Initial CSF total tau correlates with 1-year outcome in patients with traumatic brain injury

Different biochemical neuromarkers (BNMs) can reflect the extent of traumatic brain injury (TBI). Among them, S-100β and neuron-specific enolase (NSE), measured in blood, have most frequently been utilized. Although associations between concentrations of S-100B and outcome in TBI patients are reported, both S-100B and NSE have low sensitivity and specificity.¹ Focus has been oriented toward other BNMs.² We have demonstrated that elevated levels of glial fibrillary acidic protein are associated with bleak 1-year outcome in patients with TBI.³

Different types of cellular damage in TBI, involving different receptors and biochemical pathways, may be identified or at least characterized by BNMs. For prognosis, diffuse axonal injury (DAI) may be the most deleterious response in patients with TBI.⁴ DAI is predominantly noted in discrete regions of the brain following high-speed head rotation or deceleration injuries. DAI can histologically be detected in all unconscious TBI patients.⁵

Tau proteins, with a molecular mass of 48 to 67 kd and mainly found in neurons, are microtubule associated and play an important role in the assembly of tubulin monomers into microtubules and in maintaining the cytoskeleton and axonal transport. Aggregation of specific sets of tau proteins in filamentous inclusions is the common feature of intraneuronal tangles.⁶

In this study, we investigated if total tau in serum or ventricular CSF (vCSF) is associated with long-term outcome (1-year) in patients with severe TBI. We hypothesized that higher tau levels could be associated with a more severe TBI and a worse long-term outcome.

Methods.

The study was performed in accordance with the provisions of the Helsinki Declaration, and the University Hospital Medical Ethics Committee, Gothenburg, Sweden, approved the study protocol. Informed written consent was obtained from each patient's next of kin. The patients were admitted from October 2000 to December 2002 to the Neurointensive Care Unit (NICU) at Sahlgrenska University Hospital, Gothenburg, Sweden. Patients were included within 48 hours after trauma. The inclusion criteria were severe TBI (i.e., a Glasgow Coma Scale [GCS] score of ≤8 at admission), operated with placement of an intracranial ventricular catheter, and a need for artificial ventilation. Serum and vCSF was intermittently collected post TBI on days 0 (trauma day), 1 to 4, 6, 8, and once in the period of days 11 to 14.

After clinical and radiologic evaluations, the patients underwent neurosurgical intervention within hours (1 to 4 hours) after admission, to receive an indwelling ventricular catheter for intracranial pressure (ICP) monitoring/therapeutic CSF drainage. When appropriate, space-occupying lesions like hemorrhages and contusions were surgically removed. Patients were then treated in accordance with a standardized protocol, the Lund concept.⁷ Data collected included age, sex, and premorbid health status. Physiologic and laboratory measures were continuously recorded throughout the study period and concomitantly adjusted to be kept within the following limits: hemoglobin >120 g/L, serum sodium >135 to <150 mmol/L, serum potassium 4.0 to 5.0 mmol/L, serum albumin 35 to 50 g/L, core temperature 37 ± 0.5 °C, mean arterial blood pressure (MABP) between 70 and 100 mm Hg, ICP <20 mm Hg, cerebral perfusion pressure (= MABP − ICP) >60 mm Hg, Po ₂ 12 to 18 kPa, Pco ₂ around 4.5 kPa, and normalized pH. Blood glucose was kept between 4 and 6 mmol/L according to NICU routine.

CSF and blood samples were centrifuged at ×2,000 g for 10 minutes at 4 °C and frozen at –70 °C until analyzed. Occasionally vCSF samples could not be obtained because of unstable high ICP.

As a reference group, we examined vCSF from 20 patients with NPH. Details of this patient cohort have previously been presented.⁸

Results.

Fifty-six patients with TBI were primary included, but unstable ICP only allowed vCSF collection in 39 patients. These patients (women, n = 9; men, n = 30), with a median age of 49 years (range 16 to 82 years), are those presented.

