If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
2 Present address: Department of Emergency Medicine, Universitary Hospital Félix guyon, Allée des Topazes, CS 11021, 97400 La Réunion, France.
Matthieu Daniel
Correspondence
Corresponding author at: 1, rue Cabanis, 75014, Department of Neuroanesthesiology and Intensive Care Medicine, Hôpital Sainte Anne, GHU Paris Psychiatrie & Neurosciences, Paris, France.Corresponding author at: Université de Paris, Paris, France.
1 Co-first authors. 2 Present address: Department of Emergency Medicine, Universitary Hospital Félix guyon, Allée des Topazes, CS 11021, 97400 La Réunion, France.
Affiliations
Medical and Surgical Neurointensive Care Unit, Hôpital Sainte-Anne, GHU Paris Psychiatrie et Neurosciences, Paris, FranceUniversity of Paris, Paris, France
Medical and Surgical Neurointensive Care Unit, Hôpital Sainte-Anne, GHU Paris Psychiatrie et Neurosciences, Paris, FranceDepartment of Infection and Epidemiology, Pasteur Institute, University of Paris, Paris, France
1 Co-first authors. 2 Present address: Department of Emergency Medicine, Universitary Hospital Félix guyon, Allée des Topazes, CS 11021, 97400 La Réunion, France.
Pupillary parameters PD, PDV and slope were found significantly lower in COVID-19 patients who will die.
•
A reduced baseline PD <2.75 mm is the most predictive parameter of in-hospital mortality.
•
PD and PDV were also associated with the need for invasive mechanical ventilation with greater significance for PD.
•
A reduced PD could result from a central autonomic impairment related to a brainstem dysfunction in severe COVID-19 patients.
Abstract
Introduction
ICU patients with SARS-CoV-2-related pneumonia are at risk to develop a central dysautonomia which can contribute to mortality and respiratory failure. The pupillary size and its reactivity to light are controlled by the autonomic nervous system. Pupillometry parameters (PP) allow to predict outcomes in various acute brain injuries. We aim at assessing the most predictive PP of in-hospital mortality and the need for invasive mechanical ventilation (IV).
Material and methods
We led a prospective, two centers, observational study. We recruited adult patients admitted to ICU for a severe SARS-CoV-2 related pneumonia between April and August 2020. The pupillometry was performed at admission including the measurement of baseline pupillary diameter (PD), PD variations (PDV), pupillary constriction velocity (PCV) and latency (PDL).
Results
Fifty patients, 90 % males, aged 66 (60–70) years were included. Seven (14 %) patients died in hospital. The baseline PD (4.1 mm [3.5; 4.8] vs 2.6 mm [2.4; 4.0], P = 0.009), PDV (33 % [27; 39] vs 25 % [15; 36], P = 0.03) and PCV (3.5 mm.s−1 [2.8; 4.4] vs 2.0 mm.s−1 [1.9; 3.8], P = 0.02) were significantly lower in patients who will die. A PD value <2.75 mm was the most predictive parameter of in-hospital mortality, with an AUC = 0.81, CI 95 % [0.63; 0.99]. Twenty-four (48 %) patients required IV. PD and PDV were significantly lower in patients who were intubated (3.5 mm [2.8; 4.4] vs 4.2 mm [3.9; 5.2], P = 0.03; 28 % [25; 36 %] vs 35 % [32; 40], P = 0.049, respectively).
Conclusions
A reduced baseline PD is associated with bad outcomes in COVID-19 patients admitted in ICU. It is likely to reflect a brainstem autonomic dysfunction.
The brainstem may be as a main target of viral invasion via the olfactive and trigeminal pathway but also of neuro-inflammatory process triggered by the systemic inflammatory response (
Olfactory and gustatory dysfunctions as a clinical presentation of mild-to-moderate forms of the coronavirus disease (COVID-19): a multicenter european study.
). The brainstem controls sleep-wake cycle via the ascending reticular activating system (ARAS), the vital functions via the autonomic nervous system (ANS) and brainstem reflexes, including the pupillary diameter (PD) and light reflex (PLR). In parallel, more and more authors report dysautonomia in patients with COVID-19 in the acute phase and as long-term sequelae (
) and organ failures and death, irrespectively of the severity lung injury and the systemic inflammatory response. This BSD might therefore be routinely assessed in patients with severe SARS-CoV-2 pneumonia admitted in intensive care unit (ICU).
