*Perspective* **Beyond Guidelines and Reports on Bacterial Co-/Superinfections in the Context of COVID-19: Why Uniformity Matters**

**Johan Van Laethem 1,\*, Denis Piérard <sup>2</sup> and Sabine D. Allard <sup>1</sup>**


**Abstract:** Background: In the period following the declaration of the COVID-19 pandemic, more evidence became available on the epidemiology of bacterial co-/superinfections (bCSs) in hospitalized COVID-19 patients. Various European therapeutic guidelines were published, including guidance on rational antibiotic use. Methods: In this letter to the editor, we provide an overview of the largest metaanalyses or prospective studies reporting on bCS rates in COVID-19 patients and discuss why the reader should interpret the results of those reports with care. Moreover, we compare different national and international COVID-19 therapeutic guidelines from countries of the European Union. Specific attention is paid to guidance dedicated to rational antibiotic use. Results: We found a significant heterogeneity in studies reporting on the epidemiology of bCSs in COVID-19 patients. Moreover, European national and international guidelines differ strongly from each other, especially with regard to the content and extent of antibiotic guidance in hospitalized COVID-19 patients. Conclusion: A standardized way of reporting on bCSs and uniform European guidelines on rational antibiotic use in COVID-19 patients are crucial for antimicrobial stewardship teams to halt unnecessary antibiotic use in the COVID-19 setting.

**Keywords:** bacterial co-infection; bacterial superinfection; antibiotic stewardship; COVID-19

## **1. Introduction**

The emergence of multidrug-resistant (MDR) bacterial infections has resulted in scarifying future projections. A report from the World Health Organization (WHO) labeled the problem as "so serious, that it threatens the achievements of modern medicine" [1]. For the last few decades, local, national and international efforts have been endorsed by scientific and public health organizations, governments and caregivers to halt the emergence of antimicrobial resistance (AMR). Amongst other interventions, antimicrobial stewardship (AST) teams became the standard of care, guidelines and scientific publications encouraging rational antibiotic use were published, national antibiotic action plans were launched and awareness was cultivated in the general population.

In this perspective article, we first provide a summary of the impact of the COVID-19 pandemic on antimicrobial stewardship efforts, AMR and the growing knowledge on bacterial co-/superinfection epidemiology and antibiotic (over)use in this context. Second, we offer a critical analysis of the major papers reporting on bacterial co-/superinfection (bCS) rates in COVID-19 patients. Last, we discuss the variation in European guidelines for the diagnosis/treatment of these bCSs.

## **2. Antimicrobial Stewardship Applied to COVID-19 Patients: The Pursuit of Knowledge**

At the time the COVID-19 pandemic emerged, AST teams and other actors within the healthcare system constrainedly invested great amounts of time and resources in the

**Citation:** Van Laethem, J.; Piérard, D.; Allard, S.D. Beyond Guidelines and Reports on Bacterial Co-/Superinfections in the Context of COVID-19: Why Uniformity Matters. *Antibiotics* **2022**, *11*, 1446. https://doi.org/10.3390/ antibiotics11101446

Academic Editor: Marcello Covino

Received: 12 September 2022 Accepted: 18 October 2022 Published: 20 October 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

contention of the pandemic, the procuration of protective equipment and the reorganization of the healthcare system. The latter inevitably resulted in less stringent antimicrobial stewardship, leaving the battle against AMR in the background. Currently, there is insufficient evidence that the COVID-19 pandemic fueled the AMR threat, as present reports are context-specific and differ geographically. However, there are some ominous signs of increased AMR since the emergence of the COVID-19 pandemic. For the European Union, the European Antimicrobial Resistance Surveillance (EARS-Net) network reported a significant rise in carbapenem-resistant *Enterobacterales*, *Pseudomonas aeruginosa* and *Acinetobacter* species as well as vancomycin-resistant enterococci for the year 2020 [2]. Moreover, there was a significant rise in carbapenem use in that same year. The rise in MDR pathogens was most marked in the intensive care setting [3].

In the period following the declaration of the COVID-19 pandemic by the WHO on 11 March 2020, many admitted COVID-19 patients empirically received antibiotics [3,4]. This period of "antibiotic anarchy" can (partly) be explained by the lack of knowledge and lack of guidelines concerning the epidemiology and treatment of presumed bCSs in the context of COVID-19. From May 2020 onwards, the first reports and meta-analyses regarding the incidence and prevalence of bCSs showed very low rates of bacterial co-infections (2.2–8%) and low rates of bacterial superinfections (2.2–20%) in admitted COVID-19 patients. This was in contrast to disproportionally high antibiotic prescribing rates (up to 85%) [4–8]. Soon after the first published reports on bCS incidence, guidelines on the management of presumed bCSs in COVID-19 patients were published by the European Society of Clinical Microbiology and Infectious Disease (ESCMID) (in April 2020) and the WHO (May 2020) [9,10]. Later on, evidence on the good negative predictive value of procalcitonin in excluding bCSs in the context of COVID-19 became available [11,12]. Consequently, over time, a learning effect and a decrease in antibiotic use were noted in some studies [13]. However, antibiotic overprescribing and low quality of antibiotic prescriptions are still prevalent in admitted COVID-19 patients [14].

