Key Points
Question
In intensive care unit patients at risk of death by neurologic criteria (DNC), what is the diagnostic accuracy of brain computed tomography (CT) perfusion and CT angiography compared with the clinical examination?
Findings
In a multicenter diagnostic study of 282 patients (204 with DNC), qualitative brainstem CT perfusion showed 98.5% sensitivity and 74.4% specificity, whole-brain CT perfusion showed 93.6% sensitivity and 92.3% specificity, and CT angiography showed 75.5% to 87.3% sensitivity and 89.7% to 91.0% specificity.
Meaning
None of the imaging modalities met the prespecified accuracy threshold of greater than 98%, supporting their role only as ancillary rather than standalone tests for confirming DNC.
Abstract
Importance
Accurate and timely confirmation of death by neurologic criteria (DNC) is essential for clinical decision-making and organ-donation processes, yet currently available ancillary tests have suboptimal diagnostic performance or limited validation.
Objectives
To determine the diagnostic accuracy, interrater reliability, and safety of brain computed tomography (CT) perfusion and CT angiography as ancillary investigations for DNC.
Design, Setting, and Participants
Between April 25, 2017, and March 10, 2021, a prospective, multicenter, blinded diagnostic accuracy cohort study was conducted in 15 adult intensive care units across Canada. Consecutive, critically ill adults (aged ≥18 years) with a Glasgow Coma Scale score of 3 and no confounding factors who were at high risk of DNC were included. Data collection and analysis were performed from April 2021 to July 2024.
Exposure
Contrast-enhanced brain CT perfusion with CT angiography reconstructions performed within 2 hours of a blinded, standardized clinical DNC examination.
Main Outcomes and Measures
The primary outcomes were the sensitivity and specificity of qualitative and quantitative brainstem CT perfusion for DNC determination, assessed by 2 independent neuroradiologists blinded to clinical findings; the prespecified validation threshold was greater than 98%. Secondary outcomes were the diagnostic accuracy of whole-brain CT perfusion and CT angiography, interrater reliability (Cohen κ), and adverse events associated with imaging.
Results
A total of 282 patients (mean [SD] age, 57.8 [15.4] years; 133 [47%] female) completed the study protocol and were included in the primary analysis; 204 (72%) of these were ultimately declared deceased by standardized clinical criteria. Qualitative brainstem CT perfusion showed a sensitivity of 98.5% (95% CI, 95.8%-99.7%) and a specificity of 74.4% (95% CI, 63.2%-83.6%); quantitative brainstem CT perfusion was not diagnostically accurate. Qualitative whole-brain CT perfusion yielded a sensitivity of 93.6% (95% CI, 89.3%-96.6%) and a specificity of 92.3% (95% CI, 84.0%-97.1%). CT angiography sensitivity ranged from 75.5% (95% CI, 69.0%-81.2%) to 87.3% (95% CI, 81.9%-91.5%), and its specificity ranged from 89.7% (95% CI, 80.8%-95.5%) to 91.0% (95% CI, 82.4%-96.3%). Interrater reliability was excellent for all ancillary tests (κ ranged from 0.81 [95% CI, 0.73-0.89] to 0.84 [95% CI, 0.78-0.91]). Fourteen patients (5%) experienced minor, self-limited adverse events; no serious adverse events occurred.
Conclusions and Relevance
The observed sensitivity and specificity measures for CT perfusion and CT angiography as an ancillary test for DNC did not meet the prespecified validation threshold of greater than 98%. Clinical examination remains the cornerstone of DNC, and ancillary imaging should be interpreted cautiously within a comprehensive clinical assessment.
This diagnostic study evaluates the diagnostic accuracy, interrater reliability, and safety of brain computed tomography perfusion and computed tomography angiography for death by neurologic criteria.
Introduction
Determination of death by neurologic criteria (DNC) is a critical aspect of modern medicine, defining the end of life for many individuals with acute brain injury. DNC is defined by the permanent cessation of brain function and is characterized by the complete absence of consciousness and of brainstem reflexes, including the capacity to breathe, following a catastrophic brain injury.1,2,3 False-negative DNC may prolong the provision of futile organ support, and false-positive DNC may lead to an erroneous declaration of death. Although clinical evaluation forms the cornerstone of DNC, its validity can be influenced by confounding factors, such as the effects of sedative medications, traumatic injuries to the skull, or severe metabolic derangements. In these circumstances, DNC guidelines recommend additional ancillary investigation to determine the absence of brain blood flow or perfusion.1,2,3
Nevertheless, ancillary investigations do not directly assess pertinent clinical function.4 Significant variability exists in physician preferences5,6,7 and in hospital policies, national regulatory committees, or law8,9 regarding their indications and use. A recent systematic review concluded that studies assessing the diagnostic accuracy of DNC ancillary investigations are at risk of bias with variable or unclear sensitivities and specificities that could lead to misclassifications.10 Despite the widespread use of 4-vessel conventional angiography as a DNC ancillary investigation, the systematic review10 and the World Brain Death Project consensus statement2 acknowledged that the specificity of 4-vessel conventional angiography is unknown due to lack of thorough investigation in an appropriate diagnostic test accuracy study. The World Brain Death Project identified the diagnostic accuracy of novel DNC ancillary investigations, such as CT perfusion and CT angiography, as a knowledge gap and research priority in the field of death determination.2 Whereas CT angiography visualizes blood flow within large cerebral vessels, CT perfusion assesses tissue-level microvascular perfusion.11,12 Importantly, neither technique measures neuronal function itself, which is what is assessed clinically or with functional tests, such as electroencephalography or evoked potentials. Preliminary studies have indicated CT perfusion’s potential as a DNC ancillary investigation, but adequate validation is necessary.13
The primary objective of this study was to evaluate the diagnostic accuracy of brainstem CT perfusion for DNC based on qualitative and quantitative assessments. Secondary objectives were to assess the diagnostic accuracy of whole-brain CT perfusion and of CT angiography for DNC and to determine the safety and consistency of CT perfusion and CT angiography in critically ill patients.
