Table 2. Pre-ART serum biomarkers associated with subsequent paradoxical cryptococcal IRIS.
| Biomarker(pg/ml) | CM ControlMedian (IQR) | CM-IRIS CaseMedian (IQR) | Odds Ratio per 2-Fold Increase | p-Value |
| TNF-α | 73.9 (27.3–130) | 31.3 (<1.5–75.0) | 0.500 | 0.001 |
| IL-4 | 2.8 (2.2–3.5) | 3.3 (2.2–4.0) | 2.865 | 0.001 |
| IL-17 | 66.1 (27.5–125) | 47.8 (14.6–131) | 1.306 | 0.018 |
| VEGF | 82.7 (47.0–155) | 77.3 (34.4–176) | 0.688 | 0.016 |
| G-CSF | 25.0 (14.7–41.3) | 27.7 (15.0–53.2) | 0.742 | 0.025 |
| GM-CSF | 18.8 (4.0–41.4) | 8.2 (<1–24.8) | 0.913 | 0.34 |
| CCL2 (MCP-1) | 35.7 (9.4–68.8) | 24.5 (6.2–84.5) | 1.036 | 0.45 |
Biomarkers on the day of starting ART that are associated with prediction of subsequent risk of paradoxical CM-IRIS. Risk estimates are calculated by logistic regression using penalty parameters specified by the Lasso method based on log2-transformed data [41]. Thus, the odds ratio is reflective of the increased risk per 2-fold increase in the serum biomarker value, adjusting for the influence of each of the biomarkers. p-Values are calculated by a multivariate logistic regression.