Limits...
Utility of neuron-specific enolase in traumatic brain injury; relations to S100B levels, outcome, and extracranial injury severity

View Article: PubMed Central - PubMed

ABSTRACT

Background: In order to improve assessment and outcome prediction in patients suffering from traumatic brain injury (TBI), cerebral protein levels in serum have been suggested as biomarkers of injury. However, despite much investigation, biomarkers have yet to reach broad clinical utility in TBI. This study is a 9-year follow-up and clinical experience of the two most studied proteins, neuron-specific enolase (NSE) and S100B, in a neuro-intensive care TBI population. Our aims were to investigate to what extent NSE and S100B, independently and in combination, could predict outcome, assess injury severity, and to investigate if the biomarker levels were influenced by extracranial factors.

Methods: All patients treated at the neuro-intensive care unit at Karolinska University Hospital, Stockholm, Sweden between 2005 and 2013 with at least three measurements of serum S100B and NSE (sampled twice daily) were retrospectively included. In total, 417 patients fulfilled the criteria. Parameters were extracted from the computerized hospital charts. Glasgow Outcome Score (GOS) was used to assess long-term functional outcome. Univariate, and multivariate, regression models toward outcome and what explained the high levels of the biomarkers were performed. Nagelkerke’s pseudo-R2 was used to illustrate the explained variance of the different models. A sliding window assessed biomarker correlation to outcome and multitrauma over time.

Results: S100B was found a better predictor of outcome as compared to NSE (area under the curve (AUC) samples, the first 48 hours had Nagelkerke’s pseudo-R2 values of 0.132 and 0.038, respectively), where the information content of S100B peaks at approximately 1 day after trauma. In contrast, although both biomarkers were independently correlated to outcome, NSE had limited additional predictive capabilities in the presence of S100B in multivariate models, due to covariance between the two biomarkers (correlation coefficient 0.673 for AUC 48 hours). Moreover, NSE was to a greater extent correlated to multitrauma the first 48 hours following injury, whereas the effect of extracerebral trauma on S100B levels appears limited to the first 12 hours.

Conclusions: While both biomarkers are independently correlated to long-term functional outcome, S100B is found a more accurate outcome predictor and possibly a more clinically useful biomarker than NSE for TBI patients.

No MeSH data available.


Related in: MedlinePlus

a and b illustrate every patient as an individual line with the biomarker S100B (a) and neuron-specific enolase (NSE) (b) on the y-axis and time after trauma on the x-axis (hours). Colors are corresponding to outcome with darker color indicating a worse outcome, which becomes more favorable as it gets lighter. c and d are averages of the different GOS groups. As is shown by (a and b), there is limited data after 48 hours so it should be interpreted with caution. e and f are line plots indicating when to sample a biomarker after trauma to achieve maximum outcome prediction to long-term GOS1–5. The x-axis shows when in time since the trauma the sample of S100B (e) and NSE (f) was acquired (hours). The y-axis represents the Nagelkerke’s pseudo-R2 of a prediction model (proportional odds) toward GOS1–5, using either logged S100B (e) or NSE (f). The pseudo-R2 is calculated in each point using a sliding window incorporating 200 data points in chronological order. If a patient is represented more than once the sample is averaged, thus retaining independent points. The graph stops at approximately 48 hours as the later data points will be included in that final measurement. The line represents a locally weighted scatterplot smoothing (LOWESS), which is a nonlinear regression of the data points in the plots, a bootstrap confidence interval using two standard deviations is provided. Finally, in (g and h), which use the same method as in (e and f), but here the explained variance (y-axis) is how well the presence of extracranial multitrauma explains the levels of S100B (g) and NSE (h)
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC5015335&req=5

