Abstract
Objective. Predictors of remission in rheumatoid arthritis (RA) have been defined in cross-sectional analyses using the 28-joint Disease Activity Score (DAS28), but not with newer composite disease activity measures or using the more clinically relevant state of sustained remission. We have evaluated predictors of remission using cross-sectional and longitudinal durations of disease state, and by applying additional definitions of remission [American College of Rheumatology/European League Against Rheumatism Boolean, Simplified Disease Activity Index (SDAI), and Clinical Disease Activity Index (CDAI)].
Methods. Individuals in the Alberta Biologics Pharmacosurveillance Program were classified for the presence of remission (point and/or sustained > 1 yr) by each of the 4 definitions. Multivariate models were constructed including all available variables in the dataset and refined to optimize model fit and predictive ability to calculate OR for remission.
Results. Nonsmoking status independently predicted point remission by all definitions (OR range 1.20–2.71). Minority ethnicity decreased odds of remission by DAS28 (OR 0.13) and CDAI (OR 0.09) definitions. Male sex was associated with DAS28 remission (OR 2.85), whereas higher baseline physician global (OR 0.67) and erythrocyte sedimentation rate values (OR 0.98) decreased odds of DAS28 remission. Higher baseline patient global score (OR 0.77) and swollen joint counts (OR 0.93) were negative predictors for CDAI remission. Higher baseline Health Assessment Questionnaire (OR 0.62) reduced odds for remission by the SDAI definition, and educational attainment increased these odds (OR 2.13). Sustained remission was negatively predicted by baseline physician global for the DAS28 (OR 0.80), and higher tender joint count (OR 0.96) for the CDAI.
Conclusion. We demonstrate the influence of duration of remission state and remission definition on defining independent predictors for remission in RA requiring anti-tumor necrosis factor therapy. These predictors offer improved applicability for modern rheumatology practice.
Outcome measures in rheumatoid arthritis (RA) clinical trials and practice have evolved from relatively complex composite disease activity measures such as the American College of Rheumatology (ACR) response criteria1 and the 28-joint Disease Activity Score (DAS28)2 to simple Boolean or composite disease activity measures based on addition of core outcome domains. These include the ACR/EULAR (European League Against Rheumatism) Boolean definition of remission3, the Simplified Disease Activity Index (SDAI)4, and the Clinical Disease Activity Index (CDAI)5. Validated cutpoints for remission, the desirable disease activity state in RA, exist for each of these measures, and are useful for determining response to treatment in clinical trials and in day-to-day practice.
Past efforts to identify predictors for remission have largely been based on the ACR response criteria or the DAS28 definition of remission, as summarized by Katchamart, et al in a systematic review6. This review, which explicitly included only publications reporting on multivariate analyses, identified demographic factors (male sex, young age, late-onset RA, nonsmoker), disease presentation factors (short disease duration, low baseline disease activity, mild functional impairment, low baseline radiographic damage, seronegative status, low levels of acute-phase reactants), genetic and immune expression profiles, initial treatment approach (type and combination of therapeutic agents), and response factors that influence the odds of entering remission. These predictors are relevant in clinical practice because they are used to educate patients on their expected prognosis, and thereby guide treatment recommendations and inform the relevant risk-benefit profile of RA therapies and management strategies proposed.
The current literature lacks reports on the identification of predictive factors when newer definitions of remission are applied. It is repeatedly demonstrated that the DAS28 definition of remission (score < 2.6) is the least stringent of all the disease activity measures, and classifies about 3 times as many patients in remission as does the ACR/EULAR Boolean definition7,8,9. This means that the currently accepted predictors for remission perhaps include weak variables, and only the strongest predictors would emerge if the analysis considered more stringent definitions of remission. As well, many of the studies consider only a cross-sectional evaluation of remission, rather than a sustained state of disease activity. Sustained levels of remission are more beneficial for radiographic outcomes10, and predictors of sustained remission will inform prognosis even further.
The objective of our study was to compare independent predictors of DAS28 remission to those identified when applying the ACR/EULAR Boolean, SDAI, and CDAI definitions of remission. We hypothesized that the strongest predictors would be associated with all definitions for remission. We also evaluated predictors based on a cross-sectional evaluation of remission compared to a longitudinal evaluation of disease activity status to define sustained remission.