Methods
In collaboration with the Canadian Critical Care Trials Group, we conducted a prospective multicenter diagnostic accuracy study in 15 Canadian academic intensive care units (eAppendix 1 in Supplement 1). We obtained research ethics board approval at each participating site and written informed consent from each patient’s surrogate decision-maker prior to enrollment. The study was overseen by a steering committee composed of experts in critical care, neurology, and neuroradiology. Study reporting follows the Standards for Reporting of Diagnostic Accuracy (STARD) reporting guidelines.14 Data collection and analysis were performed from April 2021 to July 2024.
Study Population and Eligibility Criteria
Demographic characteristics, such as sex and race (reported by surrogate decision-makers or recorded in hospital records as Asian, Black or African American, First Nations, White, and other [Arab, Latin American, and multiracial, grouped owing to small sample sizes] or unknown [declined to report or could not be determined by the study team]), were collected per protocol at the time of enrollment to describe the included population and to inform the generalizability of our findings. To effectively evaluate the diagnostic accuracy (sensitivity and specificity) of CT perfusion and DNC, the study was designed to carefully include patients both with and without a final diagnosis of DNC. This mixed sample was essential to determine the true-positive rates (sensitivity), false-negative rates, true-negative rates (specificity), and false-positive rates. Therefore, our selection criteria were designed to enroll patients suspected of DNC, specifically encompassing individuals presumed to be at high risk of progressing to DNC. To ensure a reliable evaluation of diagnostic accuracy, we also implemented strict exclusion criteria to remove factors that could interfere with or confound the clinical diagnosis of DNC. We screened adults aged 18 years or older who were admitted to the intensive care unit with acute and devastating brain injury confirmed by neuroimaging and in whom the Glasgow Coma Scale score was 3 despite cessation of sedation for at least 6 hours. We excluded patients with contraindications to CT perfusion, hemodynamic instability that prevented safe transport to the CT scanner, and patients with any confounding factors that could have precluded a reliable clinical evaluation (eAppendix 2 in Supplement 1). Eligibility of all enrolled patients was confirmed by blind central adjudication (eAppendix 2 in Supplement 1).
Ancillary Investigation
Enrolled patients underwent whole-brain CT perfusion ensuring coverage of at least 8 cm starting at the foramen magnum (eAppendix 2 in Supplement 1). Unlike standard clinical practice, in which ancillary tests are typically performed after completion of the full clinical examination, imaging was conducted prior to the formal DNC. This timing was chosen to ensure the shortest possible interval between the ancillary investigation and the reference standard, while also allowing for real-time cancellation or rescheduling of imaging when clinically appropriate, balancing study feasibility with the realities of critical care practice. A total of 40 mL of nonionic iodinated contrast media was injected intravenously. Each study center transferred unprocessed CT perfusion source images to a central imaging core laboratory for processing and analysis. Image processing was performed using Olea Sphere version 3.0 (Olea Medical). CT angiography images were derived from the CT perfusion images using reconstructions, with a peak phase and a late-phase image defined on the time density curve. CT perfusion and CT angiography images were interpreted independently by 2 experienced neuroradiologists blinded to all clinical information, including results from the DNC clinical examination.
For qualitative CT perfusion interpretation, images were assessed for a matched decrease in cerebral blood flow and a corresponding decrease in cerebral blood volume. For qualitative brainstem CT perfusion, the test result was judged positive based on brainstem assessment alone, regardless of findings in the rest of the brain. For whole-brain CT perfusion, the test result was considered positive based on assessment of the entire brain. For quantitative brainstem CT perfusion, large regions of interest were assessed on axial slices at the medulla, pons, and midbrain, and the test result was judged positive if cerebral blood flow was below 10 mL/100 g/min and if cerebral blood volume was below 2 mL/100 g on at least 2 consecutive 5-mm axial slices of brainstem.12,13 For CT angiography, neuroradiologists assessed different segments of intracranial arteries on 3 standardized scales (10-point, 7-point, and 4-point scales), each applied separately to the early- and late-acquisition phases.11,15,16 Imaging interpretation disagreements were resolved by consensus using a third reading, blinded from the first 2 interpretations and DNC clinical examination result.
Determination of DNC
Within 2 hours of ancillary investigation, 2 clinicians (neurologists, neurosurgeons, or intensivists) performed a complete clinical DNC assessment. Both clinicians used standardized criteria based on contemporary DNC guidelines17,18 and were blinded to the results of the study’s ancillary investigation. Patients were declared deceased by neurologic criteria when the following criteria were met: (1) documentation of a devastating brain injury with an established etiology and plausibility of causing DNC; (2) absence of confounders that could affect the reliability of the clinical neurologic examination (including measure of plasma concentrations of common sedatives and analgesics19); (3) absence of brainstem reflexes and motor response to central pain; and (4) positive apnea test result (eAppendix 2 in Supplement 1). Patients were classified as clinically deceased by neurologic criteria, or not, by 2 examining physicians. For this study, the clinical neurologic evaluation was considered the reference standard for DNC.
