Secondary Logo

Journal Logo

Clinical Investigations

Glucose and Lactate Concentrations in Cerebrospinal Fluid After Traumatic Brain Injury

Lozano, Angels MD*,†; Franchi, Federico MD*,‡; Seastres, Ramon J. MD§; Oddo, Mauro MD; Lheureux, Olivier MD*; Badenes, Rafael MD, PhD; Scolletta, Sabino MD; Vincent, Jean-Louis MD, PhD*; Creteur, Jacques MD, PhD*; Taccone, Fabio S. MD, PhD*

Author Information
Journal of Neurosurgical Anesthesiology: April 2020 - Volume 32 - Issue 2 - p 162-169
doi: 10.1097/ANA.0000000000000582


Neurological outcome in patients with traumatic brain injury (TBI) can be influenced by the development of secondary brain damage. Cerebral hypoxia, as a result of hypotension, microvascular abnormalities, increased intracranial pressure (ICP) or seizures, promotes activation of neuroinflammation and of glutamate-mediated excitotoxicity, as well as increased permeability of the blood-brain barrier (BBB).1 Secondary brain injuries can also be related to systemic disturbances (eg, hypoxemia, hypocapnia, fever, anemia, hyponatremia), which may exacerbate the cerebral damage.2 Importantly, these secondary brain insults are often preventable or reversible.3

Recent data in TBI patients monitored with positron emission tomographic scan and/or cerebral microdialysis have demonstrated that the pathophysiology of secondary cerebral damage is complex and may be related to an uncoupling of cerebral blood flow with the increase in energy needs (ie, hyperglycolysis without increased tissue perfusion) of cerebral cells, which may result in energy dysfunction and reducedcerebral extracellular concentrations of glucose (ie, neuroglycopenia), irrespective of systemic glucose concentrations.4 Moreover, brain energetic supply also relies on the use of other substrates, such as lactate,5 so that the relative importance of glucose and lactate utilization in the acutely injured brain remains debated. Therefore, the high cerebral lactate concentrations reported in TBI patients may be secondary to hypoxic conditions, as with reduced tissue perfusion and glucose supply, or to nonhypoxic conditions secondary to increased metabolic activity that induces enhanced lactate uptake from the circulation as well as increased lactate production in astrocytes.6–9

Cerebral microdialysis and positron emission tomographic scans are not widely available and not routinely used in many critical care facilities. Analysis of glucose and lactate concentrations in the cerebrospinal fluid (CSF) is more easily performed. The glucose/lactate ratio may be of value in the detection of episodes of cerebral hypoxia and prediction of outcome, as shown from studies performed using microdialysis.10,11 The objective of this study was, therefore, to determine whether changes in CSF cerebral glucose and lactate over time could provide valuable information with regard to patient therapy or outcome.


Study Population

We reviewed data from all consecutive TBI patients admitted to our 35-bed medicosurgical intensive care unit (ICU) over a 4-year period (2011 to 2014). Patients were included if they met the following criteria: (a) age of above 18 y; (b) presence of an external ventricular drain (EVD) for ICP monitoring; (c) daily analysis of CSF from the EVD including glucose and lactate concentrations for at least 4 consecutive days; (d) concomitant measurements of arterial blood glucose and lactate concentrations within 30 minutes of CSF sampling. The purpose of the CSF glucose and lactate level analysis was, in general, to assess the presence of central nervous system infections and not to guide TBI management. EVD insertion was performed in TBI patients admitted with a Glasgow Coma Scale (GCS) < 9 and abnormal initial cerebral computed tomographic (CT) scan. The local Ethics Committee approved the study (Protocol number: 2017/033), but waived the need for informed consent because of its retrospective nature. The study was conducted according to the Declaration of Helsinki.

