Serum IgA rheumatoid factor and pyridinoline in very early arthritis as predictors of erosion(s) at two years: A simple model of prediction from a conservatively treated community-based inception cohort




To identify, in conservatively treated, very early arthritis patients, predictors of ≥1 erosion(s) at 2 years, and to construct a prediction model.


Community-based adults (n = 310) who had never taken disease-modifying antirheumatic drugs (DMARDs) or steroids with swelling of ≥2 joints persisting for >4 weeks and lasting <6 months were recruited. Erosion status was assessed at 0, 6, 12, and 24 months; evaluations were comprised of clinical criteria (Disease Activity Score, Health Assessment Questionnaire), C-reactive protein level, erythrocyte sedimentation rate, autoantibodies, bone and cartilage markers, hand densitometry, and HLA class II shared epitopes. Patients meeting American College of Rheumatology rheumatoid arthritis (RA) criteria or with undifferentiated arthritis (UA) were followed and treated conservatively: one-third of RA patients and three-fourths of UA patients received no DMARDs during 2 years; a biologic agent was given to 1.8% of the patients during the first year. The main judgment criterion was ≥1 erosion(s) at 2 years.


At 2 years, 219 patients were assessed; 31.3% with RA and 10.6% with UA had ≥1 erosion(s). Logistic regression analysis at that time showed erosion(s) strongly associated with serum IgA rheumatoid factor (IgA-RF) and pyridinoline levels for the 190 patients with no baseline erosions, with the corresponding receiver operating characteristic curve having an area under the curve of 0.77 (95% confidence interval 0.64–0.86). A prediction model was constructed with IgA-RF thresholds of 5 and 25 IU/ml and a pyridinoline threshold of 10 nM/liter; odds ratios ranged from 1 for IgA-RF <5 IU/ml and pyridinoline <10 nM/liter to 50.75 for the association of IgA-RF ≥5 IU/ml and pyridinoline ≥10 nM/liter.


This model, using serum IgA-RF and pyridinoline concentrations, was able to predict ≥1 erosion(s) at 2 years in very early arthritis patients.


When confronted with very early arthritis, the objective, after having excluded spondylarthritis (SpA), connective tissue disease, and a viral or microcrystal arthropathy, is to determine as rapidly as possible whether erosion(s) will develop, and if so, when and how severe. When persistent and erosive, these symptoms and signs are classically called rheumatoid arthritis (RA), which is the most frequent severe chronic inflammatory arthritis, often responsible for severe disability and deterioration of quality of life (1).

It is currently thought that to control disease activity and slow or even stop the progression of structural joint involvement, disease-modifying antirheumatic drugs (DMARDs) must be started very early (2–5). Under such treatment, prolonged benefit was obtained, particularly in preventing structural joint damage (6, 7). Several therapeutic modalities have proven efficacy: notably, the combinations of DMARDs and systemic glucocorticoids (6, 8), and more recently, biologics alone or combined with methotrexate (9–11). However, all of these treatments have side effects, sometimes severe, and the biotherapies are very expensive. Therefore, for very early arthritis, it is important to identify rapidly the factors predictive of erosion(s), thereby assuring that aggressive therapy is given only to patients requiring it, with milder treatments being prescribed to patients whose signs are considered benign, i.e., with no erosions and therefore no severe functional joint damage.

Few studies have attempted to identify factors predictive of erosion(s) in patients with early arthritis (12, 13). The authors of the Leiden study (12) constructed a model that discriminated well between patients with persistent arthritis and ≥1 erosion(s) and those with no erosions at 2 years of followup. To the best of our knowledge, no study has evaluated the value of bone and/or cartilage markers to predict erosion(s) in very early arthritis.

This study was conducted on the very early arthritis inception cohort recruited from the general population, comprised of patients with recent-onset inflammatory arthritis, untreated before inclusion, then followed and conservatively treated for 2 years. Its objectives were to identify baseline factors predictive of ≥1 erosion(s) at 2 years and to attempt to construct a simple to use prediction model for routine practice.



