Clinical Characteristics and Outcomes of Critically Ill COVID-19 Patients in Tokyo: A Single-Center Observational Study

Aya Banno (  ayabanno@luke.ac.jp ) St. Luke's International Hospital https://orcid.org/0000-0001-8279-0429 Toru Hifumi St. Luke's International Hospital Hiroshi Okamoto St. Luke's International Hospital Minori Masaki St. Luke's International Hospital Koichiro Seki St. Luke's International Hospital Shutaro Isokawa St. Luke's International Hospital Norio Otani St. Luke's International Hospital Kuniyoshi Hayashi St. Luke's International Hospital Shinichi Ishimatsu St. Luke's International Hospital

the low mortality rates remain unclear. Thus far, no studies have described the characteristics of the Japanese critically ill COVID-19 patient population who have undergone treatment at domestic hospitals.
We sought to analyze the clinical data of critically ill Japanese patients who were admitted to a general hospital in central Tokyo. [6] This study aimed to describe the clinical characteristics, laboratory trends, and outcomes among patients with COVID-19 in an intensive care unit (ICU) setting and to discuss the possible reasons for the low COVID-19 mortality rates in Japan.

Study Design and Subjects
This retrospective study enrolled patients with COVID-19 who were hospitalized in the ICU of St. Luke's International Hospital, Tokyo, Japan [6] between March 19 and April 30, 2020. Permission to collect data from electronic medical records were permitted by the hospital and the institutional review board approved this study (approval number 20-R057) and waived the requirement for informed consent in view of the retrospective study design and the urgent need for research insights into this growing pandemic.
The inclusion criteria for this study were ICU admission and laboratory con rmation of COVID-19.
Patients with 3 consecutive negative reverse-transcriptase polymerase chain reaction (RT-PCR) test results were excluded from the study. The indications for ICU admission of patients with COVID-19 were respiratory failure that required oxygen supplementation exceeding 5 L/min to maintain a percutaneous pulse saturation (SpO 2 ) of more than 94% or the treating physicians' expectation of progression to further respiratory distress. Patients were admitted to the ICU either directly from the emergency department or transferred from the inpatient wards. Patients designated with a do-not-resuscitate (DNR) order on admission were not admitted to the ICU. The diagnosis of COVID-19 was dependent on positive RT-PCR test results for the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) from samples, such as a nasal swab, pharyngeal swab, or sputum. [7] A suspected diagnosis of COVID-19 was based on the presence of bilateral ground-glass opacities on chest computed tomography (CT). [8] Treatments Treatment strategies for COVID-19 followed recommendations provided by the National Institutes of Health, Society of Critical Care Medicine, and Japanese Ministry of Health Labor and Welfare. [9,10] The options for antiviral agents in Japan were favipiravir and lopinavir/ritonavir combination drugs, both were used in clinical trials for treating SARS-CoV-2 infection. [11] Antibiotics and systemic steroids were administered at the discretion of the treating physician. Systemic steroids were administered as either high-dose (pulse therapy of 500-1,000 mg methylprednisolone per day for 3 consecutive days) or lowdose (1-2 mg/kg of methylprednisolone for 5-7 days) therapy. Prophylaxis for venous thromboembolism was provided with subcutaneous injections of unfractionated heparin. Supplemental oxygen delivery was maintained by either non-rebreathing mask with a reservoir bag or invasive mechanical ventilation via an endotracheal tube or tracheostomy. An early intubation plan was encouraged. A nasal cannula was worn covered with a surgical mask only for patients who had stabilized after intensive care treatment. [12] Data Collection of Variables We collected information of the baseline characteristics, comorbidities, symptoms, vital signs upon admission, laboratory tests, illness severity during ICU stay, mortality prediction scores (Sequential Organ Failure Assessment [SOFA] score, Acute Physiology and Chronic Health Evaluation [APACHE] II score, Simpli ed Acute Physiology Score [SAPS II]), ICU therapies (mechanical ventilation, use of vasopressors, renal replacement therapy, enteral feeding), pharmaceutical treatments, complications (arrhythmias, mediastinal emphysema, thromboembolisms), and outcomes. All of the study participants were followed up until May 26, 2020.