Data Collection

We recorded demographics and clinical characteristics, the presence of comorbid disease and important prognostic factors for the patients included in the study, including glucose and hemoglobin concentrations on admission, initial GCS, pupillary reactivity on admission, and presence of hypotension or hypoxemia before admission.12 The severity of intracranial lesions on admission was graded according to the Marshall CT-score, and the Injury Severity Score (ISS) for the head, chest, and abdomen was assessed for each patient.13 We also noted whether continuous insulin therapy (target glucose concentrations: 110 to 150 mg/dL or 6.1 to 8.3 mmol/L) was being administered at the time of CSF/blood sampling for glucose and lactate. ICP, mean arterial pressure (MAP) and cerebral perfusion pressure values, temperature and the development of complications (eg, seizures [either convulsive: clinically recognized or nonconvulsive: EEG recording], infection, intracranial hypertension) were recorded. Blood glucose and lactate concentrations were measured from the arterial blood gas analyzer (GEM Premier 4000 system with Plus technology, Instrumentation Laboratory, Bedford, MA—limits of quantification for glucose from 4 to 685 mg/dL and for lactate from 0.3 to 17.0 mmol/L). CSF glucose and lactate were measured within 60 minutes from sampling at the central laboratory using a standard enzymatic method. The ratio between CSF glucose and lactate concentrations as well as the ratios between CSF and blood glucose and CSF and blood lactate concentrations were also calculated. Data were collected daily for 4 consecutive days. CSF drainage was used intermittently only in cases of elevated ICP (ie, >20 mm Hg) as a first-line therapy, and was never used as continuous therapy.


Ventriculitis was diagnosed if a microorganism was identified in the CSF by Gram-stain and from culture of the proximal EVD tip, and the patients had fever (>38.0°C) and an increased CSF-white cell count (>6 cells/mm3). Shock was defined as the requirement of vasopressors for >6 hours despite adequate fluid resuscitation to maintain MAP>65 mm Hg, associated with increased blood lactate concentrations (>2 mmol/L) and signs of organ dysfunction. Cerebral ischemia was defined as the appearance of new infarctions on CT or MRI, independent of surgical procedures. Intracranial hypertension was defined as an increase in ICP>20 mm Hg for >30 minutes in the absence of confounders (eg, pain, agitation, hypertension, shivering, fever etc.). Hyponatremia was defined as serum sodium concentration <135 mEq/L. Hypoglycemia was defined as blood glucose <3.5 mmol/L.


Outcome variables included ICU mortality, ICU length of stay, hospital length of stay, and extended Glasgow Outcome Scale (GOS) at 3 months after injury (unfavorable outcome was defined as an extended GOS of 1 to 4), which was obtained from the medical chart of the patients or by telephone call to the general practitioner.

Data Processing and Statistical Analysis

Data are expressed as mean±SD, median and interquartile range or count (percentage), as appropriate. For continuous variables, normality assumption checking was performed by inspection of residual and normal plots, and using the Kolmogorov-Smirnov test. Differences between groups were assessed using the analysis of variance, Kruskal-Wallis test, Student’s T test, Mann-Whitney test, χ² test, or Fisher’s exact test, as appropriate. The time-courses of glucose and lactate concentrations were analyzed using generalized estimating equation models corrected for missing values and differences between survivors and nonsurvivors (ICU mortality), and patients with favorable or unfavorable neurological outcome, were reported. The differences at each time-point were evaluated using a Mann-Whitney test. Linear regression analysis was used to investigate the relationship between blood and CSF glucose and lactate concentrations. Two multivariable logistic regression analyses (ie, one with mortality and the second with unfavorable outcome as the dependent variables) were performed in all patients; colinearity between variables was excluded before modeling. Only variables associated with a higher risk of mortality or unfavorable outcome (P<0.05) on a univariate basis (reported in Table 1) were introduced in the multivariable models. Odds ratios (ORs) with 95% confidence intervals (CIs) were computed. Receiver operating characteristic curve analysis was used to estimate the optimum cut-off values for the variables retained in the multivariable models (ie, CSF glucose/lactate and CSF lactate) and their respective sensitivity (SE) and specificity (SP) for prediction of mortality and unfavorable outcome. Multivariable analyses were then performed to identify variables independently associated with CSF lactate concentrations and glucose/lactate ratio above the identified cut-offs, using only variables associated with a higher risk of mortality or unfavorable outcome (P<0.05) on a univariate basis (reported in Supplemental Table 3, Supplemental Digital Content 1, All reported P-values are 2-sided and a P-value of <0.05 was considered statistically significant. Data were analyzed using IBM SPSS Statistics software, version 22 for Windows (IBM, Armonk, NY).

