Classification of rheumatologic opinion on early inflammatory arthritis: Harmonization of a heterogeneous standard
Interrater variability limits rheumatologic opinion as a reference standard for early inflammatory arthritis (IA) classification. The study objectives were to determine whether rheumatologic opinion is associated with potential early IA classification methods despite high interrater variability, and to compare the relative strengths of those associations.
Eighteen rheumatologists independently classified 30 initial rheumatology presentation summaries as early IA or not and recommended a pharmacotherapy. Case fulfillment of the following classification methods was independently determined: early referral to rheumatology recommendation for rheumatoid arthritis (ERRR), common early IA cohort inclusion criteria (CEAC), and prevalent IA classification criteria (American College of Rheumatology [ACR]/European Spondylarthropathy Study Group [ESSG]). Associations between rheumatologic opinion, disease-modifying antirheumatic drug (DMARD) recommendation, and each classification method were determined.
Participating rheumatologists published on early IA and represented 3 continents. The median case was age 43 (interquartile range [IQR] 30–53) years, had 40 (IQR 24–104) weeks of symptoms, 60 (IQR 18–120) minutes of morning stiffness, a swollen joint count of 6 (IQR 1–13), and an erythrocyte sedimentation rate of 25 (IQR 10–51) mm/hour. The mean ± SD multiple-rater kappa for rheumatologic opinion on early IA was 0.16 ± 0.02. The common odds ratios for associations between rheumatologic opinion and ERRR, CEAC, and ACR/ESSG were 10.3 (95% confidence interval [95% CI] 4.6–23.2), 4.4 (95% CI 2.5–7.7), and 0.7 (95% CI 0.4–1.1), respectively. Odds ratios for associations between DMARD recommendation and ERRR, rheumatologic opinion, CEAC, and ACR/ESSG were 18.7 (95% CI 8.1–43.2), 10.6 (95% CI 6.0–18.8), 2.8 (95% CI 1.7–4.6), and 0.5 (95% CI 0.3–0.7), respectively.
Classification methods can be used to harmonize rheumatologic opinion of early IA despite high interrater variability. The ERRR is very strongly associated with both rheumatologic opinions of early IA and DMARD treatment recommendations.
Early inflammatory arthritis (IA) is an idiopathic syndrome assigned by rheumatologists to patients presenting with preliminary symptoms of chronic IA diseases. Using early IA as a prodromal diagnosis, rheumatologists may consider early treatment with disease-modifying antirheumatic drugs (DMARDs). Yet, high interrater variability in rheumatologic opinion on the definition of early IA (1, 2) may result in delayed treatment for subsets of this population. A clear, consensus definition of early IA may help overcome disparities in early care.
Defining early IA prospectively is challenging. Current reference standards for early IA classification are based on retrospective assignment of established diagnoses or fulfillment of classification criteria for established IA conditions in followup (3–6). The suboptimal sensitivity of laboratory markers (7–9) and questionable specificity of individual clinical parameters (10–14) render early diagnosis a challenge. In turn, combinations of laboratory and clinical markers per classification criteria for prevalent forms of specific IA diagnoses (15–17) have low sensitivity for incident disease (18–21). The definition of the term “early” is also variable (22–26). As such, limitations with the available diagnostic tools for early IA classification may contribute to the lack of consensus on the definition of early IA.
In light of these limitations, it was hypothesized that disease parameters encompassing early referral to rheumatology recommendations for IA or common inclusion criteria for early arthritis cohorts may account for the inter-rheumatologist variability and may be strongly associated with overall rheumatologic opinion. Toward establishing a nonarbitrary definition of early IA, the objective of this study was to determine if rheumatologic opinion on early IA is associated with these potential methods of defining the prodromal diagnosis despite high interrater variability.
