Author + information
- Received December 19, 2014
- Revision received February 15, 2015
- Accepted March 1, 2015
- Published online May 19, 2015.
- Pascal Stammet, MD∗∗ (, )
- Olivier Collignon, PhD†,
- Christian Hassager, MD, DMSc‡,
- Matthew P. Wise, MD, DPhil§,
- Jan Hovdenes, MD, PhD‖,
- Anders Åneman, MD, PhD¶,
- Janneke Horn, MD, PhD#,
- Yvan Devaux, PhD∗∗,
- David Erlinge, MD, PhD††,
- Jesper Kjaergaard, MD, DMSc‡,
- Yvan Gasche, MD‡‡,
- Michael Wanscher, MD, PhD§§,
- Tobias Cronberg, MD, PhD‖‖,
- Hans Friberg, MD, PhD¶¶,
- Jørn Wetterslev, MD, PhD##,
- Tommaso Pellis, MD∗∗∗,
- Michael Kuiper, MD, PhD†††,
- Georges Gilson, PhD‡‡‡,
- Niklas Nielsen, MD, PhD§§§,
- TTM-Trial Investigators
- ∗Department of Anesthesia and Intensive Care, Centre Hospitalier de Luxembourg, Luxembourg, Luxembourg
- †Competence Centre for Methodology and Statistics, Luxembourg Institute of Health, Luxembourg, Luxembourg
- ‡Department of Cardiology B, The Heart Centre, Rigshospitalet University Hospital, Copenhagen, Denmark
- §Department of Intensive Care, University Hospital of Wales, Cardiff, United Kingdom
- ‖Department of Anesthesia and Intensive Care, Oslo University Hospital, Rikshospitalet, Oslo, Norway
- ¶Department of Intensive Care, Liverpool Hospital, Sydney, Australia
- #Department of Intensive Care, Academic Medical Centrum, Amsterdam, the Netherlands
- ∗∗Laboratory of Cardiovascular Research, Luxembourg Institute of Health, Luxembourg, Luxembourg
- ††Department of Cardiology, Skåne University Hospital, Lund, Sweden
- ‡‡Department of Intensive Care, Geneva University Hospital, Geneva, Switzerland
- §§Department of Cardiothoracic Anesthesiology RT, The Heart Centre, Rigshospitalet University Hospital, Copenhagen, Denmark
- ‖‖Department of Clinical Sciences, Division of Neurology, Lund University, Lund, Sweden
- ¶¶Department of Anesthesia and Intensive Care, Skåne University Hospital, Lund University, Lund, Sweden
- ##Copenhagen Trial Unit, Centre of Clinical Intervention Research, Rigshospitalet, Copenhagen, Denmark
- ∗∗∗Department of Intensive Care, Santa Maria degli Angeli, Pordenone, Italy
- †††Department of Intensive Care, Leeuwarden Medical Centrum, Leeuwarden, the Netherlands
- ‡‡‡Department of Clinical Biology, Centre Hospitalier de Luxembourg, Luxembourg, Luxembourg
- §§§Department of Anesthesia and Intensive Care, Helsingborg Hospital, Helsingborg, Sweden
- ↵∗Reprint requests and correspondence:
Dr. Pascal Stammet, Department of Anesthesia and Intensive Care Medicine, Centre Hospitalier de Luxembourg, 4, rue Barblé, L-1210, Luxembourg, Luxembourg.
Background Neuron-specific enolase (NSE) is a widely-used biomarker for prognostication of neurological outcome after cardiac arrest, but the relevance of recommended cutoff values has been questioned due to the lack of a standardized methodology and uncertainties over the influence of temperature management.
Objectives This study investigated the role of NSE as a prognostic marker of outcome after out-of-hospital cardiac arrest (OHCA) in a contemporary setting.
Methods A total of 686 patients hospitalized after OHCA were randomized to targeted temperature management at either 33°C or 36°C. NSE levels were assessed in blood samples obtained 24, 48, and 72 h after return of spontaneous circulation. The primary outcome was neurological outcome at 6 months using the cerebral performance category score.
