Circulating dipeptidyl peptidase 3 and bio-adrenomedullin levels are associated with impaired outcomes in critically ill COVID-19 patients: a prospective international multicentre study

Abstract

Introduction

Dipeptidyl peptidase-3 (DPP3) is a protease involved in the degradation of several cardiovascular mediators. Adrenomedullin (bio-ADM) is a peptide essential for regulation of endothelial barrier function. In different shock-pathologies, both biomarkers are associated with disease severity, organ dysfunction and mortality. Associations with outcome in critically ill COVID-19 patients are unknown. The objectives of the present study were to investigate associations of bio-ADM and “circulating DPP3” (cDPP3) with short-term outcome in critically ill COVID-19 patients (n=80).

Methods

A multicentre prospective cohort study was performed. The primary end-point was 28-day mortality. Secondary end-points included different severities of acute kidney injury (AKI).

Results

cDPP3 levels were mainly associated with 28-day mortality; Area under the receiver operating characteristics (AUROCs) of 0.69 (0.56–0.82, p=0.023), 0.77 (0.64–0.90, p<0.001) and 0.81 (0.65–0.96, p<0.001) at admission, day 3 and day 7, respectively. In contrast, bio-ADM levels were mainly associated with AKI, with AUROCs of 0.64 (0.51–0.77, p=0.048), 0.75 (0.64–0.86, p<0.001) and 0.83 (0.74–0.93, p<0.001) for day 1, 3 and 7, respectively. Interestingly, patients with high levels of both cDPP3 and bio-ADM at day 7 had an additionally increased risk of 28-day mortality (hazard ratio 11.8; 95% CI 2.5–55.3, p<0.001).

Conclusions

cDPP3 and bio-ADM responses were associated with short-term mortality and AKI in critically ill COVID-19 patients, respectively. These findings suggest that treatment with specific antibodies modulating cDPP3 or bio-ADM-related pathways may improve outcome of COVID-19.

Short abstract

cDPP3 and bio-ADM responses are associated with short-term outcome in critically ill #COVID19 patients. These findings suggest that treatment with specific antibodies against these biomarkers may improve outcome of COVID-19 patients. https://bit.ly/3SWx3EE

Introduction

Since late 2019, the world has been hit by an unprecedented pandemic crisis due to SARS-CoV-2. Ever since the pandemic started, there has been a global effort to search for predictors of COVID-19 disease severity and clinical outcome [1, 2]. While early studies provided an initial understanding on broad risk factors associated with detrimental outcome in SARS-CoV-2 infection, including age and cardiovascular disease, these were all non-modifiable risk factors, not amendable to therapeutic interventions [2]. The use of interleukin-6 (IL-6) antagonist therapies in combination with steroids has effectively improved survival of critically ill COVID-19 patients [3]. Interestingly, the beneficial effects of dexamethasone appear to be most pronounced in patients with a hyperinflammatory phenotype [4], while IL-6 concentrations predict the therapeutic efficacy of tocilizumab [5]. These studies not only suggest that severe COVID-19 pathology is accompanied by elevated levels of several (inflammatory) parameters, but also that the levels of differently activated molecular pathways can predict the therapeutic efficacy of interventions.

Dipeptidyl peptidase-3 (DPP3) is a ubiquitous cytosolic metallopeptidase involved in blood pressure regulation, inflammation and pain regulation [5, 6]. When cell injury occurs, intracellular DPP3 is released into the circulation (coined cDPP3 for “circulating DPP3”). The subsequent rapid degradation of cardiovascular mediators such as angiotensin II by cDPP3 may contribute to haemodynamic instability [6]. Recent experimental and clinical studies demonstrated that high levels of cDPP3 are associated with organ dysfunction, such as renal and cardiac failure, and mortality in patients with sepsis [7, 8]. Biologically active adrenomedullin (bio-ADM) is a 52 amino-acid peptide essential for the regulation of vascular tone and endothelial barrier function [6, 9]. High concentrations of bio-ADM were observed in different shock-associated pathologies, with highest levels observed in septic shock. During sepsis, bio-ADM kinetics are strongly associated with sepsis severity and development of organ dysfunction, as well as mortality [6, 10].