De nitions
The rst day of the illness (Day 1) was considered to be the day of symptom onset as reported by the patient. All event dates are presented as days of illness throughout the article. The symptoms at presentation were based on the patients' complaints and history. Speci cally, fever was recorded if the patient claimed to have experienced fever, and this was not based on the initial body temperature measurements. Arrhythmias included new-onset atrial brillation, paroxysmal supraventricular tachycardia, and non-sustained ventricular tachycardia. Mediastinal emphysema was diagnosed on chest radiography or chest CT scanning. Thromboembolisms were investigated through ultrasonography and CT scanning.

Study Endpoints
The primary endpoint was death during hospital admission. Secondary endpoints included days of mechanical ventilation, days of oxygen supplementation, days of ICU or hospital stay, and complications.

Statistical Analysis
Continuous variables are presented as median and interquartile ranges. Categorical variables are reported as the number and percentages. We used the locally weighted scatterplot smoothing (LOWESS) curves to visualize the associations of laboratory ndings including white blood cell count (WBC), neutrophil-tolymphocyte ratio (NLR), C-reactive protein (CRP), and D-dimer with the days of illness. Statistical analysis was conducted with R: A Language and Environment for Statistical Computing, version 4.0.0. (R Foundation for Statistical Computing, Vienna, Austria). The missing data were not imputed or replaced.

Results
In total, 38 suspected cases of COVID-19, based on patient history and the ndings of chest CT imaging, were admitted to the ICU. After con rmation by the results of negative RT-PCR tests, 14 patients were excluded. Thus, 24 patients were included in the nal analysis dataset.

Baseline Characteristics
The median age was 57.5 years, and 19 (79%) patients were male. All patients were of Japanese ethnicity. Six patients worked in customer service, 3 were managers, 2 were attorneys, and 1 patient was a health care provider. Baseline characteristics, symptoms, vital signs, and laboratory ndings upon admission are listed in Table 1. The majority of patients (54%) were normal weight individuals, based on the body mass index (BMI) calculations from the weight and height measurements obtained on admission. Common symptoms upon admission were fever (96%) and dyspnea (71%). Diarrhea was observed in 25% of patients. Data are expressed as numbers (percentage) or median (interquartile range).
*Rheumatoid arthritis, hyperthyroidism, ulcerative colitis, and a past history of pulmonary embolism were present in one patient each, respectively. †Body temperature measurements were all axillary. ‡Includes measurements with room air inhalation. Initial SpO 2 measurements under the in uence of oxygen inhalation were excluded.
A high NLR was observed, with a median of 10.1. The median platelet count was decreased to 17.4 × 10 4 /µL, and 6 patients presented with a platelet count below 15.0 × 10 4 /µL. The elevation of the levels of lactic acid dehydrogenase (LDH), hepatic enzymes, creatinine kinase (CK), CRP, and D-dimer was common. Four patients presented with CK elevation over 1,000 U/L during hospitalization.

Events and Days of Illness
The dates of hospital admission ranged from March 11 to April 19, 2020. The median day of illness from the beginning of symptoms to hospital admission was Day 10. Thirteen (54%) were admitted directly to the ICU. In the 11 patients who were admitted to the wards before the ICU admission, the median duration from hospitalization to ICU admission was 6 days. The median day of illness on ICU admission for all patients was Day 11.

Laboratory Findings
The associations of WBC, CRP, NLR, and D-dimer measurements with days of illness for all patients are shown in Fig. 1. WBC, NLR, and CRP followed a similar trend and peaked on Day 10 of illness. LDH was elevated in the initial phase of illness and subsequently decreased. D-dimer was more likely to increase after Day 20 of illness.

ICU Admission and Therapy
Severity and mortality prediction scores, therapies, and pharmaceutical treatments in the ICU are described in Table 2  Data are expressed as numbers (percentage), or median (interquartile range).
* Measurements and treatments in mechanically ventilated patients are displayed.
Continuous injections of vasopressors were administered in 13 patients who underwent mechanical ventilation and 2 who did not. Two patients without previous kidney disease received renal replacement therapy for uremia and hyperkalemia, respectively.