Three patients had a history of epilepsy, one of which had been operated for meningioma, and three more patients had a previous history of other neurologic diseases like TIA, stroke, or polio. Thus, the remaining 33 patients had no history of neurologic disorders before TBI. Eight patients had measurable content of blood ethanol, and three had narcotics in the blood at the time of accident. The remaining 28 patients had no measurable drug content in the blood at the time of accident; however, 2 of these had a history of alcohol addiction. The initial trauma was traffic accident in 16 patients, fall accidents in 13 patients, and miscellaneous causes like assault in 10 patients. Twenty-six patients had isolated TBI, whereas in eight cases, other fractures than in the skull could also be detected. Finally, in five patients, internal organs were traumatized in combination with TBI.

At hospital arrival, the patients' scores ranged in GCS from 3 to 15, subdividing them as GCS 3 to 4 (n = 6), GCS 5 to 6 (n = 11), GCS 7 to 9 (n = 9), GCS 10 to 13 (n = 7), and GCS 14 to 15 (n = 6). All patients had GCS ≤8 within 48 hours after trauma. After arrival to NICU, the patients were then taken to surgery for placement of an intracranial ventricular catheter within 4 hours. The mean hospital stay was 18.9 days, and the hospital mortality was 15% (6/39).

The concentrations of vCSF total tau on days 2 to 3 post trauma correlated to morbidity and mortality at 1 year (R² = 0.17, R = 0.42, p = 0.001).

Nonsurvivors (GOSE 1) at 1-year follow-up had higher levels of vCSF total tau (8,500 pg/mL, IQR 7,638) on days 2 to 3 post trauma compared with survivors (GOSE 2 to 8), with levels of 682 pg/mL (IQR 1,155, p < 0.001) (figure 1). Patients with bad outcome (GOSE 1 to 4) had higher levels of vCSF total tau (2,580 pg/mL, IQR 7,443) vs those with good outcome (GOSE 5 to 8) (504 pg/mL, IQR 1,256) on days 2 to 3 (p < 0.01) (figure 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Tau levels in ventricular CSF days 0 to 14 in alive vs dead patients related to 1-year outcome in traumatic brain injury patients. ***p < 0.001.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Tau levels days 0 to 14 in patients with traumatic brain injury related to 1-year good (Extended Glasgow Outcome Scale [GOSE] 5 to 8) vs bad (GOSE 1 to 4) outcome. **p < 0.01.

The vCSF total tau levels in NPH patients (reference group) were 677 pg/mL (IQR 308). When comparing these levels vs those of TBI patients, those succumbing (GOSE 1) had increased vCSF total tau levels at all sample intervals, days 2 to 3 (p < 0.001), days 4 to 6 (p < 0.001), and days 8 to 14 (p < 0.001).

TBI survivors (GOSE 2 to 8) had higher levels of vCSF total tau vs NPH patients on days 4 to 6 (p < 0.05) and days 8 to 14 (p < 0.01). Patients with poor outcome (GOSE 1 to 4) had increased levels of vCSF total tau at all sample intervals, days 2 to 3 (p < 0.001), days 4 to 6 (p < 0.01), and days 8 to 14 (p < 0.001) compared with patients with NPH. In patients with good outcome (GOSE 5 to 8), similar increases of vCSF total tau could be detected only on days 8 to 14 (p < 0.001).

One-year outcome assessed with GOSE demonstrated that of the 39 patients included, 51% (20/39) had a good outcome (GOSE 5 to 8), whereas 49% (19/39) had a bad outcome (GOSE 1 to 4). The 1-year mortality (GOSE 1) was 18% (7/39).

With ROC, we detected a sensitivity of 100% and a specificity of 81.5% for 1-year mortality at a vCSF total tau level above 2,126 pg/mL on days 2 to 3 (table). The calculated AUC was 0.934 (figure 3). The specificity, sensitivity, and LR+ of different cut-off levels for mortality of vCSF total tau are depicted in the table. Patients with bad outcome (GOSE 1 to 4) had a cut-off level of vCSF total tau of >702 pg/mL, with a sensitivity of 83.3% and a specificity of 69%, with a calculated AUC of 0.814 (figure 4).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Receiver operator curve for 1-year mortality. Calculated sensitivity and specificity for tau levels in patients with traumatic brain injury on days 2 to 3.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. Receiver operator curve for 1-year bad (Extended Glasgow Outcome Scale [GOSE] score of 1 to 4) outcome. Calculated sensitivity and specificity for tau levels in patients with traumatic brain injury on days 2 to 3.