Interestingly, combining measurements of both these biomarkers in e.g. cardiac surgery [11] or septic shock patients [6] improved accuracy to predict clinical outcomes, compared to separate assessment of individual biomarkers. Lastly, both biomarkers represent promising druggable targets, as bio-ADM- and cDPP3-modulating therapies have already shown beneficial effects in animal models of cardiogenic and septic shock [12–14], with clinical studies currently planned or ongoing [15].

Understanding the temporal and quantitative response patterns of bio-ADM and cDPP3 in critically ill patients with COVID-19 may improve our understanding of COVID-19-associated molecular pathophysiology. It may also provide important information on the likelihood of therapeutic efficacy of treatments related to these pathways. The main objective of this study was to clarify the associations between bio-ADM and cDPP3 and the occurrence of 28-day mortality in critically ill patients with COVID-19. Secondary end-points included the biomarkers’ association with acute kidney injury (AKI) (as a representative of organ failure), as well as to investigate the kinetics of both biomarkers in patients with COVID-19.

Methods

Study design and population

We performed a multicentre prospective cohort study including all COVID-19 patients admitted to an intensive care unit (ICU) from March to April 2020. Participating centres included the Radboudumc University Medical Center in Nijmegen (Netherlands) and the Lariboisiere and Saint-Louis hospitals in Paris (Université de Paris Cité, assistance Publique des Hôpitaux de Paris, France). All patients admitted to the ICU for severe COVID-19 were eligible. Patients aged under 18 years or those who were pregnant were excluded from the analysis. The diagnosis of COVID-19 was based on the detection of SARS-CoV-2 from nasopharyngeal samples by PCR and the clinical picture of COVID-19.

This study was carried out in accordance with the applicable rules concerning the review of research ethics committees and informed consent. All patients or legal representatives were informed about the details of this cohort study and could decline to participate. Clinical and biological data were recorded at admission, as well as days 3, 7, 14, 21 and 28 of ICU admission. Patient discharge status and mortality were recorded up to day 28 after ICU admission, regardless of patient admission status. Primary end-points were 28-day mortality and AKI. AKI was defined and staged according to the Kidney Disease: Improving Global Outcomes (KDIGO) criteria [16]. The kinetics of both biomarkers during the course of the first week of admission represented a secondary end-point.

Measurements

Ethylenediaminetetraacetic acid (EDTA)-anticoagulated blood for determination of cDPP3 and bio-ADM concentrations was sampled within 24 h after ICU admission; follow-up samples were acquired on day 3 and 7 after ICU admission. After centrifugation, plasma was stored at −80°C until blinded bio-ADM and cDPP3 analysis using luminescence immunoassays for bio-ADM (Sphingotec GmbH, Hennigsdorf, Germany) and cDPP3 (4TEEN4 Pharmaceuticals GmbH, Hennigsdorf, Germany) [17, 18]. After patient enrolment, the following data were collected at baseline and on days 3, 7, 14 and 28: need for renal replacement therapy, need for vasopressor therapy and need for invasive mechanical ventilation. The following routine laboratory parameters were also collected: leukocytes, serum creatinine, troponin-T, creatinine-kinase, aspartyl-aminotransferase (ASAT), alanine-aminotransferase (ALAT), interleukin-6 (IL-6) and lactate.

Statistical analyses

Data are presented as median (interquartile range (IQR)) or mean (95% confidence interval (CI)), according to distribution, and as counts and percentages for categorical variables. Group comparisons of continuous variables were performed using Mann–Whitney U-tests or Kruskal–Wallis tests, depending on the number of groups. If Kruskal–Wallis tests yielded overall significance, Tukey's post hoc tests were performed. Data were log-transformed if not normally distributed. Categorical data were compared using Fisher's exact tests. Correlations between continuous variables were calculated using Pearson's correlation. If correction for repeated observations was necessary, the Bland–Altman method was used [19]. Area under the receiver operating characteristics (AUROCs) with 95% CIs are reported as an effect measure of the discriminative capacity of different predictor variables. Differences between AUROCs were assessed with the Hanley–McNeil method. To assess whether combining biomarker measurements with other known predictor variables improved predictive value, the added value of these biomarkers was evaluated using logistic regression modelling. Different variables with significant predictive value were entered as independent variables, with binary outcomes as dependent variables. To test for added predictive value, we used the likelihood ratio chi-square test for nested models to assess whether a biomarker added predictive value to a clinical model or a risk score. When dichotomising the population based on a continuous biomarker measurement was necessary, cut-offs defined by earlier studies were used for both cDPP3 (40 ng·mL⁻¹; [6, 7]) and bio-ADM (70 pg·mL⁻¹; [20, 21]). Sensitivity, specificity and accuracy implementing these cut-offs are reported as additional biomarker effect measures. Kaplan–Meier curves implementing these cut-offs were generated for illustrative purposes.

