Comparison of β2-Transferrin and β-Trace Protein for Detection of Cerebrospinal Fluid in Nasal and Ear Fluids

Cerebrospinal fluid (CSF) fistulas occur mainly in basal skull fractures or operations at the skull base entailing a dural defect; up to 40% of CSF fistulas are spontaneous (1). The treatment consists of surgical closure of the dural defect. Complications such as intracranial abscess or bacterial meningitis may develop, with a mortality rate of 25–50%. Therefore, a simple and reliable diagnosis of CSF fistulas is of great importance. β-Trace protein is synthesized mainly in the epithelial cells of the choroid plexus and is found in CSF in concentrations ∼35-fold higher than in plasma (2). Thus, β-trace protein represents a potentially useful marker for the diagnosis of CSF leakage (3), with immunoelectrophoresis having a specificity of 100% and a sensitivity of 91% for the diagnosis of CSF leakage (4). β2-Transferrin, the asialo form of transferrin, is produced by neuraminidase activity in the brain and is usually detectable only in CSF, perilymph, and aqueous humor. Thus, like β-trace protein, it is a potentially useful marker to diagnose CSF leakage (3). In the present work we compared the first two commercially available assay systems: a new, sensitive immunofixation method from Sebia GmbH for detection of β2-transferrin and an automatic, quantitative immunonephelometric assay from Dade Behring for detection of β-trace protein. We also established a cutoff value for β-trace protein and evaluated the potential of both markers as indicators of CSF leakage. The β-trace protein assay is based on latex particle enhancement using rabbit polyclonal antibodies against β-trace protein. Calibrators contain β-trace protein purified from CSF (N Protein Standard UY®; Dade Behring). Imprecision was assessed by assaying N/T protein Control LC® (Dade Behring). Measurements were performed on a BN II analyzer (Dade Behring). The assay covers a range of 0.25–15.8 mg/L in the initial 1:100 dilution. The within-day imprecision (CV) was 12% at 0.68 mg/L (n = 10) in serum and 2.8% at 18.3 mg/L (n = 10) in CSF. Between-day CV were 15% at 0.74 mg/L (n = 10) in serum and 10% at 19.4 mg/L (n = 10) in CSF. At least 200 μL (35-μL sample volume) is required for a single measurement, which requires 20 min. β2-Transferrin was detected with the Hydrasys LC (Sebia GmbH) electrophoretic system, which separates the more cathodic asialo form of transferrin from an anodic fraction containing the sialo forms. Both transferrin fractions are detected by immunofixation with anti-transferrin antibodies labeled with peroxidase. A patient’s serum sample is investigated in parallel in a 1:100 dilution to avoid false-positive results in case of alcoholism. The assay required 10 μL and 120–150 min. For comparison of the detection limits for β2-transferrin (Sebia GmbH Hydrasys LC) and β-trace protein (Dade Behring BN II), both proteins were analyzed in samples generated by dilution of a CSF pool (containing 16.3 mg/L β-trace protein). We measured β-trace protein and β2-transferrin in nasal fluids from 7 healthy volunteers and β-trace protein in nasal and ear fluids from 26 patients (17 males and 9 females; age range, 5–74 years) and β2-transferrin in samples from 25 patients. All were admitted to the Otorhinolaryngology, Neurosurgery, or Oral and Maxillofacial Surgery departments at our Hospital between July 2000 and August 2003. Laboratory investigations were performed without information about the clinical diagnosis. Rhinoliquorrhea was suspected in 22 patients, otoliquorrhea in 2 patients, and subcutaneous and transcutaneous fistulas in 2 patients (Table 1 in the Data Supplement that accompanies the online version of this Technical Brief at http://www.clinchem.org/content/vol50/issue3/). Laboratory investigations were performed as a precaution or because the patients presented with suspicious clinical symptoms or on computed tomography scans after surgical intervention, accidents, or spontaneously. Evidence for a CSF fistula was obtained in 13 patients (9 males and 4 females) by typical clinical symptoms, computed tomography, cysternography, or intraoperative findings (Table 1 in the online Data Supplement). The CSF and serum samples were collected for diagnostic workup. Nose and ear fluid was collected in the gauze of Salivettes (Sarstedt). The secretions were separated from the Salivette gauze by centrifugation at 2000g for 10 min