MATERIALS AND METHODS
A meta-analytical approach was used to investigate the relative strengths of association between rheumatologic opinion and 3 potential classification methods for early IA. Eighteen rheumatologists each evaluated 30 initial rheumatology presentation summaries of patients with established IA diagnoses at 2 years of clinical followup. The rheumatologists were blinded to the 2-year diagnoses of the cases. Case selection was conducted independently from the selection of potential classification methods. The strengths of association between rheumatologic opinion and the potential classification methods were pooled across the 18 rheumatologists. The primary reference standard for early IA classification was rheumatologic opinion (23–26). Rheumatologic recommendation to treat with DMARDs (including biologics) was used as a secondary reference standard. The relative strengths of associations between the 3 potential classification methods were compared.
A purposive sample of 30 established patients with IA was selected to represent a range of demographic and disease characteristics at rheumatology presentation using variables commonly available in the clinical records (Table 1). The cases were identified from 2 clinics: 17 from an early IA clinic and 13 from an academic general rheumatology practice. The medical history, physical examination, and clinical and laboratory investigations conducted at the initial rheumatology presentation were used to prepare anonymized structured case summaries for each patient. Case summaries were posted on SurveyMonkey (online at: SurveyMonkey.com) and accessed by the participating rheumatologists for evaluation. The working clinical diagnoses of the patients at 2 years post–initial rheumatology presentation were collected but not reported to the evaluating rheumatologists. The 2-year clinical diagnoses were collected to explore if rheumatologic opinion of early IA or DMARD or biologic treatment recommendations were influenced by early disease characteristics of specific prevalent IA diagnoses.
Table 1. Initial rheumatology presentation characteristics of the study cases*
|Case demographics|| || |
| Age, median (IQR) years||30||43 (30–53)|
| Women||30||24 (80)|
|Clinical assessment|| || |
| Constitutional symptoms||30||10 (33)|
| Enthesopathy||30||2 (7)|
| Inflammatory bowel disease||30||1 (3)|
| Morning stiffness duration, median (IQR) minutes||28||60 (18–120)|
| Nodules||30||1 (3)|
| Symptom duration, median (IQR) weeks||29||40 (24–104)|
| Response to NSAID therapy||30||21 (70)|
| Positive family history of spondylarthritis||30||11 (37)|
| Psoriasis||30||3 (10)|
| Sacroiliitis||30||1 (3)|
|Diagnostic tests|| || |
| Anti-CCP antibody positivity||30||7 (23)|
| Antinuclear antibody positivity||30||5 (17)|
| CRP level, median (IQR) mg/liter||26||5 (1–14)|
| ESR, median (IQR) mm/hour||22||25 (10–51)|
| Radiographic evidence of erosions||30||12 (40)|
| Rheumatoid factor positivity||30||14 (47)|
|Joint assessment|| || |
| Asymmetric perisynovitis||30||6 (20)|
| Asymmetric perisynovitis of the lower extremities||30||13 (43)|
| Metacarpophalangeal joint involvement||30||11 (37)|
| Metatarsophalangeal joint involvement||30||10 (33)|
| Proximal interphalangeal joint involvement||30||11 (37)|
| Swollen joint count, median (IQR)||27||6 (1–13)|
| Symmetric involvement||30||13 (43)|
| Wrist involvement||30||10 (33)|
|Pain assessment|| || |
| Alternating buttock pain||30||2 (7)|
| Inflammatory spine pain||30||5 (17)|
| Pain||30||30 (100)|
Eighteen rheumatologists participated in the study (Table 2). International rheumatologists were identified from the early IA literature and solicited for participation in the study. Additional rheumatologists were recruited to supplement the number of participants. The participating rheumatologists were blinded to the 2-year diagnosis of the cases and the study analysis plan. Each rheumatologist provided the following data for each case: 1) classification as early IA or not, 2) confidence with classification on a 5-point Likert scale, and 3) pharmacologic treatment recommendation. In total, 18 rheumatologists rated 30 cases, yielding up to 540 evaluations each of the above queries.