A two-sided p-value of <0.05 was considered statistically significant. All analyses were performed using Statistical Package for the Social Sciences (SPSS) version 25.0 (SPSS Inc., Chicago, IL, USA) and GraphPad Prism version 8.0 (GraphPad Software, La Jolla, CA, USA).

Results

Study population

cDPP3 and bio-ADM were measured in a cohort of 80 patients hospitalised in the ICU for COVID-19, from March 2020 to April 2020. Baseline characteristics of studied patients are depicted in table 1. Over the 28-day follow-up period, a total of 18 (22.5%) patients died. As patients were included during the first COVID-19 surge, none of the patients received treatment with dexamethasone or IL-6-targeted antibodies.

TABLE 1.

Admission baseline characteristics of the total cohort, as well as admission characteristics of 28-day survivors versus non-survivors

	Total	Non-survivors	Survivors	p-value
Subjects n	80	18	62
Demographics
Sex, male	60 (75.0)	14 (77.8)	46 (74.2)	1.000
Age years	65 (57–70)	64 (58–71)	65 (57–69)	0.809
BMI kg·m−2	27.5 (25.4–31.0)	29.6 (27.0–32.1)	27.0 (24.6–30.6)	0.037
Medical history
COPD	4 (5.0)	0 (0.0)	4 (6.5)	0.570
Diabetes	31 (38.8)	7 (38.9)	24 (38.7)	1.000
Hypertension	41 (51.2)	9 (50.0)	32 (51.6)	1.000
Immune-deficiency	7 (8.8)	3 (16.7)	4 (6.5)	0.185
Characteristics during hospitalisation
Leukocytes	8.5 (6.2–12.4)	9.5 (7.3–11.5)	8.4 (6.0–12.5)	0.541
Creatinineday1 µmol·L−1	80 (61–119)	125 (72–269)	77 (58–100)	0.004
Lactateday1 mmol·L−1	1.2 (1.0–1.6)	1.5 (1.0–1.9)	1.2 (1.0–1.6)	0.102
Troponin-Tday1 ng·L−1	23 (14–56)	68 (19–130)	22 (11–42)	0.013
P/F ratio	146 (96–200)	88 (66–183)	165 (118–203)	0.036
Vasopressors	38 (47.5)	7 (38.9)	31 (50.0)	0.289
Inotropics	3 (3.8)	1 (5.6)	2 (3.2)	0.540
Mechanical ventilation	66 (82.5)	15 (83.3)	51 (82.3)	1.000
AKI (any KDIGO stage)	34 (42.5)	9 (50.0)	25 (40.3)	0.590
AKI (KDIGO stage 3)	15 (18.8)	7 (38.9)	8 (12.9)	0.034
Severity scores
Admission SOFA	6 (4–7)	6 (3–9)	6 (4–7)	0.731

Data presented as n (%) or median (IQR) unless stated otherwise. BMI: body mass index; COPD: chronic obstructive pulmonary disease; P/F ratio: P_aO2/F_IO2 ratio; AKI: acute kidney injury; KDIGO: Kidney Disease: Improving Global Outcomes; SOFA: sequential organ failure assessment.

