Markers for Giant Cell Arteritis and Polymyalgia Rheumatica
Eight out of 26 serum markers were significantly modulated in newly diagnosed GCA and/or PMR patients vs healthy controls (Table 1). Serum BAFF, CXCL9 and IL-6 were increased in both newly diagnosed GCA and newly diagnosed PMR patients. Serum CCL2 and CCL11 were decreased in GCA patients only, whereas serum IL-10 and sIL-2R were increased. In contrast, serum CXCL10 was increased in PMR patients only. Thus, both overlapping and specific changes in serum markers were observed in GCA and PMR patients (supplementary Fig. S1 http://rheumatology.oxfordjournals.org/content/54/8/1397/suppl/DC1, available at Rheumatology Online).
In a ROC analysis, we next determined the diagnostic accuracy of individual markers in distinguishing newly diagnosed patients from healthy controls. Serum CXCL9 and IL-6 provided excellent discrimination of newly diagnosed GCA and PMR patients from healthy controls, as indicated by AUCs > 0.90 (Table 1 and supplementary Table S3 http://rheumatology.oxfordjournals.org/content/54/8/1397/suppl/DC1, available at Rheumatology Online). Serum BAFF also accurately distinguished newly diagnosed GCA and PMR patients from healthy controls, with AUCs > 0.80. Serum CXCL10, which was not modulated in GCA, could discriminate between PMR patients and healthy controls, as shown by an AUC > 0.80. None of the other markers provided an AUC > 0.80. Taken together, CXCL9, IL-6 and BAFF were the most accurate markers for discriminating newly diagnosed GCA and PMR patients from healthy controls. Of interest, these three serum markers were increased in most GCA and PMR patients in whom ESR levels did not fulfill the classification criteria for GCA or PMR, respectively (supplementary Table S4 http://rheumatology.oxfordjournals.org/content/54/8/1397/suppl/DC1, available at Rheumatology Online).
Next, we evaluated the eight serum markers that were significantly modulated in patients at baseline, for changes upon remission after 3 months of CS treatment. Serum levels of BAFF and IL-6 were significantly decreased in GCA and PMR patients in remission, whereas CCL11 was increased in both patient groups (Fig. 1A). In addition, serum CCL2 was increased in GCA patients in remission, but not in PMR patients. No significant modulation of CXCL9, CXCL10, IL-10 or sIL-2R was observed upon CS-induced remission. Among the serum markers that were not modulated at baseline, only a slight increase in CCL4 was observed in GCA patients in remission (data not shown).
(Enlarge Image)
Figure 1.
Correlation of serum markers with disease activity in GCA and PMR
(A) Serum levels of the eight markers that were initially modulated in newly diagnosed GCA and PMR patients (nGCA and nPMR) are shown in combination with the serum levels of the same patients in remission (rGCA and rPMR) after 3 months of corticosteroid treatment. Statistical significance is indicated as *P < 0.05. (B) Correlation coefficients between these eight serum markers and the ESR and CRP. Cell colours indicate the strength of the correlations. Statistical significance is indicated as *P < 0.05 and **P < 0.01.
Finally, we studied the serum markers that were initially modulated in newly diagnosed GCA and PMR patients for correlations with the ESR and CRP. Serum BAFF provided correlation coefficients of >0.75 with the ESR and CRP in both GCA and PMR patients (Fig. 1B and supplementary Fig. S2 http://rheumatology.oxfordjournals.org/content/54/8/1397/suppl/DC1, available at Rheumatology Online). Serum IL-6 correlated strongly with the ESR and CRP in PMR patients and moderately in GCA patients, as indicated by correlation coefficients of >0.75 and >0.50, respectively. Serum CCL2 and CCL11 showed a moderate to strong inverse correlation with the ESR and CRP in GCA patients, but not in PMR patients. Serum CXCL9 correlated moderately with the ESR and CRP in PMR patients only. Serum CXCL10, IL-10 and sIL-2R, which were modulated at baseline, did not correlate with ESR and CRP in patients with GCA or PMR. Taken together, serum BAFF and IL-6 showed stronger correlations with ESR and CRP in GCA and PMR patients than any other serum marker studied.