Fig1: a and b illustrate every patient as an individual line with the biomarker S100B (a) and neuron-specific enolase (NSE) (b) on the y-axis and time after trauma on the x-axis (hours). Colors are corresponding to outcome with darker color indicating a worse outcome, which becomes more favorable as it gets lighter. c and d are averages of the different GOS groups. As is shown by (a and b), there is limited data after 48 hours so it should be interpreted with caution. e and f are line plots indicating when to sample a biomarker after trauma to achieve maximum outcome prediction to long-term GOS1–5. The x-axis shows when in time since the trauma the sample of S100B (e) and NSE (f) was acquired (hours). The y-axis represents the Nagelkerke’s pseudo-R2 of a prediction model (proportional odds) toward GOS1–5, using either logged S100B (e) or NSE (f). The pseudo-R2 is calculated in each point using a sliding window incorporating 200 data points in chronological order. If a patient is represented more than once the sample is averaged, thus retaining independent points. The graph stops at approximately 48 hours as the later data points will be included in that final measurement. The line represents a locally weighted scatterplot smoothing (LOWESS), which is a nonlinear regression of the data points in the plots, a bootstrap confidence interval using two standard deviations is provided. Finally, in (g and h), which use the same method as in (e and f), but here the explained variance (y-axis) is how well the presence of extracranial multitrauma explains the levels of S100B (g) and NSE (h)

Mentions: Known outcome predictors of TBI such as age, pupil responsiveness, and GCS were all significant in univariate analysis and had an expected high pseudo-R2 (Table 2). S100B AUC 48 h exhibited a pseudo-R2 of 0.132, on par with Stockholm CT score and surpassed only by age (0.151). S100B showed better discrimination between all the different dichotomizations of outcome, whereas NSE is best at differentiating mortality (GOS1 vs. 2–5). The values of S100B and NSE over time for individual patients are shown in Fig. 1. In general, high or increasing levels are more correlated to a more unfavorable outcome, something that is better visualized for S100B (Fig. 1a) than for NSE (Fig. 1b). Similar to what was seen in Table 2, when the biomarker levels for the specific outcome groups were aggregated, S100B exhibits better discrimination between different levels of GOS, especially 24–36 hours after trauma (Fig. 1c), while NSE was only discriminates GOS1 as compared to GOS3–5 (Fig. 1d). GOS2 was excluded in Fig. 1c-d since only two patients were assessed as vegetative at long-term follow-up. S100B AUC 48 h (Fig. 2a) and NSE AUC 48 h (Fig. 2b) highlights the biomarker levels in different GOS groups, using conditional density plots (Fig. 2).Table 2


Utility of neuron-specific enolase in traumatic brain injury; relations to S100B levels, outcome, and extracranial injury severity
a and b illustrate every patient as an individual line with the biomarker S100B (a) and neuron-specific enolase (NSE) (b) on the y-axis and time after trauma on the x-axis (hours). Colors are corresponding to outcome with darker color indicating a worse outcome, which becomes more favorable as it gets lighter. c and d are averages of the different GOS groups. As is shown by (a and b), there is limited data after 48 hours so it should be interpreted with caution. e and f are line plots indicating when to sample a biomarker after trauma to achieve maximum outcome prediction to long-term GOS1–5. The x-axis shows when in time since the trauma the sample of S100B (e) and NSE (f) was acquired (hours). The y-axis represents the Nagelkerke’s pseudo-R2 of a prediction model (proportional odds) toward GOS1–5, using either logged S100B (e) or NSE (f). The pseudo-R2 is calculated in each point using a sliding window incorporating 200 data points in chronological order. If a patient is represented more than once the sample is averaged, thus retaining independent points. The graph stops at approximately 48 hours as the later data points will be included in that final measurement. The line represents a locally weighted scatterplot smoothing (LOWESS), which is a nonlinear regression of the data points in the plots, a bootstrap confidence interval using two standard deviations is provided. Finally, in (g and h), which use the same method as in (e and f), but here the explained variance (y-axis) is how well the presence of extracranial multitrauma explains the levels of S100B (g) and NSE (h)
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC5015335&req=5

