Initial general characteristics
From March 21st 2020, to December 22nd 2021, 769 patients admitted in ICU for ARDS were screened, of which 135 were included in this cohort, 100 with COVID-19-related ARDS and 35 with non-COVID-10-related ARDS (Figure 1). Patients baseline characteristics, blood and urinary tests performed at inclusion are summarized in Table 1. Patients were mostly male (67%). The mean age was 62 years. The most common comorbidities were hypertension (44%), obesity (39%), active smoking (36%) and diabetes (28%). In non-COVID-19-related ARDS the main causes of ARDS were bacterial pneumonia (40%) and extra-pulmonary ARDS related to septic shock (37%) and inhalation pneumonia (20%).In the COVID-19 group, patients were more frequently obese (46% vs. 7%, p=0.007) whereas patients with non-COVID-19-related ARDS presented significantly more active smoking, chronic lung disease, and active neoplasia (respectively 57% vs. 28% p=0.002; 31% vs 13% p=0.014; 14% vs 1% p=0.045, respectively). Only 4 patients had chronic kidney disease, and only two patients had received a nephrotoxic medication (non-steroidal anti-inflammatory drugs).
Eighteen patients (13%) were treated with angiotensin converting enzyme inhibitors (ACEI) and twenty-four (13%) with angiotensin receptor blockers (ARB). The mean time between onset of symptoms and ICU admission was significantly longer in the COVID-19 group (7.7 ± 4.3 vs 3.3 ± 3.9, p<0.0001). The mean time between the onset of symptoms and the MV was also significantly longer in COVID-19 cohort (9.9 ± 5.3 vs 4.2 ± 4.0, p<0.0001). At inclusion, mean SOFA score was 6.0 ± 2.9 and mean SAPS-II score was 36.6 ± 11.0, SAPS-II score tended to be higher in the non-COVID-19 group: 39.9 ± 12.5 vs 35.5 ± 10.3 p=0.078 and non-COVID-19 patients received more frequently norepinephrine at the time of inclusion (48.6% vs. 25.0%, p=0.009). The mean initial PaO2/FiO2 ratio was not significantly different between the two groups: 98.0 ± 40.3 in the COVID-19 cohort vs. 119.1 ± 61.4 (p=0.604), but ARDS were more frequently severe in the COVID-19 group (64.6% vs. 45.7%, p=0.05). There were no significant differences between the two cohorts on MV settings except for the positive end expiratory pressure (PEEP) used which was higher in the COVID-19 group (12.7 ± 2.5 vs. 9.7 ± 3.4 cmH20, p<0.0001). Initial tidal volume in overall cohort was 5.47 ± 1.30 mL/kg and static compliance was 29.4 ± 9.5 mL/cmH20. Regarding the specific treatments prescribed for COVID-19, more frequent were Dexamethasone (66%) and Azithromycin (36%). Twenty-one patients received Hydroxychloroquine, 9 Anakinra, 7 Ruxolitinib and 1 Tocilizumab (Table 1).
Initial biological characteristics
No difference was found in routine biochemical serum tests between the two groups. The mean creatinine serum level was 80.5 ± 65.6 and 30 patients (22%) displayed AKI among whom 18 stage 1 AKI. AKI at inclusion was more frequent in non-COVID-19 group: 34% vs. 18%, p=0.046 displayed AKI (Table 1). Glycosuria was detected in 94 patients (81%) and was more frequent in COVID-19 patients (83% vs. 54% p=0.0009) while mean serum glucose level was 8.8 ± 3.4 mmol/L in the overall cohort. Hematuria was present in 84.0% of COVID-19 patients and 73.5% on non-COVID patients, p=0.176. Mean UPCR and UACR were comparable between the two groups (0.765 ± 1.36 vs. 0.58 ± 0.46 g/g, p=0.292 and 187 ± 571 vs. 149 ± 233 mg/g p=0.595).