Outcome Measures
For all patients, detailed demographic characteristics, clinical information, scan-related adverse events, and clinical outcomes were collected. Because clinical DNC principally focuses on brainstem function (absence of brainstem reflexes and respiratory drive), our primary outcome was the diagnostic accuracy of brainstem perfusion assessment by CT perfusion. Secondary outcomes were diagnostic accuracy of whole-brain CT perfusion and of CT angiography, as well as their safety and consistency (interrater reliability). Ancillary investigation results were considered a true positive or a true negative when concordant with the reference standard. False-positive or false-negative results occurred when discordant with the reference standard.
Statistical Analysis
We estimated that 45% to 55% of eligible patients would satisfy bedside-determined DNC. Under this prevalence, enrolling 270 protocol-complete patients would allow estimation of brainstem CT perfusion sensitivity and specificity of greater than 98% with a margin of error of 2.5% or less at an α error of .05.
To account for unforeseen events during the study procedure, the initial expected enrollment sample size was 300. In August 2020, blinded progress reports showed a higher-than-expected rate of protocol noncompletion, so the executive committee increased the enrollment target to 330 to ensure that the planned 270 protocol-complete patients would be achieved.
We only included patients with complete CT perfusion evaluations and a clinical evaluation without confounding, as determined by adjudication of the clinical files and blinded to the imaging results. Baseline characteristics were assessed using frequency distributions. For primary analyses, we calculated qualitative and quantitative brainstem CT perfusion sensitivities, specificities, positive predictive values, negative predictive values, positive likelihood ratios, and negative likelihood ratios with corresponding 95% CIs. For secondary analyses, we assessed the respective diagnostic accuracy of qualitative whole-brain CT perfusion and of CT angiography using the same steps as brainstem CT perfusion. We also evaluated interrater agreement for each ancillary investigation using the Cohen κ statistic, and we reported descriptive frequency distribution statistics to describe safety adverse events. R version 4.3.3 software (R Foundation) was used for statistical analysis.
In preplanned subgroup analyses, we assessed diagnostic accuracy of the ancillary investigations based on age (<60 vs ≥60 years), sex (male, female), type of brain injury (traumatic brain injury, anoxic brain injury, ischemic stroke, hemorrhagic stroke, other), and presence or absence of artifacts on ancillary imaging.
We conducted 2 preplanned sensitivity analyses. First, we excluded patients without a reference standard consensus and patients who had protocol violations. Second, we tested for potential death misclassification in the reference standard by excluding patients who were considered alive based on peripheral movements alone.
Results
Between April 25, 2017, and March 10, 2021, 333 patients were enrolled, of whom 282 were included in the final analysis (Figure 1 and Table 1). Exclusion of 51 enrolled patients occurred primarily due to incomplete or confounded clinical examinations, technical issues with CT perfusion, or withdrawal of consent. Mean (SD) age was 57.8 (15.4) years, and 133 patients (47%) were female. Causes of brain injury were hemorrhagic stroke in 150 patients (53%), traumatic brain injury in 40 (14%), anoxic brain injury in 59 (21%), ischemic stroke in 15 (5%), tumor in 1 (0.4%), infection in 4 (1%), and other causes in 13 (5%). Clinical DNC assessments occurred a median (IQR) of 64.5 (40.0-87.8) minutes after ancillary testing. Central adjudication confirmed that none of the patients included in the analyses were confounded by sedative medications based on the plasma concentrations of common sedatives. Based on the clinical neurologic evaluation (reference standard), 204 patients (72%) were determined to be deceased by neurologic criteria (Table 2; eTables 1-3 in Supplement 1), with the remaining 78 (28%) classified as being alive.
Figure 1. Study Flowchart.
GCS indicates Glasgow Coma Scale; IV, intravenous.
aThe reasons for these exclusions are not mutually exclusive.
Table 1. Baseline Characteristics.