Table 2. Characteristics of the participating rheumatologists (n = 18)*
| Female||9 (50.0)|
| Male||9 (50.0)|
| Australia||3 (16.7)|
| Europe|| |
| The Netherlands||1 (5.6)|
| Romania||3 (16.7)|
| Spain||2 (11.1)|
| Switzerland||1 (5.6)|
| UK||1 (5.6)|
| North America|| |
| Canada||4 (22.2)|
| US||3 (16.7)|
|Career stage|| |
| Early||1 (5.6)|
| Mid||15 (83.3)|
| Late||2 (11.1)|
|Majority of time spent|| |
| Academic||10 (55.6)|
| Clinical||8 (44.4)|
|Geographic clinic setting|| |
| Rural||1 (5.6)|
| Urban||17 (94.4)|
|Primary setting|| |
| Institution||15 (83.3)|
| Private practice||3 (16.7)|
The primary reference standard for early IA classification was rheumatologic opinion. Given the impetus of early, aggressive therapy as an outcome for early IA classification, rheumatologic recommendation to treat with DMARDs (including biologics) was used as a secondary reference standard.
Early referral to rheumatology recommendations (27–29), common inclusion criteria for early IA cohorts (13, 23–25, 30–34), and classification criteria for prevalent IA conditions (15–17, 35, 36) were identified from the literature as potential early IA classification methods. The following 3 classification methods were selected for investigation: 1) early referral recommendation to rheumatology for rheumatoid arthritis (ERRR) (27), i.e., at least 3 swollen joints, a positive squeeze test for the metacarpophalangeal (MCP) or metatarsophalangeal (MTP) joints, or at least 30 minutes of morning stiffness; 2) common early arthritis cohort inclusion criteria (CEAC), i.e., at least 2 swollen joints (30–33) and 6–52 weeks of symptom duration (13, 23–25, 30, 34); and 3) fulfillment of either the American College of Rheumatology (ACR; formerly the American Rheumatism Association) classification criteria for rheumatoid arthritis (RA) (15) or the European Spondylarthropathy Study Group (ESSG) classification criteria for spondylarthritis (SpA) (16).
The ERRR is an RA-specific recommendation, but includes as a criterion at least 30 minutes of morning stiffness (27), a measure of inflammatory back pain included in the early referral recommendation for ankylosing spondylitis (AS) (28, 29). The early referral recommendation for AS was excluded because only 1 study case fulfilled its screening requirements. The CEAC were derived from inclusion criteria commonly used in early IA cohorts. Although the inclusion criteria for early IA cohorts are heterogeneous, at least 2 swollen joints (30–33) and a symptom duration proximate to 6–52 weeks (13, 23–25, 30, 34) were common criteria to many. The ACR and ESSG classification criteria represent widely recognized classification schemes for prevalent IA conditions and were included as a negative control for associations between rheumatologic opinion and established disease. General classification criteria for SpA were included (16) rather than diagnosis-specific criteria (17, 35, 36). The classification methods were selected independently from and blinded to the rheumatologists' case assessments.
For each case, the percentage of rheumatologists to make a classification of early IA, the percentage to report “somewhat” or greater confidence in their opinion, and the pharmacologic treatment recommendations were summarized. The multiple-rater kappa for expected agreement between the rheumatologic opinions of the cases across the participating rheumatologists was tested against the null hypothesis of chance agreement (κ = 0). The odds ratio (OR) for the associations between rheumatologic opinion, fulfillment of each classification method, and DMARD treatment recommendation was calculated for each rheumatologist. Breslow-Day heterogeneity tests were conducted to determine if differences in the strengths of association between rheumatologists existed that precluded pooling data into a single, Cochran-Mantel-Haenszel common OR. Where appropriate, the common OR was calculated. The logit common OR approximation was used where the Cochran-Mantel-Haenszel common OR could not be calculated due to “0” entries in the 2 × 2 table. The logit approximation substitutes a value of 0.5 for 0 to enable the calculation of the common OR. Measured associations were stratified by rheumatologist characteristics to explore the variability of the results across these parameters.