Characteristics of the Study Population, According to Hospital Mortality and Long-term Neurological Outcome


Study Population

During the study period, 276 patients were admitted to our ICU following TBI. Of these patients, 151 required an EVD and 56 met the inclusion criteria. The clinical data of the patients included in the study are shown in Table 1. Most (41/56) suffered from multiple trauma. Most EVDs (50/56) were placed on the day of ICU admission and EVDs were kept in situ for 10 (6 to 14) days. The median GCS on admission was 7 (3 to 10) and most initial CT scans were classified as Marshall score V or VI (30/57). Twenty patients (36%) died during the ICU stay, and 33 (59%) had an unfavorable outcome (Table 1).

Blood and CSF Glucose and Lactate

There was a modest but statistically significant correlation (R2=0.07; P=0.04) between blood and CSF glucose concentrations, and a larger correlation between blood and CSF lactate concentrations (R2=0.32, P<0.001) on day 1 (ie, the first day of CSF sampling) (Fig. 1). There were no statistically significant differences in blood and CSF glucose and lactate concentrations in patients who were receiving insulin therapy on the first day of CSF assessment (n=46) and those who were not (n=10) (Supplemental Table 1, Supplemental Digital Content 1, No patients developed hypoglycemia.

Correlation between CSF and blood glucose (A) and CSF and blood lactate (B) concentrations (N=56) on the first day on CSF sampling. CSF indicates cerebrospinal fluid.

CSF Glucose, CSF Lactate and Outcome

On day 1, CSF and blood glucose concentrations were similar in survivors and nonsurvivors, but CSF lactate concentrations (4.1 [interquartile range: 2. to 6.0] vs. 2.3 [2.0 to 3.7] mmol/L; P=0.003) and blood lactate concentrations (1.8 [1.0 to 3.1] vs. 1.1 [0.7 to 1.4] mmol/L; P=0.013) were higher in nonsurvivors than in survivors; similar patterns were seen when comparing patients with unfavorable versus those with favorable outcomes. The correlations between glucose and lactate concentrations in CSF and blood were similar in survivors and nonsurvivors, and in patients with favorable and unfavorable outcomes (data not shown). The CSF glucose/lactate ratio was lower in nonsurvivors than in survivors (1.03 [0.82 to 1.67] vs. 1.94 [1.35 to 2.34]; P=0.001) and in patients with unfavorable than in those with favorable outcomes (1.12 [0.85 to 1.77] vs. 2.12 [1.53 to 2.58]; P<0.001) (Table 2). CSF protein concentrations were significantly higher in nonsurvivors than in survivors and in patients with unfavorable than in those with favorable outcome (Table 2). Nonsurvivors had higher CSF lactate concentrations and lower CSF glucose/lactate ratios than survivors throughout the study period (Fig. 2, Supplemental Table 2, Supplemental Digital Content 1,, as did patients with unfavorable compared with those with favorable outcome.

Data on Cerebrospinal Fluid (CSF) and Blood Samples Collected on the First Day After the Placement of an Extraventricular Drainage, According to Hospital Mortality and Long-term Neurological Outcome
Cerebrospinal fluid (CSF) glucose/lactate ratio (A), CSF lactate (B), or CSF glucose (C) over 4 days, according to neurological outcome or hospital mortality. FO indicates favorable; UO, unfavorable. The P-value presented in each figure represents the 2-way analysis of variance analysis. *P<0.05 and **P<0.01 with post hoc Bonferroni analysis.

Multivariable Analyses

In the multivariable analysis, CSF glucose/lactate ratio on day 1 was the only variable independently associated with mortality (OR: 0.22, 95% CI: 0.08-0.61; P=0.003) and CSF lactate concentration on day 1 was the only variable independently associated with unfavorable outcome (OR: 4.86, 95% CI: 1.86-12.67; P=0.001). After a further adjustment for additional confounders on outcome (ICP and cerebral perfusion pressure on day 1, age and pupillary reactivity on admission), the OR of the CSF glucose/lactate for mortality was 0.18 (CI: 0.06-0.58; P=0.004), and the OR of CSF lactate for unfavorable outcome was 4.54 (CI: 1.66-12.43; P=0.003). When the lowest CSF glucose/lactate ratio or highest CSF lactate concentration over the 4 days was used in the multivariable analyses, similar results were observed (data not shown).