A total of 310 patients with early arthritis were prospectively recruited between October 1998 and January 2002 in 2 French regions, i.e., the entire Upper Normandy province (1,800,000 inhabitants) and the metropolitan Amiens area (300,000 inhabitants). All of the private rheumatologists and those in the 5 regional hospitals (Amiens, Elbeuf, Evreux, Le Havre, and Rouen) participated in this project, along with most general practitioners. All of these physicians were encouraged to notify and refer all patients with arthritis to one of these hospital clinics for assessment. To recruit a maximum number of patients and thereby obtain a representative sample of this region, an annual extensive 2-week information campaign (regional newspapers, radio, and TV) was conducted. Information meetings were organized and documentation and reminder notices were sent to pharmacies and rheumatologists', general practitioners', and physiotherapists' offices. Physicians referred patients by using a very simple printed notification form.

Inclusion criteria consisted of the following: man or woman age ≥18 years, ≥2 swollen joints confirmed by a rheumatologist, swelling persisting for >4 weeks, symptoms lasting <6 months, no previous glucocorticoid prescription (only 1 intraarticular injection >1 month before or oral prednisone [<10 mg] for 1 week >2 weeks before enrollment were tolerated) or DMARDs, no inflammatory back pain, no foreseen move during the 10 next years, and speaking fluent French. All of the hospital clinics were organized to examine eligible patients within a few days, especially to ensure that the very active arthritis patients had never taken glucocorticoids.

This study was approved by the Upper Normandy Ethics Committee (French law 88-1138; 20 December 1988; file 95/138/HP). All of the patients gave their informed written consent at the time of inclusion.

Baseline assessment.

A standard evaluation was performed during the inclusion visit. The data collected consisted of: age, sex, familial history of arthritis, duration of symptoms, duration of morning stiffness, intensity of joint pain, Ritchie articular index, and number of swollen joints of a total 44. Joint symptom duration took into account the first period during which arthritis had persisted for at least 1 month. Pain intensity was assessed using a 100-mm visual analog scale. To evaluate disease activity, we calculated the Disease Activity Score (DAS) comprising 3 variables (Ritchie articular index, number of swollen joints of a total 44, and erythrocyte sedimentation rate [ESR; mm/hour]). American College of Rheumatology (ACR; formerly the American Rheumatism Association) criteria (14), Health Assessment Questionnaire score, ESR, and C-reactive protein (CRP) concentration were also recorded. The following autoantibodies were sought: rheumatoid factors (RFs) detected by agglutination (latex and Rose-Waaler) tests and homemade enzyme-linked immunosorbent assay (ELISA; IgM, IgA, and IgG isotypes), antikeratin, antiperinuclear factor, anti–cyclic citrullinated peptide 2 (anti–CCP-2; Euroimmun), anti–α-enolase, anti–glucose-6-phosphate isomerase, anti–annexin V, and antibodies to the 27 C-terminal amino acids of calpastatin. HLA–DR4 and/or HLA–DR1 and shared epitopes associated with a predisposition for RA were determined. All of these data, obtained once at baseline, were the only factors assessed as potential predictors of erosion(s).

Wrist, hand, and foot radiographs were obtained according to our protocol. Bone status was assessed by hand bone mineral density dual x-ray absorptiometry of both hands from the wrist joint to all distal bones using the same densitometer (Hologic) in all centers, and serum bone resorption markers derived from collagen crosslinks. Morning fasting blood samples were drawn without any dietary restrictions. Serum bone resorption markers were measured by ELISA (Nordic Bioscience Diagnostics) for serum C-terminal crosslinking telopeptide of type I collagen (CTX-I) and by high-performance liquid chromatography (HPLC) after blood hydrolysis for pyridinoline and deoxypyridinoline. The reliability of those HPLC measurements was previously reported (15); in this study, their reproducibility rates, expressed as coefficients of variation, ranged between 8.5% and 12% and between 12% and 16% for pyridinoline and deoxypyridinoline, respectively, with a minimum detection limit of 0.4 nM/liter for both molecules. The YKL-40 concentration, serving as a cartilage marker, was also assessed by ELISA (Nordic Bioscience Diagnostics).

All biologic parameter determinations were centralized, with a given laboratory being responsible for a given parameter, assayed with the same technique.

Criterion of judgment at 2 years.

The primary end point was the presence of ≥1 unequivocal erosion(s) at 2 years in patients with no erosion at inclusion. Hand and foot radiographs were obtained according to our standardized, rigorous procedure taught to radiologists in the 5 evaluation centers during a training session. Each anteroposterior, frontal, hand, wrist, and foot radiograph was obtained separately. The procedure relied on precise positioning of the patient, centering of the x-ray beam, and exposure of appropriate films. The constants were the same for each series of images from a given patient. Quality was systematically controlled and any film not meeting those standards was retaken.