Case data were double entered into a Microsoft Access database, 2003 edition (Microsoft). Statistical analyses were conducted using SAS, version 9.1.3 (SAS Institute). Statistical tests were conducted using a 5% level of significance.
At rheumatology presentation, the selected cases were a median age of 43 (interquartile range [IQR] 30–53) years, had a median symptom duration of 40 (IQR 24–104) weeks, a median of 60 (IQR 18–120) minutes of morning stiffness, a median of 6 (IQR 1–13) swollen joints, a median erythrocyte sedimentation rate of 25 (IQR 10–51) mm/hour, and a median C-reactive protein level of 5 (IQR 1–14) mg/liter (Table 1). Four-fifths (80%) of the cases were female, 33% had either MCP and/or MTP joint involvement, 40% had radiographic evidence of erosions, 17% had inflammatory spinal pain, 37% had a family history of SpA, 43% had asymmetric perisynovitis of the lower extremities, and 10% had psoriasis. At 2 years of rheumatologic followup, 60% (n = 18) of the cases had an established clinical diagnosis of RA, 20% (n = 6) remained undifferentiated, 13% (n = 4) had AS, and 7% (n = 2) had psoriatic arthritis (PsA).
The 18 participating rheumatologists included a mixture of academic and community practitioners from 3 continents, including 8 countries (Table 2). The majority were in midcareer stage (age 40–60 years), were affiliated with an institution rather than in private practice, and practiced in an urban setting. Fourteen of the 18 rheumatologists evaluated at least 25 of the 30 cases. In total, 468 of the possible 540 rheumatologic evaluations of early IA classification and pharmacologic treatment recommendation were completed for the 30 cases studied. These included 293 rheumatologist classifications of early IA and 302 recommendations for DMARD treatment.
The mean ± SD multiple-rater kappa for expected agreement on the classification of the 30 cases as early IA by the 18 rheumatologists was 0.16 ± 0.02, which was significantly different than chance (P < 0.001). At least 50% of the rheumatologists assessed 70% of the cases as early IA, at least 70% of the rheumatologists assessed 50% of the cases as early IA, and at least 80% of the rheumatologists assessed 33% of the cases as early IA.
The logit common ORs for the associations between rheumatologic opinion and the ERRR, CEAC, and ACR/ESSG were 10.3 (95% confidence interval [95% CI] 4.6–23.2, P = 0.17 by Breslow-Day heterogeneity test), 4.4 (95% CI 2.5–7.7, P = 0.28 by Breslow-Day heterogeneity test), and 0.7 (95% CI 0.4–1.1, P = 0.86 by Breslow-Day heterogeneity test), respectively (Table 3). The Breslow-Day heterogeneity tests for differences in the strengths of association across the rheumatologists were not significant, indicating that it was appropriate to pool data across the rheumatologists.
Table 3. Primary and sensitivity analyses of associations between rheumatologic opinion and potential classification methods for early inflammatory arthritis*
|Primary analysis§|| || || || || || |
| All rheumatologists; cases classified by all methods|| || || || || || |
| ERRR||18||26||408||36.0 (11.0–118.5)||10.3 (4.6–23.2)||0.17|
| CEAC||18||26||408||6.3 (2.5–7.7)||4.4 (2.5–7.7)||0.28|
| ACR/ESSG||18||26||408||0.7 (0.4–1.1)||0.7 (0.4–1.1)||0.86|
|Sensitivity analyses¶|| || || || || || |
| All rheumatologists; all classified cases|| || || || || || |
| ERRR||18||30||468||17.1 (7.1–41.1)||8.9 (4.1–19.2)||0.055|
| CEAC||18||26||408||6.3 (2.5–7.7)||4.4 (2.5–7.7)||0.28|
| ACR/ESSG||18||30||468||0.7 (0.4–1.0)||0.7 (0.4–1.1)||0.90|
|Rheumatologic subset; cases classified by all methods|| || || || || || |
| ERRR||14||26||363||32.6 (9.8–108.7)||9.6 (4.1–22.1)||0.17|
| CEAC||14||26||363||6.0 (3.3–11.0)||4.5 (2.5–8.3)||0.17|
| ACR/ESSG||14||26||363||0.7 (0.4–1.2)||0.7 (0.4–1.2)||0.95|
|Rheumatologic subset; all classified cases|| || || || || || |
| ERRR||14||30||417||15.4 (6.4–37.5)||8.3 (3.7–18.3)||0.057|
| CEAC||14||26||363||6.0 (3.3–11.0)||4.5 (2.5–8.3)||0.17|
| ACR/ESSG||14||30||417||0.7 (0.5–1.2)||0.7 (0.5–1.2)||0.98|
Sensitivity analyses accounting for all of the classified cases, and excluding rheumatologists who evaluated less than 25 of the 30 cases, were conducted (Table 3). These analyses resolved no substantial differences in the relative strengths of association with each classification method. In an exploratory analysis, the results were stratified by the rheumatologist characteristics listed in Table 2 (data not shown). Although the relative mean magnitude of the strengths of association for stratified results was unchanged, this analysis was limited by few rheumatologists and evaluations per stratum.
