Prognosis Value of Plasma S100B Protein Levels after Subarachnoid Aneurysmal Hemorrhage

The study was approved by our local ethical committee (Comité de Protection des Personnes, Pitié-Salpêtrière, Paris, France). In accordance with the Helsinki Declaration, written informed consent was obtained from the patient or patient’s next of kin.

The inclusion period was from December 2003 through October 2004. Inclusion criteria were clinical history of SAH within the last 2 days before admission with evidence of bleeding on CT and presence of an aneurysm at cerebral angiography, age 18 yr or older, and treatment by surgery or coiling within the 48 h after admission. All patients were admitted to our neurosurgical intensive care unit after initial treatment. Exclusion criteria were admission later than 48 h after bleeding, surgery or coiling later than 48 h after admission, and therapeutic abstention decision.

At admission, clinical severity was assessed using the World Federation of Neurological Surgeons (WFNS) score19as follows: 1 = Glasgow coma scale (GCS) score of 15, no motor deficit; 2 = GCS score of 13–14, no motor deficit; 3 = GCS score of 13–14 and any motor deficit or aphasia; 4 = GCS score of 7–12, with or without motor deficit; and 5 = GCS score of 3–6, with or without motor deficit. Hemodynamic status was evaluated by clinical examination, arterial blood gas analysis, chest radiograph, and echocardiograph when appropriate. Cardiac troponin Ic blood concentration (Stratus Analyzer; Dade, Massy, France) was systematically measured. The initial CT was reviewed by an independent radiologist blinded to clinical history, therapeutics, and S100B values and classified according to the modified Fisher score as follows: grade 1 = no subarachnoid blood; grade 2 = broad diffusion of subarachnoid blood; grade 3 = with clots or thick layers of blood; grade 4 = intraventricular hemorrhage or intracerebral hematoma, no clot; and grade 5 = intraventricular hemorrhage or intracerebral hematoma with clot.20Neurologic outcome was assessed using the Glasgow outcome scale (GOS) score21at discharge from the intensive care unit and at 6 months by a same anesthesiologist (CC). GOS was defined as follows: 1 = death; 2 = persistent vegetative state; 3 = severe disability; 4 = moderate disability; and 5 = good recovery.

The type of treatment (surgery or coiling) was decided according to both location and size of the aneurysm by the neurosurgeon and the neuroradiologist. Coiling was preferred for patients in whom it was deemed feasible.22Surgery was performed in the remaining cases. All patients received intravenous nimodipine at a dose of 2 mg/h from admission until at least completion of the study, except during periods of uncontrolled increased intracranial pressure during which intravenous nimodipine was discontinued. Seizures were systematically prevented by gabapentin (600 mg × 3, per os). A central venous and an arterial catheter were inserted in most of the patients before and/or after surgery or coiling. Normovolemia was maintained through the systematic administration of intravenous physiologic electrolyte solution at 1,000–2,000 ml/day and colloids if necessary. Additional fluids were given in case of hypovolemia diagnosed on clinical symptoms and increased systolic pressure variation. After treatment, systolic arterial blood pressure was maintained above 130–160 mmHg by continuous infusion of norepinephrine as needed. Great caution was taken to avoid hyperthermia (above 38.5°C) and hyperglycemia (above 7.5 mm) through the administration of acetaminophen and insulin as needed. In ventilated patients, arterial carbon dioxide partial pressure (Paco²) was maintained between 35 and 40 mmHg, and peripheral oxygen saturation (Spo²) was maintained above 97%. Oral or enteral nutrition was begun as soon as possible. An external ventricular drain (Sophysa, Orsay, France) was inserted in case of hydrocephaly on CT and in patients with a high WFNS grade (WFNS score of 3–5). The line was connected to an external pressure strain gauge to continuously monitor intracranial pressure according to a recently published protocol.23Increased intracranial pressure was treated by cerebrospinal fluid drainage, mechanical ventilation, reinforcement of sedation, and, rarely, moderate hypothermia. CT was performed whenever clinical deterioration occurred to search for secondary complications such as hydrocephalus or ischemia.

Vasospasm was suspected on clinical deterioration, fever, appearance of new symptoms (cephalgia, confusion, seizure, or motor deficit), mean transcranial Doppler velocities above 120 cm/s or a daily change in mean transcranial Doppler velocities above 50 cm/s,24and confirmed by cerebral angiography. Each vasospasm episode was treated with intraarterial administration of nimodipine as recently described.25This therapy was repeated if necessary. Angioplasty was used as a second-line therapy when nimodipine was judged insufficient.

