Re-exploring the cerebrospinal fluid (CSF) to serum glucose ratio.

Abstract

Background In interpretation of lumbar puncture (LP) findings, CSF glucose is conventionally expressed as a ratio to the simultaneous level in serum. A low ratio indicates hypoglycorrhachia and pathological conditions. Method We retrospectively studied the CSF findings in consecutive LPs performed in our ward over 5 years. Patients with conditions known to cause CSF hypoglycorrhacia, uncertain diagnosis and inadequate laboratory data were excluded. Blood for simultaneous serum glucose level was sampled within one hour of LP. ANOVA, Newman-Keuls multiple comparisons test and linear regression were used for statistical analysis. Results 170 sets of samples from patients with central demyelination (45), neuropathies (66), headache (26), normal pressure hydrocephalus (11), neurodegenerative diseases (8), and others conditions (14) were studied. CSF glucose was <2.2 mmol/L in only one sample. Mean CSF to serum glucose ratio was 0.61 (range 0.21-1.00). Proportion of samples with ratios of <0.50, 0.50-0.59, 0.60-0.69 and ≥0.70 were 18.8, 27.1, 29.4 and 24.7% for all patients, and 6.8, 27.3, 38.6 and 27.3%, 60.0, 24.0, 8.0 and 8.0%, or 61.5, 30.8, 0.0 and 7.7% for patients with simultaneous serum glucose of < 7.8, 7.8-11.1 or >11.1 mmol/L, respectively. Mean CSF to serum glucose ratios were significantly different between patients with simultaneous serum glucose of <7.8, 7.8-11.1 and >11.1 mmol/L (P <0.0001), but not for different neurological conditions (P = 0.5911). Linear regression showed a significant relation between serum and CSF glucose (r = 0.769, P <0.0001) and inverse correlation between serum glucose and CSF to serum glucose ratio (r = -0.569, P <0.0001, y = 0.8 - 0.03x). Conclusions Our study demonstrates the CSF to serum glucose ratio is not constantly related to but varies inversely with the simultaneous serum glucose level. A significant proportion of patients without conditions causing CSF hypoglycorrhachia has a low ratio during hyperglycaemia. The coefficient derived from our linear regression model can be applied for adjustment of such deviations.