#### **3. Evidence on bCS Rates in COVID-19 Patients: A Critical Point of View**

Robust scientific evidence on COVID-19-related bCS epidemiology and clear therapeutic guidance regarding antibiotic use in COVID-19 patients are crucial for antimicrobial stewardship (AST) teams to prevent antibiotic overprescribing. Despite all progress made to gain expertise on bCS prevalence and antibiotic prescribing patterns in COVID-19 patients, significant knowledge gaps and flaws prevail. Moreover, practical guidance on judicious antibiotic use in COVID-19 patients should be improved, as these guidelines are very heterogeneous and lack specificity. Finally yet importantly, meta-analyses and prospective studies reporting on bCS rates in COVID-19 patients (Table 1) should be interpreted with care for several reasons.

**Table 1.** Overview of largest (n > 3000) meta-analyses or prospective studies reporting on bacterial co- /superinfection rates in COVID-19 patients (see Supplementary Material S1 for the search strategy).



#### **Table 1.** *Cont.*

\* Superinfection and secondary infection are used as synonyms. NR: not reported.

First, there is a significant heterogeneity in the used definitions of co-infection and superinfection (also referred to as "secondary infection"). Certain reports use the term coinfection as "every infection contracted before or during the first 48 h of admission", while others use 24 h of admission as a time limit to differentiate co-infection from superinfection. Some even refer to "every infection diagnosed on presentation", while in the meta-analysis of Langford et al. (2022) [5], no definition of co-infection is mentioned. Depending on the used definition, bacterial co-infection and superinfection rates can thus be under- or overestimated. Moreover, international COVID-19 therapeutic guidelines such as the World Health Organization and the European Centre for Disease Control and Prevention guidelines do not define bacterial co-infection and superinfection [10]. Second, although most studies exclusively included microbiological diagnoses, it is not always clear if clinical diagnoses, based on other criteria than microbiological documentation, were included. While the first meta-analysis of Langford et al. (2020) [4] did not mention if included diagnoses were exclusively based on microbiological criteria, their second meta-analysis stated that "presumed" or "suspected" diagnoses of infection were excluded. Depending on the used definition of "infection", the final bCS rate will be different. Third, there is also an important heterogeneity regarding the included microbiological diagnoses. For example, the meta-analysis of Lansburry et al. reports high rates of *Mycoplasma pneumoniae* infections (representing 42% of all reported bacterial infections) [6]. Although no information is provided on the used diagnostic methods, this could be an overestimation due to the inclusion of patients with aspecific serological results. The ISARIC study did not include any *Mycoplasma pneumoniae* infection, as routine testing for atypical pathogens was discontinued in most United Kingdom laboratories during the study period. Fourth, the proportion of included patients depending on the setting (ward versus ICU) and age (adult versus pediatric patients) is not always clearly mentioned. Fifth, some reports include viral and fungal co-/superinfections together with bCSs. Last, the reported infection sites can differ from one study to another. Although most studies focused on both respiratory

and bloodstream infections, some also included urinary tract infections whereas others exclusively included respiratory tract co-/superinfections. One should thus pay attention to the reported endpoint when comparing different studies.

#### **4. European Therapeutic COVID-19 Guidelines: An Emphasis on Antibiotic Guidance**

As the COVID-19 pandemic progressed, knowledge about bCS epidemiology and antibiotic prescribing in COVID-19 patients increased rapidly. This led to the publication of various antibiotic guidance guidelines. However, studies reporting on bCS rates and guidelines regarding judicious antibiotic use show a significant heterogeneity.