Results NSE was a robust predictor of neurological outcome in a baseline variable-adjusted model, and target temperature did not significantly affect NSE values. Median NSE values were 18 ng/ml versus 35 ng/ml, 15 ng/ml versus 61 ng/ml, and 12 ng/ml versus 54 ng/ml for good versus poor outcome at 24, 48, and 72 h, respectively (p < 0.001). At 48 and 72 h, NSE predicted neurological outcome with areas under the receiver-operating curve of 0.85 and 0.86, respectively. High NSE cutoff values with false positive rates ≤5% and tight 95% confidence intervals were able to reliably predict outcome.
Conclusions High, serial NSE values are strong predictors of poor outcome after OHCA. Targeted temperature management at 33°C or 36°C does not significantly affect NSE levels. (Target Temperature Management After Cardiac Arrest [TTM]; NCT01020916)
Comatose patients admitted to an intensive care unit (ICU) after an out-of-hospital cardiac arrest (OHCA) have a mortality rate of around 50%. In the majority of cases, initial ICU mortality is driven by hemodynamic failure, whereas later morbidity and mortality are due to brain damage (1). A large proportion of patients die of withdrawal of life-sustaining therapies because of presumed poor prognosis (2,3). Thus, adequate prognostication tools for neurological outcome prediction are crucial for therapeutic guidance in this severely ill population.
Biomarkers of brain damage, particularly neuron-specific enolase (NSE), have been widely studied as markers for outcome prognostication (4,5). The protein NSE is a 78kDa glycolytic enzyme involved in glucose metabolism and is mainly found in neuronal and neuroendocrine cells. Its half-life is approximately 24 h. Previous studies on patients not treated with hypothermia after cardiac arrest suggested a cutoff level of 33 ng/ml at 48 h to be predictive of death and poor neurological function (6); the American Academy of Neurology subsequently adopted this cutoff into prognostication guidelines (7). With the implementation of induced hypothermia and its assumed neuroprotective effect, the validity of this cutoff has been questioned. Subsequent studies yielded conflicting results, probably due to methodological issues and the lack of standardization of dosing methods (8). Consequently, current guidelines do not advocate NSE for outcome prediction (9), and a recent advisory statement suggests a cautious use of “high NSE levels” within a multimodal prognostication algorithm (10).
In this context of uncertainty, the TTM trial (Target Temperature Management After Out-of-Hospital Cardiac Arrest) (11), a multicenter clinical trial that included 950 patients randomized to targeted temperature management of 33°C or 36°C, provided a platform to investigate the role of NSE as a prognostic marker of outcome after OHCA in a contemporary setting.
All patients included in this study were part of the TTM trial (November 2010 to July 2013) comparing 2 temperature regimens in unconscious adult patients admitted to an ICU after an OHCA of a presumed cardiac cause. The TTM trial design, the statistical analysis plan, and the main results have been published previously (11–13). The randomization was stratified by site and performed centrally with adequate allocation concealment and sequence generation. A target temperature of 33°C or 36°C was initiated in each group according to allocation. At 28 h after start of the intervention, rewarming to 37°C was commenced at a maximum speed of 0.5°C/h. This pre-defined substudy of the TTM trial on NSE was approved by the steering committee before starting NSE analysis.
The TTM trial protocol was approved by ethical committees in each participating country, and informed consent was waived or obtained from all participants or relatives according to national legislations, in line with the Helsinki declaration (14).
Serum blood samples were taken from the patients at 24, 48, and 72 h after return of spontaneous circulation (ROSC). All samples were pre-analytically processed at the different sites, aliquoted, and frozen to −80°C before shipment to the Integrated BioBank of Luxembourg before analysis. NSE values were not available to the treating physicians during the trial.
NSE analyses were performed 6 months after trial completion at the clinical biology laboratory of the Centre Hospitalier de Luxembourg. All serum samples were tested for hemolysis using the Roche hemolysis index with measurements at 600 and 570 nm. Because of measurement interference, all samples with a positive hemolysis index (≥500 mg/l of hemoglobin) were discarded.
Determination of NSE was performed using a COBAS e601 line with an Electro-Chemi-Luminescent-Immuno-Assay (ECLIA) kit (Roche Diagnostics, Rotkreuz, Switzerland). The measuring range extended from 0.05 to 370 ng/ml. Samples with values above the measuring range had to be diluted accordingly. Functional sensitivity was at 0.25 ng/ml, and expected normal values were <17.0 ng/ml. In our laboratory, between-run precision at concentrations of 10.5 and 83.3 ng/ml was 6.8% and 5.7%, respectively.
