Background

Unresponsiveness is a challenging complaint in patients during emergency department visits and hospitalizations [1]. Various causes underlie unresponsiveness, including metabolic encephalopathy, structural brain lesions, meningoencephalitis, and seizure disorders. Early management of the underlying causes, as well as the clinical symptoms, should be promptly provided in order to avoid clinical deterioration, morbidity, and mortality [2]. Hospitalized patients generally have more comorbidities and are sicker than the general population; hence, the etiologies and outcomes of sudden unresponsiveness may differ between in-hospital patients and outpatients. Nevertheless, comprehensive studies on the early evaluations and long-term outcomes of hospital-onset unresponsiveness (HOU) are lacking.

Pupillary examinations are the gold standard in the initial evaluation of unresponsive patients [3, 4]. The assessments of pupils include diameter, shape, symmetry, and light reflexes. Evaluation of the pupillary light reflexes (PLR) is conventionally performed by shining light into the patient’s eye and subjectively described by examiners. However, such conventional assessments of PLR have low inter-rater and intra-rater correlations [5]. Recently, quantitative pupillometry (QP) was introduced to the neurocritical care field to provide objective assessments of the PLR. QP is able to detect subtle early changes in the size and light reflexes of pupils [6]. This bedside tool is becoming popular in the routine evaluation of unresponsive patients who are admitted to intensive care units and require serial measurements of pupils, especially for those with brain herniation syndrome (BHS) [7,8,9,10]. Moreover, QP was shown to be useful in neurological prognostication in unresponsive patients after cardiac arrest [11,12,13,14,15,16,17]. Because the PLR is regulated by the autonomic nervous system, QP may also have a role in the assessment of autonomic dysfunction in critically ill patients [18, 19]. However, studies on the QP findings in patients who are hospitalized outside intensive care units are lacking. Moreover, the clinical implication of the findings of sluggish pupils remains largely uninvestigated in unresponsive patients, although fixed dilated pupils inarguably suggest urgent situations and grave prognosis [3].

The aims of the study were the following: (1) to describe the etiology of HOU, (2) to clarify whether quantitative assessments of the PLR through QP measurement can be used to detect BHS, which is the serious in-hospital neurological complications requiring emergent interventions, and (3) to explore which QP parameters are related to clinical outcomes in patients who developed HOU after admission for non-neurological reasons.

Methods

Patients

This registry-based longitudinal observational study was performed between September 1, 2017 and November 30, 2018 at Asan Medical Center, a tertiary hospital located in Seoul, South Korea. All data for the current study have been collected and documented on the registry of the Neurological Alert Team (NAT), a round-the-clock neurologist-led rapid response team organized for the improvement of hospital-wide performance in response to in-patient neurological emergency [20]. For this study, we included patients who were (1) 18 years of age or older, (2) admitted to non-neurological departments, and (3) activated the NAT due to hospital-onset unresponsive state (i.e., stupor or comatose state) as assessed by the ACDU (Alert, Confused, Drowsy, and Unresponsive) scale, which is a simple four-point scale for assessing patients presenting with altered mental status [21]. The NAT neurologists or clinical nurse specialists routinely measured the patients’ pupils by using QP for initial neurological assessment during office hours. We excluded patients with no available QP measurements, incomplete demographic data, or unresponsiveness following cardiac arrest. This study was approved by the institutional review board of Asan Medical Center and the need for written informed consent was waived considering the retrospective nature of the study.

Clinical assessments

According to our protocol, the neurologists performed initial neurological assessments as soon as possible following NAT activation. Systematic evaluations, including vital signs, ACDU scale, Glasgow Coma Scale (GCS), laboratory tests, computed tomography (CT) and/or magnetic resonance imaging (MRI) of the brain, and electroencephalography (EEG) were conducted. Basal functional status before the admission was also assessed with the modified Rankin Scale (mRS) by the neurologists, based on the information from the next of kin. Clinical outcomes were assessed by the survival status at hospital discharge and the mRS at 3 months after the day of NAT activation. The mRS at 3 months was evaluated via clinical visits, telephone interviews, and medical record review and was documented on the dedicated NAT registry [20]. Neurological status was categorized into good (mRS = 0–3) and poor (mRS = 4–6) [22]. All data were prospectively documented in our registry and were reviewed and adjudicated by the investigators at weekly conferences.