Films were centralized and read independently by 2 highly experienced rheumatologists (PF and OM), blinded to the patient's identity and diagnosis but not film chronological order (16). In the case of disagreement, consensus was reached immediately. Analysis concerned erosion(s), joint space narrowing, and the modified total Sharp/van der Heijde score (17). A score of ≥1 was chosen a priori as the judgment criterion at 2 years and expert agreement on the unequivocal presence of ≥1 erosion(s) on any hand or foot joint. The interreader reproducibility for the presence of erosion(s) (yes/no) was estimated with Cohen's kappa (0.94, 95% confidence interval [95% CI] 0.86–1.00), and for the Sharp erosion score was estimated with the intraclass correlation coefficient (0.92, 95% CI 0.89–0.94) and the mean ± SD (0.04 ± 0.9) of the difference between the 2 readers' measures.

Therapeutic recommendations.

For the first 2 years, treating rheumatologists were given recommendations so that the included patients would be treated homogeneously. The guidelines had been devised before early intensive DMARD administration became the internationally accepted strategy and prior to biotherapy availability in France. Schematically, it was recommended not to use systemic glucocorticoids unless necessitated by very active disease, and then briefly at the lowest possible dose. For DMARDs, it was recommended to start with hydroxychloroquine (6 mg/kg/day), to be replaced or combined with oral methotrexate, starting at 7.5 mg/week.

Followup evaluations.

All of the patients were followed for at least 2 years, except for those with definitively (non-RA) diagnosed arthritis, such as systemic lupus erythematosus, SpA, etc., who were withdrawn from the study as soon as the entity was recognized. All other patients who subsequently fulfilled the ACR RA criteria (14) or had undifferentiated arthritis (UA) were followed, even if their disease was in spontaneously or posttreatment-induced remission.

Followup visits were conducted 3 and 6 months after inclusion and then every 6 months until 2 years. At each visit, all of the clinical and biologic parameters described above (except genetic tests) and ongoing treatment(s) were again noted by a well-trained rheumatologist (XLL, PB, SP). An earlier pilot study had been conducted to ensure the good reproducibility of joint counts among rheumatologists in the different evaluation centers. Hand and foot radiographs were obtained at inclusion and at 6, 12, and 24 months (except during pregnancy); they were scored as described above.

Statistical analyses.

Continuous variables such as age, clinical parameters, biologic inflammation parameters, bone densitometry measurements, and modified Sharp scores are expressed as the median (range) and were subjected to univariate analysis with the nonparametric Mann-Whitney U test. Comparisons of categorical variables, e.g., sex, each ACR classification criterion, or continuous variables with known positivity thresholds (autoantibodies and vascular, bone, cartilage, and genetic markers), using Fisher's exact test are reported as nonadjusted odds ratios (ORs [95% CI]).

Variables with P values less than or equal to 0.15 were entered into the step-by-step logistic regression multivariate analysis to identify independent parameters associated with ≥1 erosion(s). To test the significance of one variable x, the likelihood ratio statistic was calculated by comparing the log likelihood for the full model containing all of the variables with the log likelihood for the reduced model containing all of the variables except x. Receiver operating characteristic (ROC) curves were drawn to evaluate, by the area under the curve (AUC) and its 95% CI, the abilities of these parameters to predict erosion(s) as a function of chosen combined thresholds.

Statistical analyses were run using NCSS, version 2004 (NCSS Corporation), and StatXact software, version 4 (Cytel Software Corporation). All tests were 2-tailed and statistical significance was defined as P values less than or equal to 0.05.


Patient characteristics.

Baseline characteristics of the 310 patients included in the study are shown in Table 1. Among them, 1 (0.3%) died before completing followup, 29 (9.4%) declined to continue, and 1 (0.3%) was lost to followup (moved). In accordance with the protocol, 60 patients were withdrawn between inclusion and year 2, when they were classified as having one of the following clearly differentiated arthritides other than RA: 14 had hand osteoarthritis, 11 had psoriatic arthritis, 9 had peripheral SpA, 8 had connective tissue disease, 3 had gout, 2 had chondrocalcinosis, 2 had sarcoidosis, 2 had parvovirus arthritis, 2 had reactive arthritis, and 1 each had paraneoplastic arthritis, Behçet's syndrome, eosinophilic fasciitis, hydroxyapatite deposition, Lyme disease, algodystrophy, or Wegener's granulomatosis.