MATERIALS AND METHODS
Data sources
The Alberta Biologics Pharmacosurveillance Program was initiated in 2004 to identify the efficacy, safety, and cost-effectiveness of new biologic therapies for RA11. Patients in Alberta qualify for biologic therapy cost coverage if they have RA refractory to parenteral methotrexate (MTX), in combination with at least 1 other disease-modifying antirheumatic drug (DMARD), and leflunomide. They must achieve and retain a minimum DAS28 improvement of 1.2 units and a minimum improvement of their Health Assessment Questionnaire (HAQ) score by 0.22 units over their baseline scores at 12 weeks to continue receiving cost coverage for therapy. Patients who fail to meet these response criteria will be switched to another agent. Patients participating in the program are assessed for disease activity, adverse events, effect on function and quality of life, healthcare use, and self-reported economic effect of their disease at the start of a new biologic agent, 12 weeks after initiation of that drug, and at 6-month intervals if no treatment switches have occurred. The patients are also brought in for assessment if they contact the program reporting suspected treatment failure or adverse events, which may require a treatment switch. This analysis was restricted to patients naive to biologic therapy (to reduce confounding introduced by prior response to biologic therapy in patients switching therapy) and who had at least 2 study visits. The study period for this analysis was April 2004 to March 2011.
Ethics
All patients provided informed consent in accordance with ethical standards described in the Declaration of Helsinki. The study is approved by the University of Calgary Health Research Ethics Board (E-20424) and by the University of Alberta Research Ethics Board (Study ID Pro00000914).
Outcomes
Clinical variables collected at study visits were used to determine achievement of the ACR/EULAR Boolean definition of remission, and to calculate disease activity scores for DAS28, SDAI, and CDAI. For each study visit, based on these scores and their recognized cutpoints for remission (< 2.6, < 3.3, and < 2.8, respectively), subjects were categorized as being in remission or not for each separate definition at any single visit (point remission). We also used an ACR Boolean definition excluding the acute-phase response result, as Kuriya, et al, have done9, in sensitivity analysis. Longitudinal disease course was considered to classify patients for sustained remission, defined as patients classified in remission on at least 2 consecutive visits with no treatment changes and no steroid requirements for a minimum of 1 year.
Covariates
We included disease duration at first anti-tumor necrosis factor (TNF) start, baseline disease activity measures [tender joint count on 28 joints, swollen joint counts on 28 joints, HAQ score12, erythrocyte sedimentation rate (ESR), C-reactive protein results, patient and evaluator global scores, and the relevant composite disease activity measure]. Smoking status was characterized as current smoker or ex-/never smoker. Obesity was defined as a body mass index of ≥ 30 kg/m2 at treatment start. Serologic status was considered positive if either rheumatoid factor (RF) or anticyclic citrullinated peptide (anti-CCP) antibody were ever recorded as positive in the database. We also included sociodemographic factors, including self-reported ethnicity (white, Aboriginal, Hispanic, East Indian, Asian, African American), marital status (single, separated/widowed/divorced, or married/common-law) at anti-TNF start, and highest education level (elementary as reference, high school, college/university). We also decided to explore whether speed of attaining remission affected the ability to achieve sustained versus point remission. Subjects were classified as achieving early remission if that target was met by their first visit (< 16 weeks) after starting anti-TNF therapy.
Statistical analysis
Multivariate models for point remission and sustained remission were individually calculated for each definition to compare which variables significantly predicted attaining the outcome. All covariates were included in the initial model, with sequential removal of variables until a maximal value for the goodness-of-fit test and the percent correctly classified. All models are reported using OR. We used as-observed data only and did not impute any results. Reference categories for the sociodemographic factors were white, single marital status, and elementary education. Stata IC version 12.0 (StataCorp) was used for analyses.
RESULTS
Our population consists of 1116 biologic-naive patients treated with anti-TNF therapies who made at least 2 study visits during the study period. Details on patient demographics and baseline disease activity measures are provided in Table 1. The mean duration of followup in the cohort was 990 days (SD 745), with a median of 840 days, and a range of 44–3786 days. On average, each individual person had 6.3 visits (median 6, range 2–19).
