The soluble intercellular adhesion molecule-1 (sICAM-1) was measured in paired CSF and serum samples from 128 patients with different neurological diseases. The reference range of blood-derived sICAM-1 fractions in CSF was characterized with reference to the albumin CSF/serum quotients. The low mean concentrations of sICAM-1 of normal controls (n=33) in CSF (1.5 ng/ml; C.V.=40%) compared to serum (285.1 ng/ml; C.V.=32%) indicate that about 60% to 80% of sICAM-1 in normal lumbar CSF derives from blood. This calculation is based on the theoretically expected molecular size-dependent blood-CSF gradient between 300:1 to 250:1. In patients with non-inflammatory diseases (n=21) the sICAM-1 CSF/serum quotient increased non-linearly with increasing albumin CSF/serum quotient (blood-CSF barrier dysfunction) displaying the shape of a saturation-like curve in contrast to hyperbolic curves of other blood-derived proteins in CSF. This non-linear relation between sICAM-1 and albumin quotients does not allow a linear index evaluation reported in earlier studies. In bacterial meningitis (n=31) and viral meningoencephalitis (n=28) in addition to the increased blood-derived fraction, the brain-derived fraction of sICAM-1 in CSF was up to 12-fold higher than that in controls. The sICAM-1 CSF/serum quotients in MS (n=15) did not differ from non-inflammatory controls, i.e., there was no brain-dependent sICAM-1 fluctuation in CSF in contrast to the known fluctuations in blood. Earlier published reports on sICAM-1 have been controversial due to less sensitive assays and unsuitable linear evaluation concepts for blood-CSF barrier dysfunction.