| Characteristic | Total (N = 282) | Deceased (n = 204)a | Alive (n = 78)a |
|---|---|---|---|
| Age, mean (SD), y | 57.8 (15.4) | 57.3 (16.1) | 59.1 (13.6) |
| Sex, No. (%) | |||
| Female | 133 (47) | 110 (54) | 23 (29) |
| Male | 149 (53) | 94 (46) | 55 (71) |
| Race, No. (%) | |||
| Asian | 24 (9) | 20 (10) | 4 (5) |
| Black or African American | 4 (1) | 3 (2) | 1 (1) |
| First Nations | 11 (4) | 10 (5) | 1 (1) |
| White | 217 (77) | 153 (75) | 64 (82) |
| Other or unknownb | 26 (9) | 18 (9) | 8 (10) |
| Cause of brain injury, No. (%) | |||
| Traumatic brain injury | 40 (14) | 28 (14) | 12 (15) |
| Hemorrhagic stroke | 150 (53) | 121 (59) | 29 (37) |
| Ischemic stroke | 15 (5) | 13 (6) | 2 (3) |
| Anoxic brain injury | 59 (21) | 31 (15) | 28 (36) |
| Tumor | 1 (0.4) | 0 | 1 (1) |
| Infection | 4 (1) | 4 (2) | 0 |
| Other | 13 (5) | 7 (3) | 6 (8) |
| Weight at admission, mean (SD), kg | 80.5 (21.0) | 79.1 (21.8) | 83.9 (18.6) |
| Comorbidity, No. (%) | |||
| Hypertension treated with medication | 86 (31) | 59 (29) | 27 (34) |
| Diabetes | 50 (18) | 30 (15) | 20 (26) |
| Coronary artery disease | 40 (14) | 23 (11) | 17 (22) |
| Peripheral vascular disease | 6 (2) | 4 (2) | 2 (3) |
| Previous stroke | 27 (10) | 19 (9) | 8 (10) |
| Active smoking | 42 (15) | 28 (14) | 14 (18) |
| Chronic kidney failure | |||
| With dialysis, No. (%) | 3 (1) | 0 | 3 (4) |
| Without dialysis, No. (%) | 10 (4) | 7 (3) | 3 (4) |
| Liver disease | 25 (9) | 15 (7) | 10 (13) |
| Other | 160 (57) | 114 (56) | 46 (59) |
| Time between injury and study enrollment, median (IQR), h | 48.8 (25.3-87.8) | 43.8 (24.4-72.9) | 77.2 (33.7-107.0) |
Deceased or alive as per the reference standard (clinical examination for death by neurologic criteria).
Other or unknown includes participants identified as Arab, Latin American, or multiracial by surrogate decision-makers—or, when no surrogate was available, as recorded in hospital records—along with individuals who declined to report race and ethnicity or whose race and ethnicity could not be determined by the study team. These were grouped owing to small sample size.
Table 2. Results From Clinical Examination and Ancillary Investigation.
| Outcome | No. (%)a | ||
|---|---|---|---|
| Total (N = 282) | Deceased (n = 204)b | Alive (n = 78)b | |
| Reference standard | |||
| Aware of study ancillary investigation prior to clinical evaluation | 3 (1) | 2 (1) | 1 (1) |
| Aware of other ancillary investigation result prior to clinical evaluation | 18 (6) | 13 (6) | 5 (6) |
| Ancillary investigation | |||
| Artifacts | 62 (22) | 53 (26) | 9 (11) |
| Qualitative CT perfusion result | |||
| Brainstem | |||
| Compatible with death | 221 (78) | 201 (99) | 20 (26) |
| Not compatible with death | 61 (22) | 3 (1) | 58 (74) |
| Whole brain | |||
| Compatible with death | 197 (70) | 191 (94) | 6 (8) |
| Not compatible with death | 85 (30) | 13 (6) | 72 (92) |
| Quantitative CT perfusion result | |||
| Midbrain blood flow <10 mL/100 g/min | |||
| Compatible with death | 121 (43) | 89 (44) | 32 (41) |
| Not compatible with death | 157 (56) | 113 (55) | 44 (56) |
| Missing data | 4 (1) | 2 (1) | 2 (3) |
| Midpons blood flow <10 mL/100 g/min | |||
| Compatible with death | 116 (41) | 85 (42) | 31 (40) |
| Not compatible with death | 162 (57) | 117 (57) | 45 (58) |
| Missing data | 4 (1) | 2 (1) | 2 (3) |
| Medulla blood flow <10 mL/100 g/min | |||
| Compatible with death | 102 (36) | 77 (38) | 25 (32) |
| Not compatible with death | 176 (62) | 125 (61) | 51 (65) |
| Missing data | 4 (1) | 2 (1) | 2 (2) |
| Midbrain blood volume <2 mL/100 g | |||
| Compatible with death | 175 (62) | 151 (74) | 24 (30) |
| Not compatible with death | 103 (36) | 51 (25) | 52 (66) |
| Missing data | 4 (1) | 2 (1) | 2 (3) |
| Midpons blood volume <2 mL/100 g | |||
| Compatible with death | 173 (61) | 147 (72) | 26 (33) |
| Not compatible with death | 105 (37) | 55 (27) | 50 (64) |
| Missing data | 4 (1) | 2 (1) | 2 (3) |
| Medulla blood volume <2 mL/100 g | |||
| Compatible with death | 175 (62) | 153 (75) | 22 (28) |
| Not compatible with death | 103 (37) | 49 (24) | 54 (69) |
| Missing data | 4 (1) | 2 (1) | 2 (3) |
| CT angiography result | |||
| Peak arterial phase | |||
| 4-Point scale | |||
| Compatible with death | 186 (66) | 178 (87) | 8 (10) |
| Not compatible with death | 96 (34) | 26 (13) | 70 (90) |
| 7-Point scale | |||
| Compatible with death | 178 (63) | 170 (83) | 8 (10) |
| Not compatible with death | 104 (36) | 34 (17) | 70 (90) |
| 10-Point scale | |||
| Compatible with death | 174 (62) | 167 (82) | 7 (9) |
| Not compatible with death | 108 (38) | 37 (18) | 71 (91) |
| Late arterial phase | |||
| 4-Point scale | |||
| Compatible with death | 180 (64) | 172 (84) | 8 (10) |
| Not compatible with death | 102 (36) | 32 (16) | 70 (90) |
| 7-Point scale | |||
| Compatible with death | 163 (58) | 155 (76) | 8 (10) |
| Not compatible with death | 119 (42) | 49 (24) | 70 (90) |
| 10-Point scale | |||
| Compatible with death | 161 (57) | 154 (76) | 7 (9) |
| Not compatible with death | 121 (43) | 50 (25) | 71 (91) |
Abbreviation: CT, computed tomography.