Using the secondary reference standard, the ORs for the associations between DMARD treatment recommendation and the ERRR, CEAC, and ACR/ESSG were 18.7 (95% CI 8.1–43.2, P = 0.14 by Breslow-Day heterogeneity test), 2.8 (95% CI 1.7–4.6, P = 0.12 by Breslow-Day heterogeneity test), and 0.5 (95% CI 0.3–0.7, P = 0.54 by Breslow-Day heterogeneity test), respectively. The OR between DMARD treatment recommendation and rheumatologic opinion on early IA was 10.6 (95% CI 6.0–18.8, P = 0.24 by Breslow-Day heterogeneity test). Of 302 rheumatologic evaluations that recommended DMARD treatment, 97.4% were classified as early IA by the ERRR criteria, 53.5% by the CEAC, and 33.8% by the ACR/ESSG. Approximately three-fourths (76.2%) of the evaluations that recommended DMARD treatment corresponded to cases classified as early IA by rheumatologic opinion. Conversely, 71.9% of the pharmacologic recommendations for the ERRR-classified early IA cases were for DMARD treatment compared with 81.7% of the pharmacologic recommendations for the CEAC-classified early IA cases and 55.1% of the pharmacologic recommendations for the ACR/ESSG-classified early IA cases. Of the pharmacologic recommendations for cases classified as early IA by rheumatologic opinion, 78.5% were for DMARD treatment.
Both rheumatologic opinion of early IA and DMARD treatment recommendation were most frequently assigned to cases with 2-year diagnoses of RA, followed by undifferentiated IA, PsA, and least with AS (Table 4). The majority of cases (87% [n = 26]) were classified as early IA using the ERRR. Only the 4 AS cases were classified as non–early IA using the ERRR criteria. In turn, 40% (n = 12) of the cases were classified as early IA using either the CEAC (including 4 unclassifiable cases) or the ACR/ESSG. Missing symptom duration or unspecified tender versus swollen joint involvement prevented the classification of 4 cases by the CEAC method. Of the remaining 26 cases, the CEAC more likely classified undifferentiated IA cases as early IA than RA, and detected neither the remaining PsA case nor the 3 AS cases. Four RA cases, 3 undifferentiated IA cases, 3 AS cases, and 1 PsA case were classified as early IA by the ACR/ESSG method.
Table 4. Two-year diagnosis of study cases by potential classification methods for early IA*
|Rheumatologic opinion||Early IA||293‡||70.0||4.8||4.4||20.8|
| ||Not early IA||175‡||45.1||10.9||26.3||17.7|
A median of 13 (IQR 12–15) rheumatologists were “somewhat” to “very confident” with the classification of the case summaries as early IA or not, 1 (IQR 0–3) rheumatologist was “very unconfident” to “unsure,” and 3 (IQR 2–4) rheumatologists did not comment on their confidence. Of the cases assessed by rheumatologists as early IA, the rheumatologists were “somewhat” to “very confident” with their assessment for 88.0% as compared with 80.6% for cases assessed as non–early IA. The association between rheumatologic opinion of early IA and confidence was significantly heterogeneous across the rheumatologists (P = 0.0017) and data were not pooled. The associations between confidence and case fulfillment of either the ERRR, CEAC, or ACR/ESSG classification criteria were not significant.