Table 1. Baseline characteristics of the patients with very early arthritis recruited for this study according to their erosion status*
 All patients (n = 310)Patients without erosions (n = 258)
  • *

    Values are the median (range) unless otherwise indicated. RF = rheumatoid factor measured by latex agglutination (positivity ≥20 IU); anti–CCP-2 = anti–cyclic citrullinated peptide 2; DAS = Disease Activity Score with 3 variables; HAQ = Health Assessment Questionnaire; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein.

Age, years52 (19–84)51 (19–84)
Women, no. (%)211 (68.1)184 (71.3)
Symptom duration, months4.2 (0.9–6)4.1 (0.9–6)
No. of painful joints6 (0–58)6 (0–58)
No. of swollen joints7 (2–37)7 (2–37)
Erosions on hands and/or feet, no. (%)52 (16.8)0 (0)
RF positive, no. (%)70 (22.6)58 (22.5)
Anti–CCP-2 positive, no. (%)72 (23.2)58 (22.5)
DAS2.95 (0.45–7.53)2.93 (0.45–7.53)
HAQ score0.75 (0–2.88)0.75 (0–2.75)
ESR, mm/hour18 (1–110)16 (2–110)
CRP level, mg/liter7 (5–206)6 (5–206)
DR4 or DR1 carriers, no. (%)150 (48.4)120 (46.5)
DR4/DR4 or DR1/DR1 or DR4/DR1, no. (%)20 (6.5)12 (4.7)
≥1 shared epitope, no. (%)118 (38.1)94 (36.4)
2 shared epitopes, no. (%)11 (3.5)8 (3.1)

The distribution of arthritides and their erosion status at inclusion and 1 and 2 years of followup are reported in Table 2. At 2 years, 57 of the 219 patients had ≥1 erosion, among whom 190 had been erosion free at inclusion. At 2 years, 31.3% of the patients classified as having RA according to ACR criteria and 10.6% of the UA patients had ≥1 erosion(s). No erosion repair was observed. At 2 years, 42 (19.2%) of the 219 patients had achieved remission, defined as a DAS <1.6.

Table 2. Distributions of the classification and erosion status of the 219 very early arthritis patients at inclusion and at 1 and 2 years of followup*
Arthritis diagnosedInclusion1 year2 years
  • *

    Values are the number (percentage) of patients, taking into account death, declining to continue, lost to followup, and definitive diagnosis, unless otherwise indicated. RA = rheumatoid arthritis; ACR = American College of Rheumatology.

  • At inclusion, 29 of 219 patients had erosion(s) and 190 of 219 patients had no erosions.

RA (ACR criteria)130 (80.2)32 (19.8)162138 (78.9)37 (21.1)175110 (68.8)50 (31.3)160
 No. of patients without erosion(s) at inclusion   138914711025135
Undifferentiated arthritis99 (92.5)8 (7.5)10754 (88.5)7 (11.5)6142 (89.4)5 (10.6)47
 No. of patients without erosion(s) at inclusion   5425642244
Classified (other than RA)29 (70.7)12 (29.3)4114 (77.8)4 (22.2)1810 (83.3)2 (16.7)12
 No. of patients without erosion(s) at inclusion   1401410111
Total258 (83.2)52 (16.8)310206 (81.1)48 (18.9)254162 (74.0)57 (26.0)219
 No. of patients without erosion(s) at inclusion   2061121716228190

Treatment during the 2 years of followup.

The distributions of DMARDs and prednisone prescribed to the 219 patients (160 RA, 47 UA, and 12 classified other than RA) and their doses during the 2 years of followup were carefully collected (data not shown). More than two-thirds of the patients classified as having RA and >95% of those with UA received no glucocorticoids during followup. No DMARDs had been prescribed to approximately one-third of RA patients and three-fourths of UA patients during this period. In accordance with the initial therapeutic recommendations of the protocol, the DMARDs prescribed at 6 months to RA and UA patients consisted of: hydroxychloroquine for 32.3% and 10.2% and methotrexate for 22.3% and 14.2%, respectively. Anti–tumor necrosis factor α (anti-TNFα) was never used during the first 6 months of followup, being prescribed to only 8 RA patients (4.7%) and 1 UA patient (2.0%) after 2 years of disease progression. During the first 6 months, 164 (74.9%) of 219 patients had received no glucocorticoids and 101 (46.1%) of 219 took no DMARDs.

Univariate analysis identification of factors predictive of ≥1 erosion(s) at 2 years (n = 190).