Fig1: a and b illustrate every patient as an individual line with the biomarker S100B (a) and neuron-specific enolase (NSE) (b) on the y-axis and time after trauma on the x-axis (hours). Colors are corresponding to outcome with darker color indicating a worse outcome, which becomes more favorable as it gets lighter. c and d are averages of the different GOS groups. As is shown by (a and b), there is limited data after 48 hours so it should be interpreted with caution. e and f are line plots indicating when to sample a biomarker after trauma to achieve maximum outcome prediction to long-term GOS1–5. The x-axis shows when in time since the trauma the sample of S100B (e) and NSE (f) was acquired (hours). The y-axis represents the Nagelkerke’s pseudo-R2 of a prediction model (proportional odds) toward GOS1–5, using either logged S100B (e) or NSE (f). The pseudo-R2 is calculated in each point using a sliding window incorporating 200 data points in chronological order. If a patient is represented more than once the sample is averaged, thus retaining independent points. The graph stops at approximately 48 hours as the later data points will be included in that final measurement. The line represents a locally weighted scatterplot smoothing (LOWESS), which is a nonlinear regression of the data points in the plots, a bootstrap confidence interval using two standard deviations is provided. Finally, in (g and h), which use the same method as in (e and f), but here the explained variance (y-axis) is how well the presence of extracranial multitrauma explains the levels of S100B (g) and NSE (h)
Mentions: Known outcome predictors of TBI such as age, pupil responsiveness, and GCS were all significant in univariate analysis and had an expected high pseudo-R2 (Table 2). S100B AUC 48 h exhibited a pseudo-R2 of 0.132, on par with Stockholm CT score and surpassed only by age (0.151). S100B showed better discrimination between all the different dichotomizations of outcome, whereas NSE is best at differentiating mortality (GOS1 vs. 2–5). The values of S100B and NSE over time for individual patients are shown in Fig. 1. In general, high or increasing levels are more correlated to a more unfavorable outcome, something that is better visualized for S100B (Fig. 1a) than for NSE (Fig. 1b). Similar to what was seen in Table 2, when the biomarker levels for the specific outcome groups were aggregated, S100B exhibits better discrimination between different levels of GOS, especially 24–36 hours after trauma (Fig. 1c), while NSE was only discriminates GOS1 as compared to GOS3–5 (Fig. 1d). GOS2 was excluded in Fig. 1c-d since only two patients were assessed as vegetative at long-term follow-up. S100B AUC 48 h (Fig. 2a) and NSE AUC 48 h (Fig. 2b) highlights the biomarker levels in different GOS groups, using conditional density plots (Fig. 2).Table 2

View Article: PubMed Central - PubMed

ABSTRACT

Background: In order to improve assessment and outcome prediction in patients suffering from traumatic brain injury (TBI), cerebral protein levels in serum have been suggested as biomarkers of injury. However, despite much investigation, biomarkers have yet to reach broad clinical utility in TBI. This study is a 9-year follow-up and clinical experience of the two most studied proteins, neuron-specific enolase (NSE) and S100B, in a neuro-intensive care TBI population. Our aims were to investigate to what extent NSE and S100B, independently and in combination, could predict outcome, assess injury severity, and to investigate if the biomarker levels were influenced by extracranial factors.

Methods: All patients treated at the neuro-intensive care unit at Karolinska University Hospital, Stockholm, Sweden between 2005 and 2013 with at least three measurements of serum S100B and NSE (sampled twice daily) were retrospectively included. In total, 417 patients fulfilled the criteria. Parameters were extracted from the computerized hospital charts. Glasgow Outcome Score (GOS) was used to assess long-term functional outcome. Univariate, and multivariate, regression models toward outcome and what explained the high levels of the biomarkers were performed. Nagelkerke’s pseudo-R2 was used to illustrate the explained variance of the different models. A sliding window assessed biomarker correlation to outcome and multitrauma over time.

Results: S100B was found a better predictor of outcome as compared to NSE (area under the curve (AUC) samples, the first 48 hours had Nagelkerke’s pseudo-R2 values of 0.132 and 0.038, respectively), where the information content of S100B peaks at approximately 1 day after trauma. In contrast, although both biomarkers were independently correlated to outcome, NSE had limited additional predictive capabilities in the presence of S100B in multivariate models, due to covariance between the two biomarkers (correlation coefficient 0.673 for AUC 48 hours). Moreover, NSE was to a greater extent correlated to multitrauma the first 48 hours following injury, whereas the effect of extracerebral trauma on S100B levels appears limited to the first 12 hours.

Conclusions: While both biomarkers are independently correlated to long-term functional outcome, S100B is found a more accurate outcome predictor and possibly a more clinically useful biomarker than NSE for TBI patients.

No MeSH data available.


Related in: MedlinePlus