Article Text
Statistics from Altmetric.com
Anti-cyclic citrullinated peptide (CCP) antibodies and rheumatoid factor (RF) are recommended screening tests for rheumatoid arthritis.1 In a meta-analysis,2 the sensitivity of anti-CCP (67%) for diagnosing rheumatoid arthritis was comparable to the sensitivity of RF (69%).2 The specificity of anti-CCP for rheumatoid arthritis (95%) was higher than the specificity of RF (85%).2 Consequently, the positive likelihood ratio was higher for anti-CCP (12.46) than for RF (4.86).2
Studies that addressed the clinical usefulness of anti-CCP used a single cut-off value and, hence, likelihood ratios were calculated based on a single cut-off. In the present letter, we illustrate how likelihood ratios for anti-CCP and RF depend on the antibody level. Our calculations were based on a clinically well defined group of patients with rheumatoid arthritis (n = 85), diseased controls (n = 165) (including psoriatic arthritis, connective tissue disease and organ specific autoimmune diseases), and a group of consecutive patients for whom a rheumatologist ordered anti-CCP (n = 48). The characteristics of the patients and controls are described in detail in Coenen et al.3 Anti-CCP was assayed with Phadia Unicap (cut-off: 7 units/ml; Phadia, Uppsala, Sweden). On all samples, RF was determined by nephelometry (Immage, Beckman-Coulter, Fullerton, California, USA; cut-off: 20 IU/ml).
Table 1 summarises the sensitivities and specificities of anti-CCP and RF for several cut-offs. This illustrates that increasing the cut-off results in enhanced specificity and decreased sensitivity. Next, we calculated the likelihood for an anti-CCP test result <7 units/ml, between 7 and 25 units/ml and >25 units/ml for patients with rheumatoid arthritis and controls. For patients with rheumatoid arthritis, the likelihoods were 22.5%, 6.9% and 70.6%, respectively. For controls, the values were 96.0%, 1.5% and 2.5%, respectively. For each test result interval, the likelihood ratio (ie, the likelihood for patients with rheumatoid arthritis divided by the likelihood for controls) was calculated. The likelihood ratios for anti-CCP were 0.23 (95% CI 0.15 to 0.33), 4.5 (95% CI 0.64 to 84.2) and 27.7 (95% CI 11.6 to 97.1) for <7 units/ml, 7–25 units/ml and >25 units/ml, respectively. The results are shown in fig 1A. A similar analysis was performed for RF (fig 1B). The likelihood ratios for RF were 0.4 (95% CI 0.28 to 0.57), 2.0 (95% CI 1.0 to 4.2), 2.1 (95% CI 1.3 to 7.4) and 4.8 (95% CI 2.1 to 14.0) for <20 IU/ml, 20–100 IU/ml, 101–300 IU/ml and >300 IU/ml, respectively. These data (i) illustrate that the likelihood ratios increase with increasing antibody levels and (ii) confirm that the likelihood ratios are higher for anti-CCP than for RF. The likelihood ratio for anti-CCP between 7 and 25 units/ml (slightly elevated values) equalled the likelihood ratio of highly elevated (>300 IU/ml) RF values, which demonstrates the superior performance of anti-CCP compared to RF.
For each test result interval of anti-CCP and RF, the post test probability for rheumatoid arthritis was calculated as a function of the pretest probability. The results are shown in fig 1C,D.
In conclusion, we illustrated how likelihood ratios and post test probabilities depend on antibody level for anti-CCP and RF. Such knowledge helps with the interpretation of a specific test result. Clinical laboratories might consider providing likelihoods ratios for test result intervals.
Footnotes
Competing interests: None declared.
Ethics approval: Ethics approval was obtained.