The area under the receiver operating characteristic curve (AUC) of the CSF glucose/lactate ratio on day 1 to predict mortality was 0.78 (95% CI: 0.65-0.91; P<0.001) (Fig. 3A); the optimum cut-off was 1.47, with a SE of 70% and a SP of 72%. The AUC of CSF lactate on day 1 to predict unfavorable neurological outcome was 0.89 (95% CI: 0.80-0.97; P<0.001) (Fig. 3B). A CSF lactate >3.0 mmol/L on day 1 predicted an unfavorable neurological outcome with a SE of 70% and a SP of 87%.

Receiver operating curve (ROC) for cerebrospinal (CSF) lactate (A, area under the curve [AUC]: 0.89; 95% CI: 0.89-0.97; P<0.001) or CSF glucose/lactate ratio (B, AUC: 0.78; 95% CI: 0.65-0.91; P<0.001) to predict long-term unfavorable outcome or hospital mortality, respectively.

In multivariable analysis, high blood lactate and CSF protein concentrations and low MAP were independent predictors of a CSF glucose/lactate <1.47 on day 1 (Table 3 and Supplemental Table 3, Supplemental Digital Content 1, High blood lactate and CSF protein concentrations and low body temperature were independent predictors of a CSF lactate >3.0 mmol/L on day 1.

Multivariable Analysis to Identify Independent Predictors of Cerebrospinal Fluid (CSF) Glucose/Lactate <1.47 and CSF Lactate >3.0 mmol/L


In this study, cerebral glucose and lactate concentrations correlated modestly with blood glucose and lactate concentrations. High CSF lactate concentrations and a low CSF glucose/lactate ratio were significant predictors of poor outcome in these TBI patients. High CSF protein and blood lactate concentrations were independently associated with these predictors of outcome.

The initial systemic metabolic response to TBI includes “stress hyperglycemia,” which is directly related to TBI severity.14 Despite the association between high glucose concentrations during the hospital stay and increased mortality in critically ill patients, tight glycemic control has not been shown to improve neurological outcome in patients with TBI.15–17 As in studies evaluating cerebral glucose concentrations using microdialysis techniques, we observed that CSF glucose concentrations were weakly correlated with systemic concentrations, which confirms that the main determinant of cerebral glucose availability is systemic supply. As glucose supply to the brain is independent of insulin, but relies on specific glucose transporters located on capillaries and the BBB, monitoring glucose concentrations to avoid neuroglycopenia could be useful in TBI patients. Previous studies using microdialysis techniques have shown that “tight” blood glucose control (ie, 4.4 to 6.7 mmol/L) was associated with an increased incidence of neuroglycopenia and metabolic crisis, as suggested by increased lactate/pyruvate ratio.18,19 Most of the patients enrolled in the present study had blood glucose concentrations >6 mmol/L (108 mg/dL) regardless of insulin administration; therefore, we were unable to evaluate CSF glucose and lactate concentrations at lower glycemic ranges. Importantly, we have no reliable data to evaluate whether CSF glucose concentrations are correlated with microdialysis glucose concentrations and how monitoring of such concentrations could be helpful to better titrate insulin therapy in TBI patients.

CSF glucose concentrations were not related to outcome. Whether the blood glucose >6 mmol/L in all patients was sufficient to maintain brain tissue metabolism during the early phase after severe TBI in our patients or whether CSF glucose is not a sensitive marker of cerebral distress in this setting should be evaluated in future trials. The small number of patients included in this cohort may also have limited the power of the statistical analysis. Kurtz et al20 reported that a reduced brain-blood glucose ratio may induce cerebral metabolic distress and thereby increase mortality after severe TBI; brain glucose concentrations were more closely related to brain/serum glucose ratios than to absolute serum glucose concentrations implicating impaired glucose transport as a potentially important cause of neuroglycopenia. Another study showed that also high brain extracellular glucose concentrations were associated with unfavorable outcome.10 As it appears that both high and low cerebral glucose concentrations could be associated with unfavorable outcome,21 large studies evaluating CSF and microdialysis glucose monitoring in heterogeneous populations of TBI patients are needed to better identify optimal glucose ranges and how glycemic control may influence these ranges.