We reasoned that quantitative measurement of PD and PLR could be a marker of BSD and as such have a prognostic value (
Quantitative pupillometry and neuron-specific enolase independently predict return of spontaneous circulation following cardiogenic out-of-hospital cardiac arrest: a prospective pilot study.
Quantitative versus standard pupillary light reflex for early prognostication in comatose cardiac arrest patients: an international prospective multicenter double-blinded study.
Quantitative pupillometry and neuron-specific enolase independently predict return of spontaneous circulation following cardiogenic out-of-hospital cardiac arrest: a prospective pilot study.
The objectives of our prospective observational study were to assess whether basis PD and light-induced PD variation (PDV), pupillary constriction velocity (PCV) and latency are associated with in-hospital mortality (primary endpoint) and with the need for invasive mechanical ventilation (secondary endpoint) in patients admitted in ICU for SARS-CoV-2 pneumonia.
2. Methods
2.1 Study approval
This study was approved by independent institutional review boards (IRBN602020/CHUSTE for Saint-Etienne Academic Hospital and CPP Sud-Méditerranée V n°2020-A01003-36 for hospital Sainte-Anne). Written informed consent was obtained from all patients or a legal surrogate before enrollment in this study. All images are taken from anonymized data from patients who provided their prior consent. The CNIL (Commission Nationale Informatique et Libertés) has also given its agreement for the use of data within the framework of this study.
2.2 Design
This was a prospective, two centers, cohort observational study. We included all consecutive patients, aged 18 years or older, admitted to the ICU of Saint-Etienne University Hospital and Saint-Anne Hospital between April 14th and August 13th of 2020 for a SARS-CoV-2 infection confirmed by a nasopharyngeal Reverse Transcriptase polymerase chain reaction (RT-PCR). All included patients presented a COVID-19 associated pneumonia with severe illness, as defined in the NIH COVID-19 guidelines by any of the following criteria: respiratory rate ≥ 30 breaths per minute, SpO2 < 94 %, PaO2/FiO2 ratio < 300 and involvement of ≥50 % of the lung parenchyma as evidenced by a reference chest CT-scan (
). Exclusion criteria were represented by preexisting neurologic evolutive/degenerative disorders, pregnancy/nursing condition, mental illness or impossibility for the subject to have a good comprehension of the study, therapeutic limitation before inclusion and lack of health insurance coverage.
2.3 Pupillary variables assessment
Pupillary variables were measured with using the AlgiScan® video pupillometer (iDMed™, Marseille, France) that delivers a calibrated flashlight. The quantitative pupillometry test was performed on each eye and at time of admission in ICU (Day 1 or D1). The technique consisted of placing the tested pupil in the dark for 10 s, with contralateral eye closed, and then to record the baseline and post-flashlight pupillary variables over 3 s. The pupillary variables included the baseline and post-flashlight pupillary diameter (PD in mm), the PD variation (PDV in %), the pupillary constriction velocity (mm.s−1) and the latency (ms).
2.4 Data collection
At time of inclusion (D1), we collected the demographic characteristics (i.e. age, gender, weight and height) and also main COVID-19 associated features (i.e. history of type 2 diabetes mellitus, hypertension, dyslipidemia, overweight or obesity, immunosuppression, neoplasia, ischemic heart disease, liver failure or chronic kidney disease). The presence of multilobed involvement on CT, pulmonary embolism, septic shock and hemorrhagic stroke was assessed during hospitalization as well as the need for mechanical ventilation. We also evaluated the critical illness severity with help of SAPS-II, SOFA score and the Glasgow coma scale (GCS). The use of epinephrine, sedative and opioids, since they can modify the pupillary reactivity, were collected. Patients were followed-up to hospital discharge. The onset and duration of invasive mechanical ventilation (IV), the time and cause of death and the length of stay in ICU and in hospital were measured.
This manuscript adheres to the applicable CONSORT guidelines.