A venous blood sample for S100B level determination was systematically withdrawn each day from day 1 to day 8. The first sample was taken within 12 h after admission. By reference, the first 24 h after admission defined day 1. Venous blood was collected on blood collection devices and centrifuged within 2 h (1,800 g for 10 min at +4°C). S100B concentrations were measured with an immunoluminometric sandwich assay on a LIA-mat 300 analyzer (Byk-Sangtec France Laboratories, Le Mée sur Seine, France) using the manufacturer’s reagents.26Briefly, tubes coated with anti-S100B chain were incubated with 100 μl per tube of samples for 60 min. After the tubes were washed, aminobutylethylisoluminol-labeled anti-S100B was added for 120 min and, after a final washing step, the luminescence of aminobutylethylisoluminol when oxidized in the presence of deuteroferriheme and hydrogen peroxide was measured. The amount of S100B in the samples was calculated using standard curves prepared with calibrators with known concentrations of the proteins. The detection limit of the assay is 0.02 μg/l. Values in healthy individuals are considered to be below 0.15 μg/l. Intraassay and interassay coefficient of variation of the measurement were less than 7 and 9%, respectively.

In patients who did not complete the 8-day period because their good clinical status allowed them to be discharged from the intensive care unit to the ward or because death occurred within the study period, the last value obtained was taken to replace missing values until day 8 in the statistical analysis according to the neighboring nonmissing values method.

For statistical purposes, WFNS, Fisher, and GOS scores were dichotomized (WFNS score of 1–2 vs.  WFNS score of 3–5, Fisher score of 1–3 vs.  Fisher score of 4–5, and GOS score of 4–5 vs.  GOS score of 1–3). Initial S100B values were compared between groups by analysis of variance and the Fisher exact method. The effects of WFNS score, Fisher score, GOS score, treatment, and spasm on S100B concentrations over time were analyzed by two-way repeated-measures analysis of variance for one grouping factor, i.e. , WFNS (WFNS score of 1–2 vs.  WFNS score of 3–5) and for one within factor, i.e. , time (day 1 to day 8).

The receiver operating characteristic (ROC) curves were used to determine the best threshold for initial and mean daily values of S100B to predict poor outcome. Assessment of the diagnostic performance of initial and mean daily values of S100B was analyzed by calculating the sensitivity, specificity, positive and negative predictive values, accuracy, and their 95% confidence intervals (CIs). For clinical reasons, specificity was preferred to sensitivity to determine the best threshold, i.e. , we choose the threshold that maximally reduced the number of patients having a good outcome despite a mean daily value of S100B above threshold.

After dichotomization of the mean daily values of S100B in two groups (according to the results of the previous analysis), contingency tables were used for the following categorical variables: initial WFNS and Fisher scores, age (below and above 50 yr), aneurysm location (middle cerebral artery vs.  others), troponin Ic levels (below and above 0.10 μg/l), vasospasm (present or absent at angiography), and type of treatment (coiling vs.  surgery).

A first multivariate analysis was performed using backward stepwise logistic regression where all of the variables suggested by the univariate analysis (P < 0.2) were entered into the model to predict an increased mean daily value of S100B. A second multivariate analysis including dichotomized mean daily S100B as a variable was performed to predict outcome at 6 months. Calibration and discrimination of the logistic models were assessed using Hosmer-Lemeshow statistics and ROC curves, respectively.

Data are expressed as mean ± SD. All tests were two sided, and a P  value of less than 0.05 was considered significant. Statistical analyses were performed using JMP IN 5.1 statistical software (SAS Institute Inc., Cary, NC).

During the study period, 101 patients were admitted to our intensive care unit with an SAH diagnosis. Of these, 27 were excluded for the following reasons: delay of more than 48 h between admission and onset of symptoms (n = 16), delay of more than 48 h between admission and date of intervention (n = 7), or no surgical or endovascular treatment (n = 4). Eventually, 74 patients were included (mean age, 48 ± 11 yr; 32 men and 42 women). Among these, 53 patients (72%) underwent coiling and 21 patients (28%) underwent surgical clipping. Table 1shows the main characteristics of this population. Patients with middle cerebral artery aneurysms underwent surgery more often than coiling, whereas the opposite was true for all of the other locations.

