Added value of multiple autoantibody testing for predicting progression to inflammatory arthritis in at-risk individuals

Abstract

Background: Predicting progression to clinical arthritis in individuals at-risk of developing rheumatoid arthritis is a prerequisite to developing stratification groups for prevention strategies. Selecting accurate predictive criteria is the critical step to define the population at-risk. While positivity for anti-citrullinated protein antibodies (ACPA) remains the main recruitment biomarker, positivity for other autoantibodies (AutoAbs) identified before the onset of symptoms, may provide additional predictive accuracy for stratification.

Objective: To perform a multiple AutoAbs analysis for both the prediction and the time of progression to inflammatory arthritis (IA).

Methods: 392 individuals were recruited based on a new musculoskeletal complaint and positivity for ACPA or rheumatoid factor (RF). ELISAs were performed for ACPA, RF, anti-nuclear Ab, anti-carbamylated protein (anti-CarP) and anti-collagen AutoAbs. Logistic and COX regression were used for analysis.

Results: Progression to IA was observed in 125/392 (32%) of cases, of which 78 progressed within 12 months. The AutoAbs ACPA, RF, anti-CarP were individually associated with progression (p<0.0001) and improved prediction when combined with demographic/clinical data (Accuracy >77%; area under the curve (AUC) >0.789), compared with prediction using only demographic/clinical data (72.9%, AUC=0.760). Multiple AutoAbs testing provided added value, with +6.4% accuracy for number of positive AutoAbs (AUC=0.852); +5.4% accuracy for AutoAbs levels (ACPA/anti-CarP, AUC=0.832); and +6.2% accuracy for risk-groups based on high/low levels (ACPA/RF/anti-CarP, AUC=0.837). Time to imminent progression was best predicted using ACPA/anti-CarP levels (AUC=0.779), while the number of positive AutoAbs was/status/risk were as good (AUC=0.778).

Conclusion: We confirm added value of multiple AutoAbs testing for identifying progressors to clinical disease, allowing more specific stratification for intervention studies.

Keywords: Anti-Citrullinated Protein Antibodies; Arthritis, Rheumatoid; Rheumatoid Factor.

Figure 1

Positivity (A), as number of patients and (B) levels of AutoAb in at-risk individuals. (A) Bars represent the number of individual tested positive (black) or negative (white) for each individual. X² tests were used individually to assess associations with progression (***p<0.0001, #p<0.100). ACPA n=363, RF n=381, anti-CarP n=372, RF-IgG n=152, ANA n=290, all anti-collagen autoAbs n=71. (B) Violin plots represent the distribution of AutoAb levels observed. Solid lines across the plot indicate the positivity cut-off for each test. Medians and quartiles of distribution are indicated by dotted and dashed lines within violin plot respectively. MWU tests comparing levels were used individually to assess associations with progression (*p<0.05, ***p<0.001, #p<0.100). ACPA, anticitrullinated protein antibody; ANA, anti-nuclear Ab; RF, rheumatoid factor.

Figure 2

Modelling the predictive value of autoAbs in at-risk individuals. (A) Individual time of progression with respect to positivity for the three main autoAbs tested (red positive, blue negative). (B) Logistic regression: AUC for prediction models using clinical data and multiple autoAbs based on status, levels, high and low risk groups or autoAbs count. (C, D) Cox regression: Survival and AUC for prediction models using clinical data and multiple autoAbs based on status, levels, risk groups or autoAbs count. (B–D) Black clinical data only; green clinical data+status; red clinical data+levels; blue clinical data+risk group; purple clinical data+autoAbs count. ACPA, anticitrullinated protein antibody; AUC, area under the curve; RF, rheumatoid factor.