• Rheumatoid arthritis;
  • Diagnosis;
  • Immunology


  1. Top of page
  2. Abstract
  7. Acknowledgements


To determine which laboratory test or tests at presentation best predicted a diagnosis of rheumatoid arthritis (RA) 2 years later.


Two hundred seventy patients with early arthritis seen in 7 hospitals underwent comprehensive evaluations at 6-month intervals for 2 years, when the diagnosis of RA was assessed by 5 rheumatologists. The sensitivity and specificity of each test at the first visit for discriminating between RA (38%, n = 98) and non–RA patients were determined. Optimal cutoffs for continuous tests were derived from receiver operating characteristic curves. Sensitivity and specificity of test combinations selected by multiple logistic regression were determined.


IgM rheumatoid factor (RF) by enzyme-linked immunosorbent assay, IgG-antikeratin antibody (AKA), and latex test had the strongest associations with RA. These 3 tests formed the most powerful combination for distinguishing RA from non–RA.


IgM-RF, IgG-AKA, and the latex test are the best laboratory tests for discriminating between patients with and without RA. Combining these tests slightly improves diagnostic value.


  1. Top of page
  2. Abstract
  7. Acknowledgements

Early diagnosis of rheumatoid arthritis (RA) is essential, because there is strong evidence that early treatment with one or more disease-modifying antirheumatic drugs (DMARDs) improves the outcome (1–3). Moreover, the recent introduction of anti-tumor necrosis factor (TNF) agents makes this an opportune time for evaluating laboratory tests used to diagnose RA.

The diagnosis of rheumatoid arthritis rests on clinical grounds. The only laboratory test used in standard practice is determination of serum rheumatoid factors (RFs), which are nonspecific (4–6). However, RA is associated with several autoantibodies, some of which may be sufficiently specific to assist in the diagnosis of the disease (7). A test or test combination capable of determining which early arthritis patients have RA would be of great help to clinicians, biologists, and patients. We looked for such a test by examining data from a longitudinal cohort study of patients with recent-onset arthritis. In a cohort of 270 patients with early arthritis, we sought to determine which laboratory test or combination of laboratory tests used at study inclusion best predicted a diagnosis of RA 2 years later. Because laboratory tests may have less diagnostic value when performed routinely than when done as part of a battery of tests for a clinical study, we based our study on tests performed routinely at our laboratory.


  1. Top of page
  2. Abstract
  7. Acknowledgements

Study population

The study cohort comprised patients first evaluated for early arthritis between 1995 and 1997 in 7 hospitals in Brittany (France). All patients were referred by general practitioners or rheumatologists who had been informed of the study. Inclusion criteria were as follows: age 16 years or older, swelling of at least 1 joint, absence of a previous diagnosis of joint disease, and disease duration of no more than 1 year. The study was approved by the appropriate ethics committee, and all patients gave their written informed consent.

Baseline assessment

All patients had a standardized interview; a general physical examination; and a rheumatologic examination during which information on more than 100 parameters was collected, including present and past medical history, family history of RA or spondylarthropathy, synovitis, and American College of Rheumatology (ACR; formerly American Rheumatism Association) RA classification criteria (6). Radiographs of the chest, pelvis, hands, and feet were obtained. The following laboratory tests were performed: blood counts; serum creatinine; proteinuria; serum C-reactive protein; a latex test and enzyme-linked immunosorbent assays (ELISAs) for IgM-, IgG-, and IgA-rheumatoid factor (RFs); tests for antiperinuclear factor (APF), antikeratin antibody (AKA), anti–RA-33 antibody, antinuclear antibody (ANA), and anti-DNA antibody; ELISA for glycosylated IgG; circulating immune complexes; IgG1, IgG2, IgG3, and IgG4; and HLA–AB DR type.


Each patient was followed by an office-based rheumatologist at 6-month intervals until a clinical diagnosis of a specific joint disease was made, with fulfillment of published criteria for that joint disease. Each evaluation included a standardized interview; a general physical examination; a rheumatologic examination including evaluation of ACR criteria for RA; radiographs of the hands and feet; tests for RF (latex and ELISAs); and tests for APF, AKA, and ANA. Patients were asked to attend a final visit between June and November 1999. After this last visit, a panel of 5 rheumatologists determined whether RA was present, based on all the data available for each patient (8). This final diagnosis was used as the gold standard against which the diagnostic value of laboratory tests at the first visit was evaluated.

Laboratory methods

RF was measured using the latex test and in-house ELISAs specific for the IgG, IgM, and IgA isotypes. The latex test was performed using Rhumalatex (Fumouse, Asnière, France), and results were expressed in international units (IU) per milliliter. The in-house ELISAs used a rabbit IgG Fc as the antigen (Sigma, St. Louis, MO) and isotype-specific goat F(ab′)2 antibodies coupled to alkaline phosphatase (Jackson, Avondale, PA); results were expressed as the optical density (OD).

Although RFs are a classic marker for RA, they are now widely recognized as products of ongoing normal immune responses. It remains unclear whether distinct forms of RF exist, one abnormal and the other normal. However, in vitro studies of affinity suggest that RFs found in RA patients may differ from those of normal individuals. In vivo, the polyclonality of RFs in RA makes it difficult to characterize their qualitative properties, such as specificity or avidity, which are probably important determinants of pathogenicity. However, a reproducible method for measuring the avidity index of RFs has been developed and used to evaluate the avidity of RFs, or functional affinity (FA), in subjects with RA and in controls. In our study, the FA of IgM-RF in sera yielding a positive ELISA for this isotype was determined using a procedure described elsewhere (9). Briefly, the chaotropic agent diethylamine was used in an ELISA. Samples were diluted serially on plates (Nunc, Naperville, IL) coated with FC fragments of human IgG (Jackson), with or without 10 mM diethylamine (Sigma). The plates were incubated at 37°C for 90 minutes then washed 3 times with phosphate-buffered saline (PBS) containing 0.05% Tween 20. Horseradish peroxidase-conjugated F(ab′)2 antihuman IgM (Jackson) was added. After incubation at 37°C for 60 minutes followed by 3 washings, color was developed with H2O2 and o-phenylenediamine, and absorption was measured at 492 nm on a Titertek Multiskan (Flow, McLean, VA). Dose-response curves were plotted, and the fall in log-titer was taken as an estimate of FA. Thus, high FA interactions between RF and IgG produced a low inhibition index, and low FA interactions produced a high inhibition index.