Both cDPP3 and bio-ADM levels were associated with 28-day mortality in critically ill patients with COVID-19

At ICU admission, patients that would eventually not survive already presented with higher plasma creatinine (median: 125 (IQR: 72–269) versus 77 (58–100), p=0.004), higher troponin-T (68 (19–130) versus 22 (11–42), p=0.013), and a worse arterial oxygen tension (P_aO2)/inspiratory oxygen fraction (F_IO2) ratio (88 (66–183) versus 165 (118–203), p=0.036) (table 1). At ICU admission, median (IQR) cDDP3 and bio-ADM concentrations were 35.7 ng·mL⁻¹ (25.0–54.3) and 47.7 pg·mL⁻¹ (33.5–78.8), respectively. 28-day non-survivors presented with higher admission cDPP3 levels compared to survivors, and these differences became more pronounced over time (figure 1a) with AUROCs of 0.69 (0.56–0.82, p=0.023), 0.77 (0.64–0.90, p<0.001) and 0.81 (0.65–0.96, p<0.001) for admission, day 3 and day 7, respectively (figure 2a). In multivariate analyses (implementing creatinine and P_aO2/F_IO2 ratio, as these were found to be significantly associated with 28-day mortality as well), the more pronounced predictive value of DPP3 over time remained present: DPP3 day 1 standardised hazard ratio (HR) 2.02 (95% CI 0.77–5.29), p=0.152; day 3 standardised HR 2.49 (95% CI 0.98–6.31), p=0.054; and day 7 standardised HR 3.51 (95% CI 1.21–10.15), p=0.021. Bio-ADM concentrations did not differ significantly between non-survivors and survivors on day 1 and 3, but concentrations were higher at day 7 in non-survivors (figure 1b). Consequently, day 1 and 3 bio-ADM levels did not significantly predict 28-day mortality, with AUROCs of 0.64 (0.48–0.80, p=0.089) and 0.63 (0.47–0.79, p=0.110), respectively (figure 2a). At day 7, significant predictive capacity was present, with an AUROC of 0.70 (0.55–0.85, p=0.020) (figure 2a). Combining cDPP3 (cut-off at 40 ng·mL⁻¹) and bio-ADM (cut-off at 70 pg·mL⁻¹) measured at day 7 showed additional prognostic value to predict 28-day mortality compared to separate assessment of both biomarkers alone; HR 11.8 (2.5–55.3, p<0.001) compared to patients with one of the two, or both biomarkers below the cut-off value (figure 3). An overview of the sensitivity, specificity and accuracy of associations with 28-day mortality of both biomarkers implementing predefined cut-offs is provided in figure 4a and b.

FIGURE 1.

Biomarker measurements of a) cDPP3 and b) bio-ADM on day 1, 3 and 7 of ICU admission. Results of 28-day survivors and non-survivors are displayed, respectively. The median, interquartile ranges and Tukey's hinges are displayed. cDPP3: circulating dipeptidyl peptidase 3; bio-ADM: biologically active adrenomedullin. *: p<0.05; **: p<0.01; ***: p<0.001.

FIGURE 2.

Areas under the receiver operating characteristic curve (AUC) of cDPP3 and bio-ADM on days 1, 3 and 7 of ICU admission. Separate graphs display different study outcomes. Acute kidney injury (AKI) was defined according to the Kidney Disease: Improving Global Outcomes (KDIGO) criteria. Bio-ADM: bioactive adrenomedullin; cDPP3: circulating dipeptidyl peptidase 3.

FIGURE 3.

Kaplan–Meier analysis of 28-day mortality in critically ill COVID-19 patients on day 7 after ICU admission. Groups are based on high cDPP3 (>40 ng·mL⁻¹) versus low cDPP3 (<40 ng·mL⁻¹), as well as high bio-ADM (>70 pg·mL⁻¹) versus low bio-ADM (>70 pg·mL⁻¹). The reported hazard ratio compares the group with high cDPP3 and high bio-ADM levels to the other three groups (curves displayed separately in figure for illustrative purposes). Bio-ADM: bioactive adrenomedullin; cDPP3: circulating dipeptidyl peptidase 3.

FIGURE 4.

Graphical overview of sensitivity, specificity and accuracy for associations of cDPP3 and bio-ADM with 28-day mortality and all-stage AKI. Cut-offs for cDPP3 and bio-ADM were >40 ng·mL⁻¹ and >70 pg·mL⁻¹, respectively. Bio-ADM: bioactive adrenomedullin; cDPP3: circulating dipeptidyl peptidase 3. *: p<0.05; **: p<0.01.