2.5 Statistical analysis
Continuous data were expressed as mean and standard deviation (mean ± SD). The assumption of normality was assessed with the Shapiro-Wilk test. The comparisons between groups (death yes/no) were performed using the Student t-test or the Mann-Whitney test when assumptions required for the t-test were not met. The homoscedasticity was analyzed using the Fisher-Snedecor test. The relationships between continuous variables were explored using Pearson or Spearman correlation coefficient according to statistical distribution. Then, the multivariable analysis was carried out using generalized linear model (logistic for death as dichotomous endpoint) to take into account possible confounding factors. Hence, no specific statistical strategy approach, such as stepwise, was conducted. The covariates were chosen with caution due to sample size according to the univariate results and to clinical relevance (
). The testing and parameter estimation performed using a statistical model clearly depends on the variables included in the model. It is therefore crucial for confounding adjustment that known clinically significant variables are included in the regression model. A clinically significant variable may well be an important confounder also when it is statistically insignificant (
). Multivariable regression analyses were run with each parameter and then added together in multivariable model: age, gender, BMI, opioid drugs, SAPSII and SOFA score (
). Furthermore, particular attention was paid to possible multicollinearity using Farrar-Glauber test. The statistical analyses were performed using Stata software version 15 (StataCorp®, College Station, US). Statistical tests were two-sided with the type-I error set at 5 %. As analyses were exploratory, the individual p-values have been reported without applying systematically mathematical correction but with a specific attention paid on the magnitude of differences and to the clinical relevance (
3.1 Characteristics of study subjects and pupillary parameters on the day of admission (D1)
From April to august 2020, 50 patients were enrolled out of 65 patients admitted in ICU for a SARS-CoV-2 related pneumonia. The flow-chart of the study is presented on Fig. 1.
As one patient withdrew his consent, 49 patients were analyzed (Table 1). Patients were mostly male, hypertensive, overweighted and admitted in ICU after a median one week after COVID-19 onset. At ICU admission, patients had a moderate hypoxemia secondary to a multilobe pneumonia diagnosed on CT scan in 92 % of them. Forty-five (92 %) patients were treated with dexamethasone. Eighteen (37 %) and 28 (57 %) patients received opioids and midazolam, respectively. Thirty-five (71 %) received mechanical ventilation, among which 28 (57 %) were intubated for IV. Pupillary variables were assessed within the 12 h from ICU admission and presented in Table 1.
Table 1Demographic and clinical characteristics of included patients and values of pupillary parameters.
Characteristics
Number of patients N = 49 (100 %)
Male sex, n (%)
44 (90 %)
Age years, median (IQR)
66 (60–70)
BMI kg/m2, median (IQR)
27 (24–30)
Delay between the onset of symptoms and ICU admission, median (IQR)
8 (7–10)
Smoking history, n (%)
3 (6 %)
Overweight (25 < BMI < 30), n (%)
21 (43 %)
Obesity (BMI > 30), n (%)
14 (29 %)
Type 2 diabetes mellitus, n (%)
13 (27 %)
Hypertension, n (%)
29 (59 %)
Ischemic heart disease, n (%)
5 (10 %)
Chronic kidney disease, n (%)
2 (4 %)
Liver failure, n (%)
2 (4 %)
Multilobe involvement in CT-scan, n (%)
45 (92 %)
SOFA score
4 (3–6)
SAPSII score
40 (29–47)
Admission Glasgow Coma Scale (GCS)
14 (13–15)
Glasgow Coma Scale at time of pupillometry
14 (12–15)
D-Dimers, median (IGR)
1918 (991–2675)
Corticosteroids (dexamethasone), n (%)
45 (92 %)
Opioid drugs, n (%)
18 (37 %)
Vasopressors, n (%)
32 (65 %)
Epinephrine, n (%)
0 (0 %)
Midazolam, n (%)
28 (57 %)
Pa02/Fi02 ratio calculated the day of admission in ICU, median (IQR)
173 (104–154)
No Need for mechanical ventilation, n (%)
14 (29 %)
Need for ventilation support (invasive and/or non-invasive), n (%)
Regarding the use of ventilatory support, some patients had, throughout the hospital-stay, both mechanical ventilation techniques successively (non-invasive then invasive).
