Materials and methods

The present is a single center, retrospective cohort study in cluding adult patients with severe COVID-19 admitted to ICU from March 17th to September 13th of 2020, at the Hospital Italiano de Buenos Aires, a tertiary teaching hospital located in Buenos Aires, Argentina. Patients were eligible if they were 18 years or older, had a confirmed COVID-19 diagnosis by means of real-time-polymerase-chain-reaction (Bosphore 2019 Novel Coronavirus Detection Kit v2 from Anatolia GeneWorks, Turkey) testing from throat swab or tracheal aspirate samples, and were admitted to the ICU during hospital stay. Patients with a history of end stage kidney disease requiring renal replacement therapy (RRT) and renal transplant recipients were excluded. Patients included in the presented study were followed from COVID-19 diagnosis until hospital discharge or death.

This study was approved by the local ethics review board (protocol number 1636) and was conducted according to the amended Declaration of Helsinki. Informed consent was wai ved because of the observational nature of the study.

Demographic, history of comorbidities, severity of COVID-19 illness, scoring systems used to predict mortality (APACHE II, SOFA and Charlson score) were systematically calculated at ICU admission and laboratory data from electronic medical charts were retrieved. The comorbidities considered were diabetes mellitus, hypertension, obesity, chronic obstructive pulmonary disease (COPD), heart failure (HF), chronic kidney disease (CKD), stroke, asthma, tobacco use and active malig nancies. Laboratory assessment consisted of a complete blood count, coagulation testing, assessment of renal function, and measure of electrolytes, C-reactive protein (CRP), procalcitonin, pro b-type natriuretic peptide (pro-BNP), lactate dehydrogenase, high-sensitivity cardiac troponin, d-dimer and ferritin.

Data regarding invasive mechanical ventilation (IMV) requi rement and ventilation parameters on the first day (i.e. PEEP, PaO2/FiO2) were also captured. PaO2/FiO2 was calculated as the relation between the first partial pressure of oxygen measured after intubation and the fraction of inspired oxygen set on the ventilator. Data concerning prescribed drugs during hospital stay was extracted from medical records. Specifically, we documented the prescription of nephrotoxic drugs such as polymyxins and aminoglycosides during hospital stay. In addition, drugs from the ambulatory setting were also recorded [e.g. immunosuppressants, corticosteroids, ACE inhibitors (ACEI) and angiotensin receptor blockers (ARB)]. Finally, the length of hospital stay, length of period under IMV, days from symptom onset until ICU admission and IMV were calculated for each patient. A nephrologist and a critical care specialist (CFV and NCR) reviewed electronic medical charts in order to assess the cause of AKI.

AKI was defined following the 2012 Kidney Disease Im proving Global Outcomes (KDIGO) definitions⁸. Baseline creatinine was defined as the last creatinine obtained on the ambulatory setting within the last year prior to hospital admission. Glomerular filtration rate was estimated (eGFR) using CKD-EPI⁹ formula.

Transient AKI was defined as an episode lasting 48 hours from symptom onset with sustained reversal for at least 48 hours; and persistent AKI as an episode lasting more than 48 hours¹⁰. Sepsis-associated acute kidney injury was defined as an episode occurring in the context of sepsis or septic shock following the sepsis-3 consensus^10,11. Renal replacement thera py (RRT) requirement during hospital stay was also recorded.

As an exploratory analysis, episodes of AKI were classified occurring within the first 14 days (early AKI) and after day 14 from symptom onset (late AKI), describing epidemiologic characteristics, identifying causes of AKI, severity of disease and RRT requirements for each group. We collected PEEP and PaO2/FiO2 values on the day of the episode. Norepine phrine dose was calculated as the average dose administered on the full day on which AKI occurred in micrograms/kg of body weight/minute. The main hypothesis was that early AKI would develop in the period of more severe hypoxemia¹² and late AKI would be more probably related to intensive care unit complications such as hospital-acquired infections and nephrotoxic drug exposure.

Data conforming to a normal distribution was presented as mean ± SD, and median and quartiles were used for non-normal distribution. Rate comparisons were performed by chi-squared test. Student’s, Wilcoxon rank-sum, Wilcoxon signed-rank, and Kruskal-Wallis tests were used to compare means across groups according to the number of groups and distribution of variables.

A bivariate and multivariate logistic regression analysis was performed to identify risk factors associated with AKI. Variables were entered into the model when the p-value was lower than 0.15 in the bivariate analysis (Supplementary appendix, table s1). The final model after a stepwise approach included age, a history of hypertension or HF and IMV requirement. PaO2/ FiO2 was not included because of the high correlation with IMV (only patients with this feature had PaO2/FiO2 values available)

STATA v.14.2 (StataCorp LLC, College Station, Tx) was used for analysis. All reported p values are two-sided and p value < 0.05 was used as threshold for statistical significance.

