FormalPara Key Summary Points

Why carry out this study?

Reaching low-disease activity (LDA) is an alternative therapeutic goal, particularly in rheumatoid arthritis patients (RA) with long-standing disease.

Boolean-defined LDA are not unequivocally established, which indeed restricted the application of the definition in daily practice.

What was learned from this study?

Boolean-defined LDA with cutoff of 3 shows high consistency with SDAI- or CDAI-LDA.

Higher PGA leads to most inconsistences between Boolean-LDA with cutoff of 3 and SDAI-LDA.

Using cutoff of 3 to Boolean-LDA provides great clinical utility with index-based LDA, especially when exclusively increasing PGA cutoffs to 4.0, 4.5 or replacing PGA by EGA with cutoffs to 3.0, which should be validated in other settings and may deserve considering in clinical practice.

Digital Features

This article is published with digital features to facilitate understanding of the article. You can access the digital features on the article’s associated Figshare page. To view digital features for this article go to https://doi.org/10.6084/m9.figshare.13348442

Introduction

Rheumatoid arthritis (RA) is a chronic, systemic autoimmune disease characterized by progressive, symmetric arthritis, leading to irreversible joint destruction [1, 2]. Over the past decades, treat-to-target strategy and novel therapeutic agents have tremendously transformed the clinical trajectory of RA patients [3,4,5]. In the American College of Rheumatology (ACR) and European League Against Rheumatism (EULAR) management recommendations for RA, a state of clinical remission is placed as the primary target for RA treatment [6]. Alternatively, low disease activity (LDA) may also be an acceptable therapeutic goal, particularly in long-standing disease, in the context of complicated clinical scenario [7].

In general, the states of LDA and remission can be defined according to different composite measures with corresponding cutoffs, including the disease activity score based on 28-joint count (DAS28) [8], Simplified Disease Activity Index (SDAI) [9], and Clinical Disease Activity Index (CDAI) [10]. As for each target, each of these different composite measures or cutoffs are intended to measure the same construct. Ideally, they would all strongly agree with each other and result in the same treatment consequences. Although CDAI- and SDAI-defined LDA and remission are reported to strongly agree with each other, other measures had higher discordance [11, 12]. For example, the DAS28 remission allows for a significant number of residual swollen joints, which cannot be overcome by modifying the cutoff [13, 14]. Apart from these criteria, ACR/EULAR Boolean definition provided remission criteria for RA patients almost one decade ago, which includes swollen joint count (SJC), tender joint count (TJC), patient’s global assessment (PGA, 0–10 scale), and C-reactive protein (CRP) levels (mg/dl). A patient must have scores of 1 or less in each component to attain Boolean remission [15]. The LDA is an alternative therapeutic goal, however, Boolean-defined LDA is not unequivocally established, which indeed restricts the application of the definition in daily practice.

To fill the gap, we aimed to propose Boolean definition for LDA, and test its clinical utility with a comparator of either SDAI-LDA or CDAI-LDA, on basis of a large real-world database of RA patients.

Methods

Study Population

All patients were identified from the outpatient clinic of our department in real-world longitudinal database, which has been previously described and approved by the local ethics committee [5, 12, 16, 17]. In brief, a longitudinal database of a large cohort of RA patients had been established over the course of more than 10 years. RA was diagnosed according to the 1987 ACR classification criteria [18] and/or 2010 ACR/EULAR classification criteria for RA [19]. At the first visit in the database, demographics, symptom duration, RA core set variables (TJC28, SJC28, PGA, evaluator’s global assessment (EGA, 0–10 scale)), laboratory testing (CRP, mg/dl)) and treatment details (e.g. methotrexate (MTX); leflunomide (LEF); hydroxychloroquine (HCQ); sulfasalazine (SSZ)) were documented. After the initial visit, each patient contact was prospectively documented at least three-monthly for the patients in moderate/high diseases activity and every 3–12 months (usually 3–6 months) for the patients in remission/LDA. Extra follow-up was additionally scheduled besides those at regular intervals if clinically necessary. The treatment decisions at each visit were made at the discretion of rheumatologists based on disease activity scores for the explicit goal of clinical remission.