Pupillometry measurements

QP was measured at bedside using the NPi-100 pupillometer (NeurOptics, Irvine, CA, USA). The detection thresholds were 10.00 mm for the maximum pupil diameter (Max), 1.00 mm for the minimum pupil diameter (Min), and 0.03 mm for the change in size. The QP parameters included Max (mm), Min (mm), percentage of change (CH = 100 × [Max – Min]/Max), constriction velocity (CV, mm/sec), latency of constriction (Lat, msec), dilation velocity (DV, mm/sec), and Neurological Pupil index (NPi). NPi is a scale with values ranging from 0.0 to 5.0, a combination of z-scores of Max, Min, CH, CV, Lat, and DV measurements compared against the mean distribution of values obtained from healthy volunteers [8]. Low NPi value can be interpreted as less distinct PLR (i.e., decreased pupillary reactivity). All QP assessments were performed at least twice in each eye. Among the repetitive QP measurements, the dataset with the maximum NPi value content was documented for each eye to minimize type I error, in which normal PLR is considered sluggish. From the results measured on each eye, we selected the more inferior parameters (lower NPi, lower CH, lower CV, lower DV, higher Max, higher Min, and higher Lat) for either eye for the final analysis to detect any abnormalities of PLR of either eye sensitively. As a result, each QP parameter used in the final analysis may come from different eyes.

Etiology of hospital-onset unresponsiveness

The etiology of HOU was discussed by the neurologists at a weekly conference. The causes of HOU categorized as follows: (1) metabolic encephalopathy (e.g., sepsis, organ dysfunction, electrolyte imbalance, and drug side effects); (2) ischemic stroke; (3) hemorrhagic stroke; (4) intracranial bleeding (subdural or epidural); (5) brain tumor (primary or metastatic); (6) meningoencephalitis; (7) seizure-related events (convulsive or non-convulsive seizure, status epilepticus, or postictal status); (8) hypoxic-ischemic encephalopathy; and (9) other causes (e.g. syncope, pseudocoma, and cerebral concussion). In case an HOU had multiple causes, their contributions to the HOU were ranked. All data were prospectively documented in our registry and were reviewed and adjudicated by the investigators at weekly conferences.

The presence of BHS was radiographically and clinically evaluated. We defined BHS as the development of acute brain lesions with mass effect that were relevant to the unresponsiveness, including midline shift, brainstem compression, hydrocephalus, and diffuse cerebral edema [23]. All CT scans and MRI sequences were jointly interpreted by two investigators who were blinded to the QP data and clinical outcomes. A third investigator was consulted in cases of disagreement.

Statistical analyses

Data are summarized as mean ± standard deviation for normally distributed continuous variables, medians and interquartile ranges for non-normally distributed continuous variables, and frequencies with percentages for categorical variables. Univariable analyses were carried out to identify the relationship between each variable and BHS or clinical outcomes; categorical variables were compared with Pearson’s χ2 test or Fisher’s exact test, and continuous variables were compared with the Student’s t-test. Variables with p values < 0.1 in the univariable analysis were included as candidate variables in the multivariable logistic regression model and removed by backward stepwise selection. We further performed all analyses using a forward selection procedure to confirm the final model. All QP variables significantly associated with BHS or clinical outcomes in the univariable analysis were individually added to the final clinical models to generate the odds ratio (OR) and the 95% confidence intervals (CIs) thereof. To analyze the usefulness of QP for the prediction of BHS, we calculated the predictive accuracy by sensitivity, specificity, negative predictive value, and positive predictive value. Two-tailed p values < 0.05 were considered statistically significant. All statistical analyses were performed using R version 3.6.2 (R Foundation for Statistical Computing, Vienna, Austria) and GraphPad Prism, version 8.3.1 (GraphPad Software, La Jolla, CA, USA).