Antiperinuclear factor was assayed as described previously in detail (10). In brief, a wooden tongue depressor was used to scrape off buccal mucosal cells from the inside of both cheeks of healthy volunteers. The cells were washed 3 times in PBS, pH 7.4, resuspended in PBS containing colimycin and sodium azide, and transferred dropwise on multispot slides, with approximately 5,000 cells per well. Titers of 1:80 or more were considered significant.

IgG, IgM, and IgA AKA were detected by indirect immunofluorescence using a section of the middle third of a rat esophagus as the substrate. Titers of 1:10 or more were considered significant.

Anti–RA-33 antibody was detected using a nuclear extract containing 7–10 mg/ml of protein, prepared from freshly grown HeLa cells as previously reported (11). The nuclear extract was subjected to 12% sodium-dodecyl-sulfate polyacrylamide gel electrophoresis as described by Laemmli (12), and the antigen was transferred to nitrocellulose. The nitrocellulose strips were incubated for 120 minutes with sera diluted 1:25 in incubation buffer, with constant shaking. Bound autoantibodies were detected using goat antihuman antibodies coupled to alkaline phosphatase.

ANAs were sought using a standard immunofluorescence test on HEp-2 cells. Sera with an ANA titer of 1:20 or more were tested for antibodies to the following antigens: double-stranded DNA (dsDNA), using Crithidia luciliae; Sm, RNP, SSA, and SSB, using an extractible nuclear antibody profile microplate (EIA, Kallestad, Sanofi-Pasteur, MN); and Jo-1 and Scl-70, using an ELISA kit (BMD, Marne la Vallée, France).

Assay of IgG sialylation.Complexed IgGs were precipitated by 20% polyethylene glycol (PEG) 6000 (Sigma) in PBS. Glycosylation of free IgGs in the supernatant was analyzed. Sialic acid was detected using a previously described ELISA (13). Briefly, plates were coated with F(ab′)2 anti-IgG (Jackson), washed 3 times with PBS supplemented with 0.05% Tween 20, and blocked by incubation for 90 minutes at 37°C with PBS containing 3% bovine serum albumin. To eliminate any background reaction between the capturing agent and the detecting lectins, an oxidation step was incorporated into the assay. Following 5 washes, the plates were oxidized with 200 μl of 50 mM sodium periodate (Sigma) in 0.05M citrate buffer, pH 4.0, for 3 minutes, then washed 5 more times. The plates were saturated with free IgGs and incubated for 60 minutes at 37°C. Biotinylated Sambucus nigra agglutinin (SNA) (Vector, Burlingame, CA) diluted 1:3000 was added and left for 1 hour at 37°C. After washing, the plates were incubated for 1 hour at 37°C with 100 μl of horseradish peroxidase-conjugated streptavidin (Amersham, Arlington Heights, IL) diluted 1:4000. Then the plates were washed before development with the substrate 1,2 o-phenylenediamine (Dakopotts, Copenhagen, Denmark).

Galactose (Gal) was detected using an ELISA with biotinylated Ricinus communis agglutinin (RCAI) (Vector), and N-acetylglucosamine (GlcNAc) using an ELISA with biotinylated Bandeiraea simplicifolia II agglutinin (BSDII), as described above. To differentiate lectin binding caused by masking of Gal by sialic acid from lectin binding caused by absence of Gal, Gal was exposed by treatment of the sera with neuraminidase: Sera were incubated for 18 hours at 37°C with Clostridium perfringens type F neuraminidase (Sigma), with an enzyme/substrate ratio of 0.1 U/20 μl of serum. The enzyme was inactivated by heating at 56°C for 60 minutes. Release of sialic acid was then verified using the assays indicated above, with SNA, RCAI, and BSDII as the revealing agents for sialic acid, Gal, and GlcNAc, respectively.

  • Circulating immune complexes (CICs) were detected using an ELISA as previously described. IgG- and IgM-containing CICs were measured using a modification of the solid-phase C1q technique (14). To evaluate the IgA content of CICs, an ELISA was used according to the method of Rodriguez and colleagues (15). CICs were precipitated from sera by 2% PEG 6000, and the precipitate was dissolved in veronal buffered saline. Then the samples were incubated on plates coated with an F(ab′)2 anti-IgA (Dakopotts). Revelation was with horseradish–peroxidase-labeled F(ab′)2.

  • IgG1, IgG2, IgG3, and IgG4 were quantitated using a human IgG subclass ELISA kit (Peliclass, CLB, Amsterdam, Netherlands).

  • HLA–AB typing was performed using a standard microcytotoxicity test on B lymphocytes, and HLA DR typing was performed using molecular biology techniques.

Statistical analysis

Distributions of laboratory test results at the first visit were described as percentage per category for qualitative items and as median (range) for quantitative items. The distributions of laboratory test results were compared between patients with and without RA at the final visit using the chi-square test (or the Fisher's exact test, when necessary) for qualitative items and the Mann–Whitney U test (16) for quantitative items. For each test with a P value <0.20 in the previous comparison, diagnostic value was evaluated by determining sensitivity (proportion of positive tests among patients with RA, i.e., proportion of true-positive tests) and specificity (proportion of negative tests among patients without RA, i.e., proportion of true-negative tests). For quantitative items, diagnostic value was described by plotting sensitivity against 1-specificity to obtain the receiver-operating characteristic (ROC) curve (17). The ROC curve was plotted for each continuous laboratory test by varying the cutoffs (a value lower than the cutoff was considered negative and other values positive). For these items, diagnostic value was described by determining the sensitivity and specificity for various cutoffs. The optimal cutoff for a continuous laboratory test was defined as the value nearest to the northwest point of the ROC curve. Sensitivity, specificity, and estimate with its 95% confidence interval (95% CI) were calculated. In addition, for descriptive purposes, positive predictive value (PPV, true positives among positives), negative predictive value (PPV, true negatives among negatives), and accuracy (true positives and true negatives among all patients tested) were determined.

To evaluate how well combinations of the laboratory tests discriminated between patients with and without RA, 2 different procedures were developed. First, all laboratory tests with P values lower than 0.2 in the univariate analysis were combined by pairs; then the pair with the best diagnostic value was combined with a third selected laboratory test to determine the best combination of 3 laboratory tests. Second, a multiple logistic regression procedure with backward selection using the likelihood ratio test (18) was applied to all laboratory tests with P values lower than 0.20 in the univariate analysis. Data were analyzed using the Statistical Package for the Social Sciences version 9.0, 1999 (Chicago, IL).