Therefore, we analyzed and compared all published national and international guidelines on COVID-19 therapeutic guidance in the European Union (EU) (see Supplementary Materials for methods, complete data and references). Most countries have published their own national therapeutic guideline (see Figure 1), while others refer to international scientific guidelines, such as those of the World Health Organization [10]. Certain countries, such as Austria, refer to the guidelines of neighboring countries. The majority of the EU countries have also included specific guidance on rational antibiotic use in COVID-19 patients. However, there is a large variability in the extent and content of provided guidance regarding antibiotic use in this setting (Supplementary Material S3, Figure S1). For example, the Dutch Working Party on Antibiotic Policy, as well as the health authorities of Bulgaria, dedicated specific attention to rational antibiotic use in the COVID-19 setting, while in the national guidelines from other countries, such as Belgium, France, Italy, Poland and Spain, only a few sentences on rational AB are found [15–21]. While bacterial co-infections are rare in admitted COVID-19 patients, secondary infections are more prevalent in patients with severe COVID-19. This is probably why most guidelines recommend initiating empiric antibiotics exclusively in patients with severe infection, provided that the need for antibiotics would be regularly evaluated. This is in contrast with the Polish national guidelines, which strongly advise against antibiotics in cases of acute respiratory distress syndrome (ARDS) unless there are evident signs of secondary bacterial infection [20]. Some guidelines limit themselves to recommending antibiotics in cases of suspected bCSs, without elaborating on how to diagnose bCSs [19,20]. The ESCMID guidelines state that only patients with clinical or radiological suspicion of bacterial co-/superinfection should receive empirical antibiotics [9]. However, this is quite vague, as radiological consolidations and clinical signs, such as fever and elevated inflammatory markers, are often present in the context of COVID-19. Therefore, the ECDC guidelines advocate for more clarity in defining secondary bacterial infections in COVID-19 patients [21]. Although the Croatian guidelines advocate that bacterial infection is likely in case of leukocytosis and/or a neutrophil left shift with increased procalcitonin concentration and very high CRP and IL-6 levels, elevated procalcitonin and IL-6 levels have low positive predictive value for bacterial infection in the COVID-19 setting and are also observed in the context of COVID-19 sepsis. Those same guidelines, together with the Danish guidelines, suggest following the "sepsis campaign" guidelines in cases of COVID-19 sepsis [22,23]. However, as COVID-19 sepsis is due to a hyperinflammatory state with a potential cytokine storm, this does not necessarily reflect bacterial sepsis. Yet most sepsis campaign guidelines focus on bacterial sepsis, and this includes the empiric use of antibiotics. Nevertheless, one could agree to empirically start antibiotics in severe and degrading presentations of COVID-19. While the German COVID-19 guidelines recommend antibiotic prescribing at admission in the intensive care unit, those same guidelines paradoxically state that there is no place for prophylactic antibiotics [24,25]. Despite the good negative predictive value of low procalcitonin levels for bCSs [11,12], the use of this predictor is only incorporated in the Latvian guidelines [26].

this predictor is only incorporated in the Latvian guidelines [26].

concentration and very high CRP and IL-6 levels, elevated procalcitonin and IL-6 levels have low positive predictive value for bacterial infection in the COVID-19 setting and are also observed in the context of COVID-19 sepsis. Those same guidelines, together with the Danish guidelines, suggest following the "sepsis campaign" guidelines in cases of COVID-19 sepsis [22,23]. However, as COVID-19 sepsis is due to a hyperinflammatory state with a potential cytokine storm, this does not necessarily reflect bacterial sepsis. Yet most sepsis campaign guidelines focus on bacterial sepsis, and this includes the empiric use of antibiotics. Nevertheless, one could agree to empirically start antibiotics in severe and degrading presentations of COVID-19. While the German COVID-19 guidelines recommend antibiotic prescribing at admission in the intensive care unit, those same guidelines paradoxically state that there is no place for prophylactic antibiotics [24,25]. Despite the good negative predictive value of low procalcitonin levels for bCSs [11,12], the use of

**Figure 1.** Availability of national therapeutic COVID-19 guidelines on rational antibiotic use. Created with Mapchart: https://www.mapchart.net/europe.html; accessed on Oktober 10th 2022. **Figure 1.** Availability of national therapeutic COVID-19 guidelines on rational antibiotic use. Created with Mapchart: https://www.mapchart.net/europe.html; accessed on 10 October 2022.

#### **5. Conclusions 5. Conclusions**

In conclusion, studies reporting on bCS rates and guidelines regarding judicious antibiotic use show significant heterogeneity. Antibiotic prescribing guidelines depend too much on clinical judgment and should instead take variables into account that have proven to be good predictors or excluders of bCSs, such as procalcitonin. The roles of other potential markers and predictors of bCSs, such as certain comorbidities, the presence of immune suppression and the presence of dense radiological consolidations, are still unclear and should be further investigated. In conclusion, studies reporting on bCS rates and guidelines regarding judicious antibiotic use show significant heterogeneity. Antibiotic prescribing guidelines depend too much on clinical judgment and should instead take variables into account that have proven to be good predictors or excluders of bCSs, such as procalcitonin. The roles of other potential markers and predictors of bCSs, such as certain comorbidities, the presence of immune suppression and the presence of dense radiological consolidations, are still unclear and should be further investigated.

A standardized way of reporting on bCSs in the context of COVID-19 is the only way to obtain more robust and precise evidence on their incidence and associated risk factors. We therefore strongly advocate for the implementation of international diagnostic guidelines, using predictors and excluders of bCSs and standardized definitions of bCSs. These definitions and guidelines should be dynamic and more detailed. Guidelines should be based on different clinical situations and could indicate the level of diagnostic certainty, and they should evolve according to the best available evidence. A standardized way of reporting on bCSs in the context of COVID-19 is the only way to obtain more robust and precise evidence on their incidence and associated risk factors. We therefore strongly advocate for the implementation of international diagnostic guidelines, using predictors and excluders of bCSs and standardized definitions of bCSs. These definitions and guidelines should be dynamic and more detailed. Guidelines should be based on different clinical situations and could indicate the level of diagnostic certainty, and they should evolve according to the best available evidence.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/antibiotics11101446/s1, Supplementary Material S1: Methods. Supplementary Material S2, Table S1: National and international guidelines on COVID-19 therapeutic guidance in the European Union (EU). Supplementary Material S3, Figure S1: National therapeutic COVID-19 guidelines: number of words dedicated to rational antibiotic use per country. Refs. [9,14–34] cited in Supplementary Materials.