Results

Patient characteristics

A total of 1442 patients activated the NAT during the study period. We included 404 (28%) patients who were in an unresponsive state as assessed by the ACDU scale at the time of on-call consultations to the NAT; and excluded 190 patients who either did not have QP measurements due to NAT activation during after-hours (n = 129), showed unresponsiveness following cardiac arrest (n = 46), or had incomplete demographic data (n = 15). Thus, the remaining 214 patients were included in the final analysis; their mean age was 63.9 ± 14.9 years, and 127 (59%) were men. The baseline characteristics of the 214 patients are presented in Table 1.

Table 1 Baseline Characteristics
Full size table

Etiology of hospital-onset unresponsiveness

Of the included patients, 195 (91%) underwent CT and/or MRI of the brain, which showed acute intracranial lesions in 104 (53%) patients. Of them, 91 patients had brain lesions relevant to the HOU, and 13 patients had brain lesions that were irrelevant to the HOU (acute small infarcts). EEG was performed in 174 (81%) patients, including continuous EEG monitoring in 25 patients. The identified causes of HOU in 214 patients are shown in Fig. 1. The most common cause of HOU was metabolic encephalopathy (95/214), followed by seizure-related events (61/214) and ischemic stroke (19/214).

Fig. 1
figure 1

Etiologies of hospital-onset unresponsiveness. Etiologies were categorized into primary and contributing causes in 214 patients who were admitted to the hospital for non-neurological diseases

Full size image

Brain herniation syndrome

BHS was found in 37 (17%) patients, who showed higher in-hospital mortality rate compared with those without BHS (57% [21/37] vs. 21% [38/177]; p <  0.001). The diagnoses of the patients with BHS were hemorrhagic stroke (n = 10), ischemic stroke (n = 6), acute liver failure (n = 6), brain tumor or leptomeningeal seeding with acute hydrocephalus (n = 6), hypoxic-ischemic brain injury (n = 3), meningoencephalitis (n = 2), subdural hemorrhage (n = 2), cerebral venous thrombosis (n = 1), and unknown etiology of diffuse cerebral edema (n = 1). In univariable analysis, younger age, good functional status before admission, absence of a history of hypertension or diabetics, and lower GCS score at NAT activation were significantly associated with the risk of BHS (Table 2, Additional File 1: Table S1). Among the QP parameters, lower NPi, larger Max, and larger Min showed independent association with the risk of BHS on multivariable analysis. The NPi cut-off value of < 1.6 provided the maximum area under the receiver operating characteristic (ROC) curve (0.750) for predicting BHS, with a sensitivity of 49% (95% CI = 32–66) and a specificity of 91% (95% CI = 86–95) (Additional File 1: Figure S1). In addition, the positive predictive value and negative predictive value of this NPi cut-off value for predicting BHS were 53% (95% CI = 39–67) and 89% (95% CI = 86–92), respectively.

Table 2 Factors Associated with the Presence of Brain Herniation Syndrome
Full size table

In-hospital mortality

In-hospital mortality was noted in 59 (28%) patients. The presence of hemorrhagic stroke (12% [7/59] vs. 3.2% [5/155], p = 0.034) and presence of BHS (36% [21/59] vs. 10% [16/155], p <  0.001) were significantly associated with in-hospital mortality. NPi was significantly lower in patients with in-hospital mortality than in those without (3.2 [1.2–4.2] vs. 4.0 [3.2–4.4], p = 0.001). The following QP variables were also significantly associated with in-hospital mortality: smaller CH (14.5 ± 9.3 vs. 20.0 ± 10.5%, p = 0.001), slower CV (1.0 ± 0.8 vs. 1.4 ± 0.9 mm/sec, p = 0.004), longer Lat (331.8 ± 69.0 vs. 297.4 ± 51.9 msec, p = 0.002), and slower DV (0.4 ± 0.4 vs. 0.6 ± 0.4 mm/sec, p = 0.005). In multivariable analysis, the presence of BHS (OR = 4.80, 95% CI = 2.28–10.09, p <  0.001), NPi (OR = 0.77, 95% CI = 0.62–0.96, p = 0.019), CH (OR = 0.95, 95% CI = 0.92–0.98, p = 0.004), CV (OR = 0.56, 95% CI = 0.36–0.86, p = 0.008), Lat (OR = 1.01, 95% CI = 1.00–1.02, p = 0.001), and DV (OR = 0.32, 95% CI = 0.13–0.78, p = 0.011) were independently associated with in-hospital mortality.

