%, 16.7–27.1%, and 23.8% patients according to DAS28, SDAI, CDAI, and ACR/EULAR remission criteria.

Table 3 Characteristics of studies of sustained remission and corresponding predictors
Full size table

Discussion

It is important to know how real-world RA patients are managed as reaching the target of clinical remission or LDA is associated with better long-term prognosis [39]. Many studies have reported remission rates and relevant predictors in individual studies, but to our knowledge, this is the first systematic review and meta-analysis to investigate the pooled remission rates in patients with RA from real-world longitudinal studies. Scott et al. [40] reviewed remission of patients with early RA from 17 observational studies and 20 RCTs, and found a successive decrease in remission rate through patients from RCTs receiving DMARD combination therapies (42% for DAS criteria), patients from observational studies (33%), and patients from RCTs receiving DMARD monotherapy (26%). Though carried out in early RA patients, the observational studies group showed similar remission rate with our study. Patients in RCTs may have higher remission rate than observational studies due to reasons like better compliance, regularly arranged follow-ups, better access to certain high-cost drugs, and less heterogeneity. But results from RCTs are strictly restricted within certain patients that cannot be generalized. In observational studies, a trend of increasing remission rate with follow-up duration was seen in both longitudinal (Table S3) and consecutive cross-sectional cohorts (Fig. 3). Remission is a reachable target in real-life patients and real-world studies are providing tremendous evidence.

Fig. 3
figure 3

Remission rate in all included consecutive cross-sectional studies

Full size image

We have tried to find potential sources of heterogeneity by sensitivity analysis and subgroup analysis according to follow-up duration, remission criteria, geographical region, and treatment strategies. In both 6- and 12-month group, patients treated with TNFi had higher remission rate than non-TNFi group. Both studies [22, 24] included patients who had previous TNFi failures for non-TNFi treatment. Though only one tenth of patients reached remission, it proved the efficacy of non-TNFi biologics on patients inadequately responded to TNFi. This is consistent with another meta-analysis which showed that switching to non-TNFi was more effective than cycling among TNFis [41].

Our study approved traditional predictors of remission like male, higher educational level, lower TJC28, lower baseline disease activity, and lower HAQ. But we also found controversies. Three studies reported that younger age was a predictor of remission [16, 29, 31] while Lahaye et al. [24] found that younger age does not predict remission but less adverse effects and complications. Smoking has also been a traditional negative predictor of remission [42], but we found that there may be no difference between smokers and non-smokers. Canhao et al. [30] reported that smoking predicted reduced remission rate while Kuriya et al. [5] found no significant difference among current or ex-smokers and never smokers. Patients with negative predictors, like older age and lower educational level, should receive strengthened therapies as they are less likely to achieve remission. Stratified therapies for individual patient should be noted on the basis of international guidelines. Three studies showed that initial use of GCs predicts worse prognosis [5, 29, 30]. All three studies explained that initial use of GCs may introduce channeling bias as doctors tended to prescribe GCs to patients less likely to achieve remission with higher baseline disease activity [43].

Sustained remission was rare in included studies regardless of different definitions. In the VARA registry, probability of remaining in remission for 3, 12 and 24 months after the first remission were 86.3%, 24.7%, and 8.1% (DAS28), respectively. They also reported that probability of a remission lasting for 2 years was 6.0–14.1%. We have worked so hard to treat patients to remission, but transient remission is not our target. More attention should be paid on maintaining remission and enhancing sustained remission rate. Predictors for sustained remission (SR) were also summarized. Some SR predictors were in accordance with that of remission: male gender, lower baseline disease activity [31], lower PhGA [23], and initial use of combination DMARDs [5]. Darawankul et al. [38] reported that in multivariate regression analysis, only a good EULAR response at the first year was significantly associated with SR (OR 3.1, 95% CI 1.15–8.36, p = 0.03). Negatively associated predictors were higher TJC [23], initial use of oral steroids, baseline pain, and longer time to remission [5]. Kuriya et al. [5] found that being obese (OR 0.30 95% CI [0.10–0.90], p = 0.002) predicted worse outcomes in patients with RA, while association of BMI and remission was not identified in other included studies. Association of other variates, including early diagnosis, RF negativity, no erosion on X-ray at baseline, and ever prednisolone use with SR, were not significant.

Some registries illustrated how disease activity changed over time by reporting consecutive cross-sectional remission rates, though not included in quantitative analysis. For each registry, cross-sectional remission rate (DAS28) increased with years (Fig. 3). Yamanaka et al. [7] reported that from 2000 to 2006, remission rate defined by DAS28 increased from 8.5 to 21.5%, and patients in LDA increased from 22.2 to 38.8%. Not only mean DAS28, but also all four components TJC, SJC, ESR, and PtGA had also improved from 2000 to 2006. In the report of Hyrich et al. [4], both the proportions of 1-year EULAR good responders and remission rate increased from 2001 to 2008 (8% vs. 17%, p < 0.001). Aga et al. [14] reported a twofold increase of remission rate in the 6-month follow-up group from 2000 to 2010, and RA disease activity level at MTX and TNFi+MTX initiation decreased from high to moderate at the mean time. Littlejohn et al. [11] reported that the remission rate increased from 36.7% in 2009 to 53.5% in 2014 (DAS28 criteria). They also addressed that most disease activity change happened in the first 3 years of diagnosis. Greenberg et al. [36] pointed out that according to their two-phase (2005–2007 and 2010–2012) study, improvements in disease activity and remission rate were seen among all racial and ethnic groups.

This study has several limitations. First, as studies were designed, carried out, and reported differently, heterogeneity among studies was significant. Verstappen et al. [9] compared demographic and clinical characteristics in five registries (CORRONA, SRR, NOAR, IORRA, and CORRONA International) and found it difficult to compare data across registries. They have found a solution of collecting original data and redivide patients into subgroups by treatment response. Some registries that recruit international patients (like the METEOR registry in 32 countries) found that geographical variance was significant in patients with RA [13]. Radner et al. [44] have recently reported 2017 EULAR recommendations for a core data set in RA observational research, which has listed the least content variables so that data setup will be more standardized and comparable in RA registries. Second, generalizability is limited as some large, bDMARDs-based registries may introduce channeling bias as more patients receiving biologics were included [6, 29, 37, 45]. Third, as we only included patients with at least 3 months follow-up, some well-designed registries with only one cross-sectional remission rate reported have been ruled out and a lot of patients may be lost in this way.

Conclusions

Rate of remission increased with longer follow-up duration. Though remission rate has been elevated in recent years, sustained remission is still rare. Remission was more frequently achieved in patients who were male, with younger age, higher education level, and lower baseline disease activity, while initial use of corticosteroids reduced the likelihood of remission. Remission is a reachable target for patients with RA, and further efforts on maintaining remission and achieving sustained remission should be paid.