  1. Top of page
  2. Abstract
  7. Acknowledgements

Study population

Two hundred seventy patients (184 women) with arthritis of less than 1 year duration constituted the study cohort. Median followup was 30 months. Followup was shorter than 1 year in 16 cases (6%) and was 1 to 2 years in 21 cases (8%), 2 to 3 years in 97 cases (36%), 3 to 4 years in 81 cases (30%), and 4 years or longer in 55 cases (20%). Ninety-eight patients (36%) were given a diagnosis of probable or definite RA by the panel of 5 rheumatologists at the last visit (Table 1).

Table 1. Demographic and clinical data (office-based rheumatologist diagnoses) at the last visit of patients presenting with arthritis at the first visit
  • *

    Among the cases of undifferentiated arthritis, 19 were considered by the office-based rheumatologists as “probably not RA.”

  • RA = rheumatoid arthritis; RS3PE = remitting seronegative symmetrical synovitis with pitting edema.

Age (years)49.5 ± 16.3
Sex ratio (female/male)184/86
Number of joints with synovitis4.4 ± −5.9
Diagnosis of RA for the panel of 5 rheumatologists98/270
Diagnoses of the office-based rheumatologists
 RA + spondylarthropathy1
 Gouty arthritis4
 Erythema nodosum–sarcoidosis5
 Septic arthritis5
  Lyme disease1
  Neisseria gonorrhoeae1
  Viral infection3
 Still's disease2
 Sjögren's syndrome6
 Systemic lupus erythematosus5
 Giant cell arteritis- polymyalgia rheumatica3
 Polymyalgia rheumatica or  RA?1

Diagnostic value of laboratory tests

Among the laboratory tests performed at the first visit, result distributions of all RF assays, APF, IgG-AKA, ANA, and HLA type were significantly different at the last visit in patients with and without RA (Table 2). For each of these tests except HLA type, diagnostic value was described by the ROC curve (Figure 1), sensitivity and specificity, positive and negative predictive values, and accuracy, with various cutoffs (Table 3).

Table 2. Laboratory data at the first visit in 270 patients with early arthritis according to the final diagnosis at the end of followup
Final diagnosisQuantitative data (median [range])P
RA (n = 98)nNot RA (n = 172)n
  1. ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; RF = rheumatoid factor; OD = optical density; APF = antiperinuclear factor; AKA = antikeratin antibodies; CIC = circulating immune complexes; NS = not significant.

ESR (mm/hour)22.5 (2–128)9222 (1–136)158NS
CRP (mg/liter)15 (2–182)958.3 (2–384)166NS
Latex test (IU/ml)20 (0–1600)970 (0–640)169< 0.001
IgM-RF (OD)0.241 (0–0.790)920.045 (0–0.851)162< 0.001
IgG-RF (OD)0.098 (0–0.387)0.007 (0–0.414)0.001
IgA-RF (OD)0.040 (0–0.800)0.017 (0–0.492)< 0.001
APF1/80 (0–10 000)930 (0–1000)160< 0.001
IgM-AKA0 (0–0)900 (0–0)158NS
IgG-AKA0 (0–640)0 (0–320)< 0.001
IgA-AKA0 (0–10)0 (0–10)NS
Antinuclear antibodies1/20 (0–1/1000)970 (0–1/1000)168< 0.01
Sialic acid (OD)1.57 (0.71–2.17)851.56 (0.51–2.12)141NS
Galactose (OD)1.74 (1.11–2.51)1.66 (0.68–2.44)
N-acetyl-glucosamine (OD)1.24 (0.015–1.97)1.2 (0.028–2.03)
IgM-CIC (OD)0.04 (0.001–0.257)890.03 (0.002–0.305)149NS
IgG-CIC (OD)0.076 (0.01–0.372)0.07 (0.01–1.41)
IgA-CIC (OD)0.12 (0.01–0.639)0.11 (0.002–1.72)
IgG1 (mg/ml)6.96 (2.35–31.2)856.96 (0.64–21.36)140NS
IgG2 (mg/ml)3.9 (0–12.15)4.12 (0.110–12.9)
IgG3 (mg/ml)0.45 (0.15–1.15)0.39 (0.09–2.72)
IgG4 (mg/ml)0.45 (0.054–2.76)0.48 (0.03–3)
Qualitative data (%)
HLA DR459%9639%1580.002
Shared epitope
 At least one (single or  double dose)53 (40 + 13)68 (60 + 8)0.01
Table 3. Value of enzyme-linked immunosorbent assay (ELISA) for IgM, IgG, and IgA rheumatoid factor (RF); latex test; test for APF; and test for IgG-AKA in discriminating between patients with and without rheumatoid arthritis
Sensitivity (%)Specificity (%)Positive predictive value (%)Negative predictive value (%)Accuracy (%)
  • APF = antiperinuclear factor; AKA = antikeratin antibodies; ANA = antinuclear antibodies; IU = international units; OD = optical density.

  • *

    Optimal cutoff (see Material and Methods section for definition).

  • Various cutoff titers were used (patients with titers lower than the cutoff were classified in the non-RA group and those with titers at or above the cutoff in the RA group).

Latex testn = 97n = 167n = 264
 ≥ 20 IU*5593827880
 ≥ 40 IU4395847576
 ≥ 80 IU3497877274
IgM-RF ELISAn = 92n = 162n = 254
 ≥ 0.110 OD7282638478
 ≥ 0.150*6587747379
 ≥ 0.2005891787979
 ≥ 0.2265592807879
 ≥ 0.2604594827576
 ≥ 0.3503296819670
IgG-RF ELISAn = 92n = 162n = 254
 ≥ 0.090 OD*5961467260
 ≥ 0.1502391596867
IgA-RF ELISAn = 92n = 162n = 254
 ≥ 0.050 OD*4485637370
 ≥ 0.1003094767171
APFn = 93n = 160n = 253
 ≥ 1:80*5279677467
 ≥ 1:2004289707372
IgG-AKAn = 90n = 158n = 248
 ≥ 1:10*4794827677
 ≥ 1:203997887376
 ≥ 1:403798927376
 ≥ 1:802899967173
ANAn = 97n = 168n = 265
≥ 1:100*4371466861
thumbnail image

Figure 1. Receiver operating characteristic curve for rheumatoid factor (RF) assays, antinuclear antibodies, antiperinuclear factor (APF), and IgG antikeratin antibodies (AKA) evaluated at the first visit. The curves show the relationships between sensitivity and 1-specificity of each test in discriminating between patients with and without rheumatoid arthritis at the end of followup. ELISA = enzyme-linked immunosorbent assay.