Bio-ADM levels were associated with the development of AKI

An important proportion of COVID-19 patients (n=34, 42.5%) developed AKI during the 28-day follow-up (median diagnosis day 3 (3–7) days). A total of 16 (20.0%), 3 (3.8%) and 15 (18.8%) patients developed KDIGO stage 1, 2 and 3, respectively. Of note, patients that developed KDIGO stage 3 AKI were more likely to die during admission; 7 out of 15 (46.7%) versus 11 out of 65 (16.9%) (p=0.034), while no differences in survival status were present in patients that developed KDIGO stage 1 or 2 AKI (table 1). No associations between cDPP3 and creatinine were found (table 2). As depicted by figure 2b and c, cDPP3 concentrations at day 1 and day 3 were not associated with the development of AKI. Associations with day 7 cDPP3 levels and AKI were present, albeit only weakly; AUROC of 0.66 (0.53–0.76, p=0.019) for all stage AKI, and AUROC of 0.69 (0.52–0.85, p=0.032) for stage 3 AKI, respectively (figure 2b and c).

TABLE 2.

Correlation coefficients between different predictor variables compared to cDPP3 and bio-ADM measurements

Variable	r-coefficient	p-value
cDPP3 correlation
Bio-ADM	0.144	0.089
Creatinine	0.162	0.058
Troponin-T	0.176	0.103
ALAT	0.530	<0.001
ASAT	0.714	<0.001
Bilirubin	0.181	0.065
Lactate	0.374	<0.001
P/F ratio	−0.138	0.138
Urine output	−0.310	0.027
IL-6	0.277	0.075
Bio-ADM correlation
cDPP3	0.144	0.089
Creatinine	0.617	<0.001
Troponin-T	0.215	0.046
ALAT	0.005	0.910
ASAT	0.008	0.363
Bilirubin	0.196	<0.001
Lactate	0.334	<0.001
P/F ratio	0.104	0.268
Urine output	−0.009	0.937
IL-6	0.353	0.023

cDPP3: circulating dipeptidyl peptidase 3; Bio-ADM: biologically active adrenomedullin; ALAT: alanine-aminotransferase; ASAT: aspartyl-aminotransferase; P/F ratio: P_aO2/F_IO2 ratio; IL-6: interleukin-6.

By contrast, bio-ADM levels were associated with the occurrence of all-stage AKI (figure 2b). Bio-ADM levels correlated most strongly with creatinine (r=0.62, p<0.001). Association between bio-ADM levels and all stage AKI became more pronounced from day 1 to day 7 (Hanley–McNeil p=0.022), with AUROCs of 0.64 (0.51–0.77, p=0.048), 0.75 (0.64–0.86, p<0.001) and 0.83 (0.74–0.93, p<0.001) for day 1, 3 and 7, respectively (figure 2b). An overview of the sensitivity, specificity and accuracy of associations with all-stage AKI of both biomarkers implementing predefined cut-offs is provided in figure 4c and d. Associations between bio-ADM and stage 3 AKI were also present for all time points, with good discriminative capacity already being present for day 1 and 3 time points (figure 2c). Lastly, to further investigate factors influencing biomarker kinetics, a combined approach (with subgroups based on the occurrence of both 28-day mortality and AKI) was also performed. An overview of these analyses is presented in figure 5. Results should be interpreted with caution, as further subgroup stratification severely limits statistical power given the study cohort size.

FIGURE 5.

Biomarker kinetics of cDPP3 and bio-ADM based on subgroups combining study outcomes of 28-day mortality and AKI. a) Subgroups based on survival status and occurrence of KDIGO stage 1 AKI. b) Subgroups based on survival status and occurrence of KDIGO stage 3 AKI. Median and interquartile ranges are displayed. AKI: acute kidney injury; bio-ADM: bioactive adrenomedullin; cDPP3: circulating dipeptidyl peptidase 3; ICU: intensive care unit. *: p<0.05; **: p<0.01; ***: p<0.001.

Additional association analyses with other measurements of disease severity

At corresponding time points, cDPP3 levels correlated most strongly with ALAT (r=0.53) and ASAT (r=0.71), as well as serum lactate (r=0.37) (all p<0.001) (table 2). cDPP3 levels did not correlate with time-corresponding measurements of bio-ADM (table 2). Correlations of bio-ADM with troponin (r=0.22, p=0.046), lactate (r=0.33, p<0.001), IL-6 (r=0.35, p=0.023) and bilirubin (r=0.196, p<0.001) were also present (table 2).