No correlation could be found between NIHSS, investigating neurologic status 1 year post trauma, and initial vCSF total tau levels. Correspondingly, there was no correlation between initial vCSF total tau levels and ADL, assessed by Bartel Scale, 1 year post trauma.

Throughout the study, there were no detectable levels of total tau in serum.

Discussion.

Using tau as a marker for axonal damage studying patients with severe TBI, we could verify that axonal damage is linked to 1-year morbidity (GOSE 2 to 4) as well as mortality (GOSE 1). We could also demonstrate a correlation between initially low vCSF total tau and good 1-year outcome (GOSE 5 to 8). The results indicate a cut-off level of vCSF total tau of >2,126 pg/mL related to 1-year mortality (GOSE 1), with a sensitivity of 100% and a specificity of 81.5%.

In the current study, vCSF was exclusively obtained for analysis. We have previously demonstrated a difference between vCSF and lumbar CSF (lCSF) total tau concentrations, with levels being higher in vCSF.¹¹ The way of obtaining CSF may thus be very important. This notion may not previously been appreciated as a mix of both vCSF and lCSF has been used in both reference and treatment groups in many investigations.

Thus, in a small (n = 15), retrospective study in TBI patients,¹² the authors presented similar vCSF tau levels as ours. They could, however, in contrast to our results, not detect a correlation between vCSF tau and in-house outcome. Their control group was a mixture of patients with dementia, headache, and NPH with analyses made on lCSF, which may explain their lower levels of lCSF tau compared with our homogeneous reference group of NPH patients, with total tau analyzed exclusively from vCSF. In a recent article about cleaved tau (c-tau) in TBI patients, the authors similarly did not state how CSF was collected.¹³

Tau protein is a small phosphoprotein found in the axonal compartments of neurons binding to microtubules promoting their stability and assembly.¹² Human brain tau protein has six isoforms, with molecular mass of approximately 48 to 67 kd on sodium dodecyl sulfate polyacrylamide gel electrophoresis.¹⁴ Increased levels of CSF tau are probably a sign of axonal injury and have been reported in Alzheimer disease (AD),^15,16 Creutzfeldt–Jakob disease,¹⁷ Guillain–Barré syndrome,¹⁶ as well as in TBI.¹⁸ Axonal injury is a common feature of TBI damaging intracellular microtubules, and proteins like tau are then supposedly released into vCSF. Thus, speculatively increased levels of vCSF total tau may signal the development of an intracranial lesion, with higher levels of vCSF total tau correlating to poor outcome, results in accordance with the current investigation.

When analyzing, we could not find any detectable serum levels of total tau. We therefore present only total tau analyzed in vCSF. On the contrary, in a previous investigation, arterial blood levels of c-tau ranging from 5 to 12 ng/mL were detected.¹⁹

In 1999, using monoclonal antibodies developed by immunizing mice with tau proteins and screening hybridomas with CSF from brain trauma patients, one research group reported the presence of a set of 30- to 50-kd proteins in the CSF of brain trauma patients,²⁰ suggested to be c-tau, lacking the N- and C-terminal domains. These antibodies show poor reactivity with recombinant tau,²⁰ and no sequencing data on the proteins they recognize have been presented. The antibodies used in the current work recognize tau proteins of the expected size in CSF,²¹ similar to the molecular mass of CSF tau proteins reported by other groups.^22,23 The relation between the tau proteins detected by our method and c-tau cannot be determined at present.

Tau is normal in lCSF of patients with NPH, indicating no or minor cortical degeneration in this disorder. In vCSF, total tau levels in patients with NPH levels have been estimated to 550 ng/L.²⁴ We have not found any investigations measuring vCSF total tau in healthy volunteers, but it does not seem pertinent to assume that volunteers may have even lower vCSF total tau values than patients with NPH.