Download figure to PowerPoint

  • Rheumatoid factors. A high level in any RF assay was related to a diagnosis of RA. IgM-RF ELISA was the most sensitive test; titers between 0.150 and 0.226 offered a good tradeoff between sensitivity and specificity. In contrast, IgA-RF and IgG-RF ELISAs were of limited diagnostic value (Table 3). The optimal cutoff ODs for IgG-RF, IgA-RF, and IgM-RF were 0.09, 0.05, and 0.150, respectively. For IgM-RF, with 0.150 as the cutoff, sensitivity was 65% (95% CI 54.6–74.9) and specificity 87% (95% CI 80.9–91.8). For the latex test, 20 IU was the best cutoff, with a sensitivity of 55% (95% CI 44.2–64.8) and a specificity of 93% (95% CI 87.9–96.3) (Table 3).

  • Diagnostic value of FA in RF-positive patients. In patients with IgM-RF ELISA titers >0.200, the sensitivity and specificity of IgM-RF FA with a cutoff of 0.41 were 22% and 75%, respectively. Thus, FA did not seem useful for the diagnosis of RA.

  • APF and AKA. A high level of APF or IgG-AKA was strongly associated with RA (Table 2). These 2 parameters were not perfectly correlated with each other (Table 4). The optimal cutoff titers for APF and IgG-AKA were ≥1:80 and ≥1:10, respectively. IgG-AKA was the most specific test for the diagnosis of RA, and titers ≥1:10 gave a good tradeoff between sensitivity (47% [95% CI 36.1–57.5]) and specificity (94% [95% CI 89.5–97.4]).

  • ANA and anti–RA-33 antibodies. A high level of ANAs was significantly associated with RA, in contrast to anti–RA-33 antibodies (Table 2). The optimal cutoff for ANA was 1:100. However, the diagnostic value of ANA was low. Conversely, specific ANAs were helpful indicators of systemic lupus erythematosus (SLE) and scleroderma (Scl), although a few RA patients had specific ANAs (1 patient without clinical symptoms of SLE had antibody to dsDNA, 1 with secondary Sjögren's syndrome had antibody to SSA, and 1 with morphea had antibody to Scl). In the non-RA group, all patients with specific ANAs (to centromeres, dsDNA, RNP, or SSA) received a diagnosis of connective tissue disease.

  • Igs and CICs. There were no differences in the distribution of IgG subclass serum concentrations between RA and non-RA patients (Table 2). Similarly, neither glycosylation changes nor the amounts of IgG-, IgA-, or IgM-containing CICs were associated with RA.

  • HLA alleles. HLA–DR4 was significantly more common in RA patients than in non-RA patients (59% versus 39%, P < 0.002). The shared epitope allele, which was tested in 212 patients, seemed more common in the RA group than in the non-RA group (Table 2).

Table 4. Tests for antiperinuclear factor (APF) and antikeratin antibodies (AKA) in 248 patients with early arthritis
0[0–80]> 80
 > 10733040

Diagnostic value of laboratory test combinations

Using the optimal cutoff values of each continuous item, the diagnostic values of the “x and y” combination (both tests are positive) and of the “x or y” combination (either test is positive) were studied. The best pairs of tests combined 2 of the following: IgM-RF ELISA, latex test, and IgG-AKA (Table 5). Consequently, we studied the diagnostic value of the combination of these 3 tests. When we defined a positive combination as having all 3 tests positive, sensitivity and specificity were 33% (95% CI 23.6–44.3) and 99% (95% CI 96.4–99.9), respectively. A positive combination had performance characteristics similar to those of a positive latex and IgG-AKA pair. When we considered the combination positive if at least 2 of the 3 tests were positive, diagnostic performance was high, with 56% sensitivity (95% CI 45.3–66.9) and 94% specificity (95% CI 88.4–96.8). In our sample, having 2 or 3 positive tests was highly predictive of RA (83% positive predictive value [PPV]), whereas having 2 or 3 negative tests indicated that RA was unlikely (79% negative predictive value [NPV]). A positive result of at least 1 of the 3 tests yielded the best trade-off between sensitivity and specificity (75% [95% CI 64.3–83.4] and 82% [95% CI 74.8–87.6], respectively). The diagnostic value of the 3-test combination was slightly better than that of its components used alone or in pairs, and this combination performed better than the 1987 ACR criteria at the first visit (8).

Table 5. Sensitivity and specificity of the combination of 2 tests in discriminating between patients with and without rheumatoid arthritis
 Sensitivity (%)Specificity (%)
  1. “x + y” = positive result only if both tests are positive; “x or y” = diagnosis is positive if either test is positive.

  2. RF = rheumatoid factor; ANA = antinuclear antibody; APF = antiperinuclear factor; AKA = antikeratin antibodies.

Latex testn = 91n = 91n = 96n = 92n = 89n = 95n = 160n = 160n = 168n = 158n = 156n = 156
 x + y52352383435949599989999
 x or y676356676577878390748856
IgM RFn = 92n = 91n = 90n = 88n = 90n = 162n = 159n = 157n = 156n = 151
 x + y3924335429398969597
 x or y71677273807884708351
IgA RFn = 91n = 90n = 88n = 90n = 159n = 157n = 156n = 151
 x + y3031303099939993
 x or y4764617381728152
ANAn = 92n = 89n = 95n = 157n = 157n = 157
 x + y20310010099
 x or y535160748958
APFn = 90n = 91n = 158n = 148
 x + y36389889
 x or y61747548
IgG AKAn = 88n = 146
 x + y3596
 x or y7357

A multiple logistic regression analysis with backward selection using the likelihood ratio test (Table 6) also selected IgM-RF ELISA, the latex test, and IgG-AKA. The performance of either the 3-test combination or the multiple logistic regression function is shown in Figure 2. Table 7 summarizes the best strategies for predicting RA based on combinations of IgM-RF, latex, and IgG-AKA.