The abilities of baseline continuous and categorical parameters to predict erosion(s) in the 190 erosion-free patients at inclusion are reported in Tables 3 and 4. The only continuous clinical variable associated with erosion appearance was morning stiffness. Male sex and morning stiffness lasting ≥1 hour, RF, and radiologic changes (ACR criteria 1, 6, and 7, respectively) were significantly associated with erosion(s). The number of ACR criteria (Table 3) met was significantly associated with erosion(s) (P = 0.0006). Many biologic parameters were significantly associated with erosion(s): inflammatory markers ESR and CRP level (P = 0.003 and 0.02, respectively); RFs, regardless of the method used to detect them; all autoantibodies to filaggrin family proteins (antikeratin and antiperinuclear factor) or citrullinated peptides (anti–CCP-2; P = 0.003); and the bone and cartilage markers: pyridinoline (P = 0.0004) but not deoxypyridinoline and serum CTX-I. Genetic markers were not associated with erosion(s).

Table 3. Erosion-free patients at baseline (n = 190): univariate analysis of continuous and discrete clinical and laboratory variables associated with nonerosive or erosive arthritis at 2 years*
 Nonerosive (n = 162)Erosive (n = 28)P
  • *

    Values are the median (range). VAS = visual analog scale; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; DAS = Disease Activity Score with 3 variables; ACR criteria = 1987 classification criteria of the American College of Rheumatology (formerly the American Rheumatism Association); HAQ = Health Assessment Questionnaire; BMC = bone mineral content; BMD = bone mineral density.

Age, years52 (19–83)54.5 (24–75)0.22
Symptom duration, months4 (0.9–6)3.8 (1.3–5.9)0.68
Morning stiffness, minutes60 (0 to ∞)120 (0–240)0.04
Pain VAS, mm (range 0–100)40 (0–96)40 (10–100)0.40
Ritchie articular index6 (0–48)6.5 (0–58)0.45
No. of swollen joints of a total 447 (2–37)8 (2–34)0.23
ESR, mm/hour15 (2–110)28 (6–110)0.003
CRP level, mg/liter6 (5–185)15.5 (5–206)0.02
DAS3 (0.6–7.5)3.3 (1.1–7.5)0.12
No. of ACR criteria4 (0–6)5 (0–7)0.0006
HAQ score (range 0–3)0.75 (0–2.75)0.94 (0–2.63)0.21
Dominant hand BMC, gm/cm225.2 (12.4–46.8)26.9 (18.3–42.3)0.46
Dominant hand BMD, gm/cm20.37 (0.25–0.49)0.37 (0.27–0.43)0.62
Total modified Sharp score2 (0–25)2 (0–15)0.61
Table 4. Erosion-free patients at baseline (n = 190): univariate analysis of categorical clinical and laboratory variables associated with a risk of erosion(s) at 2 years*
 OR (95% CI)P
  • *

    OR = odds ratio; 95% CI = 95% confidence interval; ACR = American College of Rheumatology; AU = arbitrary units; anti–CCP-2 = anti–cyclic citrullinated peptide 2; ACAST-C27 = autoantibodies directed against the 27 C-terminal amino acids of calpastatin; ACAST-D1 = autobodies directed against domain 1 of calpastatin; anti-G6PI = anti–glucose-6-phosphate isomerase; CTX-I = C-terminal crosslinking telopeptide of type I collagen.