Outcomes
In this study population of 24 patients, 2 (8%) died. One patient was designated with a DNR order after ICU admission and died within 24 hours of entry. In total, 18 patients were con rmed to have been discharged and had returned home. Four patients remain hospitalized. The median lengths of ICU and hospital stay were 6 and 22 days, respectively. With regard to patients who received invasive mechanical ventilation, the median duration of ventilator dependence was 7 days. Thirteen patients were successfully extubated. Among the 4 patients who underwent tracheostomy, 3 were successfully weaned off from the mechanical ventilator, and 1 patient died. Figure 2 illustrates the clinical course of patients who received invasive mechanical ventilation. The association between hospitalization or mechanical ventilation duration and days of illness is displayed in the chronological order of hospitalization based on the patients' admission dates. The distribution of the duration of mechanical ventilation is shown in Fig. 3. Supplemental oxygen was administered in all patients. The median duration of oxygen supplementation was 13 days. The incidence of complications, including arrhythmia, mediastinal emphysema, and thromboembolisms, is listed in Table 3.  (17) ICU: intensive care unit. Data are expressed as numbers (percentage), or median (interquartile range).
* Variables in patients receiving invasive mechanical ventilation. †Includes four new onset arterial brillation, two paroxysmal supraventricular tachycardia, and one non-sustained ventricular tachycardia. ‡Mediastinal emphysema was diagnosed prior to endotracheal intubation in one patient and during mechanical ventilation in two patients with radiological ndings. §Deep vein thromboembolisms were evident in three patients with ultrasound investigations. One other patient developed multifocal cerebral embolisms.
Locally weighted scatterplot smoothing (LOWESS) curves showing the associations of WBC, CRP, NLR, and D-dimer with days of illness in critically ill COVID-19 patients. The shaded area represents the 95% con dence intervals. Each patients' laboratory tests were obtained on the day of admission but are plotted according to the day of illness.

