Figure 5.

Association of Treg function and immune markers with outcomes in sarcoidosis. (A), Treg suppressive function, evaluated at study entry, was significantly reduced in subjects who required treatment at one year follow-up (n = 6) compared to patients who did not require treatment (n = 10). Plasma levels of CXCL10 (n = 6 on therapy, n = 19 no therapy required) and sCD25/IL-2Rα, TNFR I and II (n = 7 on therapy, n = 19 no therapy required) were compared. DLCO at enrolment also was associated with treatment need at follow-up (n = 7 on therapy, n = 14 no therapy required). (B), Six probit regression models with k-Fold cross validation were run using need for therapy as the dependent variable and the inflammatory score (derived from Treg function, sCD25, TNFR I and II, TNFα, CXCL10 and Ki-67⁺ in PBMC) and DLCO (% pred) as predictors. The fitted regression equation was: probit (P_{on therapy)} = 7.17*inflammatory score – 9.76*DLCO%. Both coefficients were significant at p = 0.007 and p = 0.002 levels, respectively. The overall regression was statistically significant (likelihood ratio Chi-Square χ²(2) = 10.96, p = 0.004). Correctly classified (90%, blue dots) and misclassified (red dots) cases are shown with predictive probability of need for therapy for validation set of patients as a function of DLCO, % pred (left) and the inflammatory score (right). Further details are provided in Supplementary Table S5. (C), Comparison of probit models with and without Treg function as a part of inflammatory scores are detailed in Supplementary Table S6. (A) (except DLCO) – Mann-Whitney test; (A) (DLCO) – unpaired T test, (C) – Pearson’s correlation.