Table 6. Multiple logistic regression with backward selection using the likelihood ratio test
 Coefficient mean ± SEPOR estimate 95% CI
  1. OR = odds ratio; CI = confidence interval; IU = international units; RF = rheumatoid factor; ELISA = enzyme-linked immunosorbent assay; AKA = antikeratin antibodies. All first-visit laboratory tests with P lower than 0.20 in the univariate analysis were included in the multiple logistic regression model. The goal was to determine which tests discriminated between patients with and without rheumatoid arthritis at the last visit.

Latex test ≥ 20 IU1.28 ± 0.540.0203.6 (1.2–10.4)
IgM RF ELISA ≥ 0.1501.39 ± 0.470.0034.0 (1.6–10.1)
IgG AKA ≥ 1:101.89 ± 0.450.0016.6 (2.7–16.0)
Table 7. Sensitivity and specificity of the best strategies for predicting rheumatoid arthritis (RA) by using combinations of IgM-RF, latex, and IgG-AKA test results*
RA diagnosis ifSensitivity % (Estimate 95% CI)Specificity % (Estimate 95% CI)
  • *

    The following cutoffs were used: IgM RF > 0.150 OD, latex test ≥ 1/20 IU, IgG AKA ≥ 1:10. CI = confidence interval; OD = optical density; RF = rheumatoid factor; AKA = antikeratin antibody.

IgM RF and latex52 (40.9–62.2)94 (88.8–96.9)
IgM RF or latex67 (56.4–76.5)87 (80.6–91.7)
IgG AKA or IgM RF73 (62.2–81.7)83 (75.8–88.3)
IgG AKA and latex34 (24–44.5)99 (95.4–99.8)
IgG AKA, latex, IgM RF
 At least 1 of 375 (64.3–83.4)82 (74.8–87.6)
 At least 2 of 356 (45.3–66.9)94 (88.4–96.8)
 All 333 (23.6–44.3)99 (96.4–99.9)
thumbnail image

Figure 2. Receiver operating characteristic curves for the combination of IgG-antikeratin antibody (IgG-AKA ≥1:10), IgM-rheumatoid factor by enzyme-linked immunosorbent assay (ELISA) ≥ 0.150, and latex test >20 IU at the first visit in discriminating between patients with and without rheumatoid arthritis. The curve shows the relationships between sensitivity and 1-specificity for either the combination latex test + IgM-RF + IgG-AKA or the multiple logistic regression function (1.3 latex test + 1.4 IgM-RF + 1.9 IgG-AKA) in discriminating between patients with and without rheumatoid arthritis at the end of followup. (A lighter “ghost” curve of the usual care standard of latex testing is shown.)

Download figure to PowerPoint


  1. Top of page
  2. Abstract
  7. Acknowledgements

Rheumatoid factor is the autoantibody most often used as a diagnostic marker for RA. Agglutination tests for RF such as the Rose–Waaler and latex tests are widely used and are of proven diagnostic value (4, 5). An abnormal serum RF titer by any method that has yielded a positive result in fewer than 5% of normal controls was included in the ACR criteria in 1987 (6). Agglutination tests fail to discriminate among RF isotypes, which can be separated by radioimmunoassay, nephelometry, or ELISA. These methods have been found helpful when used routinely (19, 20), despite variations across laboratories (21).

Most studies of RF isotypes looked for associations with specific clinical manifestations or with outcomes of recent-onset RA; few focused on differences in diagnostic value across assays (22–29), and none evaluated the diagnostic value of specific RF isotypes in a prospective cohort of patients with recent-onset arthritis. Classically, IgM-RF is the isotype most strongly associated with a diagnosis of RA. The functional affinity of RF has been found to be higher in RA patients than in healthy controls (30) and higher in recent-onset RA than in advanced RA (9), suggesting that functional affinity may help to discriminate between RF associated with RA and RF not associated with RA. The diagnostic value of the other RF isotypes remains unclear but may be low, according to a previous study (29). However, it has been shown that increases in IgA-RF and IgG-RF can precede the onset of clinical RA.

Other autoantibody–antigen systems may be useful in the diagnosis of RA, including antibodies to citrulline-containing peptides (called anticitrulline antibodies, antifilaggrin antibodies, antiperinuclear factors, or antikeratin antibodies, in reference to the different methods used for their detection), RA-33, Sa, p68 (or Bip), and calpastatin. The diagnostic value of anticalpastatin (31, 32) and anti–RA-33 seems low (11, 33), and that of anti-p68 has not been evaluated. Many studies have investigated the diagnostic value of antibodies to citrulline-containing antibodies (34–38). Two recent reports suggest that anti-Sa may be a useful diagnostic marker (39, 40).

Apart from autoantibodies, several laboratory tests have been suggested (but not evaluated) as markers for early RA. In particular, certain immunoglobulin glycosylation alterations (41, 42) and IgG subclass imbalances (43, 44) may be specific for RA. Also, assays for immune complexes have been developed (45), and circulating immune complexes have been detected in most patients with RA (46).

HLA typing is not only costly but also lacks specificity for the diagnosis of early RA. However, 2 reports suggest that HLA alleles may contribute to shape the clinical pattern of early inflammatory arthritis (47, 48). Thus, combining HLA typing with other tests may assist in the diagnosis of early arthritis.

The present study was designed to determine which combination of laboratory tests best discriminates between early RA and other forms of arthritis at presentation. Laboratory tests may have less diagnostic value when performed routinely than when done as part of a battery of tests for a clinical study, because variations may be observed across test kits. Consequently, we based our study on tests performed routinely at our laboratory.

To produce valid results, a study of the diagnostic value of laboratory tests in RA must meet 3 prerequisites: Disease duration must be short at the time the laboratory tests are performed, followup must be long enough to ensure that few patients have unclassifiable disease at the end of followup, and the test cutoffs must be determined in patients with inflammatory joint disease. Accordingly, we included only patients with recent-onset inflammatory joint disease. These patients were recruited by office-based physicians to ensure that the population was representative of patients seen in everyday clinical practice. They were followed up prospectively. Median followup was 30 months; to compensate for this fairly short duration, we collected a vast array of clinical, radiographic, and laboratory data. Because the diagnosis of RA is difficult, it was made in our study by a panel of 5 rheumatologists, based on all available data. This diagnosis was used as the benchmark for evaluating the value of first-visit laboratory tests for predicting RA 2 years later. The best tests were IgG-AKA, IgM-RF by ELISA, and the latex test. IgG-AKA was the most specific test and IgM-RF ELISA the most sensitive test; the latex test offered the best tradeoff between sensitivity and specificity.