and stored at 4 °C at most for 1 day. Serum was collected into serum separator tubes (Sarstedt) and centrifuged at 2000g for 10 min immediately after arriving in our laboratory. CSF was centrifuged for 10 min at 2000g. Statistical analysis was performed with SigmaPlot 2001 for Windows. Parallel measurements of β-trace protein and β2-transferrin in samples from a CSF pool (16.3 mg/L β-trace protein) diluted down to 0.1 mg/L β-trace protein (1:128 dilution) revealed linear quantification of β-trace protein (r2 = 0.999; y-intercept, 0.0286 mg/L), whereas the detection limit for β2-transferrin was at the 1:32 dilution. In nasal fluids from the seven volunteers, β-trace protein concentrations were <0.25 mg/L, and β2-transferrin was detected in none. In patients without CSF fistula, β-trace protein concentrations were <0.25–0.95 mg/L. In one patient with nasal secretion during provocation testing (bowing during abdominal press), a typical symptom for rhinoliquorrhea, no β2-transferrin was detected and β-trace protein tested 3 months later was <0.25 mg/L (patient 13 in Table 1 in the online Data Supplement). These results might be attributable to collection of nasal fluid without actual CSF leakage in an intermittent rhinoliquorrhea. In 12 patients with confirmed CSF fistulas, β-trace protein concentrations were 2.5–35 mg/L. At a cutoff of 1 mg/L, the sensitivity was 100% (statistical uncertainty, 78–100%; P >0.05), and the specificity was 100% (86–100%; P >0.05) for the diagnosis of CSF leakage. β2-Transferrin was detected in 8 of 11 patients (73%) with CSF leakage. In 37% of all investigated patients (three patients with and six without CSF fistulas), no interpretation of the immunofixation was possible because of an overlap of the transferrin band with the β2-transferrin band. The detailed data are presented in Table 1 in the online Data Supplement and in Fig. 1B . Electrophoretic detection of β2-transferrin in a CSF pool (A), and representative results of electrophoretic detection of β2-transferrin in nasal fluid/serum pairs from four patients suspected for rhinoliquorrhea and CSF as positive control (B). (A), the pool was variably diluted with saline from 1:1.5 to 1:128. (B), the numbering corresponds to that of Table 1 in the online Data Supplement. N, nasal fluid; S, serum. Diagnosing CSF fistulas is always challenging. Particularly in subclinical liquorrhea with a loss of small quantities of CSF, the fistula may not detectable by radiologic imaging. Because of the risk of developing meningitis from a CSF leak, a reliable diagnostic tool to identify trace amounts of CSF is required. Chemical identification of glucose and protein is obsolete for this purpose (5)(6), but β2-transferrin assays are widely used. This protein is detected by immunoblotting (7) or immunofixation and silver staining (8), both of which are time-consuming and labor-intensive. The immunoblot assay on the Hydrasys LC takes 2–2.5 h and requires many manual steps. Moreover, interpretation of β2-transferrin electrophoresis is complicated by the overlap of the application area and the detection point for β2-transferrin. Finally, the main disadvantage is the interference of higher concentrations of transferrin with the β2-transferrin band, which prevents interpretation. No improvement is achieved by sample dilution because β2-transferrin is diluted to concentrations below the detection limit. β-Trace protein has been identified by rocket immunoelectrophoresis, which has a reported sensitivity of 91% in diagnosing CSF leakage (4). For the automated assay from Dade Behring, a range of 0.22–1.69 mg/L for healthy individuals was reported, and a cutoff at 1.31 mg/L (97.5 percentile) was proposed (9). At concentrations between 1.31 and 1.69 mg/L, however, there is overlap of individuals without and patients with CSF leakage (9). In our study, β-trace protein measured without predilution was <1 mg/L in individuals without CSF leakage. In the earlier study, nasal fluids were partly prediluted 1:100. At a cutoff of 1 mg/L, the Dade Behring assay had a 100% sensitivity (78–100%; P >0.05) and specificity (86–100%; P >0.05) for diagnosing a CSF leakage. Thus, in our opinion the nephelometric quantification of β-trace protein is preferable, particularly because this method is less time-consuming and labor-intensive than the electrophoretic method. We gratefully acknowledge the excellent technical assistance of Jutta Breidenich and Ursula Emmler.