Urinary protein electrophoresis was performed in 108 patients and showed a profile of tubular dysfunction in 58 patients (54%). This tubular dysfunction was three times more frequent in COVID-19 patients (65% vs. 21% p<0.0001) than in non-COVID-19 patients, and non-COVID-19 patients displayed more frequently a normal profile (43% vs. 11% p=0.003). Only 4 patients displayed a glomerular profile (3 COVID-19 patients and 1 non-COVID-19 patient) (Figure 2A).
In the overall population, analysis of urinary proteins showed an important leakage of alpha-1-microglobulin (α1m 108.3 ± 84.5 mg/L), Beta-2-microglobulin (ß2m 40.8 ± 52.3 mg/L), FLLC and FKLC (respectively 83.3 ± 67.1 and 464.0 ± 429.6 mg/L). This low molecular weight proteins leakage was higher in COVID-19 patients for ß2m, FLLC and FKLC (respectively 47.6 ± 53.7 vs. 22.5 ± 44.4 mg/L p=0.022; 95.5 ± 65.0 vs. 51.1 ± 62.4, p=0.003 and 554.1 ± 435.1 vs. 225.7 ± 312.1 mg/L, p<0.0001) (Table 1, Figure 2B).
Treatments and outcomes
Outcomes and patients’ evolution are provided in Table 2. Median follow-up was 48 days. In patients with COVID-19 ARDS, after a median ICU length of stay of 26 days, 25% died in the ICU, and 51% had experienced AKI (17% stage 3), requiring RRT in 8 patients.
Compared to non-COVID ARDS, COVID-19 patients tended to experience more AKI (51% vs 34% p=0.088). There was no difference in RRT requirement. COVID-19 patients required more frequently specific ARDS treatments like methylprednisolone, inhaled nitric oxide (iNO) and almitrine (respectively 28% vs 3% p=0.001; 37% vs 20% p=0.06; 16% vs. 0% p=0.012; with longer iNO requiring (7 vs. 3 days p=0.001), suggesting a more severe or longer severe respiratory condition during the ICU stay. COVID-19 patients also had longer invasive ventilation (median 20 vs. 9 days p<0.0001) and longer ICU stay (26 vs. 15 days, p<0.0001) but similar mortality at 28 days (14% vs. 11%, p=0.999) and at ICU discharge (25% vs. 17% p=0.341). ICU-free days at D21 and D28 were lower in COVID-19 patients (p=0.016 and p=0.005 respectively) (Table 2).
Predictive value of proximal tubular injury on main outcomes
In COVID-19 ARDS, AKI KDIGO ≥ 2 was associated with D28 mortality (30.0% vs. 7.1%, p=0.003), leakage of FLLC (p=0.042) and iNO exposure (p=0.008) (Table 3, Figure 2C). In multivariate analysis, urinary FLLC (OR: 1.014, 95%CI [1.003-1.025], p=0.011) and iNO exposure (OR: 3.88 95%CI [1.127-13.355], p=0.032) remained the only predictive factors associated with AKI. In the overall ARDS population, KDIGO ≥2 AKI was associated with D28 mortality (27.0% vs. 82%, p=0.004) and ICU-free days at D28 (3.7±6.9 vs. 7.5±7.6, p=0.008) (Table 3). In multivariate analysis, AKI was independently associated with higher initial SOFA score, FLLC and ICU-free days.
In COVID-19 ARDS, patients deceased at D28 had more frequently experienced AKI (85.7% vs. 45.4%, p=0.008), especially stage 3 (35.7% vs. 14.0%, p=0.04), and had been more frequently exposed to almitrine and iNO (p<0.0001 and p=0.0002, respectively) than those who survived (Table 4). AKI was associated with decreased overall survival in patients with COVID-19-related ARDS (p=0.029, Figure 3A). In the overall ARDS population, in multivariate analysis, higher age, active neoplasia, and iNO exposure were independently associated with mortality. AKI was also associated with decreased overall survival (p=0.002, Figure 3B).