IgG-AKA had a better diagnostic value than other AKA isotypes or APF. The best cutoff for IgG-AKA in our study was lower (≥ 1:10) than the cutoff established in our laboratory as 2 standard deviations under the mean in a control group (> 1:10). This may explain the discrepancy between our results and those of previous studies reporting low sensitivity of AKA in early RA (49). We did not look for antifilaggrin antibodies (36, 37) or anticitrulline antibodies (38). Although these antibodies are not perfectly correlated with each other, their diagnostic value seems similar; furthermore, assays for both antibodies are expensive.

The second and third most useful tests, even when combined with AKA, were the IgM-RF ELISA and the latex test. Although highly significant correlations were found among RF assays, in keeping with earlier studies (23, 49–53), differences in diagnostic value were apparent, and combining these tests produced a small improvement in diagnostic value.

Previous studies found that the latex test and IgM-RF ELISA were useful for diagnosing RA. However, these studies did not use cutoffs determined in patients with inflammatory joint disease. In our study, the ROC curve of the IgM-RF ELISA suggested that the 0.150 OD cutoff provided both good specificity and high sensitivity. The best cutoff for the latex test on the ROC curve was 20 IU, and any deviation from this value was associated with a dramatic decrease in diagnostic value. These cutoffs differ slightly from those established in our laboratory as 2 standard deviations under the mean in a control group; i.e., 0.200 OD for the IgM-RF ELISA and >20 IU for the latex test. With these cutoffs, the diagnostic value of the 2 tests was quite similar but lower than that observed with the cutoffs determined in the present study. However, the ROC curves showed a range of optimal IgM-RF ELISA cutoffs, whereas a single cutoff value was optimal for the latex test.

Previous studies suggested that IgG-RF and IgA-RF may be more specific than other RF isotypes and may assist in the diagnosis of RA (23, 49, 54). In the present study, both isotypes were of limited diagnostic value. The findings from the only previous study (29) of the diagnostic value of RF in early arthritis agree in part with our results: ELISAs for IgG-RF and IgA-RF were not useful in diagnosing RA, whereas the IgM-RF ELISA emerged as a reasonable alternative to the latex test. Data from several previous studies suggest that age at disease onset (29, 55–57) or sex (29, 58) may influence the diagnostic value of RF. This was not the case in our study (data not shown). Differences in patient population, followup duration, or sensitivity of the RF test used may explain these discrepancies.

Presence of ANA was significantly associated with RA but was not useful as a diagnostic tool for RA. HLA–DR4 and shared epitopes were associated with RA. However, their diagnostic value was limited in our population of patients with early arthritis. Anti–RA-33, CICs, and serum IgG levels were of no assistance in making the diagnosis of RA in our study. The same was true of IgG glycosylation. However, this finding may be ascribable to the difficulty of standardizing the IgG glycosylation assay in routine laboratory practice, as well as to the heterogeneity of our population in terms of age and disease activity.

In conclusion, combined positivity of IgG-AKA, IgM-RF by ELISA, and the latex test indicated RA with nearly complete certainty in our population characterized by a high prevalence of RA (36%): Sensitivity was 33%, specificity 99%, PPV 97%, and NPV 73%. Moreover, positivity of at least 2 of these 3 tests indicated that RA was extremely likely (PPV 79%) and negativity of at least 2 of these 3 tests that RA was extremely unlikely (NPV 80%). Positivity of 1 or more of the 3 tests was the criterion with the best tradeoff between sensitivity and specificity (sensitivity 75%, specificity 82%) in a strategy emphasizing sensitivity. The diagnostic value of this 3-test combination was slightly better than that of the constituent tests used alone or in pairs, and this combination performed better than the 1987 ACR criteria at the first visit (8). However, it is important to note that our results may not apply to populations in which the prevalence of RA is different from that in our study cohort. Studies in various settings are warranted. We are currently investigating the diagnostic value of these tests used in combination with clinical and radiographic data.


  1. Top of page
  2. Abstract
  7. Acknowledgements

We are grateful to the following rheumatologists for referring their patients to us: E. Blat, P. Busson, A. Castagné, J. P. Caumon, P. Chicault, V. Desmas, J. P. Elie, X. Filliol, C. Gauthier, J. Glemarec, J. Y. Grolleau, M. N. Guillermit, R. Guyader, M. Hamidou, P. Herrou, G. Lavel, F. LeJean, R. Lemaître, M. C. Lheveder, A. Martin, Y. Maugars, I. Nouy Trolle, J. Olivry, C. Paturel, N. Paugam, A. Prost, D. Rault, B. Ribeyrol, A. Rossard, D. Rodet, I. Valls, P. Vilon, and P. Voisin. We thank O. Meyer (Paris, France) for his help with anti-RA-33 detection. We are also grateful to A. Wolfe for reviewing the English in this article.