Totals for some categories may exceed 100% because each percentage is calculated independently and rounded to the nearest whole number.
Deceased or alive as per the reference standard (clinical examination for death by neurologic criteria).
Protocol violations occurred in 14 patients (5%) (eTable 4 in Supplement 1). Clinicians performing DNC evaluation were aware of results from other ancillary investigations for 18 patients (6% of deceased patients and 6% of living patients) and were mistakenly unblinded to the study CT perfusion scan for 3 patients (1% of deceased patients and 1% of living patients). The delay between ancillary investigations and clinical examinations exceeded 2 hours for 12 patients (4%). Patient clinical outcomes at hospital discharge are provided in eTable 5 in Supplement 1. No patient declared deceased by neurologic criteria was alive at hospital discharge. Six patients (8%) classified as alive eventually survived to hospital discharge.
Qualitative brainstem CT perfusion had a sensitivity of 98.5% (95% CI, 95.8%-99.7%) and a specificity of 74.4% (95% CI, 63.2%-83.6%) (Figure 2). Quantitative brainstem CT perfusion yielded poor sensitivity and specificity estimates. Qualitative whole-brain CT perfusion had a sensitivity of 93.6% (95% CI, 89.3%-96.6%) and a specificity of 92.3% (95% CI, 84.0%-97.1%). CT angiography sensitivity ranged from 81.9% (95% CI, 75.9%-86.9%) to 87.3% (95% CI, 81.9%-91.5%) for peak-phase examinations and from 75.5% (95% CI, 69.0%-81.2%) to 84.3% (95% CI, 78.6%-89.0%) for late-phase examinations, and specificity ranged from 89.7% (95% CI, 80.8%-95.5%) to 91.0% (95% CI, 82.4%-96.3%) for peak-phase and late-phase examinations. The Cohen κ for interrater reliability was 0.81 (95% CI, 0.73-0.89) for brainstem qualitative CT perfusion, 0.82 (95% CI, 0.74-0.89) for whole-brain qualitative CT perfusion, and 0.82 (95% CI, 0.75-0.89) to 0.84 (95% CI, 0.78-0.91) for CT angiography (Table 3). CT perfusion–related adverse events occurred in 14 patients (5%), with no serious adverse events (eTable 6 in Supplement 1).
Figure 2. Sensitivity and Specificity of Computed Tomography (CT) Perfusion and of CT Angiography for Death by Neurologic Criteria.
Table 3. Ancillary Investigation Diagnostic Accuracy and Interrater Reliability.
| Ancillary investigation | Sensitivity, % (95% CI) | Specificity, % (95% CI) | Accuracy, % (95% CI) | Positive predictive value, % (95% CI) | Negative predictive value, % (95% CI) | Positive likelihood ratio (95% CI) | Negative likelihood ratio (95% CI) | Cohen κ (95% CI) |
|---|---|---|---|---|---|---|---|---|
| Primary outcomes | ||||||||
| Brainstem CT perfusion | ||||||||
| Qualitative brainstem CT perfusion | 98.5 (95.8-99.7) | 74.4 (63.2-83.6) | 91.8 (88.0-94.8) | 91.0 (86.4-94.4) | 95.1 (86.3-99.0) | 3.84 (2.63-5.61) | 0.02 (0.01-0.06) | 0.81 (0.73-0.89) |
| Quantitative CT perfusion | ||||||||
| Midbrain | 38.1 (31.4-45.2) | 81.6 (71.0-89.5) | 50.0 (44.0-56.0) | 84.6 (75.5-91.3) | 33.2 (26.5-40.4) | 2.07 (1.25-3.43) | 0.76 (0.65-0.88) | NA |
| Midpons | 36.1 (29.5-43.2) | 81.6 (71.0-89.5) | 48.6 (42.5-54.6) | 83.9 (74.5-90.9) | 32.5 (25.9-39.6) | 1.96 (1.18-3.26) | 0.78 (0.67-0.91) | NA |
| Medulla | 32.7 (26.3-39.6) | 90.8 (81.9-96.2) | 48.6 (42.5-54.6) | 90.4 (81.2-96.1) | 33.7 (27.2-40.6) | 3.55 (1.70-7.38) | 0.74 (0.66-0.84) | NA |
| Secondary outcomes | ||||||||
| Qualitative whole-brain CT perfusion | 93.6 (89.3-96.6) | 92.3 (84.0-97.1) | 93.3 (89.7-95.9) | 97.0 (93.5-98.9) | 84.7 (75.3-91.6) | 12.17 (5.64-26.28) | 0.07 (0.04-0.12) | 0.82 (0.74-0.89) |
| Peak-phase CT angiography | ||||||||
| 4-Point scale | 87.3 (81.9-91.5) | 89.7 (80.8-95.5) | 87.9 (83.6-91.5) | 95.7 (91.7-98.1) | 72.9 (62.9-81.5) | 8.51 (4.40-16.44) | 0.14 (0.10-0.20) | 0.84 (0.77-0.90) |
| 7-Point scale | 83.3 (77.5-88.2) | 89.7 (80.8-95.5) | 85.1 (80.4-89.1) | 95.5 (91.3-98.0) | 67.3 (57.4-76.2) | 8.13 (4.20-15.71) | 0.19 (0.14-0.25) | 0.84 (0.78-0.91) |
| 10-Point scale | 81.9 (75.9-86.9) | 91.0 (82.4-96.3) | 84.4 (79.6-88.4) | 96.0 (91.9-98.4) | 65.7 (56.0-74.6) | 9.12 (4.49-18.55) | 0.20 (0.15-0.27) | 0.82 (0.75-0.89) |
| Late-phase CT angiography | ||||||||
| 4-Point scale | 84.3 (78.6-89.0) | 89.7 (80.8-95.5) | 85.8 (81.2-89.7) | 95.6 (91.4-98.1) | 68.6 (58.7-77.5) | 8.22 (4.25-15.89) | 0.17 (0.13-0.24) | 0.83 (0.78-0.90) |
| 7-Point scale | 76.0 (69.5-81.7) | 89.7 (80.8-95.5) | 79.8 (74.6-84.3) | 95.1 (90.6-97.9) | 58.8 (49.4-67.8) | 7.41 (3.83-14.45) | 0.27 (0.21-0.35) | 0.84 (0.78-0.90) |
| 10-Point scale | 75.5 (69.0-81.2) | 91.0 (82.4-96.3) | 79.8 (74.6-84.3) | 95.7 (91.2-98.2) | 58.7 (49.4-67.6) | 8.41 (4.13-17.13) | 0.27 (0.21-0.35) | 0.84 (0.78-0.90) |
Abbreviation: NA, not applicable.