Neurological outcomes

A total of 201 (94%) patients had available data on mRS at 3 months. The median mRS at 3 months was 5 (interquartile ranges 3–5). Poor neurological outcome was noted in 152 (76%) patients. Table 3 and Additional File 1: Table S2 show the factors associated with poor neurological outcomes in patients with HOU, which included age, pre-admission mRS ≥ 4, diabetes mellitus, cancer, and previous stroke. The following QP Parameters were also associated with poor neurological outcomes at 3 months: NPi, CH, CV, Lat, and DV. Multivariable analysis showed that pre-admission mRS ≥ 4, cancer, previous stroke, NPi, CH, Lat, and DV were significantly associated with poor neurological outcomes at 3 months. The area under the ROC curve for poor neurological outcomes at 3 months was 0.762 [95% CI = 0.694–0.830] in the model with clinical variables (pre-admission mRS ≥ 4, cancer, and previous stroke). Although not statistically significant, the area under the ROC curve for poor neurological outcomes at 3 months was increased to 0.837 [95% CI = 0.784–0.894] upon addition of the QP variables (NPi, CH, Lat, and DV) (Fig. 2). The optimal cut-off value of NPi for the poor neurological outcomes at 3 months was 3.0, with sensitivity of 33% (95% CI = 26–41), specificity of 82% (95% CI = 68–91), positive predictive value of 85% (95% CI = 75–91), and negative predictive value of 28% (95% CI = 25–32).

Table 3 Factors Associated with Poor Neurological Outcomes at 3-months
Full size table
Fig. 2
figure 2

Receiver operator characteristic curves showing accuracy for the prediction of poor neurological outcomes at 3 months. The AUC for clinical variables (pre-admission modified Rankin Scale ≥4, cancer, and previous stroke) were 0.762 (a dash line); and the AUC for QP parameters (Neurological Pupil index, percentage of change, latency of constriction, and dilation velocity) in addition to aforementioned clinical variables were 0.837 (a green line). AUC area under the curve, QP quantitative pupillometry

Full size image

Discussion

In this registry-based observational study, we showed that the application of QP was useful for detecting BHS and predicting clinical outcomes in patients with HOU. When the patients’ altered mental status activated the neurological rapid response team, the initial detection of reduced pupillary reactivity (lower values of NPi) as well as enlarged pupillary size (Max and Min) were significantly associated with later identifications of BHS. The specificity and negative predictive value of the finding of NPi < 1.6 were as high as ~ 90% in predicting the presence of BHS, while the sensitivity and positive predictive value were relatively low at ~ 50%. A previous study examined the predictive values of PLR using a conventional penlight method in patients presenting to an emergency room in comatose states (GCS score less than eight) [4]. The absence of manual PLR in at least one eye showed a high sensitivity of 83% (95% CI = 76–90) but a relatively low specificity of 77% (95% CI = 69–85) for predicting structural causes of coma. In addition, the sensitivity of anisocoria more than 1 mm in this population was only 39% (95% CI = 30–48), while the specificity was at 96% (95% CI = 79–92) [4]. QP parameters indicating reduced pupillary reactivity were significantly associated with in-hospital mortality (lower values of NPi, CH, CV, and DV, and higher values of Lat) and poor neurological outcome at 3 months (lower values of NPi, CH, and DV, and higher values of Lat). The addition of QP in considering the patients’ medical conditions enhanced the prediction value (larger area under the ROC curve) of mRS at 3 months. To the best of our knowledge, this is the first study to show the prognostic value of measuring QP in patients who are admitted for non-neurological illness and develop sudden unresponsiveness.