Discussion

The main finding of this multicentre prospective cohort study is that cDPP3 and bio-ADM show distinctly different kinetics and associations with outcome in critically ill COVID-19 patients. cDPP3 levels were mainly associated with short-term mortality, with measurements performed later during ICU admission showing the strongest associations. In contrast, the association of bio-ADM concentrations was most pronounced with the development of AKI, with measurements performed early during ICU admission already strongly associating with development of severe AKI. Interestingly, patients with high levels of both biomarkers 1 week into ICU admission demonstrated an incrementally increased risk of 28-day mortality compared to patients with either high levels of only one biomarker or low levels of both of them. These observed differences in outcome prediction in critically ill COVID-19 patients might be explained by the different pathophysiological pathways involved in the release of these biomarkers.

DPP3 only reaches high levels in the circulation if significant cell injury occurs [6]. Interestingly, until now, high cDPP3 levels were mainly described in different etiologies of circulatory shock, including sepsis, cardiogenic shock and severe burn injury [7, 8, 22, 23]. Apart from circulatory shock, high cDPP3 levels also appear to be induced by major surgical trauma, albeit less sustained than during circulatory shock [11, 24]. However, both these mechanisms cannot readily explain the sustained high cDPP3 levels observed in our critically ill COVID-19 cohort. While close to 50% of patients required a low dose of vasopressor therapy, the majority of the cohort did not meet circulatory shock definitions. Moreover, observational studies describe circulatory shock in <20% of critically ill COVID-19 patients [25]. Thus, the absence of shock makes it unlikely that cellular lysis induced by tissue hypoperfusion is the main cause of cDPP3 release in our cohort. DPP3 is however also expressed in different pulmonary tissues [26]. Putatively, the increasing cDPP3 levels observed in our study are associated with pulmonary cell damage, incurred because of the ongoing infection, as well as ventilator-associated pulmonary injury. Interestingly, a recent study that investigated endothelial pulmonary injury biomarkers described markers already peaking at ICU admission [27]. All these endothelial injury markers subsequently only decreased up until the second week of ICU admission [27]. These results make it unlikely that COVID-19-induced pulmonary endothelial injury alone can explain the further increase in cDPP3 levels observed after 1 week of ICU admission. Another possible source of high cDPP3 levels could be liver cell damage. Hepatocytes are known to contain large amounts of DPP3 [26]. Moreover, cDPP3 levels were already found to correlate with the severity of liver failure in a cohort of liver cirrhosis patients [28], as well as liver sequential organ failure assessment (SOFA) scores in sepsis patients [7, 29]. In accordance, we describe notable associations between cDPP3 levels and the transaminases ALAT and ASAT. However, transaminases have been described as poor measures of liver function disorder during critical illness, rather reflecting overall disease severity [30]. Also, significant correlations between cDPP3 and bilirubin, a measure of hepatic dysfunction with higher specificity, were not present. Therefore, correlations between cDPP3 and transaminases might well be explained because they are markers of general cellular damage. This would mean they are associated with disease severity, rather than being associated with a specific hepatic origin.

Regardless of aetiology, there are multiple ways by which high cDPP3 levels would be able to cause further clinical deterioration in COVID-19 patients. Circulating DPP3 is able to rapidly metabolise essential cardiovascular mediators, most notably angiotensin II [6]. Recent research has already demonstrated markedly reduced angiotensin-II levels in critically ill COVID-19 patients, also compared to other acute respiratory distress syndrome etiologies [31, 32]. Based on our results, it might well be that high cDPP3 levels represent an additional pathway leading to failing angiotensin-II responses in critically ill COVID-19 patients.

Several studies have already assessed associations of mid-regional-pro-adrenomedullin (MR-pro-ADM), another biomarker of the adrenomedullin pathway, with outcome in COVID-19 patients [33–35]. Interestingly, while most of these studies performed only one measurement within a short timeframe after admission, they report moderate associations with short-term mortality [33–35]. Interestingly, the strength of these reported associations appears quite similar to those we report. At the same time, they contrast with the weaker associations we report for the two measurements performed early during admission.