35 (71 %)
- Mechanical invasive ventilation, n (%)
28 (57 %)
- Mechanical invasive ventilation before enrollment, n (%)
4 (8 %)
- Non-invasive ventilation, n (%)
12 (24 %)
Nasal High Flow Oxygen therapy, n (%)
9 (18 %)
Non-Invasive ventilation with facial mask, n (%)
4 (8 %)
Days on mechanical ventilation (invasive and/or non-invasive), median (IQR)
22 (13–41)
Pulmonary embolism, n (%)
7 (14 %)
Septic shock, n (%)
3 (6 %)
Death during the stay in ICU, n (%)
7 (14 %)
Pupillary parameters measured the day of ICU admission
Mean ± SD
PD (mm)
4.1 ± 1.2
PDV (%)
33.8 ± 7.6
PCV (mm.s−1)
3.7 ± 1.0
Latency (ms)
273.1 ± 36.6
BMI: Body Mass Index; Multilobe involvement: involvement of >2 pulmonary lobes on the CT-scan realized the day of ICU admission; GCS: Glasgow Coma Scale; ICU: Intensive Care Unit; IV: Invasive Ventilation; NIV: Non-Invasive Ventilation (Nasal High Flow Oxygenotherapy or non-invasive ventilation).
a Regarding the use of ventilatory support, some patients had, throughout the hospital-stay, both mechanical ventilation techniques successively (non-invasive then invasive).
Seven (14 %) patients died in hospital, all during their stay in ICU. Univariate analysis revealed that, in comparison to patients who will survive (Table 2), the SAPS-II score (51 [44; 59] vs 38 [29; 45], p = 0.03), SOFA score (11 [6; 11] vs 4 [2; 6], p = 0.002) and GCS score at time of pupillometry (14 [9; 15] vs 4 [3; 11], p = 0.012) were significantly greater in patients who died. The PD (2.6 mm [2.4; 4.0] vs 4.1 mm [3.5; 4.8], p = 0.009), PDV (33 % [27; 39] vs 25 % [15; 36], P = 0.03) and PCV (3.5 mm.s−1 [2.8; 4.4] vs 2.0 mm.s−1 [1.9; 3.8], P = 0.02) were significantly lower in dead patients in univariate analysis. The PD latency (298 ms [237; 322] vs 274 ms [250; 301], p = 0.68) was not significantly associated with in-hospital mortality.
Table 2Relationships between severity scores, pupillary variables and in-hospital mortality and between need for intubation for mechanical ventilation and pupillary parameters measured at the day of admission (D1) (univariate analysis).
Parameters
Alive n = 42 (100 %)
Death n = 7 (100 %)
P
Gender (male)
37 (88 %)
7 (100 %)
1.00
Age
66 [58; 70]
70 [65; 78]
0.05
BMI
27.5 [24.2; 30.5]
26.4 [23.9; 28.0]
0.61
Diabetes mellitus
12 (29 %)
1 (14 %)
0.66
SOFA
4 [2; 6]
11 [6; 11]
0.002*
SAPS-II
38 [29; 45]
51 [44; 59]
0.03*
GCS at ICU admission
14 [14; 15]
14 [6; 15]
0.26
GCS at time of pupillometry
14 [9; 15]
4 [3; 11]
0.012*
Opioids (y/n)
13 (31 %)
5 (71 %)
0.08
PD (mm)
4.1 [3.5; 4.8]
2.6 [2.4; 4.0]
0.009*
PDV (%)
33 [27; 39]
25 [15; 36]
0.03*
PCV (mm.s−1)
3.5 [2.8; 4.4]
2.0 [1.9; 3.8]
0.02*
Latency (ms)
274 [250; 301]
298 [237; 322]
0.68
Pupillary parameters
Need for intubation n = 24
No need for intubation n = 24
P
PD (mm)
3.5 [2.8; 4.4]
4.2 [3.9; 5.2]
0.03*
PDV (%)
27.5 [24.5; 36]
35.0 [31.5; 39.5]
0.049*
PCV (mm.s−1)
3.1 [2.4; 4.0]
3.9 [3.0; 4.3]
0.17
The asterisk represents the values of p < 0.05 and therefore a statistically significant difference between the 2 groups (p-values < 0.05).
Data are expressed in mean ± SD of the different pupillary parameters unless otherwise indicated. P-value for multivariate analysis. GCS: Glasgow Coma Scale; PD: Pupillary diameter, PDV: Pupillary Diameter Variation, PCV: Pupillary Constriction Velocity.