This study verified earlier findings of low interrater agreement for rheumatologic opinion on early IA (1). Although the interrater agreement was greater than could be expected by chance, the effect size represents slight agreement (37). Here it was demonstrated that disease characteristics developed for related IA interventions may be used to overcome the heterogeneity in rheumatologic opinion on early IA to produce a single, harmonized, nonarbitrary reference standard. An early IA reference standard with such properties may facilitate 1) adoption of early, aggressive treatment guidelines; 2) knowledge translation of appropriate early referrals to rheumatology; 3) implementation of common eligibility criteria for studies of this population; and 4) uptake of early IA study data for clinical decision making.
Specifically, 3 potential classification methods were investigated. Rheumatologic opinion on early IA was very strongly (38) associated with the ERRR: at least 3 swollen joints, a positive composite compression test of the MCP or MTP joints, or a duration of at least 30 minutes of morning stiffness (27). A strong association (38) of lesser strength was determined between rheumatologic opinion and the CEAC. The association with the ACR/ESSG was not significant. The relative strengths of these associations were corroborated by associations between the classification methods and rheumatologic recommendations for DMARD treatment, the use of which accounted for rheumatologists' classifications of “non–early IA” for reasons of either prolonged symptom duration or interpretation of early IA as “early RA.”
The ERRR was developed using an evidence-based approach to provide primary care practitioners with guidance on appropriate, early referral of suspected RA to rheumatology (27). The grade C evidence on which the recommendation was based supports the discriminant validity of at least 3 swollen joints (15, 39), positive compression tests of either the MCP or MTP joints (39), and at least 30 minutes of morning stiffness (40, 41). These criteria discriminate between persistent, erosive disease and other common musculoskeletal disorders (27). The relationships between the study reference standards and the 2-year clinical diagnosis of the cases help to explain the current findings. Both rheumatologic opinions on early IA and DMARD treatment recommendations most strongly agreed with RA, which the ERRR was designed to detect.
Despite capturing parameters for both RA (27) and SpA (28, 29, 42), the ERRR did not detect any of the 4 AS cases in the study. The CEAC were also insensitive to the detection of the AS cases. Clinical onset of AS may be better characterized by inflammatory back pain and stiffness (29). Despite the omission of the AS cases from these 2 potential classification methods, 78% of rheumatologic evaluations for AS cases were classified as non–early IA. As such, rheumatologic opinion on early IA may exclude certain SpA, especially AS.
Rheumatologic opinion on early IA agreed most with early RA characteristics, despite the majority of evaluators being active, academic, clinical researchers with an interest in early IA. This sample of evaluators was purposely skewed toward early IA experts because they inform the broader rheumatology population. Participating rheumatologists were blinded to the analysis plan and classification methods investigated. Their evaluations were completely independent of study investigator influence. Conversely, the selection of potential classification methods by study investigators was blinded to case evaluations. These methodologic measures mitigated circular reasoning and channeling bias. Low agreement among expert rheumatologists in this setting highlights the heterogeneity of opinion on early IA in the field.
A total of 468 rheumatologic evaluations were collected and the primary analysis included the 26 cases classified by all 3 methods. The relative effect sizes determined from the primary analysis remained unchanged in sensitivity analyses that included all of the classified cases. Precise estimates of both expected agreement and effect sizes for the associations investigated were obtained with the 18 rheumatologists and 30 cases used. The number of rheumatologist evaluators and cases used also compares favorably with other studies that have investigated interrater agreement (1, 43, 44).