Male sex2.60 (1.02–6.46)0.03
ACR criterion  
 12.75 (1.01–8.70)0.04
 21.57 (0.54–5.63)0.48
 32.17 (0.30–96.4)0.70
 41.05 (0.32–4.54)1
 55.93 (0.07–4.68)0.28
 63.89 (1.48–9.91)0.002
 73.89 (1.25–11.25)0.009
Rheumatoid factor  
 IgM by latex test, IU/liter (≥20)4.01 (1.60–10.01)0.001
 IgM by Rose-Waaler test, IU/liter (≥12)4.96 (1.93–12.58)0.0003
 IgM isotype, AU/liter (≥16)3.74 (1.48–9.31)0.002
 IgA isotype, AU/liter  
  5 ≤ IgA < 254.35 (1.43–13.26)0.01
  IgA ≥2510.63 (3.38–33.42)0.0001
 IgG isotype, AU/liter (≥53)4.83 (1.93–12.36)0.0002
 Antikeratin, IU/liter (0/1)3.51 (1.26–9.25)0.007
 Antiperinuclear factor (0/1)3.13 (1.25–7.73)0.007
 Anti–CCP-2, AU/liter (≥5)3.47 (1.38–8.62)0.003
 ACAST-C27, AU/liter (≥27)1.37 (0.23–5.48)0.71
 ACAST-D1, AU/liter (≥12.5)0.88 (0.09–4.23)1
 Anti–annexin V, AU/liter (≥25)0.57 (0.10–2.08)0.43
 Anti-G6PI, AU/liter (22.9)0.69 (0.12–2.82)0.76
 Anti–α-enolase, AU/liter (≥20)0.84 (0.23–2.51)0.81
Pyridinoline, nM/liter (≥10)6.77 (2.14–20.61)0.0004
Deoxypyridinoline, nM/liter (≥1.2)2.61 (0.55–9.97)0.12
Serum CTX-I, mg/liter (≥22.56)1.88 (0.76–4.88)0.15
YKL-40, mg/liter (≥69)1.33 (0.55–3.35)0.54
DR4 or DR11.92 (0.78–4.93)0.15
DR4 or DR1/DR4 or DR12.87 (0.45–13.17)0.15
Shared epitope(s)  
 ≥12.61 (0.55–9.97)0.68
 21.99 (0.19–11.86)0.34

Multivariate analysis retention of factors predictive of ≥1 erosion(s) at 2 years (n = 190).

At 2 years, logistic regression analysis was performed on the continuous and categorical parameters of the 190 patients without erosion at inclusion that had achieved a P value ≤0.15 in the univariate analysis. The multivariate analysis showed that erosion(s) was strongly associated with pyridinoline and IgA-RF concentrations. An ROC curve was drawn for these parameters (AUC 0.77, 95% CI 0.64–0.86), with a slightly but not significantly different (P = 0.054) curve obtained by substituting latex test–determined IgM-RF for IgA-RF (AUC 0.72, 95% CI 0.57–0.82) (Figure 1).

Figure 1.

Receiver operating characteristic curves. Rheumatoid factor (RF) levels (enzyme-linked immunosorbent assay–determined IgA [1] and latex test–determined IgM [2]) and pyridinoline concentrations retained by logistic regression analysis as predicting erosion or nonerosion status at 2 years for 190 patients without erosions at baseline. The curves plot the relationship between sensitivity and 1 − specificity for different cutoff levels of IgA-RF and pyridinoline combinations given by the stepwise logistic regression.

Erosion prediction model.

To construct an erosion prediction simple model for routine use, 2 IgA-RF thresholds (5 and 25 IU/ml) and 1 pyridinoline threshold (10 nM/liter) were selected (Table 5). The significance levels of the likelihood ratios were P = 5 × 10−3 for the paired IgA-RF and pyridinoline levels and P = 0.20 for each of the other variables. The OR ranged from 4.61 for low IgA-RF and pyridinoline thresholds to 50.75 for very high thresholds.

Table 5. Model predicting erosion(s) at 2 years of very early erosion-free arthritis according to different IgA-RF levels and pyridinoline concentrations*
 Regression coefficient (β)SEPOR (95% CI)Positive predictive value, % (ratio)
  • *

    IgA-RF = IgA rheumatoid factor; OR = odds ratio; 95% CI = 95% confidence interval.

  • Number of patients with erosion(s)/all patients with or without erosion(s) at the given level.

  • The 2 subgroups are combined to obtain a sufficient number of patients.

IgA-RF <5 IU/ml and pyridinoline <10 nM/liter   13.3 (3/90)
5 ≤ IgA-RF < 25 IU/ml and pyridinoline <10 nM/liter1.530.710.034.61 (1.14–17.71)13.7 (7/51)
IgA-RF <5 IU/ml and pyridinoline ≥10 nM/liter2.110.990.038.29 (1.18–58.11)22.2 (2/9)
IgA-RF ≥25 IU/ml and pyridinoline <10 nM/liter2.570.710.000313.05 (3.24–52.61)31.0 (9/29)
5 ≤ IgA-RF < 25 IU/ml and IgA-RF ≥25 IU/ml and pyridinoline ≥10 nM/liter3.930.865 × 10−650.75 (9.43–273.22)63.6 (7/11)


The primary objective of this study was to test the ability of baseline clinical and laboratory parameters, including numerous autoantibodies and cartilage and bone markers, to predict ≥1 unequivocal erosion(s) at 2 years in patients with very early arthritis recruited for an inception community-based cohort.