  1. Top of page
  2. Abstract
  7. Acknowledgements
  • 1
    Irvine S, Munro R, Porter D. Early referral, diagnosis, and treatment of rheumatoid arthritis: evidence for changing medical practice. Ann Rheum Dis 1999; 58: 5103.
  • 2
    Van de putte LBA, Van Gestel AM, Van Riel PLCM. Early treatment in rheumatoid arthritis: rationale, evidence and implications. Ann Rheum Dis 1998; 57: 5112.
  • 3
    Mikuls TR, O'Dell J. The changing face of rheumatoid arthritis therapy: results of serial surveys. Arthritis Rheum 2000; 43: 4645.
  • 4
    Walker DJ, Pound JD, Griffiths ID, Powell RJ. Rheumatoid factor tests in the diagnosis and prediction of rheumatoid arthritis. Ann Rheum Dis 1986; 45: 68490.
  • 5
    Wolfe F, Cathey MA, Roberts FK. The latex test revisited: rheumatoid factor testing in 8287 rheumatic disease patients. Arthritis Rheum 1991; 34: 95160.
  • 6
    Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF, Cooper NS, et al. The American Rheumatism Association revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum 1988; 31: 31524.
  • 7
    Bläß S, Engel JM, Burmester GR. Review: the immunologic homunculus in rheumatoid arthritis. Arthritis Rheum 1999; 42: 2499506.
  • 8
    Saraux A, Berthelot JM, Le Henaff C, Chales G, Thorel JB, Hoang S, et al. Classification and diagnostic value of the 1987 ACR criteria for rheumatoid arthritis (RA) in early arthritis (EA). Arthritis Rheum 2000; 43 Suppl 9: S154.
  • 9
    Saraux A, Bendaoud B, Dueymes M, LeGoff P, Youinou P. Functional affinity of IgM rheumatoid factor in rheumatoid arthritis. Ann Rheum Dis 1997; 56: 1269.
  • 10
    Youinou P, Le Goff P, Dumay A, Lelong A, Fauquert P, Jouquan J. The antiperinuclear factor. I. Clinical and serological associations. Clin Exp Rheumatol 1990; 8: 16.
  • 11
    Meyer O, Tauxe F, Fabregas D, Gabay C, Goycochea M, Haim T, Elias A, Kahn MF. Anti-RA 33 antinuclear antibody in rheumatoid arthritis and mixed connective tissue disease: comparison with antikeratin and antiperinuclear antibodies. Clin Exp Rheumatol 1993; 11: 4738.
  • 12
    Laemmli UK. Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 1979; 227: 6805.
  • 13
    Youinou P, Pennec YL, Casurd-Budd R, Dueymes M, Lamour A. Galactose terminating oligosaccharides of IgG in patients with primary Sjögren's syndrome. J Autoimmun 1992; 5: 393400.
  • 14
    Mihailov C, Lamour A, Bendaoud B, Dueymes M, Pennec YL, Mottier D, et al. Isotype-specific assays for circulating complexes are of value in distinguishing various connective tissue diseases. Med Sci Res 1992; 20: 74951.
  • 15
    Rodriguez MA, Bankurst AD, William RC, Troup GM, Stastny P. Studies on the relationship between HLA DR4 and in vitro IgM factor production. Clin Immunol Immunopathol 1983; 27: 96109.
  • 16
    Siegel S. Nonparametric statistics for the behavioural sciences. New York: McGraw Hill; 1956. p. 11627.
  • 17
    Metz LE. Basic principles of ROC analysis. Semin Nucl Med 1978; 8: 28398.
  • 18
    Mosmer DW, Lemeshow S. Applied regression analysis. New York: John Wiley & Sons; 1989. p. 10610.
  • 19
    Adebajo AO, Cawtson TE, Hazleman BL. Routine quantitation of rheumatoid factors: ELISA or nephelometry? Scand J Rheumatol 1992; 21: 3024.
  • 20
    Svedler W, Wallman J, Froelich CJ, Teodorescu M. Routine measurement of IgM, IgG and IgA rheumatoid factors: high sensitivity, specificity, and predictive value for rheumatoid arthritis. J Rheumatol 1997; 24: 103744.
  • 21
    Roberts-Thompson PJ, Langhans T, McEvoy R, Hollingsworth PN, Bonifacio E, Dawkins RL, et al. SEAPAL quality control analysis of rheumatoid factor measurement in 29 diagnostic laboratories. Ann Rheum Dis 1987; 46: 41720.
  • 22
    Silvestris F, Goodwin JS, Williams RC Jr IgM, IgA and IgG rheumatoid factors in patients with rheumatoid arthritis and normal donors. Clin Rheumatol 1985; 4: 3928.
  • 23
    Highton J, Hessian PA, Small B, Palmer DG. An assessment of the diagnostic value of quantitative measurements of IgA rheumatoid factor. J Rheumatol 1985; 12: 8548.
  • 24
    Powell RJ, Leyiand AM, Pound JD, Bossingham DH. An improved assay for IgG rheumatoid factor: its value in the diagnosis of rheumatoid arthritis. J Rheumatol 1985; 12: 42731.
  • 25
    Gioud-Paquet M, Auvinet M, Raffin T, Girard P, Bouvier M, Lejeune E, Monier JC. IgM rheumatoid factor (RF), IgA RF, IgE RF, and IgG RF detected by ELISA in rheumatoid arthritis. Ann Rheum Dis 1987; 46: 6571.
  • 26
    Van Leeuwen MA, Westra J, Limburg PC, De Jong HJ, Marrink J, Van Rijswijk MH. Quantitation of IgM, IgA and IgG rheumatoid factors by ELISA in rheumatoid arthritis and other rheumatic disorders. Scand J Rheumatol 1988; Suppl 75: 2531.
  • 27
    Kleveland G, Egeland T, Lea T. Quantitation rheumatoid factors (RF) of IgM, IgA, and IgG isotypes by a simple and sensitive ELISA: discrimination between false and true IgG-RF. Scand J Rheumatol 1988; 75 Suppl: 1524.
  • 28
    March RE, Kirwan JR, Davies PG, Winrow VR, Holborow EJ. Sensitive ELISA measurement of IgA, IgG and IgM antiglobulins and other diseases and a comparison of their specificity in rheumatoid arthritis and infectious mononucleosis. Scand J Rheumatol 1987; 16: 4459.
  • 29
    Visser H, Gelink LBS, Kaamphraath AH, Breedvald FC, Hazes JMW. Diagnostic and prognostic characteristics of the enzyme linked immunosorbent rheumatoid factors assays in rheumatoid arthritis. Ann Rheum Dis 1996; 55: 15761.
  • 30
    Newkirk MM, Fournier MJ, Shiroky J. Rheumatoid factor avidity in patients with rheumatoid arthritis: identification of pathogenic RFs which correlate with disease parameters and with the Gal(0) glycoform of IgG. J Clin Immunol 1995; 15: 2507.
  • 31