Patient characteristics. All values mean (SD) unless specified.
Table 2 presents point remission rates, and sustained remission rates in patients who had attained point remission. Point remission rates varied to be as low as 26.2% for SDAI, and as high as 61.6% for DAS28. This variability was less pronounced when examining sustained remission, with the lowest rates for SDAI (19.9%) and highest for DAS28 (37.0%).
Remission rates, by disease activity definition. Data are n (%).
Tables detailing all covariates and OR for each remission of definition for both point and sustained durations appear in the appendices. Predictors of remission varied between disease activity measure definitions, and when comparing point and sustained remissions (Table 3). Ex-smokers and never smokers had higher odds of point remission by all definitions (OR range 1.20–2.71). Minority ethnicity decreased odds of remission by DAS28 (OR 0.13, 95% CI 0.02–0.86, p = 0.034) and CDAI (OR 0.09, 95% CI 0.01–0.80, p = 0.030). Male sex was associated with DAS28 remission (OR 2.85, 95% CI 1.59–5.10, p < 0.001), whereas higher baseline physician global [OR 0.67 per unit on a 0–10 visual analog scale (VAS), 95% CI 0.51–0.88, p = 0.004] and ESR values (OR 0.98 per mm/h, 95% CI 0.96–1.00, p = 0.030) decreased odds of DAS28 remission. Higher baseline patient global score (OR 0.77 per unit on a 0–10 VAS, 95% CI 0.60–0.99, p = 0.038) and swollen joint counts (OR 0.93 per swollen joint, 95% CI 0.87–1.00, p = 0.05) were negative predictors for CDAI remission. Higher baseline HAQ (OR 0.62 per unit, 95% CI 0.38–1.00, p = 0.005) reduced odds for remission by the SDAI definition, and educational attainment increased these odds (OR 2.13 for university or college training, 95% CI 0.98–4.65, p = 0.05). Sustained remission was negatively predicted by the baseline physician global for the DAS28 (OR 0.80 per unit on a 0–10 VAS, 95% CI 0.66–0.99, p = 0.036), and higher tender joint count (OR 0.96 per tender joint, 95% CI 0.92–1.00, p = 0.05) for the CDAI. In the analysis of predictors of point remission in the ACR/EULAR Boolean definition excluding the acute-phase reactants, never smoking or quitting smoking increased odds of remission (OR 1.83, 95% CI 1.04–3.22, p = 0.037) and East Indian ethnicity was associated with decreased odds of remission (OR 0.11, 95% CI 0.01–0.89, p = 0.038). In the sustained remission model, being obese was associated with decreased odds of remission (OR 0.30, 95% CI 0.10–0.90, p = 0.032). RF status did not influence remission by any definition in the multivariate models.
Independent predictors of remission, by definition and duration of remission. Exact estimates and all model variables are in the supplemental data.
We also explored whether early response to treatment, defined as achieving remission in the first 16 weeks after anti-TNF agent initiation, could predict achieving sustained remission. In the multivariate model for the DAS28 definition, early remission was an independent predictor for sustained remission (OR 1.88, 95% CI 1.27–2.78, p = 0.002). No other variables independently predicted sustained remission.
DISCUSSION
Using a population-based biologics registry in Canada, we have determined that predictors of remission vary by both the disease activity measure used to define remission, and by whether the remission is determined cross-sectionally or longitudinally. Consistent across the definitions of point remission was the finding of ex-/never smokers having improved odds of remission, and East Indian patients having lower odds of remission. Aboriginal, Hispanic, and Asian patients also demonstrated a trend to lower odds of remission, but not reaching statistical significance. Additional predictors increasing the odds of point remission were male sex (by DAS28 remission) and university/college education (by SDAI remission). Disease activity did affect the odds of attaining point remission, although with variations by the measure used. High baseline physician global and higher ESR decreased the odds of DAS28 remission, whereas high baseline patient global and higher swollen joint counts decreased the odds of CDAI remission. High baseline HAQ scores decreased the odds of SDAI remission.
