Figure 4.

Composite scores of multiple CSF parameters allow differentiating inflammatory neuropathies with high sensitivity and specificity. Two to five CSF parameters were combined by a generalized linear model with logistic regression. Receiver operating characteristics (ROC) analysis was performed. We increased the number of parameters sequentially. The model selection was performed by an exhaustive search. The maximal number of parameters was determined by the Bayesian information criteria. The individual parameters that form the composite score and the resulting AUC are given in the respective figure. The confidence was calculated using De Long test and is indicated in brackets. (A) Guillain–Barré syndrome (GBS) vs. idiopathic intracranial hypertension (IIH). (B) Chronic inflammatory demyelinating neuropathy (CIDP) vs. IIH. (C) CIDP vs. GBS. BBB, blood-brain barrier.