A retrospective analysis of prospectively collected data was performed for all participants in the real-world longitudinal database with RA. To confirm discovery, we randomly classified RA patients in the database at a ratio of 3:2 into either analysis cohort or validation cohort. Early or established RA was defined as the duration from onset of first arthritic symptom to the first visiting in our center with definite diagnosis of RA ≤ 1 year or > 1 year, respectively. A patient who never received disease-modifying antirheumatic drugs (DMARDs) treatment or was on DMARD treatment for less than 3 months at first visit was defined as a DMARD-naive patient. This study was approved by the Institutional Review Board (IRB) of the Peking University First Hospital (IRB number 2014–785). The participants provided their written informed consent to participate in this study. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Definitions of LDA

At present, various composite instruments were available to measure disease activity for RA patients, however no universal consensus on the assessment of LDA. In this study, the SDAI and CDAI index-based definitions of LDA were applied as comparators. The SDAI is obtained from a linear sum of SJC28, TJC28, PGA, EGA, and CRP (mg/dl), and SDAI ≤ 11 is defined as LDA [9]. Similarly, the CDAI is the linear sum of SJC28, TJC28, PGA, and EGA, with ≤ 10 as the definition of LDA [10].

The Boolean remission includes SJC28, TJC28, PGA, and CRP. To satisfy the remission criteria, all of these components must be ≤ 1. Since no definition of LDA exists for Boolean, we initially put forward the candidates for Boolean-LDA by increasing the cutoff of the four core set variables in steps of 1 from 2 to 3, 4, and 5. We respectively referred to them as Boolean-LDA2, Boolean-LDA3, Boolean-LDA4, and Boolean-LDA5 thereafter. Furthermore, we investigated the reasons for discordance between the initial Boolean-LDA proposals and SDAI index-based LDA, and further modifications on the impeditive variables were made to increase the agreement.

Statistical Analysis

The data were presented as means (standard deviation (SD)) or median (interquartile range (IQR)) depending on the level of resemblance to the normal distribution. Absolute and relative frequencies were reported for categorical variables. The frequencies of LDA were measured by each definition, reported as percentages with 95% confidence intervals (95% CIs). We assessed the agreement of proposed Boolean-LDA with the SDAI definition of LDA by k-coefficient for agreement at all documented visits and initial visits of included patients, respectively. The kappa values were interpreted according to Landis and Koch (agreement: poor < 0.00; slight 0.00–0.20; fair 0.21–0.40; moderate 0.41–0.60; substantial 0.61–0.80; almost perfect 0.81–1.00) [20]. Besides, previous reports showed the swollen joint was most significantly associated with radiographic progression in RA patients [13, 14, 21, 22]. Therefore, to reflect the impact of different Boolean-LDA on radiographic progression, in part, we compared the distribution of SJC across the different LDA definitions. All the analyses were conducted using SPSS version 22.0.

Results

Characteristics of Patients at Baseline

In total, 1121 patients with 8187 clinic visits in the longitudinal database were included and randomly grouped into analysis cohort or validation cohort, respectively (Table 1). In the analysis cohort, there were 672 participants comprising 4881 visits, with a mean onset age of 49 (SD 15) years, 77.2% females (Table 1). Most of the patients were rheumatoid factor (76.1%), anti-cyclic citrullinated peptides (85.6%) positive. 61.0% of all patients were DMARD-naïve at baseline. Of included patients with a mean disease duration of 24 (6–84) months, there were 314 (46.7%) early RA and 358 (53.3%) established RA patients. The median (IQR) SDAI and CDAI of patients at cohort entry was 16.1 (9.1–27.1) and 14.0 (8.0–24.0). DMARDs were initiated to all patients after enrollment to the cohort, with MTX (80.4%), LEF (45.2%), and HCQ (40.5%) as the top three. Glucocorticoids were applied as part of initial therapeutic strategies in 283 (42.1%) patients. The validation cohort comprising 449 participants with 3306 clinic visits possessed similar demographics, clinical characteristics, and treatment details to the analysis cohort (Table 1).