**Author Contributions:** Conceptualization, J.V.L.; methodology, J.V.L.; investigation, J.V.L.; writing original draft preparation, J.V.L., S.D.A. and D.P.; writing—review and editing, J.V.L., S.D.A. and D.P.; supervision, D.P. and S.D.A. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** See Supplementary Material for additional data.

**Acknowledgments:** Special thanks to following persons for having provided their guidelines or having answered our requests to collect different guidelines: Mary Vaarpu (Finland), Hanna Rätsep (Estonia), Jorge Ruivo and Nuno Cerca (Portugal), Vladimir Krajinoviç and Neven Papic (Croatia), Ivan Ivanov and Ivailo Alexiev (Bulgaria), L. Viksna (Latvia), Oana Sandulescu and Hristea Adriana (Romania), Sotirios Tsiodras and Spyridon Pournaras (Greece), Guenter Weiss (Austria), Jakub Hurych (Czech Republic), George Petrikkos (Cyprus), Tadeja Matos (Slovenia), Stefan Tyski (Poland), Theodoros Kalpakos (for the translation of the Greek guidelines).

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


## *Article* **Mid-Regional Pro-Adrenomedullin as a Prognostic Factor for Severe COVID-19 ARDS**

**Etienne de Montmollin 1,2 , Katell Peoc'h 3,4 , Mehdi Marzouk <sup>2</sup> , Stéphane Ruckly <sup>1</sup> , Paul-Henri Wicky <sup>2</sup> , Juliette Patrier <sup>2</sup> , Pierre Jaquet <sup>2</sup> , Romain Sonneville 1,2 , Lila Bouadma 1,2 and Jean-François Timsit 1,2,\***


**Abstract:** Mid-regional proadrenomedullin (MR-proADM) protects against endothelial permeability and has been associated with prognosis in bacterial sepsis. As endothelial dysfunction is central in the pathophysiology of severe SARS-CoV-2 infection, we sought to evaluate MR-proADM both as a prognostic biomarker and as a marker of bacterial superinfection. Consecutive patients admitted to the ICU for severe SARS-CoV-2 pneumonia were prospectively included and serum was bio-banked on days 1, 3, and 7. MR-proADM levels were measured blindly from clinical outcomes in batches at the end of follow-up. Among the 135 patients included between April 2020 and May 2021, 46 (34.1%) had died at day 60. MR-proADM levels on days 1, 3, and 7 were significantly higher in day-60 nonsurvivors. The area under the curve (AUC) of the receiver operating characteristic (ROC) curve (0.744, *p* < 0.001) of day-1 MR-proADM compared favorably with the AUC ROC curve of day-1 procalcitonin (0.691, *p* < 0.001). Serial MR-proADM measurements on days 3 and 7 may add prognostic information. After adjusting for CRP, LDH, and lymphocyte values, day-1 MR-proADM remained significantly associated with day-60 mortality. MR-proADM concentrations were significantly higher in patients with respiratory superinfections (on days 3 and 7) and bloodstream infections (on days 1, 3, and 7) than in patients without infection. Our results suggest that MR-proADM is a good predictor of outcome in severe SARS-CoV-2 infection and could be a useful tool to assess bacterial superinfection in COVID-19 patients.

**Keywords:** Mid-regional Pro-adrenomedullin; COVID-19; SARS-CoV-2; ARDS; biomarker; prognosis; superinfection

#### **1. Introduction**

Coronavirus Disease (COVID-19) can present with a wide range of clinical severity, from asymptomatic infection to acute respiratory distress syndrome (ARDS) and death. As of August 2022, 591.7 million cumulative cases and 6.5 million deaths due to COVID-19 have been reported to the World Health Organization [1]. Acute hypoxemic respiratory failure is the main reason for intensive care unit (ICU) admission, and critically ill COVID-19 patients show mortality rates up to 40% [2]. The progression from a mild to severe disease is multifactorial but appears mainly driven by significant inflammation and microvascular thrombosis with evidence of endotheliitis [3–5].

Adrenomedullin is an endogenous vasoregulatory peptide that has been shown to play a role in preserving the integrity and stability of the endothelium after severe infection [6]. Mid-regional proadrenomedullin (MR-proADM) is used as a surrogate marker for adrenomedullin, as its levels are directly proportional to adrenomedullin which has a short half-life. Higher levels of MR-proADM in septic critically ill patients have been associated with disease severity [7] and mortality [8,9]. MR-proADM has also shown potential as a

**Citation:** de Montmollin, E.; Peoc'h, K.; Marzouk, M.; Ruckly, S.; Wicky, P.-H.; Patrier, J.; Jaquet, P.; Sonneville, R.; Bouadma, L.; Timsit, J.-F. Mid-Regional Pro-Adrenomedullin as a Prognostic Factor for Severe COVID-19 ARDS. *Antibiotics* **2022**, *11*, 1166. https://doi.org/10.3390/ antibiotics11091166

Academic Editors: Sabine Danielle Allard and Johan Van Laethem

Received: 31 July 2022 Accepted: 26 August 2022 Published: 29 August 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

risk stratification biomarker for bacterial community-acquired pneumonia [10]. Finally, MR-proADM also appears to be a useful diagnostic tool for sepsis [11].