In the current study, we have analyzed total tau, that is, including subforms. Other research groups have utilized various methods in quantifying tau. Groups from Cincinnati, OH, have measured c-tau in CSF in nanograms per milliliter, whereas more recent investigations, similar to the current one, analyze total tau measured in picograms per milliliter.^13,14 c-tau cannot be found in CSF in NPH patients, whereas we and others measuring total tau also in NPH patients can detect this in both lCSF and vCSF.²⁵

Previous investigations in mice with closed head injury have demonstrated an early rise in CSF tau within hours after TBI, followed by a decline within 24 to 48 hours.²⁶ This decline is then followed by a secondary increase in CSF tau levels on days 2 to 4. A new CSF tau decline is again noted followed by a third increase in CSF tau levels, peaking on day 10 to 14, returning to normal levels on day 43.¹²

Most of our patients were not admitted to our tertiary university hospital and received a ventricular catheter until 24 to 48 hours post trauma. One could question if the insertion of a ventricular catheter by itself would influence these levels. However, tau levels do not increase after insertion (patients get a catheter within 4 hours from admittance). The higher tau levels seen in this study are noted on day 2 to 3 independent of insertion of a ventricular catheter.

Thus, we could not obtain any really early vCSF tau samples for analysis. In our investigation, we could however confirm, from day 2 and on, the previously experimentally noted variations¹² in vCSF total tau levels in patients with bad outcome (GOSE 1 to 4) but not in patients with good outcome (GOSE 5 to 8). However, on days 8 to 14, there was a rise in vCSF total tau in all patients compared with those on days 0 to 1.

The initial vCSF total tau rise was very prominent in those dying (GOSE 1), with values more than 10-fold higher than in patients with good outcome (GOSE 5 to 8). All patients with lethal injury (GOSE 1) had vCSF total tau levels above 2,126 pg/L on days 2 to 3, that is, 100% sensitivity for death. We could also detect differences of vCSF total tau levels on days 2 to 3 between patients with bad outcome (GOSE 1 to 4) and those with good outcome (GOSE 5 to 8). The temporal variations in vCSF total tau levels during the first week, most prominent in patients with bleak 1-year outcome, may reflect a secondary CNS injury.¹⁷ In the current investigation, we detected increases in vCSF total tau on days 8 to 14 in all patients compared with the initial days post trauma.

A correlation between TBI and AD has been noted particularly.²⁷ The intracellular neuropathologic component of AD diagnosis is the presence of neurofibrillary tangles consisting of subunits of hyperphosphorylated tau.²⁸ AD is neuropathologically defined by the presence of both amyloid β-protein (Aβ) as well as tau, and novel research now links Aβ and neurofibrillary tangles containing tau. It seems that Aβ promotes caspase cleavage of tau. Caspase enzymes are connected to apoptosis, at least in experimental neonatal studies.²⁹ One can only speculate if the tau peaks noted by others and us are induced by caspase and represent the ongoing apoptosis noted after TBI.

Thus, in the current study, 1-year neurologic assessment was chosen. To capture even small differences in neurologic status, a battery of well-known tests were utilized. We used 1) GOSE, reflecting an overall measure of outcome; 2) NIHSS, evaluating neurologic status; and 3) the Bartel Index, rating ADL. To further strengthen our outcome measurements, the assessment time (1 year) was very fixed, with variations only by days.

A limitation of the study is that of all included patients (n = 55), 16 were excluded as having an unstable ICP preventing collection of even small volumes of vCSF. Presumably these patients would have had a high vCSF total tau, and if so, these values would be in concert with the presented values. Unstable ICP also prevented occasional CSF samples in included patients. Further, patients were not transferred to our tertiary unit rapidly enough, rendering a late enclosure and vCSF collection until days 1 to 2.

Acknowledgment

The authors thank Ingrid Eiving, Ingrid Pettersson, and Catherine Ritzén for their contribution to computer assistance and for collecting laboratory and physical data.