Results

A total of 171 patients with COVID-19 diagnosis admit ted to ICU were initially evaluated. Nine subjects were excluded because they were kidney transplant recipients or presented end stage kidney disease (Figure 1). Overall, 162 were included in the final analysis. Some patients of the cohort were already described by Carboni Bisso et al¹³ and Simonovich et al¹⁴. Baseline clinical, demographic and laboratory data is presented in Table 1. The cohort was composed of older adults (mean age 66 ± 13.2 years) and 109 (67.3 %) were male. It is worth mention ing that hypertension, diabetes mellitus and coronary artery disease were the most prevalent comorbidities and 88.3 % (n = 143) of patients presented with at least one comorbidity.

Fig. 1 Study population 

Table 1 Baseline characteristics of patients in the study 

Overall, 70/162 (43.2%) patients developed AKI dur ing hospital stay. Differences between groups were rec ognizable. Patients with AKI were older, more prone to present a history of hypertension, heart failure and ARB/ ACEI treatment. Baseline eGFR was lower for patients that subsequently developed AKI during hospital stay. In particular, non AKI patients had baseline KDIGO 1 or 2, whereas AKI patients had mostly KDIGO 2, 3a and 3b (Table 1). In addition, creatinine and urea values on ICU admission were significantly higher on patients with AKI with a median of 1.08 (IQR 0.85-1.47) vs. 0.76 (IQR 0.7-1.15) mg/dL, and 48 (IQR 37.0-74.0) vs. 37.0 (IQR 27.0-46.0) mg/ml, respectively. Regarding inflammatory parameters, no differences were found between groups on D-dimer, C-reactive protein, ferritin, lactate dehydro genase and procalcitonin levels. Troponin and pro-BNP levels were higher among patients with AKI. The KDIGO AKI stage during hospital stay was distributed as follows (Table 2): 26 (37.1 %) subjects in stage 1, 16 (23.2%) in stage 2, 28 (40.6%, second events are included) in stage 3 and 18 (25.7 %) developed AKI-RRT (11.1 % of the total ICU cohort).

Table 2 Severity of disease and outcomes 

In-hospital mortality was higher in the AKI group (58.6% vs. 13%, p <0.001). It varied widely according to the previous eGFR rate, occurring in 20.5, 32.5, 47.1, 75.0 and 50.0 % of patients in stages I, II, III, IV and V of CKD disease, respectively. In-hospital mortality in AKI patients who required RRT was 77.7% and 51.9 % in those who did not (p = 0.06).

Overall, individuals with AKI during hospital stay presented more severe disease, with higher APACHE II and SOFA scores at ICU admission. Additionally, they were more prone to undergo IMV during ICU stay. Regarding ventilatory parameters, no differences were found between groups on PaO2/FiO2 ratio and PEEP levels on day-1 of IMV. Higher neuromuscular blockade was required in the AKI group. Length of ICU and hospital stay was higher in AKI patients. Finally, it was observed a roughly four-fold increase in the incidence of death during hospital stay among patients developing AKI.

In a prespecified bivariate analysis, older age, history of hypertension and COPD, and ARB/ACEI use were as sociated with AKI development in patients with COVID-19 admitted to the ICU (Supplementary appendix, Table s1). Results of multivariate analysis are reported in the supplementary appendix (Table s2). Independent risk factors for AKI development included age, hypertension, IMV requirement during hospitalization and a history of heart failure.

Among the 70 patients developing AKI during ICU stay, 45 and 25 (64.28 and 35.71 %) patients developed early and late AKI as a first episode, respectively. No differences between groups were found in demographic characteristics (age, sex, BMI), comorbidities (Diabetes, hypertension, cerebrovascular disease, coronary artery disease, COPD, Asthma, tobacco use, a history of heart failure, chronic kidney disease and malignancies) and severity scores (APACHE, SOFA). Median time from symptom onset to first AKI diagnosis was 10 days (IQR 5.5 to 17.5). Late AKI episodes were more frequently persistent and had a higher KDIGO stage. They were also more frequently associated with sepsis and with nephrotoxic drug exposure (Table 3). Regarding vasopressor therapy and serum lactate, there were no differences between groups. Median time from IMV initiation to early AKI development was 1 (IQR: 0 - 2) day, while this time on late AKI patients was 11 (IQR 1 - 25) days (p <0.001). Higher PEEP levels were seen in this group on the day of AKI onset, too. Finally, RRT rates and mortality were similar for both groups.