High serum lactate concentrations have also been associated with poor outcome in critically ill patients and mainly reflect tissue hypoperfusion, hyperglycolysis, or adrenergic stimulation.22 We observed that only CSF lactate and CSF glucose/lactate ratio were independently associated with poor outcomes. More specifically, CSF glucose concentrations were similar in patients with favorable and unfavorable outcomes, whereas lactate concentrations were significantly higher in those with poor neurological outcome. Some small studies have assessed the relationship of CSF lactate concentrations in the acute phase after TBI with patient outcome. In an experimental study, increased brain lactate concentration was considered to be secondary to the degree of brain injury and deranged cerebral metabolism in the absence of substrate limitations (ie, normal glucose concentrations).23 One potential explanation for this finding is mitochondrial dysfunction post-TBI, which could result in lactate overproduction and reduced lactate clearance from the injured brain tissue via the choroidal plexus.24 More recently, Sala and colleagues showed that >40% of cerebral microdialysis samples in TBI patients had elevated brain lactate concentrations (>4 mmol/L), which were predominantly associated with hyperglycolysis and normal cerebral oxygenation or perfusion, suggesting a predominant non-ischemic extracellular cerebral lactate release after TBI.9 As we did not assess pyruvate concentrations or brain tissue oxygen partial pressure, we cannot determine whether the reduced glucose/lactate ratio in our patients with poor outcome was due to a hyperglycolytic status or anaerobic metabolism. Moreover, similar CSF changes may be observed in case of severe neuroinflammatory response and BBB breakdown, which is common after TBI. The association of elevated CSF proteins with high CSF lactate and low glucose/lactate ratio may suggest this mechanism, although CSF proteins could also partly arise from blood into the CSF. Interestingly, blood lactate concentrations were also associated with CSF lactate and glucose/lactate ratio, suggesting potential cerebral uptake of circulating lactate because of increased brain energetic demand.10

This study has several limitations. First, the retrospective analysis, the small study cohort and the lack of specific interventions limit the generalizability of these findings as well as the interpretation of CSF glucose and lactate concentrations in this setting. CSF glucose and lactate values were available only for one-third of the patients, which may have resulted in a selection bias. Second, we only considered one daily measurement, whereas microdialysis techniques usually include hourly measurements. One should further define whether changes in CSF metabolites are comparable with microdialysis data and which “normal range” cut-off values should be used in TBI. Moreover, samples were not centrifugated to remove blood cells before CSF analysis and cells lysis may have affected measurements of the glucose and lactate. Third, our mortality rate was quite high compared with those reported in recent trials on TBI, which may suggest the selection of severely ill patients. Fourth, GOSE assessment was retrospective from medical chart and its reliability in this setting might be questionable. Fifth, there was no specific correction for CSF metabolites for potential hemolysis, which may have increased lactate levels in some samples. Moreover, CSF drainage rate may vary among patients and potentially influence some CSF-derived biomarkers.25 Sixth, we could not identify from retrospective data collection any complication associated with repeated CSF sampling. Finally, most of the definitions of complications (ie, ventriculitis, shock, ischemia) we used were not really validated in large cohorts; however, none of these variables was considered as a part of the primary outcome assessment in this study.

In summary, monitoring of CSF glucose and lactate concentrations may provide prognostic information in TBI patients. More specifically, elevated CSF lactate and low CSF glucose/lactate ratio were associated with poor outcome and may indicate an ongoing hypoxic or inflammatory process. These variables are easily accessible in clinical practice and could be used in the initial assessment of TBI patients with EVD monitoring.