Table 1 Demographics and clinical characteristics of RA patients

The Concordance Across LDA Definitions

The rate of reaching LDA assessed by SDAI and CDAI in all included visits was 71.9% (3510) and 73.6% (3593), respectively. For initial proposals of Boolean-LDA definitions, the frequency of LDA was as follows: 52.8% (2577) for Boolean-LDA2, 65.2% (3182) for Boolean-LDA3, 73.5% (3588) for Boolean-LDA4, and 80.7% (3937) for Boolean-LDA5 (Fig. 1). When assessing the best cutoff for concordance of index-based LDA and Boolean-defined LDA, we found that high concordance rates with SDAI- or CDAI-based LDA were observed in Boolean-LDA3 (SDAI: kappa = 0.796; CDAI: kappa = 0.771) and Boolean-LDA4 (SDAI: kappa = 0.807; CDAI: kappa = 0.794) (Fig. 2). Similar findings were observed in separate analysis of either early or established RA patients (Fig. 2). Besides, we also compared the concordance across LDA criteria based on 500 initial visits among RA patients in the analysis cohort, yielding similar results (Supplementary Fig. S1).

Fig. 1
figure 1

Rates of low disease activity assessed by SDAI, CDAI, initial proposed Boolean definition, using a cutoff in steps of 1 from 2 to 3, 4, 5 (Boolean-LDA2/3/4/5), and modified Boolean definition by increasing the cutoff of the PGA by 0.5-cm increments from 3.5 cm to 5.0 cm (Boolean-LDA3(PGA3.5/4.0/4.5/5.0)) or replacement of PGA with EGA using cutoff of 3.0 cm (Boolean-LDA3(EGA3.0)). SDAI Simplified Disease Activity Index, CDAI Clinical Disease Activity Index, PGA patient’s global assessment, EGA evaluator’s global assessment

Fig. 2
figure 2

The agreement in included visits between initial proposed Boolean definition of LDA, using a cutoff in steps of 1 from 2 to 3, 4, 5 (Boolean-LDA2/3/4/5) and SDAI- or CDAI-based LDA. a The concordances and discordances. b Kappa value. LDA low disease activity, SDAI Simplified Disease Activity Index, CDAI Clinical Disease Activity Index

Previous studies have indicated residual  SJC was most closely associated with the radiographical progression in RA patients [13, 14, 21, 22]. We therefore compared the distribution of residual SJC between SDAI or CDAI-based LDA and proposed Boolean definition of LDA (Fig. 3). When SDAI (CDAI)-defined LDA was achieved, the proportion of SJC = 1, SJC = 2 and SJC ≥ 3 was 14.16% (14.22%), 6.67% (7.26%), and 2.59% (2.95%), respectively. As expected, the residual SJC of Boolean-LDA increased with the cutoff increment and most similar distribution of SJC was observed between index-based LDA and Boolean-LDA3 (13.55% of SJC = 1; 7.17% of SJC = 2; 2.36% of SJC = 3) (Fig. 3). Taken together, Boolean-LDA3 showed great performance in defining LDA.

Fig. 3
figure 3

The distribution of swollen joint count in the included clinic visits meeting low disease activity based on SDAI, CDAI, and initial proposed Boolean definition, using a cutoff in steps of 1 from 2 to 3, 4, 5 (Boolean-LDA2/3/4/5). LDA low disease activity, SDAI Simplified Disease Activity Index, CDAI Clinical Disease Activity Index

The Reasons for Discordance Between Boolean-LDA3 and SDAI-LDA

Of 4881 included visits, a total of 431 discordances were observed, with meeting SDAI-LDA but not Boolean-LDA3 in 380 visits and meeting Boolean-LDA3 but not SDAI-LDA 51 visits. Subsequently, the reasons for meeting SDAI-LDA but not Boolean-LDA3 were analyzed. Of 380 discordances, the most common reason was higher PGA score (239; 62.9%), followed by higher TJC (87; 23.0%), CRP (39; 10.3%), and SJC (36; 9.5%) (Supplementary Fig. S2). For the higher PGA score, there were 137 times of 3.1–4.0, 86 times of 4.1–5.0, and 16 times of over 5.0 (Supplementary Fig. S2). On the other hand, there was a total of 51 visits meeting Boolean LDA3 criteria but not SDAI-LDA. The mean SDAI for the visits meeting Boolean LDA3 criteria but not SDAI-LDA was 12.2, ranging from 11.1 (SJC28 of 1, TJC28 of 2, PGA of 3, EGA of 3, and CRP of 1.18) to 14.6 (SJC28 of 3, TJC28 of 3, PGA of 3, EGA of 3, and CRP of 2.61).