Taking these factors into account makes it possible to adapt initial treatment, which markedly influences structural joint outcome (6, 7). Based on recruitment methods, this study population was representative of very early arthritis seen in the primary care setting in Northwestern France. Patients with severe disease were not excluded because short-term glucocorticoid use was tolerated for highly active disease, if necessary, and because the arthritis onset inclusion interval was very short. Therefore, all but a few patients had never taken glucocorticoids and none had ever taken DMARDs, to avoid any impact on certain initial parameters. All of the patients were subsequently treated homogeneously in a nonaggressive manner, in accordance with protocol recommendations and the therapeutic approach practiced in France at the time the protocol was written. Therefore, none of the cohort patients received a biotherapy during the first 6 months of followup and only 1.8% diagnosed with RA were administered a biologic agent during the first year.

In our experience, monarticular arthritis does not pose the same diagnostic difficulties as oligo- or polyarthritis and was excluded a priori; its progression toward chronic arthritis, such as RA, is unusual (12). We selected 2 years of followup because it represents the minimal time required to draw any conclusion about the likelihood of erosion occurrence (18, 19).

When this article was written, it was routine practice to initiate very early a systemic DMARD/glucocorticoid regimen for early active RA, considering that disease activity would be better controlled in the short term and long-lasting structural joint benefit would be obtained (6, 7, 20). Early strong inhibition of active RA, using prednisone combined with DMARD(s) or a TNFα blocker and methotrexate, has been advocated to capitalize on therapeutic impact during this critical period (5, 6, 10). However, that approach is probably not adapted to all RA and, a fortiori, to all very early arthritides, as long as we do not know whether or not erosions will develop. Therefore, to avoid overtreatment of patients not requiring such intensive therapy, it has become essential to identify markers predictive of that structural joint damage (21). Therefore, 69% of our patients diagnosed with RA and 89% of those with UA had no erosions at 2 years of followup. This finding seems even more pertinent because it was obtained for conservatively treated patients, i.e., in a setting closer to the natural history of very early arthritis. Considering the RA subgroup, the erosion frequency was lower than that usually observed in RA cohorts (19), probably reflecting our population-based recruitment.

Numerous clinical and biologic parameters were assessed as predictors for erosion(s) at 2 years, along with a panel of autoantibodies, some of which had not been tested previously in this context, i.e., serum bone and cartilage markers, genetic traits, and bone mineral density of the hands.

Many investigators have attempted to identify markers predictive of structural joint damage in RA, but only a few involved early arthritis (19, 22–29). In RA, predictive factors retained by multivariate analyses varied from one study to another. Our judgment criterion differed from those used previously, e.g., progression of either radiologic severity and/or structural joint involvement, which are prognostic and not diagnostic criteria. RF, in conjunction with initial radiologic involvement, was the early factor most frequently identified. An elevated CRP concentration is usually predictive (24, 27, 29), even more often than the initial ESR (23). Even so, no isolated criterion or composite index is able to predict, at disease onset, the severity or progression of structural joint involvement in a given RA patient (12).

Few cohort studies were devoted to structural joint involvement in patients with early arthritis recruited prospectively. Compilers of the Norfolk Arthritis Register wondered if it is indeed possible to predict the transformation of UA into RA (13). The inclusion criteria and recruitment modalities that we used are close to those applied for that Register. However, because initial radiographs were lacking, the analyses of that English cohort were based on factors associated with the severity of structural joint involvement, i.e., a prognostic criterion, not the presence of ≥1 erosion(s) at 2 years. Visser et al constructed a model for persistent erosive arthritis at 2 years (12). Some of their inclusion criteria differed from ours: monarticular arthritides were accepted and the joint symptoms could have lasted for as long as 2 years. Despite these differences, the baseline erosive arthritis rate for their cohort was comparable (15% versus 17% for ours). The Leiden model consisted of 7 variables: symptom duration, morning stiffness lasting ≥1 hour, arthritis affecting >3 joints, squeeze test, RF positivity, anti–CPP-1 positivity, and erosions. However, their model applied to our cohort, replacing the squeeze test by ≥1 painful metatarsophalangeal joint, was inoperative, whether all of the patients or only those whose arthritis persisted at 2 years were considered (data not shown). Their model's inefficacy is probably attributable to the different inclusion criteria used (e.g., exclusion of monarticular arthritis from our cohort), but also to the different therapeutic modalities prescribed to the 2 cohorts. In addition, our sole objective was to identify factors predictive of erosion(s), not arthritis persistence, at 2 years.