Being a marker of endothelial dysfunction, MR-proADM has been investigated in SARS-CoV-2 infection. Higher levels have been associated with mortality in the general population [12–15] and MR-proADM appeared to be a good risk stratification tool in the specific settings of the Emergency Room [16,17] and the ICU [18–20]. However, studies performed in the ICU setting used small patient samples with high heterogeneity of patient severity [21] and measured outcomes ≤ 30 days after ICU admission, which may be insufficient for such a study population.

Severe COVID-19 patients are at increased risk of bacterial superinfections, which contribute to ICU mortality [22,23]. These infections may be difficult to diagnose and lead to antibiotic overuse, which has led to a spread of antimicrobial resistance since the beginning of the pandemic [24]. In this context, MR-proADM could be an interesting biomarker for the diagnosis of bacterial superinfection. Surprisingly, no study evaluated MR-proADM as a diagnostic tool for bacterial sepsis in COVID-19 patients.

In this study, we aimed to evaluate MR-proADM as a prognostic biomarker in critically ill patients with severe SARS-CoV-2 pneumonia, and as a diagnostic tool for bacterial superinfection.

#### **2. Results**

#### *2.1. Population Characteristics*

Between April 2020 and May 2021, among 1294 admissions to the ICU of Bichat-Claude Bernard university hospital, 358 patients had severe SARS-CoV-2 pneumonia confirmed by polymerase chain reaction (PCR) and 135 patients with systematic bio-banking, and at least one measurement at day 1, 3, or 7 was included in the analysis (Supplementary Figure S1) and followed-up for 39 (13–126) days. Day-1, day-3, and day-7 MR-proADM measurements were available in 120 out of 135 (88.9%), 119 out of 126 (94.4%), and 83 out of 86 (96.5%) patients, respectively. Among the three proADM measurement time points, 69 patients completed three, while 49 patients completed two, and 17 patients completed one.

Included patients were females in 43 (31.9%) cases, with a median age of 62.7 (51.6–71.2) years and an admission Simplified Acute Physiology Score (SAPS) II score of 27 (21–39) (Table 1). ICU admission occurred 9 (7–12) days after symptom onset. On day 1, 11 (8.1%) patients were under veno-venous extracorporeal membrane oxygenation, 19 (14.1%) patients were under invasive mechanical ventilation, 78 (57.8%) had non-invasive oxygenation techniques (non-invasive ventilation, continuous positive airway pressure or high flow nasal oxygen), and 27 (20%) had standard oxygen support. Antiviral treatment consisted of remdesivir for 65 (48.1%) patients. Immunomodulating treatments consisted of steroids in 127 (94.1%) cases, and anti-IL6 (tocilizumab) or anti-IL-1 (anakinra) in three (2.2%) cases. During ICU stay, 59 (44%) patients required invasive mechanical ventilation, 27 (20%) vasopressors, and 30 (22%) renal replacement therapy. ICU, hospital, and day-60 mortality rates were 30.4%, 36.3%, and 34.1%, respectively. ICU and hospital lengths of stay were 10 (6–22) and 16 (10–31) days, respectively.

#### *2.2. Mid-Regional Pro-Adrenomedullin and Day-60 Mortality*

Values of MR-proADM and several other biological markers are presented in Table 1. In patients that died before day 60, MR-proADM levels were significantly higher at all time points compared to survivors (Table 1 and Figure 1a). The areas under the receiver operating characteristic curve (AUROC) of MR-proADM for predicting day-60 mortality were 0.74 on day 1, 0.73 on day 3, and 0.74 on day 7 (Figure 1b). When choosing a cut-point of 1 nmol/L for day-1 MR-proADM (median value of the study population), sensitivity and specificity for predicting day-60 mortality were 77.5% (95% confidence interval (CI) 62.5–87.7) and 68.8% (95% CI 57.9–77.9), respectively.


**Table 1.** Population characteristics at ICU admission.

Abbreviations: ICU, Intensive care Unit; SAPS, Simplified Acute Physiology Score; SOFA, Sequential Organ Failure Assessment; NIV, Non-invasive ventilation; HFNC, High Flow Nasal Canula; CPAP, Continuous Positive Airway Pressure; IMV, Invasive Mechanical Ventilation; ECMO, Extra-Corporeal Membrane Oxygenation, MR-proADM, Mid-regional proadrenomedullin; CRP, C-reactive protein; LDH, Lactate Dehydrogenase. *Antibiotics* **2022**, *11*, 1166 4 of 11

**Figure 1.** (**a**) MR-proADM concentrations according to day-60 survival; (**b**) ROC curves for day-60 survival of MR-proADM on days 1, 3, and 7. \* *p* < 0.05. **Figure 2.** Survival curves according to a cut-point of 1 nmol/L of MR-proADM on day 1. **Figure 1.** (**a**) MR-proADM concentrations according to day-60 survival; (**b**) ROC curves for day-60 survival of MR-proADM on days 1, 3, and 7. \* *p* < 0.05.

and *p* = 0.002 on day 7). In a landmark analysis on day 3, the delta between day-3 and day-1 MR-proADM was significantly associated with day-60 mortality (HR 1.20, 95% CI 1.01– 1.43, *p* = 0.04) (Table S1). On day 7, the delta between day-7 and day-1 MR-proADM concentrations was not associated with day-60 mortality (HR 1.26, 95% CI 0.90–1.76, *p* = 0.16).

