Table 3.
Logistic regression model comparing baseline CRP and baseline combined biomarker panels in predicting clinical response to golimumab treatment at week 14
| ACR20* response | DAS28† response | PASI75* response | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Model (true/predicted)† | NR | R | Accuracy‡ | NR | R | Accuracy‡ | N | R | Accuracy‡ |
| Baseline biomarker panel† | NR | R | Accuracy† | NR | R | Accuracy† | N | R | Accuracy† |
| Non-responder | 26 | 14 | Specificity: 65% | 17 | 7 | Specificity: 71% | 31 | 18 | Specificity: 63% |
| Responder | 5 | 29 | Sensitivity: 85% | 5 | 45 | Sensitivity: 90% | 6 | 19 | Sensitivity: 76% |
| Predictive value | NPV 84% | PPV 67% | NPV 77% | PPV 87% | NPV 84% | PPV 51% | |||
| Baseline CRP only† | |||||||||
| Non-responder | 28 | 12 | Specificity: 70% | 17 | 7 | Specificity: 71% | 21 | 28 | Specificity: 43% |
| Responder | 23 | 11 | Sensitivity: 32% | 25 | 25 | Sensitivity: 50% | 11 | 14 | Sensitivity: 56% |
| Predictive value | NPV 55% | PPV 48% | NPV 40% | PPV 78% | NPV 66% | PPV 33% | |||
| Predictive panel§ | Weight | OR | p Value | Weight | OR | p Value | Weight | OR | p Value |
| Pyridinoline | −1.89 | 0.15 | 0.009 | – | – | – | – | – | |
| Adiponectin | 1.00 | 2.72 | 0.010 | 1.19 | 3.29 | 0.015 | 0.69 | 1.99 | 0.046 |
| Prostatic acid Phosphatase | 1.51 | 4.53 | 0.011 | – | – | – | – | – | – |
| Factor VII | −0.76 | 0.47 | 0.039 | 1.25 | 3.49 | 0.038 | 0.73 | 2.08 | 0.043 |
| IL-16 | – | – | – | – | – | – | −1.75 | 0.17 | 0.006 |
| SGOT | – | – | – | −2.07 | 0.13 | 0.001 | – | – | – |
| IgA | – | – | – | 0.92 | 2.51 | 0.010 | – | – | – |
| Leptin | – | – | – | −0.66 | 0.52 | 0.016 | – | – | – |
*ACR20, PASI75 and DAS-28 were treated as categorical variables in this analysis. To be classified as a DAS28 responder, patients had to have a good or moderate response.10
†All biomarker values were log2-transformed before inclusion in the models.
‡Model accuracy (sensitivity, specificity, NPV, PPV) was estimated using cross-validation.
§Weights are the coefficients in the logistic regression model. OR are the exponential of the weights. Multivariate p values are based on inclusion of all terms in the model.
ACR20, American College of Rheumatology 20% improvement response; CRP, C-reactive protein; DAS28, disease activity score in 28 joints; Ig, immunoglobulin; IL, interleukin; NPV, negative predictive value; NR ,non-responders; PASI75, 75% improvement in the psoriasis area and severity index score; PPV, positive predictive value; R, responders; SGOT, serum glutamic oxaloacetic transaminase.