Of these 74 patients, 14 stayed less than 8 days in the intensive care unit because of death (n = 5) or early discharge to ward (n = 9). Among these, 4 patients had values until day 7, 5 until day 6, 3 until day 5, and 2 until day 4. Values of S100B continued to increase in patients who finally died contrary to those who were discharged from the intensive care unit before day 8 who maintained low values over time (below 0.4 μg/l; P < 0.001; fig. 1). Figure 2shows S100B values according to WFNS score, Fisher score, and location of the aneurysm. For the initial S100B value, there were significant differences between groups for both WFNS and Fisher scores but not for location of the aneurysm. The time course of S100B was significantly different between patients with low and high WFNS and Fisher scores (P < 0.01). Significant differences were observed in the time course of S100B between the three locations of the aneurysm (P < 0.02), the highest values being observed in patients with a middle cerebral artery aneurysm, intermediate values being observed in patients with aneurysms of the anterior circulation, and the lowest values being observed in patients with internal carotid or posterior communicating artery aneurysms.

The initial troponin Ic value was significantly linked to the S100B initial value stratified into three groups (initial value < 0.2 μg/l, between 0.2 and 0.4 μg/l, and > 0.4 μg/l; fig. 3A). There were significant differences in S100B time course between patients with low and high initial troponin Ic values (P = 0.03; fig. 3B).

S100B time course values were significantly higher in patients who underwent surgical clipping than in those who underwent coiling, and this difference remained true after stratification for WFNS and Fisher scores (figs. 4A and B). As shown in table 2, the percentage of patients having a mean daily value of S100B above the threshold of 0.4 μg/l was higher in patients with high WFNS and Fisher scores, with a middle cerebral artery aneurysm, with a high troponin Ic value at admission, and treated surgically. In the multivariate analysis, only surgery (odds ratio = 6.1; 95% CI, 1.7–21.9) remained independently associated with a mean daily value of S100B above the threshold of 0.4 μg/l.

As shown in figure 5A, S100B initial values were significantly correlated with outcomes at discharge and at 6 months (both P < 0.03). As shown in figure 5B, S100B time course was also significantly correlated with outcome scores at discharge from the intensive care unit and at 6 months. WFNS score, Fisher score, mean daily S100B values above 0.4 μg/l (all P < 0.001), initial S100B values above 0.4 μg/l (P < 0.01), age (P < 0.05), and initial troponin value (P < 0.05) were significantly correlated with outcome at 6 months in the univariate model. Neither the location nor the type of treatment and the occurrence of spasm were predictive factors. As shown in table 3, after multivariate analysis, the initial WFNS score, a mean daily S100B value above 0.4 μg/l, and age were significantly predictive of outcome at 6 months. Calibration and discrimination of the model were appropriate as shown by the Hosmer-Lemeshow statistic (4.17; P = 0.84) and the area under the ROC curve (0.88; 95% CI, 0.80–0.96).

The ROC curve showed that initial and mean daily values of S100B significantly predicted a poor outcome (fig. 6). The area under the curve was significantly higher for the mean daily value of S100B. The diagnostic performances of initial and mean daily values of S100B above 0.4 μg/l in predicting a poor outcome are shown in table 4. This threshold of 0.4 μg/l was chosen to provide a high specificity. In contrast, the threshold of 0.2 μg/l was the one that minimized the distance to the ideal point (sensitivity = specificity = 1) on the ROC curve. For this threshold of 0.2 μg/l, specificity and sensitivity were both of 0.62 for the initial value and of 0.72 and 0.79 for the mean daily value, respectively. A threshold of 0.5 μg/l for the mean daily value increased specificity to 0.92 but reduced sensitivity to 0.29.

Finally, the S100B time course was not significantly different in patients with (n = 27, 36%) and without vasospasm (n = 47, 64%), although this was not true when this analysis was performed in the subgroup of patients having an initial S100B value below 0.4 μg/l (data not shown). Among patients with a diagnosed vasospasm, 17 patients maintained daily S100B below 0.4 μg/l over time, and none of these patients died. Ten patients had at least one value of S100B above this threshold, and five of them died.

Our data confirm that S100B serum concentrations correlate well with initial SAH severity evaluated either clinically (WFNS grading scale) or by CT (Fisher score). Furthermore, the mean value computed during the first 8 days was an independent predictive factor for GOS outcome at 6 months.