Twenty patients who did not fulfill clinical criteria for DNC had qualitative brainstem CT perfusion compatible with death (false-positive results; eTable 7 in Supplement 1), whereas 3 patients who were declared deceased clinically had brainstem perfusion (false-negative results; eTable 8 in Supplement 1). When qualitatively assessing the whole brain by CT perfusion, 6 patients who did not fulfill clinical criteria for DNC had whole-brain CT perfusion compatible with death (false-positive results; eTable 9 in Supplement 1), whereas 13 patients who were declared deceased clinically had brain perfusion (false-negatives results; eTable 10 in Supplement 1). Age, sex, type of brain injury, presence of artifacts, and sensitivity analyses were not significantly associated with outcome measures (eFigures 1-4 and eTable 11 in Supplement 1).
Discussion
In this prospective, multicenter diagnostic accuracy study of patients with a high pretest probability of meeting DNC, brainstem CT perfusion, whether evaluated quantitatively or qualitatively, was not specific for DNC, leading to a noteworthy risk of false-positive death determinations. Qualitative whole-brain CT perfusion provided high but imperfect sensitivity and specificity for DNC. CT angiography demonstrated comparable specificity to CT perfusion but lower sensitivity, increasing the chance of false-negative classifications without reducing the risk of false-positive results. Both ancillary investigations had excellent interrater reliability.
Current DNC guidelines focus on demonstrating permanent loss of brain function, including respiratory drive. However, some jurisdictions also expect demonstration of global intracranial circulatory arrest, especially in the presence of confounding factors or clinical uncertainty. This nuance matters clinically, as a patient with no demonstrable clinical brainstem activity may still have partial blood flow in the brain, making the choice of ancillary test critical. Our findings suggest that qualitative whole-brain CT perfusion poses the lowest risk of falsely classifying a patient who meets bedside DNC criteria as alive, while its risk of misclassifying a living patient as deceased is comparable to that of CT angiography.
These results support the potential role of CT perfusion and CT angiography as ancillary investigations for DNC when such a test is indicated in the comprehensive evaluation of death in a patient with a devastating brain injury. However, imperfect test specificity reinforces the importance of thoroughly evaluating patients’ neurologic status through history and clinical examination prior to conducting ancillary investigations, such as in circumstances where confounding factors cannot be eliminated or a complete clinical examination is not possible. This maximizes ancillary investigation pretest probability, which subsequently increases posttest probability, and minimizes the risk of false-positive death determination. Despite a uniform imaging protocol and postprocessing software, minor scanner-specific variations likely occurred and automated software underperformed relative to expert qualitative reads, consistent with standard neuroimaging practice. Diagnostic accuracy may therefore differ on other CT perfusion platforms that use different perfusion algorithms. Likewise, our CT angiography results were reconstructed from the CT perfusion source data rather than from a dedicated CT angiography protocol, which may potentially influence performance estimates. Finally, our data support previous observations that a subset of patients fulfilling clinical DNC can exhibit residual intracranial blood flow or perfusion.10,20 Intracranial vessel opacification was observed in up to 25% of patients meeting clinical DNC. However, cerebral perfusion was observed in only 6% of patients, demonstrating that blood flow in intracranial arteries does not systematically infer capillary-level perfusion.4 These observations emphasize the need for a comprehensive approach for accurate and ethical decision-making in DNC that integrates robust clinical evaluation, neuroimaging findings, and understanding of the patient’s overall condition.20
The strengths of our study include a clinically relevant heterogeneous sample of patients with brain injury in diverse hospital settings with a study design that permitted robust evaluation of sensitivity and specificity. We minimized biases by using duplicate, fully blinded assessments: clinicians evaluating DNC were blinded to imaging results, and neuroradiologists interpreting the scans were blinded to clinical data. The delay between imaging and clinical assessments was intentionally short, leading to a more accurate assessment of the diagnostic validity of ancillary investigations. We also observed excellent interrater agreement, further supporting the robustness of our results.