Survival curves according to this cut-point are presented in Figure 2. Survival curves according to the same cut-point for day-3 and day-7 MR-proADM are presented in Figure

Survival curves according to this cut-point are presented in Figure 2. Survival curves according to the same cut-point for day-3 and day-7 MR-proADM are presented in Figure S2 and show a significant difference in day-60 mortality (log-rank test, *p* < 0.001 on day 3 and *p* = 0.002 on day 7). In a landmark analysis on day 3, the delta between day-3 and day-1 MR-proADM was significantly associated with day-60 mortality (HR 1.20, 95% CI 1.01–1.43, *p* = 0.04) (Table S1). On day 7, the delta between day-7 and day-1 MR-proADM concentrations was not associated with day-60 mortality (HR 1.26, 95% CI 0.90–1.76, *p* = 0.16). survival of MR-proADM on days 1, 3, and 7. \* *p* < 0.05. Survival curves according to this cut-point are presented in Figure 2. Survival curves according to the same cut-point for day-3 and day-7 MR-proADM are presented in Figure S2 and show a significant difference in day-60 mortality (log-rank test, *p* < 0.001 on day 3 and *p* = 0.002 on day 7). In a landmark analysis on day 3, the delta between day-3 and day-1 MR-proADM was significantly associated with day-60 mortality (HR 1.20, 95% CI 1.01– 1.43, *p* = 0.04) (Table S1). On day 7, the delta between day-7 and day-1 MR-proADM concentrations was not associated with day-60 mortality (HR 1.26, 95% CI 0.90–1.76, *p* = 0.16).

**Figure 1.** (**a**) MR-proADM concentrations according to day-60 survival; (**b**) ROC curves for day-60

*Antibiotics* **2022**, *11*, 1166 4 of 11

**Figure 2.** Survival curves according to a cut-point of 1 nmol/L of MR-proADM on day 1. **Figure 2.** Survival curves according to a cut-point of 1 nmol/L of MR-proADM on day 1.

On day 1, MR-proADM compared favorably to other prognostic biomarkers identified in the literature, AUROC for procalcitonin, ferritin, d-dimers, and C-reactive protein being 0.69, 0.63, 0.60, and 0.55, respectively (Figure 3). It also compared favorably to the Sequential Organ Failure Assessment (SOFA) score on day 1 (AUROC 0.65). The combination of MR-proADM and procalcitonin improved only slightly prognostic accuracy, with an AUROC of 0.76. In multivariate analysis and at all time points, after adjusting for C-reactive protein, lactate dehydrogenase, and lymphocyte count, MR-proADM remained significantly associated with day-60 mortality (Table 2).

#### *2.3. Mid-Regional Pro-Adrenomedullin and Bacterial Infections*

During ICU stay, 52 (38.5%) patients presented with bacterial nosocomial pneumonia with a delay of 7 (4.5–9) days, of which 36 (26.7%) patients were ventilator-acquired pneumonia. Bacteremia occurred in 34 (25.2%) patients during the same period, with a delay of 9.5 (6–13) days. MR-proADM levels on day 3 (1.2 (0.8–2.1) vs. 0.9 (0.7–1.5), *p* < 0.01) and day 7 (1.2 (0.9–2.5) vs. 0.9 (0.6–1.2), *p* < 0.01) were significantly higher in patients with bacterial pneumonia, but not on day 1 (1.1 (0.7–1.6) vs. 0.8 (0.7–1.5), *p* = 0.22). MR-proADM levels were significantly higher in patients with bacteremia on day 1 (1.3 (1.0–2.4) vs. 0.8 (0.7–1.3), *p* < 0.01), day 3 (1.4 (0.9–2.9) vs. 0.9 (0.7–1.5), *p* < 0.01) and day 7 (1.1 (0.9–2.6) vs. 0.9 (0.7–1.4), *p* < 0.01). MR-proADM levels according to the occurrence of bacteremia or bacterial pneumonia, on days 1, 3, and 7, are presented in Figure 4.

**Table 2.** Adjusted landmark analysis of the association of MR-proADM with day-60 mortality on

mained significantly associated with day-60 mortality (Table 2).

On day 1, MR-proADM compared favorably to other prognostic biomarkers identified in the literature, AUROC for procalcitonin, ferritin, d-dimers, and C-reactive protein being 0.69, 0.63, 0.60, and 0.55, respectively (Figure 3). It also compared favorably to the Sequential Organ Failure Assessment (SOFA) score on day 1 (AUROC 0.65). The combination of MR-proADM and procalcitonin improved only slightly prognostic accuracy, with an AUROC of 0.76. In multivariate analysis and at all time points, after adjusting for C-reactive protein, lactate dehydrogenase, and lymphocyte count, MR-proADM re-

**Figure 3.** ROC curves of the association of day-60 mortality with various biomarkers on day 1 and the SOFA score. Abbreviations: MR-proADM, Mid-regional proadrenomedullin; SOFA, Sequential Organ Failure Assessment; CRP, C-reactive protein. **Figure 3.** ROC curves of the association of day-60 mortality with various biomarkers on day 1 and the SOFA score. Abbreviations: MR-proADM, Mid-regional proadrenomedullin; SOFA, Sequential Organ Failure Assessment; CRP, C-reactive protein.