Further Modification of Boolean-LDA3

Accordingly, the further modifications of PGA for Boolean-LDA3 were made by increasing the cutoff of the PGA criterion stepwise by 0.5 increments from 3.5 to 5.0 (Boolean-LDA3(PGA3.5/4.0/4.5/5.0)) or replacement of PGA with EGA using cutoff of 3.0 (Boolean-LDA3 (EGA3.0)).

For modified proposals of Boolean-LDA3, the frequency of included visits was 65.4% for Boolean-LDA3 (PGA3.5), 69.1% for Boolean-LDA3 (PGA4.0), 69.2% for Boolean-LDA3 (PGA4.5), 72.3% for Boolean-LDA3 (PGA5.0) and 70.7% for Boolean-LDA3 (EGA3.0) (Fig. 1). Highest consistency was reached with Boolean-LDA3 (EGA3.0) (SDAI: kappa = 0.851; CDAI: kappa = 0.825) (Fig. 4). Similar findings were detected either in early RA (SDAI: kappa = 0.833; CDAI: kappa = 0.805) or established (SDAI: kappa = 0.863; CDAI: kappa = 0.837) (Fig. 4). The additional comparisons of the concordance across LDA criteria based on 672 initial visits among RA patients were made in analysis cohort, yielding similar results (Supplementary Fig. S3). Meanwhile, exclusively increasing PGA cutoffs or replacement of PGA with EGA using cutoff of 3.0 on basis of Boolean-LDA3 did not substantially impact the pattern of residual SJC compared with Boolean-LDA3 (Fig. 5). Therefore, Boolean-LDA3 (EGA3.0) exhibited the greatest clinical utility in defining LDA.

Fig. 4
figure 4

The agreement in included visits between SDAI- or CDAI-based LDA and modified Boolean definition by increasing the cutoff of the PGA by 0.5 cm increments from 3.5 cm to 5.0 cm (Boolean-LDA3(PGA3.5/4.0/4.5/5.0)) or replacement of PGA with EGA using cutoff of 3.0 cm (Boolean-LDA3(EGA3.0)). a The concordances and discordances; b kappa value. LDA low disease activity, SDAI Simplified Disease Activity Index, CDAI Clinical Disease Activity Index, PGA patient’s global assessment, EGA evaluator’s global assessment

Fig. 5
figure 5

The distribution of swollen joint count in the included clinic visits meeting low disease activity based on SDAI, CDAI, and modified Boolean definition by increasing the cutoff of the PGA by 0.5-cm increments from 3.5 to 5.0 cm (Boolean-LDA3(PGA3.5/4.0/4.5/5.0)) or replacement of PGA with EGA using cutoff of 3.0 cm (Boolean-LDA3(EGA3.0)). LDA low disease activity, SDAI Simplified Disease Activity Index, CDAI Clinical Disease Activity Index, PGA patient’s global assessment, EGA evaluator’s global assessment

Validation Analysis

Finally, we validated our results in a randomly generated internal cohort, which included 449 RA patients with 3306 visits. Their baseline characteristics are shown in Table 1. In the validation cohort, high concordance rates and most similar pattern of SJC with SDAI or CDAI definition of LDA were achieved with Boolean-LDA3 (SDAI: kappa = 0.802; CDAI: kappa = 0.791 for all patients; SDAI: kappa = 0.785; CDAI: kappa = 0.780 for early RA; SDAI: kappa = 0.810; CDAI: kappa = 0.797 for established RA) and further modification with a substitute EGA for PGA provided best agreement with either SDAI or CDAI definition of LDA (SDAI: kappa = 0.849; CDAI: kappa = 0.836 for all patients; SDAI: kappa = 0.830; CDAI: kappa = 0.819 for early RA; SDAI: kappa = 0.861; CDAI: kappa = 0.846 for established RA) (Supplementary Fig. S4–5).