Neurological prognostication as well as withdrawal of life-supporting therapies were standardized and reported according to the trial protocol (12,13,15).
We aimed to investigate NSE as a predictor of death and cerebral performance after OHCA in 2 targeted temperature groups as well as in a pooled sample. We studied the influence of the targeted temperature, evolution over time, predictive power of NSE, and cutoff values, including a multivariable analysis. We defined high NSE cutoff values as having a false positive rate of ≤5%.
The primary outcome in this study was neurological function at 6 months, dichotomized to good or poor outcome according to the Cerebral Performance Category (CPC) scale (16). The CPC score classifies patients into 5 categories: CPC 1 (no neurological disability); CPC 2 (minor neurological deficit); CPC 3 (severe neurological impairment, dependent in everyday life); CPC 4 (coma); and CPC 5 (brain death). Secondary outcomes were an assessment of disability according to modified Rankin scale (mRS) at 6 months and all-cause mortality at the end of the trial. Scores on the mRS range from 0 to 6, with 0 representing no symptoms, 1 no clinically significant disability, 2 slight disability, 3 moderate disability, 4 moderately severe disability, 5 severe disability, and 6 death.
CPC scores 1 or 2 and mRS 0 to 3 were considered a good outcome, whereas CPC 3 to 5 and mRS 4 to 6 were considered a poor outcome.
Comparisons of patients’ clinical characteristics between temperature groups were performed using the Wilcoxon rank sum test for continuous variables and chi-square or Fisher exact test for categorical variables. Medians with interquartile range (IQR) and mean ± SD are presented.
Changes of NSE concentrations over time were tested for significance using the Wilcoxon signed rank test. Comparison of NSE distribution between CPC groups was performed with the Wilcoxon rank sum test.
At each time point, receiver-operating characteristic curves were plotted and corresponding areas under the curve (AUCs) were determined to evaluate the predictive power of NSE on CPC. Cutoffs were provided as a compromise between sensitivity and specificity by maximizing the Youden index, as defined by sensitivity + specificity – 100%, and by providing 95% to 100% specificity. The same analyses were then performed on NSE change from 24 to 48 h and from 48 to 72 h. All sensitivity and specificity values were corrected for optimism using bootstrap internal validation (100-fold) to avoid overfitting (17). When possible, the normal approximation was used to obtain confidence intervals (CIs); otherwise, the Wilson formula was used (18).
Finally, NSE at 24, 48, and 72 h were added to a clinical multivariable logistic model containing temperature allocation, age, sex, bystander cardiopulmonary resuscitation, first monitored rhythm, time from cardiac arrest to ROSC, lactate levels, and circulatory shock on admission. The relationship between NSE and CPC was supposed to be linear; Pearson residuals were plotted and did not reveal any strong pattern. Restricted cubic splines were also used to model the nonlinear relationship between NSE and CPC, but the findings were not markedly different (data not shown). The additional predictive power brought by NSE to these markers was evaluated by computing the continuous net reclassification index (NRI) and the integrated discrimination improvement (IDI) (19). In the multivariable analysis, missing values were accounted for using 10-fold multiple imputations.
Computations were performed using the R software, version 2.15.2, packages ROCR, pROC, Hmisc, and rms (R Foundation for Statistical Computing, Wien, Austria). A p value <0.05 was considered to indicate statistical significance.
The TTM trial investigated 939 patients with no difference in mortality or neurological function between the 33°C and 36°C groups (11). Overall, 700 consecutive patients from 29 different sites participated in the biomarker substudy (Figure 1A). A total of 1,823 serum samples from 686 different patients were analyzed (Figure 1B).
Main patient characteristics are shown in Table 1. There were no significant differences between our study population and the main TTM trial population or in neurological outcome between temperature groups (p = 0.90) (Figure 1A).
Median NSE values were 18 ng/ml (IQR: 12 to 27 ng/ml) versus 35 ng/ml (IQR: 21 to 58 ng/ml), 15 ng/ml (IQR: 10 to 2 ng/ml 1) versus 61 ng/ml (IQR: 24 to 125 ng/ml), and 12 ng/ml (IQR: 9 to 16 ng/ml) versus 54 ng/ml (IQR: 19 to 132 ng/ml) for good versus poor outcome at 24, 48, and 72 h, respectively (p < 0.001). NSE values in both temperature groups were higher in the poor versus the good outcome group at each time point (Figure 2). In both good and poor outcome groups, levels of NSE were not significantly affected by the target temperature level.