Fewer predictors could be identified in the sustained remission analysis. Only a higher baseline physician global was determined to reduce the odds of remission by the DAS28 definition, and similarly for a higher tender joint count by the CDAI definition. Using an alternative ACR/EULAR Boolean definition excluding the laboratory results, we also determined obesity to reduce odds of attaining remission. This finding supports an emerging literature on the consequences of obesity in treatment response13.
The systematic review by Katchamart, et al6 identified 18 applicable studies on predictors for cross-sectional remission. Independent variables for RA remission were male sex, young age, late-onset RA, short disease duration, nonsmoker, low baseline disease activity, mild functional impairment, low baseline radiographic damage, absence of RF and anti-CCP, low serum level of acute-phase reactant, interleukin 2 and receptor activator of nuclear factor-κB ligand at baseline, particular alleles for MTX receptors, early nonbiologic DMARD combinations, anti-TNF use, and moderate or good response to treatments in the first 6 months of disease. The included studies did not require that remission be sustained. Our analysis did not necessarily confirm all of these expected predictors, possibly related to distribution of frequency of particular patient characteristics. For example, the majority of the cohort had longstanding disease, reducing the power to detect differences in odds for patients with short disease duration; similarly the high seropositivity rate for RF means we had little power to examine the effect of being seronegative.
There are few data published on predictors of sustained remission in RA from observational cohorts. Patients able to achieve sustained remission offer critical information in the pursuit of personalized medicine and therapeutic selection. There is no accepted definition for sustained remission at this time either, other than the common definition of having achieved that state at multiple consecutive visits over a 1-year period14,15,16. Jayakumar, et al studied predictors for sustained remission in early (< 2 yrs) RA16. Point remission was defined as a DAS < 1.6; and sustained remission if DAS < 1.6 was met at all 3-, 4-, and 5-year followup visits. Only 11% of this cohort attained sustained remission, independently predicted by male sex, short duration of symptoms at treatment start, and having fewer tender joints at baseline. On the importance of sustained remission, less radiographic progression occurred and HAQ scores were significantly lower (0.13 vs 1.1) at the end of 5 years of followup if sustained remission was achieved. Reports of predictors of remission considering shorter periods of observation are available. In a Belgian cohort, remission at 4 months along with male sex, age, and disease duration at treatment start predicted remission at 1 year17. In the Danish DANBIO cohort, concomitant corticosteroid treatment, older age, low functional status at baseline, and the number of previous DMARD were negative predictors for both DAS28 and CDAI remission at 1 year18.
This paper highlights important gaps in the current reporting of outcome measures. There is clearly a need to achieve consensus on a definition of sustained remission. This would allow clinicians and administrators to better estimate anticipated treatment outcomes and plan resource allocation. From a research perspective, there are likely better ways of defining disease activity states considering the duration of existence in various states. Perhaps models to define an “area under the curve” approach to evaluate longitudinal disease activity adapted for clinical practice would be the optimal solution for the many cohorts and disease registries in existence. Other methods, such as the multistate modeling approach as described by Tom and Farewell19 or latent growth models as used by Norton, et al20 are options for this type of evaluation. Although it is beyond the scope of our current paper to compare results obtained with these different methods and we used standard analysis methods to maintain familiarity of interpretation for the reader, we are encouraged to pursue this line of investigation.
Finally, our study provides further proof of the influence of determinants of health on RA outcomes. We have identified that people of an ethnic minority are less likely to achieve the desirable treatment outcome of remission. Patients with higher education levels do better than those with primary school and high school completion. Innovations in the delivery of healthcare to reach out to populations found to have disparate outcomes are needed to create an environment where equitable health outcomes are realized.
Acknowledgment
We thank our colleagues, research assistants, and nurses in the Divisions of Rheumatology at the Universities of Calgary and Alberta.
APPENDIX 1.
OR for attaining point remission.
APPENDIX 2.
OR for attaining sustained remission.
Footnotes
-
The Alberta Biologics Pharmacosurveillance Program (ABioPharm) was supported by a grant from Alberta Health and Wellness for the study period reported. Data analyst costs were paid through the Department of Medicine and the Arthur J.E. Child Chair in Rheumatology, University of Calgary.
- Accepted for publication April 3, 2014.