Limitations
This study has several limitations. Some centers experienced challenges using the CT perfusion protocol, leading to the exclusion of several patients from the final analysis due to a lack of interpretable images. We also encountered difficulties in postprocessing CT perfusion data, highlighting an important area of training at implementation. In addition, we observed potential for diagnostic errors in both ancillary investigations and clinical assessments, as well as occasional ambiguity in clinical evaluations, reflecting the real-world complexity of DNC. We used the clinical determination of DNC as the reference standard, which, although widely accepted, has known limitations. DNC is a clinical diagnosis based on the permanent loss of brainstem reflexes and the capacity to breathe, serving as a legally and medically accepted proxy for death. However, it remains only a surrogate marker for the complete cessation of all intracranial neuronal activity. Consequently, imperfections in this reference standard may affect the apparent diagnostic accuracy of ancillary investigations that do not directly measure global neuronal function. Our study therefore compares ancillary investigations with this established, yet imperfect, clinical benchmark rather than with an ideal, but currently unachievable, criterion standard of absolute cessation of all brain functions. Blinded adjudication was performed to reduce uncertainty in reference standards. Nevertheless, some patients were included in the final analysis despite residual ambiguity, which may have contributed to false-positive or false-negative classifications. Most ancillary investigations were performed within the first few days after injury; it remains uncertain whether the accuracy we observed would hold for patients assessed later in their intensive care unit course after prolonged life-sustaining therapies. These issues, however, mirror real-world challenges and ultimately strengthen the credibility and generalizability of our findings. Additionally, clinical and imaging evaluations were performed by experts, an important consideration at implementation at centers without such experience.
Conclusions
Neither CT perfusion nor CT angiography met the prespecified threshold of greater than 98% for both sensitivity and specificity. Consequently, they should not be used as stand-alone tests to establish DNC. However, they may offer supportive evidence in situations where the bedside examination is incomplete or confounded, but their findings must be weighed against a thorough clinical assessment to minimize false-positive determinations.
eAppendix 1. INDex Trial Investigators, Collaborators, and Committees
eAppendix 2. Methods
eFigure 1. Subgroup Analyses for Age
eFigure 2. Subgroup Analyses for Sex
eFigure 3. Subgroup Analyses for Type of Brain Injury
eFigure 4. Subgroup Analyses for Artifacts
eTable 1. Vital Signs and Biologic Parameters at Admission to the Intensive Care Unit
eTable 2. Clinical Evaluation for Death Determination by Neurologic Criteria – Patient Flow
eTable 3. Clinical Evaluation for Death Determination by Neurologic Criteria – Results
eTable 4. Violations – Descriptions
eTable 5. Clinical Outcomes
eTable 6. Ancillary Investigation-Related Adverse Events
eTable 7. Details on False Positive Cases on Qualitative Brainstem CT-Perfusion (n=20)
eTable 8. Details on False Negative Cases on Qualitative Brainstem CT-Perfusion (n=3)
eTable 9. Details on False Positive Cases on Qualitative Whole-Brain CT-Perfusion (n=6)
eTable 10. Details on False Negative Cases on Qualitative Whole-Brain CT-Perfusion (n=13)
eTable 11. Sensitivity Analysis: Exclusion of Patients Considered Alive Based on Peripheral Movements Alone (n = 279)
INDex Investigators and Canadian Critical Care Trials Group Members
Data Sharing Statement
References
- 1.Greer DM, Kirschen MP, Lewis A, et al. Pediatric and adult brain death/death by neurologic criteria consensus guideline. Neurology. 2023;101(24):1112-1132. doi: 10.1212/WNL.0000000000207740 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Greer DM, Shemie SD, Lewis A, et al. Determination of brain death/death by neurologic criteria: the World Brain Death Project. JAMA. 2020;324(11):1078-1097. doi: 10.1001/jama.2020.11586 [DOI] [PubMed] [Google Scholar]
- 3.Shemie SD, Wilson LC, Hornby L, et al. A brain-based definition of death and criteria for its determination after arrest of circulation or neurologic function in Canada: a 2023 clinical practice guideline. Can J Anaesth. 2023;70(4):483-557. doi: 10.1007/s12630-023-02431-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Plourde G, Briard JN, Shemie SD, Shankar JJS, Chassé M. Flow is not perfusion, and perfusion is not function: ancillary testing for the diagnosis of brain death. Can J Anaesth. 2021;68(7):953-961. doi: 10.1007/s12630-021-01988-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Braksick SA, Robinson CP, Gronseth GS, Hocker S, Wijdicks EFM, Rabinstein AA. Variability in reported physician practices for brain death determination. Neurology. 2019;92(9):e888-e894. doi: 10.1212/WNL.0000000000007009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Chassé M, Neves Briard J, Yu M, et al. ; Canadian Critical Care Trials Group . Clinical evaluation and ancillary testing for the diagnosis of death by neurologic criteria: a cross-sectional survey of Canadian intensivists. Can J Anaesth. 2022;69(3):353-363. doi: 10.1007/s12630-021-02166-0 [DOI] [PubMed] [Google Scholar]