We then constructed the ROC curves corresponding to each parameter (Fig. 2 panel A, B, C and D and Table 3). The baseline PD has the highest AUC, sensitivity, and specificity (Table 3 and Fig. 2). A PDV <26.5 % and PCV <2.5 mm.s−1 were statistically associated with in-hospital mortality but a PD threshold value of 2.75 mm was found to be the most discriminative and was reported in 5/7 (71 %) who will die and 3/42 (5 %) who will survive.
Fig. 2Receiver operating characteristic (ROC) curve for pupillary parameters recorded on the day of ICU admission and in-hospital mortality: panel A for PD, panel B for PDV, panel C for PCV and panel D for latency, respectively. PD: Pupillary Diameter; PDV: Pupillary Diameter Variations; PCV: Pupillary constriction Velocity; AUC: air under the curve.
We performed a multivariable analysis presented on the Fig. 3 (Supplemental Digit Content: Forest-plot representation). We showed that a PD < 2.75 mm remained associated with in-hospital mortality after adjustment to the SAPS-II and SOFA scores, use of opioids or vasopressors, BMI, age and diabetes.
Fig. 3Forest-plot representation of the association between pupillary diameter (PD), treated as a continuous variable, and in-hospital mortality, adjusted on demographic characteristics, opioids, SAPS-II score, SOFA score and GCS score (univariate analysis). Subtype-specific odds-ratios (95 % CI) are denoted by black boxes (black lines). GCS: Glasgow Coma Scale; GCSa: GCS at ICU admission; GCSp: GCS at time of pupillometry.
3.3 Pupillary parameters and prediction of need for mechanical ventilation
Twenty-four (48 %) patients required invasive mechanical ventilation after their ICU-admission. Baseline PD (3.5 mm [2.8; 4.4] vs 4.2 mm [3.9; 5.2], P = 0.03) and PDV (27.5 % [24.5; 36] vs 35.0 % [31.5; 39.5], P = 0.049) were significantly lower in patients who will be subsequently intubated than in those who will not during their ICU stay (Table 3).
4. Discussion
Our study shows that assessment of pupillary diameter was useful for predicting mortality and need for mechanical ventilation in COVID-19 patients admitted in ICU, independently of critical illness severity score (i.e. SAPS-II and SOFA), use of opioids or sedation. We found that a decrease in baseline PD was the most associated with bad outcome, indicating an impairment of sympathetic/parasympathetic control balance. A summary of these observations is represented on the Supplemental Fig. 1.
The discrimination value and reliability of the clinical signs and the scores for assessing dysautonomia are limited in critically ill patient are limited. In contrast, pupillometry is simple and validated methods for investigating the ANS dysfunction of the ANS in ICU patients (
). The iris muscle is indeed innervated by the sympathetic and parasympathetic systems, which induce pupillary dilatation and contraction, respectively. The decreases in baseline PD and in pupillary reactivity to light (i.e. in PDV and Pupillary constriction velocity PCV) rather support a sympathetic impairment. The latter could be then a component of a broader central autonomic dysfunction (
). The main mechanisms would be a viral invasion or a non-viral neuro-inflammation of the brainstem autonomic centers. In addition to dysautonomia, there are neurological and neurophysiological arguments for a brainstem dysfunction (
). More recently, Karahan et al. showed the interest of pupillometry as a non-invasive tool to assess the ANS and for the long-term follow-up of patients after SARS-CoV-2 infection (
). An impaired blink reflex, involving ponto-medullary circuits and therefore the brainstem, seems to be associated with respiratory failure in COVID-19 patients (
Quantitative assessment of pupillary light reflex for early prediction of outcomes after out-of-hospital cardiac arrest: a multicentre prospective observational study.
Quantitative versus standard pupillary light reflex for early prognostication in comatose cardiac arrest patients: an international prospective multicenter double-blinded study.
Quantitative assessment of pupillary light reflex for early prediction of outcomes after out-of-hospital cardiac arrest: a multicentre prospective observational study.