In the poor outcome groups, we observed a significant increase of median NSE values between 24 and 48 h in both temperature groups: from 35 ng/ml (IQR: 21 to 56 ng/ml) to 60 ng/ml (IQR: 22 to 119 ng/ml) in 33°C (p < 0.001) and from 34 ng/ml (IQR: 21 to 62 ng/ml) to 66 ng/ml (IQR: 24 to 137 ng/ml) in 36°C (p < 0.001). Between 48 and 72 h, median NSE values decreased in the 33°C group from 60 ng/ml (IQR: 22 to 119 ng/ml) to 52 ng/ml (IQR: 20 to 147 ng/ml) (p = 0.029) and in the 36°C group from 66 ng/ml (IQR: 24 to 137 ng/ml) to 56 ng/ml (IQR: 19 to 123 ng/ml) (p = 0.75).
In the good outcome groups, we detected a significant decrease of approximately 3 to 4 ng/ml between 2 consecutive time points, with median NSE values at 24, 48, and 72 h at 33°C of 18 ng/ml (IQR: 12 to 27 ng/ml), 15 ng/ml (IQR: 11 to 22 ng/ml), 13 ng/ml (IQR: 9 to 18 ng/ml), respectively (p < 0.001), and at 36°C of 18 ng/ml (IQR: 12 to 26 ng/ml), 14 ng/ml (IQR: 10 to 20 ng/ml), 11 ng/ml (IQR: 8 to 15 ng/ml), respectively (p < 0.001).
The capacity of NSE to predict CPC at 6 months was first determined using receiver-operating characteristic curves (Figures 3A to 3C). Twenty-four hours after cardiac arrest, NSE predicted 6-month CPC with an AUC of 0.75. At 48 and 72 h, AUCs were 0.85 and 0.86, respectively. The AUCs obtained at 33°C and 36°C groups were similar.
The change of NSE between 24 and 48 h had an AUC of 0.80 (33°C group) and 0.84 (36°C group), and between 48 and 72 h, the AUC was lower than 0.70 for both groups. An increase of NSE of 6 ng/ml between any of the time points, regardless of the target temperature, was also predictive of a poor outcome (specificity 94% and sensitivity 64% between 24 and 48 h; specificity 93% and sensitivity 39% between 48 and 72 h).
In our cohort, the previously recommended (7) cutoff value of 33 ng/ml at 48 h yielded a specificity of 0.91 and a sensitivity of 0.65.
By maximizing the Youden index, cutoff values for NSE in the pooled patient group were 27, 29, and 23 ng/ml at 24, 48, and 72 h, respectively (Table 2). NSE cutoff values with false positive rates (FPRs) from 5 to 1 range from 49 to 76 ng/ml, 42 to 68 ng/ml, and 33 to 45 ng/ml at 24, 48, and 72 h, respectively (Table 2). No patient with a good outcome had an NSE value at or above the cutoff reported with an FPR of zero (“NSE 0” values in Table 2).
Kaplan-Meier curves showed that survival was significantly lower in groups with higher NSE levels as defined by quartiles (Central Illustration). NSE at each time point was an efficient predictor of survival in both temperature groups (all p < 0.05).
In multivariable analysis including serial NSE, target temperature, and baseline variables (age, sex, bystander cardiopulmonary resuscitation, first monitored rhythm, time to ROSC, lactate levels on admission, and circulatory shock), NSE was a strong predictor of neurological outcome at each time point (Table 3). Our model integrating NSE measures at 3 time points had a specificity of 0.88 and a sensitivity of 0.84. Continuous NRI (1.29; p < 0.001) and IDI (0.37; p < 0.001) showed that NSE significantly improved classification compared with a model with clinical parameters alone. When modeling CPC using the mean and the trend effects of NSE values, which are independent, we found the same results. When adjusted for centers, NSE remained a highly significant outcome predictor, and no site effect was observed (data not shown).
When analyzing the capacity of NSE to predict mRS and death at 6 months as well as death at the end of the trial, we found similar results to those referring to CPC at 6 months and with no influence of target temperature (data not shown).