Interestingly, none of the aforementioned studies investigated associations between adrenomedullin and development of AKI, despite its frequent occurrence in COVID-19 patients [36]. Despite numerous recent studies on the potential mechanisms of COVID-19-associated AKI, its pathophysiology remains elusive [36]. It is thought to involve both local and systemic inflammatory responses, endothelial injury and a dysregulation of the renin–angiotensin system [36]. Adrenomedullin is well known as a hormone essential for endothelial homeostasis, aimed at maintaining endothelial integrity during inflammation [9]. Increased levels of adrenomedullin are also found during acute episodes of capillary leak syndrome, a disease aetiology which is mainly characterised by endotheliitis [37]. Considering the strong relation between bio-ADM and the development or presence of AKI found in our study, it might well be that these levels are mainly related to COVID-associated endothelial injury as well. As bio-ADM mitigates the loss of endothelial barrier function present during septic shock, it is thought that increased levels represent a compensatory response [9]. However, because circulatory shock was not present in the majority of patients with AKI, haemodynamic compromise alone cannot serve as an explanation for the high bio-ADM levels found in our cohort.

The results of our study may have several clinically relevant implications. First, both cDPP3 and bio-ADM could act as potential stratifying biomarkers during the first week of ICU admission, identifying COVID-19 patients with an increased risk of unfavourable outcome. Assessment of cDPP3 in conjunction with bio-ADM identifies patients with an even higher risk of impaired short-term outcomes. These results emphasise that the effects of multiple pathophysiological processes are associated with markedly heterogeneric disease phenotypes in COVID-19. This heterogeneity, already well known from different etiologies of critical illness like sepsis, means that assessment of one biomarker, as well as conventional disease severity indexes, are unlikely to accurately predict short-term outcome [38].

Both cDPP3- (preclinical stage) and bio-ADM (clinical stage [15])-associated therapeutics are being developed. The availability of specific therapeutics able to correct the dysregulated pathways associated with these biomarkers greatly increases their clinical relevance. Based on our results, it appears plausible that cDPP3 inhibition might serve to mitigate clinical outcomes in COVID-19 patients. Also, bio-ADM modulators might serve as an additional therapeutic intervention, aimed at preventing COVID-associated AKI. Of interest, the derangements in both pathways were not already present at ICU admission, but rather became apparent during the first week of admission [7, 10]. This would mean that compared to bacterial sepsis, there may be a wider therapeutic window in which bio-ADM- and cDPP3-associated therapies could exert therapeutic efficacy in COVID-19 patients. Clearly, the observational nature of the current study does not imply that modulation of these pathways would translate into a clinical benefit for the patient. It does however warrant confirmation in interventional studies. Correspondingly, a phase 2 study investigating the adrenomedullin-modulating antibody Adrecizumab in COVID-19 patients is currently being planned (https://clinicaltrials.gov/ct2/show/NCT05156671).

Our study has several limitations. One major limitation is the small sample size, as only 80 patients were recruited during a short time interval. This means our results should be interpreted as hypothesis-generating. However, because we included only critically ill COVID-19 patients, unfavourable outcomes occurred frequently, explaining why we were already able to find significant associations with cDPP3 and bio-ADM in this relatively small cohort. Second, between patient's recruitment and final results of our analysis, the clinical management of COVID-19 has markedly changed, most notably related to the immunomodulatory treatment (corticosteroids and tocilizumab). Future studies are needed to demonstrate to what extent biomarker kinetics are influenced by these immunomodulatory treatments. Lastly, we did not directly compare the prognostic performance of both cDPP3 and bio-ADM to other candidate biomarkers. Since other candidate biomarkers may be more readily available at the bedside, follow-up studies with larger sample sizes should assess whether cDPP3 and/or bio-ADM are able to further improve the prediction already provided by these biomarkers.

In conclusion, our study suggests that cDPP3 and bio-ADM, two biomarkers previously used to predict outcomes in bacterial septic shock, are associated with short-term mortality and development of AKI in critically ill COVID-19 patients, respectively. The two biomarkers show different kinetics and different predictive values for organ dysfunction or mortality, while an approach combining both biomarkers identified patients with the highest risk of unfavourable outcome. Based on our results, the therapeutics associated with both these biomarkers may hold promise at improving outcome of critically ill COVID-19 patients.