). If most of the studies reports an improvement of prognostic performance using quantitative PLR or Neurological Pupil index (NPi) in comparison with qualitative assessment of PLR, the characteristics of each pupillary parameter or the relationship between parameters and in-hospital outcomes were never studied in those patients. Our study highlights the interest of simple measurements performed in a few seconds at the bedside to predict mortality of patients underwent a SARS-CoV-2 infection.
Our prospective observational study has several limitations. First, its small sample size has limited the identification of potential confounders. Secondly, the general ICU care and specific treatment of COVID-19 patients since the study period have dramatically changed. In addition, our study does not compare the parameters studied in COVID-19 patients with a control group of uninfected patients. Currently, only the team of Vrettou et al. studied ANS dysfunction by comparing pupillary parameters of patients with SARS-CoV-2 infection with a control group (
). They did not find any significant difference between the 2 groups. However, these parameters were collected from patients already intubated for >48 h and their association with morbidity and mortality was not part of the objectives. Finally, we are not able to confirm a structural insult of the brainstem, as neither a brain MRI nor an autopsy have never been performed in our patients.
For instance, criteria for invasive mechanical ventilation are established nowadays while they were debated at time of our study. These limits prompt us not to draw a definitive conclusion on prognosis value of pupillometry in COVID-19 patients. Furthermore, the low values of PD and PDV and the presence of overlapping confidence intervals for these 2 parameters in our study also limit, for the time being, the use of such measures to help in the decision of goals of care in these patients.
Finally, our study was not designed with long-term follow-up, making it impossible to collect pupillary parameters far from the acute phase. However, our results would be useful for designing a validation study that would require a standardization of care and well-defined outcomes. Second, ICU physicians were not blinded to the results of quantitative pupillary responses. However, the pupillometry has never been considered in care of COVID-19 patients, minimizing then the risk of fulfilling self-prophecy. Therefore, only a large cohort study would enable to confirm their predictive value but also to develop a new algorithm that could be tested against NPi.
5. Conclusion
In conclusion, our observational study suggests that a reduced baseline PD is associated with in-hospital mortality and with the need for invasive mechanical ventilation in COVID-19 patients admitted in ICU. A reduced PD is likely to result from a central autonomic impairment and support the existence of a brainstem dysfunction in severe COVID-19.
The following is the supplementary data related to this article.
Graphic representation summarizing the various pupillary parameters analyzed and their values found significantly associated with in hospital mortality (multivariate analysis). PD: Pupillary Diameter; PDV: Pupillary Diameter Variation; SOFA: Sepsis-related Organ Failure Assessment.
CRediT authorship contribution statement
DC and MD: study design, data acquisition, quality assessment, data interpretation, statistical analysis and manuscript drafting. BP: statistical analysis, manuscript drafting and quality assessment. MP: data acquisition and quality assessment. SM, TS: study design, data interpretation and manuscript drafting. All authors provided critical reviews of the manuscript and approved the final version. MD is the guarantor of the content of this manuscript.
Funding sources
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Data availability
Data will be made available on request.
Declaration of competing interest
None.
Acknowledgements
We would like to thank the medical and paramedical staff of the ICUs of the Saint Etienne University Hospital and of the GHU Paris Psychiatrie et Neurosciences for their help in this work and their involvement in the care provided to patients during the COVID-19 crisis.
References
Aghagoli G.
Gallo Marin B.
Katchur N.J.
Chaves-Sell F.
Asaad W.F.
Murphy S.A.
Neurological involvement in COVID-19 and potential mechanisms: a review.
Olfactory and gustatory dysfunctions as a clinical presentation of mild-to-moderate forms of the coronavirus disease (COVID-19): a multicenter european study.
Quantitative versus standard pupillary light reflex for early prognostication in comatose cardiac arrest patients: an international prospective multicenter double-blinded study.
Quantitative assessment of pupillary light reflex for early prediction of outcomes after out-of-hospital cardiac arrest: a multicentre prospective observational study.
Quantitative pupillometry and neuron-specific enolase independently predict return of spontaneous circulation following cardiogenic out-of-hospital cardiac arrest: a prospective pilot study.
00116-3&id=ga1.jpg){kind=link}
00116-3&id=gr1.jpg){kind=link}
00116-3&id=gr2.jpg){kind=link}
00116-3&id=gr3.jpg){kind=link}