The cases in this purposive sample were selected from 2 clinical settings: an academic general rheumatology practice and an early arthritis clinic. More than half of the cases (57% [n = 17]) were derived from the latter, a cohort with inclusion criteria similar to the CEAC. Despite this, the strongest associations measured were with the ERRR. To an extent, the CEAC were more general than the ERRR, incorporating a lower threshold for swollen joint count. On the other hand, the symptom duration requirement in the CEAC was more stringent than in the ERRR, which did not limit symptom duration. Restricting symptom duration may have the effect of excluding patients with insidious symptom onset. For example, the percentage of RA patients with symptom durations greater than 1 year at rheumatology presentation has been estimated to range from 40–60% (45, 46). For SpA, with an average symptom duration of 5 years or more (47), a much larger proportion of patients may be excluded by restricting symptom duration to less than 52 weeks. Yet, long-term outcomes for insidious-onset disease may not differ from acute- and severe-onset IA (48, 49). Nonetheless, since early IA cohort inclusion criteria are heterogeneous, these findings do not apply to cohorts with markedly different eligibility criteria.
Case fulfillment of the ACR/ESSG was included as a third comparator in this study to determine if rheumatologists could differentiate early IA from established IA. A nonsignificant inverse association was observed between rheumatologists' assessments of early IA and case fulfillment of the ACR/ESSG. These findings corroborate other research (18, 19) and suggest that rheumatologic assessments in this study were specific for early IA rather than established disease.
Despite the above considerations, the very strong association between rheumatologic opinion on early IA and the ERRR needs to be interpreted with caution. Rheumatologic assessments of case summaries are imperfectly correlated with the clinical management of real patients (5). Validation of these findings in the clinical setting would be ideal but poses additional logistical challenges relative to existing examples of related efforts with prevalent conditions (44). Future research on early IA should include the ERRR as a standardized adjunct to rheumatologic opinion to circumvent interstudy variability of the definition of the study population. Ultimately, one of the major advantages of harmonizing rheumatologic opinion on early IA is the standardization of early, aggressive therapy. The performance properties of the ERRR need to be investigated in this context. The utility of additional parameters and scoring algorithms to optimize the association between rheumatologic opinion and potential classification methods needs to be researched. Measures of function, inflammatory back pain, and laboratory and diagnostic imaging tests are components of neither the ERRR nor the CEAC, and their utility warrants investigation, as evident by disease-specific modifications of the Health Assessment Questionnaire (50, 51); however, any measurement tools investigated should be validated for use on the heterogeneous early IA population. Predictive models, using prevalent IA diagnoses and classification criteria for established diseases as outcomes, need to be further explored. Knowledge translation activities focused on differences between such models and the results presented here should be pursued to ensure that physicians are optimally detecting and treating early IA.
Rheumatologic opinion on early IA is variable. Despite this variability, case fulfillment of at least 3 swollen joints, a positive composite compression test for MCP or MTP joints, or at least 30 minutes of morning stiffness was found to be very strongly associated with rheumatologic classification of early IA. This association was stronger than that found with common inclusion criteria for early IA cohorts. The association with classification criteria for prevalent IA was not significant. These findings were corroborated by associations between each of these classification methods and rheumatologic recommendations for DMARD treatment. Further research into optimizing these associations with additional disease parameters, scoring algorithms, and using prevalent diagnoses and classification criteria for established diseases as outcomes is necessary.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Mr. Tavares 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. Tavares, Wells, Bell.
Acquisition of data. Scarf, Cagaanan, Bykerk, Bell.
Analysis and interpretation of data. Tavares, Wells, Bell.
The authors acknowledge the following rheumatologists for evaluating the cases: Maarten Boers, Rachelle Buchbinder, Loreto Carmona, Alfred Cividino, Les Cleland, Catalin Codreanu, John Cush, Axel Finckh, Liviu Macovei, Susanna Proudman, Sharron Sandhu, Vibeke Strand, Deborah Symmons, Andy Thompson, and 4 others who remain anonymous.