1. Mazzeo AT, Filippini C, Rosato R, et al. Multivariate projection method to investigate inflammation associated with secondary insults and outcome after human traumatic brain injury: a pilot study. J Neuroinflammation. 2016;13:157.
2. Yan EB, Hellewell SC, Bellander BM, et al. Post-traumatic hypoxia exacerbates neurological deficit, neuroinflammation and cerebral metabolism in rats with diffuse traumatic brain injury. J Neuroinflammation. 2011;8:147.
3. Robertson CS, Valadka AB, Hannay HJ, et al. Prevention of secondary ischemic insults after severe head injury. Crit Care Med. 1999;27:2086–2095.
4. Bergsneider M, Hovda DA, Shalmon E, et al. Cerebral hyperglycolysis following severe traumatic brain injury in humans: a positron emission tomography study. J Neurosurg. 1997;86:241–251.
5. Jalloh I, Helmy A, Howe DJ, et al. A comparison of oxidative lactate metabolism in traumatically injured brain and control brain. J Neurotrauma. 2018;35:2025–2035.
6. Bouzat P, Sala N, Suys T, et al. Cerebral metabolic effects of exogenous lactate supplementation on the injured human brain. Intensive Care Med. 2014;40:412–421.
7. Jalloh I, Helmy A, Shannon RJ, et al. Lactate uptake by the injured human brain: evidence from an arteriovenous gradient and cerebral microdialysis study. J Neurotrauma. 2013;30:2031–2037.
8. Sala N, Suys T, Zerlauth JB, et al. Cerebral extracellular lactate increase is predominantly nonischemic in patients with severe traumatic brain injury. J Cereb Blood Flow Metab. 2013;33:1815–1822.
9. Pellerin L, Bergersen LH, Halestrap AP, et al. Cellular and subcellular distribution of monocarboxylate transporters in cultured brain cells and in the adult brain. J Neurosci Res. 2005;79:55–64.
10. Timofeev I, Carpenter KL, Nortje J, et al. Cerebral extracellular chemistry and outcome following traumatic brain injury: a microdialysis study of 223 patients. Brain. 2011;134(pt 2):484–494.
11. Goodman JC, Valadka AB, Gopinath SP, et al. Extracellular lactate and glucose alterations in the brain after head injury measured by microdialysis. Crit Care Med. 1999;27:1965–1973.
12. Manley G, Knudson MM, Morabito D, et al. Hypotension, hypoxia, and head injury: frequency, duration, and consequences. Arch Surg. 2001;136:1118–1123.
13. Marmarou A, Lu J, Butcher I, et al. IMPACT database of traumatic brain injury: design and description. J Neurotrauma. 2007;24:239–250.
14. Marik PE, Bellomo R. Stress hyperglycemia: an essential survival response!. Crit Care. 2013;17:305.
15. Bilotta F, Spinelli A, Giovannini F, et al. The effect of intensive insulin therapy on infection rate, vasospasm, neurologic outcome, and mortality in neurointensive care unit after intracranial aneurysm clipping in patients with acute subarachnoid hemorrhage: a randomized prospective pilot trial. J Neurosurg Anesthesiol. 2007;19:156–160.
16. Van den Berghe G, Schoonheydt K, Becx P, et al. Insulin therapy protects the central and peripheral nervous system of intensive care patients. Neurology. 2005;64:1348–1353.
17. Finfer S, Chittock D, Li Y, et al. Intensive versus conventional glucose control in critically ill patients with TBI: long-term follow up of subgroup of patients from the NICE-SUGAR study. Intensive Care Med. 2015;41:1037–1047.
18. Oddo M, Schmidt JM, Carrera E, et al. Impact of tight glycemic control on cerebral glucose metabolism after severe brain injury: a microdialysis study. Crit Care Med. 2008;36:3233–3238.
19. Vespa P, McArthur DL, Stein N, et al. Tight glycemic control increases metabolic distress in traumatic brain injury: a randomized controlled within-subjects trial. Crit Care Med. 2012;40:1923–1929.
20. Kurtz P, Claassen J, Schmidt JM, et al. Reduced brain/serum glucose ratios predict cerebral metabolic distress and mortality after severe brain injury. Neurocrit Care. 2013;19:311–319.
21. Hutchinson PJ, Jalloh I, Helmy A, et al. Consensus statement from the 2014 International Microdialysis Forum. Intensive Care Med. 2015;41:1517–1528.
22. Ho KM, Lan NS, Williams TA, et al. A comparison of prognostic significance of strong ion gap (SIG) with other acid-base markers in the critically ill: a cohort study. J Intensive Care. 2016;29:43.
23. Yang MS, DeWitt DS, Becker DP, et al. Regional brain metabolite levels following mild experimental head injury in the cat. J Neurosurg. 1985;63:617–621.
24. DeSalles AA, Kontos HA, Becker DP, et al. Prognostic significance of ventricular CSF lactic acidosis in severe head injury. J Neurosurg. 1986;65:615–624.
25. Shore PM, Thomas NJ, Clark RS, et al. Continuous versus intermittent cerebrospinal fluid drainage after severe traumatic brain injury in children: effect on biochemical markers. J Neurotrauma. 2004;21:1113–1122.

brain monitoring; cerebral metabolism; lactate; glucose; traumatic brain injury

Supplemental Digital Content

Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.