LDH, Lactate Dehydrogenase. Missing data imputed by multiple imputation. Hazard ratios are computed per one point of each variable. Abbreviations: MR-proADM, Mid-regional proadrenomedullin; CRP, C-reactive protein; LDH, Lactate Dehydrogenase. bacterial pneumonia, on days 1, 3, and 7, are presented in Figure 4.

**Figure 4.** (**a**) MR-proADM concentrations according to the occurrence of pneumonia on days 1, 3, and 7. (**b**) MR-proADM concentrations according to the occurrence of pneumonia on days 1, 3, and 7. Abbreviations: MR-proADM, Mid-regional proadrenomedullin. \* *p* < 0.05. **Figure 4.** (**a**) MR-proADM concentrations according to the occurrence of pneumonia on days 1, 3, and 7. (**b**) MR-proADM concentrations according to the occurrence of pneumonia on days 1, 3, and 7. Abbreviations: MR-proADM, Mid-regional proadrenomedullin. \* *p* < 0.05.

admitted to a large French COVID-19 reference center, we showed that MR-proADM concentrations were strongly associated with day-60 mortality. The prognostic accuracy of baseline MR-proADM was higher than commonly measured laboratory parameters and the SOFA score. We also showed that the AUROC for the prediction of day-60 mortality remained high on days 3 and 7, but that serial measurements might not be useful at all time points. When choosing a cut-point of 1 nmol/L, sensitivity and specificity for predicting day-60 mortality were 77.5% and 68.8, respectively, with good discrimination of survival curves. When evaluating the predictive accuracy of respiratory bacterial superinfec-

Regarding mortality risk stratification, our results are in line with previous studies performed in the ICU population showing AUROCs between 0.73 and 0.85 for 28-day mortality, with optimal cut-points between 1 and 1.8 nmol/L [18–20]. We used cut-points according to the MR-proADM distribution and previous studies to avoid overfitting [25]. Thus, the value of 1 nmol/L we chose was close to the median value of our sample and is in the range of published cut-points for mortality. Comparatively, in the general ward [12–15] and the emergency room [16,17] settings, MR-proADM also showed interesting

tion, MR-proADM concentrations were not significantly higher at baseline.

**3. Discussion**  *3.1. Main Findings* 

*3.2. Interpretation* 

## **3. Discussion**

#### *3.1. Main Findings*

Using high-quality prospectively collected data from critically ill COVID-19 patients admitted to a large French COVID-19 reference center, we showed that MR-proADM concentrations were strongly associated with day-60 mortality. The prognostic accuracy of baseline MR-proADM was higher than commonly measured laboratory parameters and the SOFA score. We also showed that the AUROC for the prediction of day-60 mortality remained high on days 3 and 7, but that serial measurements might not be useful at all time points. When choosing a cut-point of 1 nmol/L, sensitivity and specificity for predicting day-60 mortality were 77.5% and 68.8, respectively, with good discrimination of survival curves. When evaluating the predictive accuracy of respiratory bacterial superinfection, MR-proADM concentrations were not significantly higher at baseline.

#### *3.2. Interpretation*

Regarding mortality risk stratification, our results are in line with previous studies performed in the ICU population showing AUROCs between 0.73 and 0.85 for 28-day mortality, with optimal cut-points between 1 and 1.8 nmol/L [18–20]. We used cut-points according to the MR-proADM distribution and previous studies to avoid overfitting [25]. Thus, the value of 1 nmol/L we chose was close to the median value of our sample and is in the range of published cut-points for mortality. Comparatively, in the general ward [12–15] and the emergency room [16,17] settings, MR-proADM also showed interesting risk stratification capabilities, with risk prediction of ICU admission, need for invasive mechanical ventilation, or death. These results have been confirmed in a pooled analysis of 6 studies and 487 patients, where MR-proADM values were increased by 74% (95% CI 46–103) in COVID-19 patients with critical illness compared to those without [21]. We believe risk stratification to be of paramount importance for COVID-19 patients, as patients with the highest values may benefit most from anti-viral or anti-inflammatory therapies such as steroids, interleukin-6 receptor antagonists, or anti-JAK molecules. In the context of COVID-19 patients, we show that MR-proADM has better prognostic capabilities than procalcitonin. These results are in accordance with published literature regarding the general sepsis population, where MR-proADM appeared to be a prognostic biomarker superior to procalcitonin [26,27].