In terms of the etiologies of HOU, we found that metabolic encephalopathy was the most common cause, followed by seizure-related events, ischemic and hemorrhagic strokes, and metastatic and primary brain tumors. Approximately 43% of our study population had acute diseases of the brain such as stroke, intracranial bleeding, brain tumor, meningoencephalitis, and hypoxic-ischemic brain injury. BHS, one of the most severe life-threatening neurologic emergencies, was found in 17% of patients with HOU and was caused by massive cerebral infarcts, subdural hemorrhage, or diffuse brain swelling related to metabolic conditions (eg, acute liver failure). Seizure-related events including convulsive seizure, nonconvulsive status epilepticus, and postictal state were also as frequent as 29%.

We found that the use of QP may be useful for the early detection of BHS. The pupillary examination is a crucial component in the initial evaluation of unresponsive patients, but early signs of pupils relevant to BHS can be nonspecific or elusive despite careful examination. In emergency situations, it is challenging to quickly perform and properly interpret neurological examinations, especially in hospitalized patients with multiple comorbidities. According to the experience of our neurological rapid response team, QP is a practical tool when used in addition to routine neurological examination, because it is portable and can be quickly performed at bedside. In a previous preliminary observational study, midline shift and increased intracranial pressure were associated with a decrease of CV below 0.6 mm/sec; however, patients with diffuse brain edema and without midline distortion did not show such a decrease in CV until the intracranial pressure exceeded 30 mmHg [10]. In our study, 13 of 37 patients with BHS had diffuse brain swelling but did not have midline shifts, which may account for the absence of differences in CV between the BHS and non-BHS groups. Nevertheless, we demonstrated that the NPi value of less than 1.6 has a high specificity and a high negative predictive value for detecting BHS in patients with HOU. Collectively, these findings suggest that patients who develop unresponsiveness during the hospital stay and show decreased NPi value on QP measurements would benefit from undergoing expeditious neuroimaging studies to detect BHS.

Traditionally, the PLR has been rather subjectively assessed by using a variety of non-standardized light sources by practitioners with varying levels of skills in neurological examinations. Recent studies have shown that such subjective assessments of pupillary reactivity have subpar reliability, with low inter-rater and intra-rater correlations [5]. Recently, the QP has become popular as it provides a non-invasive, hand-held implementation of neuro-monitoring. QP also offers an objective and standardized measurement of the PLR, and mean QP values in healthy volunteers (Max 3.5–5.3 mm, Min 2.4–3.4 mm, CH 29–36%, Lat 0.22–0.27 s, CV 1.5–2.9 mm/sec, and DV 0.9–2.2 mm/sec) have been suggested in prior studies [7, 10, 24,25,26]. Additionally, the automated algorithm of QP provides an NPi value derived from a combination of QP variables. An NPi score below 3.0 is generally considered an abnormal finding of sluggish light reflex [8]. Preliminary studies conducted in intensive care units reported the usefulness of QP for detecting a wide range of conditions including increased intracranial pressure [7,8,9,10], response to osmotherapy [27], discrimination between compressive lesions and microvascular ischemic oculomotor nerve palsy [28], assessment of disease severity of aneurysmal subarachnoid hemorrhage [29], the depth of sedation and analgesia [30], and neurological prognostication in traumatic brain injury and comatose resuscitation-of-spontaneous-circulation following cardiac arrest [11,12,13,14,15,16,17, 31]. The prognostic implication of QP has been investigated in patients with cardiac arrest [11,12,13,14,15,16,17], in whom a decrease in NPi values less than 2.0 was associated with unfavorable neurological outcomes [13]. Our results showed that the QP was useful in detecting BHS and predicting clinical outcomes in patients with HOU, and that decreased NPi values and increased pupillary sizes were significantly related to BHS. Decreased NPi values were also associated with in-hospital mortality and unfavorable neurological outcomes at 3-months. In terms of QP variables, the pupillary size (Max and Min) was not significantly related to mortality rate or neurological outcomes, whereas other QP variables such as CH, Lat, and DV were associated with clinical outcomes including in-hospital mortality and neurological status at 3-months. Therefore, a combination of QP variables may be more reliable than the NPi alone in the prediction of neurological outcomes, despite the high predictive value of the NPi. In summary, our results show that measurements of PLR with QP may be useful in the early detection of life-threatening neurological conditions and neuro-prognostication for patients with HOU.