    Després N, Talbot G, Plouffe B, Boire G, Ménard HA. Detection and expression of a cDNA clone that encodes a polypeptide containing two inhibitory domains of human calpastatin and its recognition by rheumatoid arthritis sera. J Clin Invest 1995; 95: 18916.
  • 32
    Mimori T, Suganuma K, Tanami Y, Jojima T, Matsumura M, Fujii T, et al. Autoantibodies to calpastatin (an endogenous inhibitor for calcium-dependant neutral protease, calpain) in systemic rheumatic diseases. Med Sci 1995; 92: 726771.
  • 33
    Hassfeld W, Steiner G, Hartmuth K, Kolarz G, Scherak O, Graninger W, et al. Demonstration of a new antinuclear antibody (anti-RA33) that is highly specific for rheumatoid arthritis. Arthritis Rheum 1989; 32: 151520.
  • 34
    Hoet RM, Van Venrooij WJ. The antiperinuclear factor and antikeratin antibodies in rheumatoid arthritis. In: SmolenJ, KaldenJ, editors. Rheumatoid arthritis. Berlin: Springer-Verlag; 1992. p. 299318.
  • 35
    Berthelot JM, Garnier P, Glemarec J, Flipo RM. Diagnostic value for rheumatoid arthritis of antiperinuclear factor at the 1:100 threshold. Rev Rhum Engl Ed 1998; 65: 914.
  • 36
    Vincent C, Simon M, Sebbag M, Girbal-Neuhauser E, Durieux JJ, Cantagrel A, et al. Immunoblotting detection of autoantibodies to human epidermidis filaggrin: a new diagnostic test for rheumatoid arthritis. J Rheumatol 1998; 25: 83846.
  • 37
    Slack SL, Mannik M, Dale BA. Diagnostic value of antibodies to filaggrin in rheumatoid arthritis. J Rheumatol 1998; 25: 84751.
  • 38
    Schellekens GA, Visser H, de Jong BAW, van den Hoogen FHJ, Hazes JMW, Breedveld FC, et al. The diagnostic properties of rheumatoid arthritis antibodies recognizing a cyclic citrullinated peptide. Arthritis Rheum 2000; 43: 155163.
  • 39
    Després N, Boire G, Lopez-Longo FJ, Ménard HA. The Sa system: a novel antigen-antibody system specific for rheumatoid arthritis. J Rheumatol 1994; 21: 102733.
  • 40
    Hayem G, Chazerain P, Combe B, Elias A, Haim T, Nicaise P, et al. Anti-Sa antibody is an accurate diagnostic and prognostic marker in adult rheumatoid arthritis. J Rheumatol 1999; 26: 713.
  • 41
    Bond A, Alavi A, Axford JS, Bourke BE, Bruckner FE, Kerr MA, et al. A detailed lectin analysis of IgG glycosylation, demonstrating disease specific changes in terminal galactose and N-acetylglucosamine. J Autoimmun 1997; 10: 7785.
  • 42
    Young A, Sumar N, Bodman K, Goyal S, Sinclair H, Roitt I, et al. Agalactosyl IgG: an aid to differential diagnosis in early synovitis. Arthritis Rheum 1991; 34: 14259.
  • 43
    Aho K, Heliovaara M, Knekt P, Reunanen A, Aromaa A, Leino A, et al. Serum immunoglobulin and the risk of rheumatoid arthritis. Ann Rheum Dis 1997; 56: 3516.
  • 44
    Kay RA, Wood KJ, Bernstein RM, Holt PJL, Pumphrey RSH. An IgG subclass imbalance in connective tissue disease. Ann Rheum Dis 1988; 47: 53641.
  • 45
    McDougal JS, Hubbard M, McDuffie F, Strobel PL, Smith SJ, Bass N, et al. Comparison of five assays for immune complexes in the rheumatic diseases: an assessment of their validity for rheumatoid arthritis. Arthritis Rheum 1982; 25: 115666.
  • 46
    Lessard J, Nunnery E, Cecerc F, McDuffy S, Pope RM. Relationship between the articular manifestations of rheumatoid arthritis and circulating immune complexes detected by three methods and specific classes of rheumatoid factors. J Rheumatol 1983; 10: 4117.
  • 47
    El Gabalawy HS, Goldbach-Mansky R, Smith D, Arayssi T, Bale S, Gulko P, et al. Association of HLA alleles and clinical features in patients with synovitis of recent onset. Arthritis Rheum 1999; 42: 1696705.
  • 48
    Saraux A, Valls I, Voisin V, Koreichi A, Baron D, Youinou P, et al. How useful are tests for rheumatoid factors, antiperinuclear factors, antikeratin antibody, and the HLA DR4 antigen for the diagnosis of rheumatoid arthritis? Rev Rhum Engl Ed 1995; 62: 1620.
  • 49
    Eberhardt KB, Svensson B, Truedsson L, WoUheim FA. The occurrence of rheumatoid factor isotypes in early definite rheumatoid arthritis: no relationship with erosions or disease activity. J Rheumatol 1988; 15: 10704.
  • 50
    Möttönen T, Hannonen P, Jokinen I, Arvilommi M, Oka M. Relation between bone erosions and rheumatoid factor IgA and IgM isotypes in recent-onset rheumatoid arthritis. Scand J Rheumatol 1988; 75 Suppl: 2449.
  • 51
    Van Zeben D, Hazes JMW, Zwinderman AH, Cais A, van der Voort EAM, Breedveld FC. Clinical significance of rheumatoid factors in early rheumatoid arthritis: results of a follow up study. Ann Rheum Dis 1992; 51: 102935.
  • 52
    Withrington RH, Seifert MH. Predominant wrist disease in rheumatoid arthritis associated with high concentration of IgA rheumatoid factor. BMJ 1985; 291: 1388.
  • 53
    Robbins DL, Feigal DW Jr, Leek JC. Relationship of serum IgG rheumatoid factor to IgM rheumatoid factor and disease activity in rheumatoid arthritis. J Rheumatol 1986; 13: 25962.
  • 54
    Jonsson T, Valdimarsson H. Is measurement of rheumatoid factor isotypes clinically useful? Ann Rheum Dis 1993; 52: 1614.
  • 55
    Van Schaardenburg D, Lagaaij AM, Otten HG, Breedveld FC. The relation between class-specific serum rheumatoid factors and age in the general population. Br J Rheumatol 1993; 32: 1049.
  • 56
    Terkeltaub R, Esdaile J, Decary F, Tannenbaum H. A clinical study of older age rheumatoid arthritis with comparison to a younger onset group. J Rheumatol 1983; 10: 41824.
  • 57
    Pease CT, Bhakta BB, Devlin J, Emery P. Does the age of onset of rheumatoid arthritis influence phenotype? A prospective study of outcome and prognostic factors. Rheumatology 1999; 38: 22834.
  • 58
    Weyand CM, Schmidt D, Wagner U, Goronzy JJ. The influence of sex on the phenotype of rheumatoid arthritis. Arthritis Rheum 1998; 41: 81722.