In a large international trial of patients treated with targeted temperature after out-of-hospital cardiac arrest, NSE was a strong and robust predictor of outcome (Central Illustration). Target temperature level did not significantly influence NSE values.
Although median NSE values declined between 48 and 72 h in all groups, we confirmed that an increase of NSE between any 2 time points was associated with poor outcome (20–24). There was no significant difference between temperature groups at any time point for any of our outcome measures, substantiating previous studies reporting no statistically significant differences in NSE values between temperatures (21,25–27). Other studies reporting lower NSE values in 33°C-treated patients suffered from limitations, notably due to the comparison of patients treated at 33°C to historical control subjects (28) or to a small sample size in a population without fever management in the control group (20).
The cut-off values at 48 and 72 h after ROSC provided the best capacity to predict outcome when referring to the highest sensitivities and specificities. At 24 h, sensitivity was too low to be of clinical interest. Deliberately, we presented FPRs of 5% or lower, as no compromise in the literature exists that defines the absolute best characteristics of a biomarker cutoff value. As such, we showed that “high” NSE cutoff values (with ≤5% FPR and tight 95% CIs) offer reliable prediction of poor outcome with sufficient sensitivity to remain clinically useful within a multimodal prognostication package, including clinical examination, imaging, neurophysiology, and biomarkers (10,29). Notwithstanding the low FPR and narrow 95% CIs of cutoff values in our sample, no single test, even with high specificity, should be considered for withdrawal of life-sustaining therapies. Also, by looking for cutoffs with an FPR of 0, indicating absolute poor outcome prediction, values around 100 ng/ml might have a too low sensitivity to be of clinical utility.
Our cutoff values are higher than the formerly reported 33 ng/ml at 48 h (6,7). Several explanations exist for discrepancies with previous studies. First, the assay we used differed from some of the previous reports, and variability among NSE assays is well described (30). A recent publication by Rundgren et al. (8) showed that NSE values can vary by 15% to 36% based on the assay used and whether fresh or frozen samples were analyzed. Outcome measurements in previous publications also differed in time to follow-up, ranging from ICU discharge to 6 months, and some studies categorized neurological outcome differently (5). We used the most common follow-up period of 6 months, the CPC 1 or 2 score for good outcome, the CPC 3 to 5 scores for poor outcome, and, most importantly, a blinded outcome assessment with face-to-face interviews (11,12,31,32). Another explanation for our reported discrepancies might be that the TTM trial prognostication and, when indicated, subsequent withdrawal of life-sustaining therapy were well codified and delayed. Furthermore, our sample included more than twice the number of patients compared with the 272 normothermic individuals in the PROPAC (PROgnosis after PostAnoxic Coma) trial (6). The latter served as a basis of the American Academy of Neurology guidelines for outcome prediction, which fixed the NSE cutoff as 33 ng/ml with an FPR of 0 (7). In our cohort, an FPR of 0 could not be verified at the 33 ng/ml cutoff, which yielded an FPR of 9%. Zellner et al. (33) had similar findings as ours in patients at 33°C with 10% FPR at cutoff values of 41 ng/ml at 48 h.
Our multivariable model, integrating NSE at the 3 time points, confirmed NSE as a predictor of CPC in this set of patients as shown by the highly significant NRI and IDI. These findings are in line with previous studies and strengthen the position of NSE as a robust and clinically-useful outcome predictor (21).
Study strengths and limitations
Although being a pre-defined substudy of the TTM trial, not all sites enrolling in the main trial participated in biomarker sampling. However, as the trial was stratified for sites, the balanced design tends to be preserved in all comparisons between the temperature groups. Indeed, our population did not differ significantly from the TTM trial population. Not all patients had blood samples taken at every time point, and there was no external quality control at each participating site where samples were collected and pre-analytically processed.
Biomarkers, unlike some prognostic neurophysiology tests, are unaltered by sedation and may, therefore, be a more objective marker of brain injury. One general limitation of biomarkers is that their measurement is punctual, whereas production or secretion is a dynamic process, highlighting the importance of serial measurements taking into account the absolute values, their changes over time, and serial cutoffs to best predict outcome (4). Brain biomarkers measured in circulating blood might have some additional weaknesses as the integrity of the blood brain barrier after ischemia-reperfusion injury in individuals cannot be measured and may vary substantially. In the case of NSE, which is predominantly released from neural and neuroendocrine cells, caution is warranted as serum levels might reflect variable degrees of brain damage, disruption of the blood brain barrier, or—albeit rarely—NSE from extracerebral origins as seen in small-cell lung cancer and neuroendocrine tumors (24).