Table 3 Clinical characteristics and outcomes of patients presenting with early and late acute kidney injury during hospitalization 

Discussion

In this retrospective cohort study conducted in a tertiary teaching hospital in Buenos Aires, a high cumulative incidence of AKI was found among patients with severe COVID-19. Further, patients developing AKI presented higher mortality. Comprehensive data regarding baseline characteristics of patient’s risk factors for AKI during hospital stay was gathered. Additionally, characteris tics of AKI were captured in order to shed light on the underlying physiopathology of kidney injury in patients with COVID-19. To our knowledge, this is the first report describing the timing and precipitating factors for AKI among severely ill patients with COVID-19 in Argentina.

According to previous studies involving adult patients with severe COVID-19, study subjects with AKI presented multiple comorbidities¹⁵, including a history of hyperten sion, higher Charlson score, CKD, COPD and a history of HF. It is worth mentioning that other critical care condi tions leading to AKI present a similar set of risk factors¹⁶. Regarding laboratory measures, no association between inflammatory parameters and the occurrence of AKI was found. Of note, troponin and pro-BNP levels on ICU admission were higher in AKI patients. This could be a manifestation of multiorgan involvement or right ventricular overload due to secondary pulmonary hypertension^17,18. However, reverse causation cannot be ruled out, since AKI often leads to fluid overload.

The incidence of AKI in our cohort is comparable to that reported elsewhere^3,19-22. Nevertheless, we found a low incidence of AKI-RRT compared with those reports. This disparity could be explained by differences in comorbidities and in the severity of the disease between the cohorts or by differences on RRT initiation criteria.

Noteworthy, in-hospital mortality was higher in patients presenting with AKI and significantly higher in those requir ing RRT. This may reflect a higher burden of comorbidities rather than a deleterious effect of RRT, specifically age, severity of disease marked by higher APACHE II and SOFA scores, and higher IMV requirement. In line with these findings, a retrospective cohort study of patients with COVID-19 published by Fisher et al.²³ compared COVID-19 associated AKI with a historical cohort. They reported that COVID-19 associated AKI had a higher rate of severe AKI and worsening clinical outcomes when compared to a historical cohort, i.e. RRT, IMV, ICU ad mission and in-hospital mortality. This suggests a more severe course of AKI among patients with COVID-19 when compared to other conditions requiring critical care.

Timing on which AKI occurred was particularly con sidered as an important clinical feature to elucidate the effect of COVID-19 on kidney function. For this purpose, a period of time in the course of the disease in which the most severe manifestations (particularly hypoxemia) occur was identified, e.g., before day 14 from symptom onset¹². AKI onset coincided with this period of time in our study (median of 10 days from symptom onset; IQR 5-17), and was strongly related with the moment of IMV initiation, mainly on early episodes (Table 3). IMV and neuromuscular blockade requirements were significantly higher in the AKI group (Table 2). Nevertheless, no direct association with hypoxemia or PEEP was found. Other mechanisms for AKI not measured in this study such as inflammation mediated by IL-6, direct effect of SARS CoV-2, renin-angiotensin-al dosterone system disbalance or microvascular thrombosis may have been playing a role⁷. In this setting, AKI could be another target of multiorgan dysfunction due to COVID-19, occurring in the period of more severe respiratory mani festations. In keeping with our findings, recently Hirsh et al. found that 52.2% of subjects with severe COVID-19 presenting AKI during hospitalization developed it within 24 hours of intubation²⁴.

When considering the underlying causes of AKI in the cohort, the majority of late AKI episodes were associ ated with sepsis and nephrotoxic drug use. We could not filiate any risk factor to early AKI but it seems that it is a severe event, showing high in hospital mortality and RRT requirement.

Numerous limitations must be weighted when consider ing our findings. Since the present study was retrospective, data was already registered in secondary databases. The previous must be considered especially when correlating AKI causes. Missing data (specifically on CPK levels, abdominal hypertension, obstructive AKI and diuretic use) could potentially threaten the validity of our findings. In this line, data regarding urinalysis and inflammatory param eters such as IL-6 could not be captured because it is not routinely obtained for all patients in our institution. Finally, since patients were followed up until hospital discharge, information regarding renal recovery in the ambulatory setting was not retrieved.

However, some strengths must also be noted. Compre hensive data was gathered from baseline characteristics and relevant clinical outcomes for all study patients. Major associated factors such as sepsis, nephrotoxic drug ex posure and vasopressor therapy were recorded. In addi tion, the transient nature of AKI episodes was thoroughly assessed by an expert nephrologist.

In conclusion, AKI is a frequent condition in patients with severe COVID-19, associated with higher in hospital mortality. Several risk factors are associated with AKI development in this setting, including increasing age, a history of hypertension or HF and IMV requirement. AKI is an early complication of severe COVID-19 adding significant morbidity and mortality, with a still unknown physiopathology. Late AKI is a condition associated with ICU complications such as sepsis and nephrotoxic exposure. In this setting, prevention and treatment of infections adequately could play a relevant role. Further studies should confirm our findings and shed light on the underpinnings of these associations.