- 7.Robbins NM, Bernat JL. Practice current: when do you order ancillary tests to determine brain death? Neurol Clin Pract. 2018;8(3):266-274. doi: 10.1212/CPJ.0000000000000473 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Greer DM, Wang HH, Robinson JD, Varelas PN, Henderson GV, Wijdicks EF. Variability of brain death policies in the United States. JAMA Neurol. 2016;73(2):213-218. doi: 10.1001/jamaneurol.2015.3943 [DOI] [PubMed] [Google Scholar]
- 9.Lewis A, Liebman J, Kreiger-Benson E, et al. Ancillary testing for determination of death by neurologic criteria around the world. Neurocrit Care. 2021;34(2):473-484. doi: 10.1007/s12028-020-01039-6 [DOI] [PubMed] [Google Scholar]
- 10.Neves Briard J, Nitulescu R, Lemoine É, et al. Diagnostic accuracy of ancillary tests for death by neurologic criteria: a systematic review and meta-analysis. Can J Anaesth. 2023;70(4):736-748. doi: 10.1007/s12630-023-02426-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Aziz Rizk A, Farhani N, Shankar J. Computed tomography perfusion for the diagnosis of brain death: a technical review. Can J Neurol Sci. 2024;51(2):173-178. doi: 10.1017/cjn.2023.242 [DOI] [PubMed] [Google Scholar]
- 12.Shankar JJ, Vandorpe R. CT perfusion for confirmation of brain death. AJNR Am J Neuroradiol. 2013;34(6):1175-1179. doi: 10.3174/ajnr.A3376 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.MacDonald D, Stewart-Perrin B, Shankar JJS. The role of neuroimaging in the determination of brain death. J Neuroimaging. 2018;28(4):374-379. doi: 10.1111/jon.12516 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Cohen JF, Korevaar DA, Altman DG, et al. STARD 2015 guidelines for reporting diagnostic accuracy studies: explanation and elaboration. BMJ Open. 2016;6(11):e012799. doi: 10.1136/bmjopen-2016-012799 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Dupas B, Gayet-Delacroix M, Villers D, Antonioli D, Veccherini MF, Soulillou JP. Diagnosis of brain death using two-phase spiral CT. AJNR Am J Neuroradiol. 1998;19(4):641-647. [PMC free article] [PubMed] [Google Scholar]
- 16.Frampas E, Videcoq M, de Kerviler E, et al. CT angiography for brain death diagnosis. AJNR Am J Neuroradiol. 2009;30(8):1566-1570. doi: 10.3174/ajnr.A1614 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Shemie SD, Doig C, Dickens B, et al. ; Pediatric Reference Group; Neonatal Reference Group . Severe brain injury to neurological determination of death: Canadian forum recommendations. CMAJ. 2006;174(6):S1-S13. doi: 10.1503/cmaj.045142 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Wijdicks E, Varelas P, Gronseth GS, Greer DM. Evidence-based guideline update: determining brain death in adults. Neurology. 2010;74(23):1911-1918. doi: 10.1212/WNL.0b013e3181e242a8 [DOI] [PubMed] [Google Scholar]
- 19.Jutras M, Williamson D, Chassé M, Leclair G. Development and validation of a liquid chromatography coupled to tandem mass spectrometry method for the simultaneous quantification of five analgesics and sedatives, and six of their active metabolites in human plasma: application to a clinical study on the determination of neurological death in the intensive care unit. J Pharm Biomed Anal. 2020;190:113521. doi: 10.1016/j.jpba.2020.113521 [DOI] [PubMed] [Google Scholar]
- 20.Neves Briard J, Plourde G, Nitulescu R, et al. Infratentorial brain injury among patients suspected of death by neurologic criteria: a systematic review and meta-analysis. Neurology. 2023;100(4):e443-e453. doi: 10.1212/WNL.0000000000201449 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eAppendix 1. INDex Trial Investigators, Collaborators, and Committees
eAppendix 2. Methods
eFigure 1. Subgroup Analyses for Age
eFigure 2. Subgroup Analyses for Sex
eFigure 3. Subgroup Analyses for Type of Brain Injury
eFigure 4. Subgroup Analyses for Artifacts
eTable 1. Vital Signs and Biologic Parameters at Admission to the Intensive Care Unit
eTable 2. Clinical Evaluation for Death Determination by Neurologic Criteria – Patient Flow
eTable 3. Clinical Evaluation for Death Determination by Neurologic Criteria – Results
eTable 4. Violations – Descriptions
eTable 5. Clinical Outcomes
eTable 6. Ancillary Investigation-Related Adverse Events
eTable 7. Details on False Positive Cases on Qualitative Brainstem CT-Perfusion (n=20)
eTable 8. Details on False Negative Cases on Qualitative Brainstem CT-Perfusion (n=3)
eTable 9. Details on False Positive Cases on Qualitative Whole-Brain CT-Perfusion (n=6)
eTable 10. Details on False Negative Cases on Qualitative Whole-Brain CT-Perfusion (n=13)
eTable 11. Sensitivity Analysis: Exclusion of Patients Considered Alive Based on Peripheral Movements Alone (n = 279)
INDex Investigators and Canadian Critical Care Trials Group Members
Data Sharing Statement