The major strength of this investigation is that it was a pre-defined substudy investigating a serum biomarker for prognostication after OHCA within the largest multicenter randomized clinical trial studying 2 target temperature regimens in comatose cardiac arrest patients. It represents the largest prospective study of its kind. All analyses were performed in a single core laboratory, limiting the influence of assay variability and laboratory processing. The results of NSE values were not available to the treating physicians during the trial and, therefore, did not influence prognostication of patients, reducing the risk of “self-fulfilling prophecy.” A unique feature of the TTM trial is that prognostication and withdrawal were standardized, which increases the validity of our results (34).
As is the case with other prognostic tools, the current study demonstrated that the functional consequences of brain injury cannot be predicted by NSE alone. When using NSE, we recommend a dynamic approach with serial measurements within a prognostication protocol including other methods, such as clinical examination, electroencephalogram, brain imaging, and somatosensory-evoked potentials, for the most accurate outcome prediction (10,29).
Serial, high NSE values have a high predictive value of poor outcome in comatose out-of-hospital cardiac arrest patients. This predictive value of NSE is not significantly affected by target temperature at either 33°C or 36°C.
COMPETENCY IN MEDICAL KNOWLEDGE: Cardiac arrest is associated with 50% mortality in those patients admitted to hospital after resuscitation in the field, and the predominant cause of death in these cases is severe neurological injury.
COMPETENCY IN PATIENT CARE: Consistently high levels of brain-oriented biomarkers, like NSE, may identify patients prone to poor outcomes after resuscitation from OHCA.
TRANSLATIONAL OUTLOOK: Identification of combinations of variables that accurately correlate with more or less favorable outcomes could lead to the development of more effective therapeutic strategies for victims of cardiac arrest.
The authors thank Jacqueline Kieffer, the team of the Integrated BioBank of Luxembourg, the staff of the biochemistry laboratory of the Centre Hospitalier de Luxembourg, and the staff from all of the sites involved in the biomarker collection and handling.
The TTM-Trial was funded by independent research grants from the Swedish Heart Lung Foundation; Arbetsmarknadens Försäkringsaktiebolag Insurance Foundation; Swedish Research Council; regional research support, Region Skåne; governmental funding of clinical research within the Swedish National Health Services; Thelma Zoega Foundation; Krapperup Foundation; Thure Carlsson Foundation; Hans-Gabriel and Alice Trolle-Wachtmeister Foundation for Medical Research; Skåne University Hospital, Sweden; TrygFonden, Denmark; the European Clinical Research Infrastructures Network; and the European Critical Care Research Network. There was no commercial funding. Funding organizations neither had any access to the data nor had any influence on the analysis or interpretation. Drs. Collignon and Devaux have received support from the Ministry of Higher Education and Research of Luxembourg and National Research Fund. Dr. Wise has served on the advisory board of Bard Medical. Dr. Erlinge has received speakers fees from ZOLL and Philips. Drs. Nielsen and Pellis have received speaker fees from Bard Medical. Dr. Erlinge has received speaker fees from Zoll and Philips. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- cerebral performance category
- false positive rate
- intensive care unit
- integrated discrimination improvement
- net reclassification index
- neuron-specific enolase
- out-of-hospital cardiac arrest
- return of spontaneous circulation
- Received December 19, 2014.
- Revision received February 15, 2015.
- Accepted March 1, 2015.
- 2015 American College of Cardiology Foundation
- Wijdicks E.F.,
- Hijdra A.,
- Young G.B.,
- Bassetti C.L.,
- Wiebe S.,
- Quality Standards Subcommittee of the American Academy of Neurology
- Harrell F.E.
- Tiainen M.,
- Roine R.O.,
- Pettila V.,
- Takkunen O.
- Reisinger J.,
- Hollinger K.,
- Lang W.,
- et al.
- Becker L.B.,
- Aufderheide T.P.,
- Geocadin R.G.,
- et al.