Discussion

To our knowledge, this is the first study aiming to propose Boolean-defined LDA and test its utility in RA. Based on the large real-world cohort of early and established RA patients in our center, we proposed the initial and modified Boolean-LDA, and further tested their performance in comparison with SDAI- and CDAI-based LDA, which are frequently applied in current daily practice. Moreover, we also compared the SJC patterns in the clinic visits meeting the corresponding LDA definitions.

Generally, Boolean-defined LDA with a cutoff of 3 showed high consistency and similar SJC distribution to the SDAI- or CDAI-defined LDA in our cohort. Further analysis of the reasons for the discordances between Boolean-LDA3 and SDAI-LDA suggested over half of the discordances were attributable to the higher PGA, mostly ranging from 3.1 to 5.0. Accordingly, the targeted modifications of PGA were designed by increasing the cutoff of the PGA criterion stepwise by 0.5 increments from 3.5 to 5.0, and replacing PGA by EGA using cutoff of 3.0. The results showed great concordance rates with similar pattern of residual SJC were achieved when exclusively increasing PGA cutoffs to 4.0, 4.5 or replacing PGA by EGA with cutoffs to 3.0. In fact, although Boolean definition of remission seems to be most strict, it has also been criticized. Patients with no inflamed joints and a normal CRP often however with PGA exceeding the cutoff of 1 are rated as non-remission [23]. Previous clinical research found patients in SDAI or CDAI remission however not Boolean remission had higher PGA compared with patients in Boolean remission [24]. In a recent study pooling six different large clinical trials of 1680 early and established RA, Studenic et al. found increasing PGA cutoff to 1.5 or 2.0 would provide high consistency between Boolean remission and SDAI-based remission [25]. Similarly, in the present study, we found the most common reason for meeting SDAI-LDA but not Boolean-LDA3 was higher PGA, mostly ranging from 3.1 to 5.0. Further modifications indicated great agreement with a similar pattern of SJC was reached when increasing PGA cutoffs to 4.0, 4.5 or replacing PGA by EGA with cutoffs to 3.0.

LDA is an acceptable therapeutic goal, particularly in long-standing disease. The corresponding analyses were separately performed in both early and established RA population and yielded similar findings, which supports the potential clinical utility of the Boolean-defined LDA presented in the present study. In addition, subsequent validation in a randomly generated internal cohort of 449 RA patients with similar findings supported the validity of the Boolean-LDA with a substitute EGA for PGA using cutoff of 3. In summary, the agreement was almost great and the distribution of residual SJC was similar among CDAI-LDA, SDAI-LDA, and Boolean-LDA with cutoff of 3, especially increasing PGA cutoffs to 4.0, 4.5 or replacing PGA by EGA with cutoffs to 3.0. Our study, for the first time, proposed the definition of Boolean LDA, and could help promote the adoption of Boolean definition in daily practice.

There are several potential weaknesses of this study that should be acknowledged. No radiographic or functional assessment was analyzed in this study, which may weaken the strength of the proposed Boolean definition of LDA. In consideration of the association between residual SJC and radiographic progression according to previous studies [13, 14, 21, 22], we further compared the distribution of SJC in the clinic visits meeting LDA based on the corresponding LDA definitions, and yielded similar pattern of residual SJC. This finding may strengthen the clinical utilization of the proposed Boolean-defined LDA. Of course, the impact of proposed Boolean-LDA on radiographic and physical outcomes definitely needs to be further validated in future studies. On the other hand, although subgroup analyses according to diseases course and internal validation analysis were performed in the present study, the findings were solely on basis of single-center longitudinal cohort of unselected RA population, which may limit the external generalization of the present results. Therefore, extended studies are warranted to prove these findings in other ethnicities or regions in the future. In addition, this paper represents a proof-of-concept study. The weight of the core set variables of diseases activity requires more discussion. Therefore, future studies are expected to optimize Boolean LDA definition on the basis of radiographic and functional outcomes.

Conclusions

In conclusion, our findings suggest that Boolean-LDA with cutoff of 3 leads to great agreement and most similar residual SJC pattern with the SDAI- and CDAI-based definition of LDA, especially when exclusively increasing PGA cutoffs to 4.0, 4.5 or replacing PGA by EGA with cutoffs to 3.0. This needs to be replicated in other settings and can be used in the management of RA patients in daily practice.