Discussion
This study evaluated the mortality rate and clinical outcomes in 24 critically ill Japanese patients with COVID-19 who were admitted to the ICU between late March and mid-April 2020. The majority of patients were male, most were normal weight, and two-thirds were former or current smokers. Mortality rates in the current study were 8% for overall ICU admissions and 6% for mechanically ventilated patients.
The highlight of this study was its focus on a population in East Asia besides China, which was the initial epicenter of COVID-19. East Asia is reported to have a lower COVID-19 mortality than that in the American or the European countries.
[13] As of July 1, 2020 we were unable to nd published reports of critically ill patients in East Asia, and the reasons underpinning these regional differences are unclear. The mortality rate was low in our study, including survivors who underwent prolonged mechanical ventilation for more than 3 weeks. The clinical course of patients who were intubated was described based on days of illness in this study. Furthermore, we reported the trends of laboratory ndings such as NLR, CRP, and D-dimer throughout the disease course. The use of the antiviral agents of favipiravir and lopinavir/ritonavir and the high frequency of systemic steroid administration may have been practices that are unique to Japan and requires further assessment and documentation.
In this study, 17 of the 24 patients received mechanical ventilation for less than 2 weeks, with the exception of 4 patients that required tracheostomy and prolonged ventilation. The intubation rate of 71% in the ICU setting was lower than the 88% reported in a multicenter study in Italy [4] and 93% reported in a large single institutional study in New York, [14] but was similar to the rates of 71 to 79% reported in multiple multicenter studies in the United States [15][16][17] and was higher than the rates of 30 to 47% reported during the early phase of the endemic in Wuhan, China. [18][19][20] The median PaO 2 /F I O 2 ratio of the intubated patients was 169, which was close to the value of 160 in the Italian study [4] although higher than that reported in most American studies. [15,17] The median duration of mechanical ventilation was 7 days, and the extubation rate was 77%. The 8% mortality rate in the current study was signi cantly lower than the rates of 26 to 78% reported in other affected cities or regions [4, 14-17, 19, 21, 22].
An early intubation strategy was enforced, and the decision to adopt invasive mechanical ventilation was made promptly. Thresholds for intubation were lowered to avoid the use of high-ow nasal-cannula or non-invasive positive pressure ventilation therapy, which carry the risk of aerosol contamination that may increase the SARS-CoV-2 transmission risk for health care providers. [12] Patients who were at risk were closely monitored and transferred to the ICU without delay. This practice may explain the relatively low SOFA, APACHE-, and SAPS-scores on ICU admission. Unlike in the Lombardy Region or New York, [4,15] Tokyo did not experience a severe outbreak of COVID-19. Therefore, the ICU capacity and the need for ventilators were not overwhelmed. These factors may have permitted early intervention before the pointof-no-return deterioration or death.
The median BMI of 24.7 kg/m 2 in our study was signi cantly lower than the mean of 30 kg/m 2 reported in New York [14,15] although it was similar to that of Chinese patients [23]. According to previous studies, obesity has been associated with worse outcomes. [24,25] The lower BMI of our study subjects may have been protective. Further investigations are required to understand the in uence of body weight on outcomes in Japanese patients.
Elucidating the underlying in ammatory mechanism of COVID-19 may aid clinicians' comprehension of the disease and clarify the deterioration in clinical course on Day 10 of illness as well as trends of laboratory ndings. [7,26] In our study, the median day of illness on ICU admission and intubation was Day 11. Laboratory ndings such as WBC, NLR, and CRP consistently peaked at this phase. NLR and CRP are well-known indicators of in ammatory damage and have been associated with worse outcomes. [19,27,28] However, changes in laboratory ndings throughout the clinical course of the disease have not been delineated. Combined assessment of the clinical timeframe and laboratory trends may be bene cial for predicting patients' course of COVID-19. Furthermore, coagulation abnormalities are predominant features of COVID-19, and high D-dimer levels are reported on initial laboratory testing. [29] In this study, D-dimer peaked on Day 20 of illness. Recognition of these changes guided the institutional clinical practice for active surveillance for the detection of thromboembolism and early administration of prophylaxis. [30] Corticosteroid administration in critically ill patients with COVID-19 remains controversial. [31,32] In previous studies, the use of systemic steroids in ICU patients varied from 26-50%.[14, 15, 19, 21] Low dose steroid treatment in the range of 1 to 2 mg/kg were frequently administered. [31,33] In this study, 88% of patients used systemic steroids, including 16 survivors that received high dosage (500 to 1,000 mg) of methylprednisolone. Given the underlying in ammatory mechanisms of COVID-19 [34,35] and the trend of in ammatory laboratory ndings, systemic steroids may play a role in the suppression of in ammatory responses when used judiciously. [36] Further studies are necessary to evaluate the e cacy of systemic steroid therapy in COVID-19.
Favipiravir, a u drug developed in Japan [37] was the main antiviral agent predominantly used during the observation period, after April 1, 2020. Moreover, 9 of 10 patients were survivors, and no adverse events were observed with use in this study population. Notably, patients in the wards were not frequently transferred to the ICU in April 2020 despite the surge of infections in Tokyo. [1] The association between favipiravir use and reduction in inpatient deteriorations remains speculative, and the results of clinical trials are anticipated. [38] Limitations The institutional review board approved this study (approval number 20-R057) and waived the requirement for informed consent in view of the retrospective study design and the urgent need for research insights into this growing pandemic.

Consent for publication
Not applicable.

Availability of data and materials
The datasets generated and analyzed during the current study are not publicly available due to them containing information that could compromise individual privacy but are available from the corresponding author on reasonable request.  The median days of illness on hospital admission, intubation, and extubation/discontinuation of mechanical ventilation were 10, 11, and 17 days, respectively, and are displayed in the box and whisker plot in yellow, navy and blue. *Still hospitalized, †Discontinuation of mechanical ventilation on Day 51 and still hospitalized, ‡Discharged on Day 47, §Discharged on Day 51, ||Still hospitalized.