Based on the characteristics of our conservatively treated, community-based cohort, our multivariate analysis was able to identify a strong association between 2 independent variables, serum IgA-RF and pyridinoline concentrations, and ≥1 unequivocal erosion(s) at 2 years. Consequently, we were able to construct a simple model using 2 readily available parameters: ELISA-determined IgA-RF level and serum pyridinoline concentration. IgA-RF can be replaced by ELISA-determined IgM-RF or IgG-RF (data not shown) or latex test–detected IgM-RF level without significantly modifying the results. Obviously, the positive predictive value does not reach 100%. According to the ROC curve, the false-negative rate was approximately 25%, in which case, the patient could risk being undertreated. Indeed, this model should serve only as a guide, indicating “risk factors” of developing ≥1 unequivocal erosion(s) within 2 years. It does not eliminate the need to monitor repeatedly functional joint status during the first 2 years of followup (21).

The fact that RF is predictive of erosion was expected because it was included as the ELISA-determined IgM-RF level in the Leiden model for persistent arthritis (12). In contrast to their observations, our multivariate analysis retained the IgA-RF isotype but not autoantibodies to CCP-2 as being predictive of erosion. In our study, anti–CCP-2 was effectively linked to erosion in univariate analysis (P = 0.003). But anti–CCP-2 also being strongly linked to IgA-RF (P < 10−6) in univariate analysis did not persist in multivariate analysis. It cannot be formally excluded that this discrepancy reflects the antibody isotype detected because the Leiden group sought only IgM-RF and used the first-generation anti-CCP detection kit, whereas we looked for all RF isotypes and used a second-generation test for CCP.

We included several markers of bone, cartilage, and synovium anabolism and catabolism (15, 30): pyridinoline, a molecule bridging collagen types I and II, is a marker of bone, cartilage, and synovium degradation; deoxypyridinoline indicates bone degradation; and serum CTX-I is a marker of bone and synovium degeneration, whereas cartilage and synovium syntheses are identified by YKL-40 (glycoprotein-39). Serum pyridinoline was the only marker independently predictive of ≥1 erosion(s) that, to the best of our knowledge, has never been reported previously. Studies designed to identify relationship(s) between erosion(s) and serum bone, cartilage, and/or synovium factors have been scarce. They generally concerned patients with proven RA and the potential link between those factors and the severity of structural joint involvement. Lindqvist et al used logistic regression analysis of the characteristics of patients with RA of a duration of <2 years to show that ESR, IgA-RF, anti-CCP, and anti–interleukin-1 receptor antagonists were associated with the severity of structural damage at 5 years (31). A satellite study of the COBRA (Combinatietherapie Bij Reumatoïde Artritis) therapeutic trial on RA evolving for <2 years showed that high baseline urinary CTX-I and CTX-II levels independently predicted a higher risk of structural damage progression over 4 years (32). When patients were grouped according to the presence or absence of initial joint damage, those molecules were predictive only for those without baseline structural joint involvement.

Several other studies examined the potential relationships between certain bone, cartilage, or synovium markers and RA activity: concentrations of urinary CTX-II and, to a lesser extent, CTX-I, serum matrix metalloproteinase 3, and pyridinoline varied according to disease activity (33, 34). Jansen et al (35) studied homogeneously treated patients with peripheral arthritis affecting ≥2 joints and a duration of symptoms of <2 years, and found that higher serum CTX-I levels were associated with radiographic progression at 2 years, but less predictive of factors already being used. To our knowledge, no study has determined the abilities of bone and cartilage markers to predict erosion(s) in an early arthritis cohort.

To conclude, 2 easily obtainable parameters, serum IgA-RF and pyridinoline levels, constitute risk factors for erosion(s). When their levels were simultaneously elevated, their OR for ≥1 erosion(s) at 2 years reached 51. It is now necessary to validate this prediction model on other population cohorts.


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. Dr. Le Loët 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. Le Loët, Menard.

Acquisition of data. Le Loët, Brazier, Mejjad, Boumier, Daragon, Gayet, Pouplin, Tron, Zarnitsky, Vittecoq, Menard, Fardellone.

Analysis and interpretation of data. Le Loët, Menard, Fardellone.


The authors thank Janet Jacobson for editorial assistance in the preparation of the manuscript and Sandrine Parisse for typing it.