To our knowledge, except for three studies in which SAH patients were analyzed together with various neurologic disorders,27–29only four studies11–14and a recent case report16have specifically addressed the usefulness of S100B level measurement in patients with SAH. These studies had variable S100B sampling (with sometimes a lack of precision regarding timing and number of samples) as well as heterogeneous clinical evaluation and treatment procedures. For all of these reasons, comparison with the current set of data is difficult. The only study with an important series of SAH patients using plasma rather than cerebrospinal fluid S100B or extracellular fluid (microdialysis) measurements was performed by Wiesmann et al.  11Our study confirms their observation correlating initial S100B values and clinical condition. In addition, our results show that the time course of S100B measured over an 8-day period correlates with WFNS and Fisher scores indicating a long-lasting effect of the initial neurologic injury induced by bleeding. Together, patients with middle cerebral artery aneurysm had higher S100B concentrations over time as compared with other locations. This is concordant with our current knowledge regarding the influence of aneurysm location on outcome.30,31Indeed, in the retrospective study from Kopera et al. ,3132% of the patients with a middle cerebral aneurysm had intracranial hemorrhage, a known predictive factor for poor outcome.32S100B time course correlated better with WFNS and Fisher scores as well as location than the initial value per se .

S100B time course was significantly correlated with initial troponin Ic values: Patients with high initial troponin Ic (> 0.10 μg/l) had also higher levels of S100B over time. This result is in accordance with previous studies where high troponin levels have been linked with the occurrence of neurologic deficit.33,34This also confirms the fact that SAH is a multiorgan dysfunction entity whose consequences may involve other organ systems, such as the cardiovascular system.35 

Major differences were found in S100B time course between patients treated by coiling compared with those who underwent surgery. Indeed, although departing at equivalent values, S100B plasma concentration decreased significantly more rapidly in coiled patients compared with those who underwent clipping. This was all the more true after stratification according to the initial WFNS or Fisher grading. If these results were to be confirmed, they would argue for a neuroprotective effect of coiling in SAH. These data are concordant with the results of the International Study of Aneurysm Treatment trial22that demonstrated that coiling improved by 23% the chances for an independent survival at 1 yr as compared with clipping in patients with ruptured aneurysms, mostly in good clinical grades. Nevertheless, it should be pointed out that the decision to perform coiling whenever possible may have introduced a selection bias in our study. Regarding the multivariate analysis that considered all factors potentially correlated with S100B mean daily values, only the type of treatment (surgery vs.  coiling) was significant. This suggests that coiling reduces brain damage as compared with surgical clipping. Some authors have suggested that surgery by itself may increase S100B through the nonspecific secretion of an astrocyte.36However, this explanation seems incompatible with the duration of increased S100B values after surgery, lasting the entire study period.

Our results show that a time course evaluation of S100B has a useful and independent prognostic value to assess outcome in SAH patients. Indeed, initial and mean daily values above 0.4 μg/l significantly predicted a poor outcome. This threshold is close to the one defined by Raabe et al.  36at 0.5 μg/l for various neurologic disorders and by Foerch et al.  37at 0.35 μg/l in ischemic stroke. Decreasing the cutoff to 0.2 μg/l improved the sensitivity at the price of the specificity. Moreover, age, WFNS initial score, and mean daily S100B value above 0.4 μg/l were the three independent predictors of outcome at 6 months. This is in accordance with a recent published case report suggesting S100B protein as a surrogate marker of brain injury.16 

In our study, direct relation between S100B values and vasospasm episodes remain unclear. However, the power of our study may have been insufficient to detect a significant association. Moreover, it should be pointed out that the initial surge of S100B could “mask” in some manner the evidence of secondary ischemic complications as assessed by a secondary increase in S100B.

Regarding the limitations of our study, a more prolonged study period might have likely increase the prognostic value of S100B time course because some patients might have experienced late-onset vasospastic episodes that could also impair the prognosis. It must also be stated that because of the very short half-life of S100B, estimated at 2 h,38we cannot exclude that transient surges in S100B were missed with the once-daily measure that we used in our study.

In conclusion, measurement of S100B plasma concentration is a good indicator of severity and an independent predictor of 6-month outcome in SAH patients. Multiple daily measurements with rapid results are already available, thus giving the opportunity for clinicians to strictly monitor brain ischemia in SAH patients. Similar to troponin and heart disease,39S100B may be an important biologic marker for the future improvement of clinical management and outcome in SAH patients.