Results
Serum Markers Discriminating Newly Diagnosed GCA and PMR Patients From Healthy Controls
Eight out of 26 serum markers were significantly modulated in newly diagnosed GCA and/or PMR patients vs healthy controls (Table 1). Serum BAFF, CXCL9 and IL-6 were increased in both newly diagnosed GCA and newly diagnosed PMR patients. Serum CCL2 and CCL11 were decreased in GCA patients only, whereas serum IL-10 and sIL-2R were increased. In contrast, serum CXCL10 was increased in PMR patients only. Thus, both overlapping and specific changes in serum markers were observed in GCA and PMR patients (supplementary Fig. S1 http://rheumatology.oxfordjournals.org/content/54/8/1397/suppl/DC1, available at Rheumatology Online).
In a ROC analysis, we next determined the diagnostic accuracy of individual markers in distinguishing newly diagnosed patients from healthy controls. Serum CXCL9 and IL-6 provided excellent discrimination of newly diagnosed GCA and PMR patients from healthy controls, as indicated by AUCs > 0.90 (Table 1 and supplementary Table S3 http://rheumatology.oxfordjournals.org/content/54/8/1397/suppl/DC1, available at Rheumatology Online). Serum BAFF also accurately distinguished newly diagnosed GCA and PMR patients from healthy controls, with AUCs > 0.80. Serum CXCL10, which was not modulated in GCA, could discriminate between PMR patients and healthy controls, as shown by an AUC > 0.80. None of the other markers provided an AUC > 0.80. Taken together, CXCL9, IL-6 and BAFF were the most accurate markers for discriminating newly diagnosed GCA and PMR patients from healthy controls. Of interest, these three serum markers were increased in most GCA and PMR patients in whom ESR levels did not fulfill the classification criteria for GCA or PMR, respectively (supplementary Table S4 http://rheumatology.oxfordjournals.org/content/54/8/1397/suppl/DC1, available at Rheumatology Online).
Serum markers Modulating Upon CS-induced Remission in GCA and PMR
Next, we evaluated the eight serum markers that were significantly modulated in patients at baseline, for changes upon remission after 3 months of CS treatment. Serum levels of BAFF and IL-6 were significantly decreased in GCA and PMR patients in remission, whereas CCL11 was increased in both patient groups (Fig. 1A). In addition, serum CCL2 was increased in GCA patients in remission, but not in PMR patients. No significant modulation of CXCL9, CXCL10, IL-10 or sIL-2R was observed upon CS-induced remission. Among the serum markers that were not modulated at baseline, only a slight increase in CCL4 was observed in GCA patients in remission (data not shown).
(Enlarge Image)
Figure 1.
Correlation of serum markers with disease activity in GCA and PMR
(A) Serum levels of the eight markers that were initially modulated in newly diagnosed GCA and PMR patients (nGCA and nPMR) are shown in combination with the serum levels of the same patients in remission (rGCA and rPMR) after 3 months of corticosteroid treatment. Statistical significance is indicated as *P < 0.05. (B) Correlation coefficients between these eight serum markers and the ESR and CRP. Cell colours indicate the strength of the correlations. Statistical significance is indicated as *P < 0.05 and **P < 0.01.
Serum Markers Correlating With the ESR and CRP in GCA and PMR
Finally, we studied the serum markers that were initially modulated in newly diagnosed GCA and PMR patients for correlations with the ESR and CRP. Serum BAFF provided correlation coefficients of >0.75 with the ESR and CRP in both GCA and PMR patients (Fig. 1B and supplementary Fig. S2 http://rheumatology.oxfordjournals.org/content/54/8/1397/suppl/DC1, available at Rheumatology Online). Serum IL-6 correlated strongly with the ESR and CRP in PMR patients and moderately in GCA patients, as indicated by correlation coefficients of >0.75 and >0.50, respectively. Serum CCL2 and CCL11 showed a moderate to strong inverse correlation with the ESR and CRP in GCA patients, but not in PMR patients. Serum CXCL9 correlated moderately with the ESR and CRP in PMR patients only. Serum CXCL10, IL-10 and sIL-2R, which were modulated at baseline, did not correlate with ESR and CRP in patients with GCA or PMR. Taken together, serum BAFF and IL-6 showed stronger correlations with ESR and CRP in GCA and PMR patients than any other serum marker studied.