Fixed or dilated pupils are generally ominous neurological signs, as sluggish or absent PLR may indicate the compression or stretching of the dorsal midbrain in which the Edinger-Westphal nuclei is located, or of the efferent oculomotor nerve that carries parasympathetic fibers [32]. Some studies suggested that the integrity and function of PLR may also be affected by the perfusion defect to the brainstem or alterations of neurotransmitter release [33, 34]. Furthermore, even though pupillary constriction by light stimulus is predominantly integrated by the parasympathetic nervous system, it is possible that sympathetic activities are also engaged in the regulation of the pupillary reactivity. First, the sympathetic nerves are involved in the dilation phase of PLR: the supranuclear inhibition via sympathetic neurons suppresses the pre-ganglionic parasympathetic neurons at the Edinger-Westphal nucleus, resulting in relaxation of the pupil sphincter muscle; also, the sympathetic neurons contract the iris dilator muscle via peripheral sympathetic innervation [32]. Second, the reticular activating system affects the pupil size and PLR by tonic inhibitory input of the Edinger-Westphal complex through releasing norepinephrine [35]. Third, cognitive and emotional processes may result in mydriatic reaction by the input of cortical innervation into the brainstem, although the exact circuits remain poorly understood [35]. Thus, the dynamics of the pupil reactivity may be indicative of lesions or dysfunctions of the cortex, subcortex, and brainstem that affect the parasympathetic system, sympathetic system, neurotransmitters, and their complex interactions.

Further studies are needed to clarify the mechanism underlying the association between QP values and clinical outcomes in patients with HOU in the absence of BHS. It is possible that autonomic dysfunction, as well as multiorgan dysfunction and brain dysfunction (e.g., metabolic encephalopathy and BHS) may underlie such association. Autonomic dysfunction mediated by inflammatory response likely has an important role in the pathogenesis of the dysfunction of the brain or other organs [36]. Neurotransmitter imbalance such as cholinergic deficiency due to inflammation and multiorgan dysfunction in critically ill patients can lead to pupillary dysfunction [18, 33]. In this context, although there is limited evidence on the association between QP values and the severity of encephalopathy [37], QP as an indicator of autonomic and brain dysfunction may play a role in assessing the severity of metabolic encephalopathy as the cause of HOU.

In addition to the inherent limitations of its single-center retrospective design, the present study has the following limitations. First, we did not assess the pupillary dilation reflex, which may be evoked by sensory stimulation and predominantly mediated by the sympathetic nervous system. Instead, we measured the DV, which could reflect sympathetic activity during the dilation phase of the PLR. Second, concurrent use of drugs that could confound the evaluation of the PLR and responsiveness such as opioids, anticholinergics, or sedative agents was not evaluated. Although a previous study showed that the use of these medications within therapeutic doses does not significantly suppress the PLR [38], we cannot exclude the possibility that other medications such as propofol may have affected the PLR [39]. Third, we did not analyze confounders such as underlying pathology of the retina or optic nerve as well as the influence of ambient lights on the PLR [40]. Fourth, we did not assess validated scales such as the Coma Recovery Scale-Revised to distinguish coma and other diagnostic entities in the field of consciousness research (e.g. unresponsive wakefulness syndrome, minimally conscious state and locked-in syndrome). However, we believe that simplified bedside assessment of the level of consciousness such as ACDU scale is more suitable for emergent situations such as in response to in-patient neurological emergencies. Lastly, we did not analyze the duration of HOU, which may have an association with outcomes.

Conclusions

In conclusion, reduced reactivity and increased size of pupils were related to BHS in patients with HOU. Moreover, reduced pupillary reactivity was associated with in-hospital mortality and poor neurological outcomes at 3-months. These findings suggest that QP measurement may be useful for early detection of potentially life-threatening neurological conditions in patients who develop unresponsiveness after admission for non-neurological reasons.