The analysis of serial measurements of MR-proADM on days 1, 3, and 7 brings valuable information. First, we showed that prognostic accuracy for day-60 mortality is equivalent at each time point, meaning that MR-proADM can be measured at any time during the first few days of ICU admission. Second, the delta between day-3 and day-1 MR-proADM was significantly associated with day-60 mortality, suggesting that MR-proADM could be used to monitor COVID-19 patients. The fact that the delta between day-7 and day-1 MR-proADM was not significantly associated with day-60 mortality may be related to a loss of power due to a smaller patient sample on day 7. In a general population of 89 COVID-19 patients, Gregoriano et al. also found that MR-proADM remained low during the whole follow-up period in survivors, whereas non-survivors had a step-wise increase from baseline [14]. Given the small patient sample, our results warrant further studies to evaluate MR-proADM as a monitoring biomarker of disease progression.

MR-proADM has been proven a performing biomarker for the diagnosis of sepsis, with a calculated AUROC of 0.91 in a recent meta-analysis, and an optimal cut-point value of 1–1.5 nmol/L [11]. It has also shown a good diagnostic accuracy in specific infections, such as complicated urinary tract infections [28] and spontaneous bacterial peritonitis [29]. Thus, we sought to evaluate the diagnostic performance of MR-proADM for bacterial superinfection. Interestingly, we found that MR-proADM concentrations were significantly higher in patients with bacteremia and bacterial pneumonia. Bacterial pneumonia is frequent in severe COVID-19 ARDS, with up to 44% of mechanically ventilated patients developing ventilator-acquired pneumonia [30]. These infections can be difficult to diagnose, as COVID-19 patients may exert persistent systemic inflammation, and chest x-rays

may be difficult to analyze due to the underlying viral pneumonia. While at the time of intubation less than 25% of patients present bacterial superinfection, ICU patients are frequently given systematic empiric antibiotic therapy. This strategy has led to an increase in antimicrobial resistance [24], and tools to identify patients with a high probability of bacterial superinfection are dearly needed. As such, we show that MR-proADM could be a useful marker to monitor bacterial superinfection in these patients, but our results need validation in larger cohorts.

#### *3.3. Strenghts and Limitations*

The strengths of our study are the prospective design and quality of collected data, including a follow up of 60 days, relevant for severe COVID-19 patients with extensive lengths of stay. The limitations of our study are: (1) a monocentric study design, (2) the small patient sample, despite being the largest published cohort in the literature, (3) the lack of external validation, and (4) the span of the study, covering different COVID-19 waves and SARS-CoV2 variants. Indeed, during each wave, the dominant variant presented distinct clinical and biological characteristics, including inflammatory response profile [31,32]. Hence, the prognostic accuracy of MR-proADM for each variant may have differed, but this has not been evaluated due to an insufficient patient sample.

#### **4. Materials and Methods**

#### *4.1. Study Population*

From April 2020 to May 2021, we included all adult patients that were admitted to the medical ICU of our hospital for severe SARS-CoV2 pneumonia and had had prospective serum bio-banking in the context of the OUTCOMEREA database. The bio-banking was conducted with the understanding and consent of each participant or surrogate. The OUTCOMEREA database has been approved by the French Advisory Committee for Data Processing in Health Research and the French Informatics and Liberty Commission (CNIL, registration no. 8999262). The database protocol was submitted to the Institutional Review Board of the French society of intensive care (CE-SRLF 22-76) on 12 September 2021. There were no exclusion criteria.

#### *4.2. Data Collection and Definitions*

The primary endpoint was the survival rate 60 days after ICU admission, and patients were followed-up to this time point or death. Data were prospectively collected at admission (demographics, chronic diseases, admission features, baseline severity indexes) and daily throughout the ICU stay (specific SARS-CoV2 treatments, need for invasive mechanical ventilation, need for vasopressors, need for renal replacement therapy, bacterial pneumonia and bacteremia, length of stay (LOS) and vital status at ICU and hospital discharge), using an anonymized electronic case report form. Severity of illness was graded at ICU admission with the use of the SAPS II [33] and the SOFA scores [34]. Immunodepression was defined as the use of long-term (>3 months) steroids, use of other immunosuppressant drugs, solid organ transplantation, solid tumor requiring chemotherapy in the last 5 years, hematologic malignancy, or HIV infection.

Serums were systematically collected on days 1, 3, and 7 (unless the patient was discharged from the ICU), allowed to clot at room temperature for 45 min, and then aliquoted and stored at −80 ◦C until assayed. Day-1, day-3, and day-7 MR-proADM concentrations were then measured blindly from clinical outcomes in batches at the end of follow-up, using an immunological assay with the TRACE technology (B·R·A·H·M·S MR-proADM KRYPTOR assay).

#### *4.3. Statistical Analysis*

Quantitative variables are presented as median, 1st, and 3rd quartiles, and compared between groups with the Mann–Whitney test or t-test, as appropriate. Qualitative variables are presented as frequency and percentage and compared with the Chi-square test or

Fisher exact test as appropriate. For the analysis of bacterial superinfections, only episodes occurring after each time point analysis were considered. The association of MR-proADM with day-60 mortality was determined using a Cox proportional hazard model, adjusted on biomarkers associated with the outcome in the literature. Landmark analysis was used to evaluate this association at each time point (day 1, day 3, and day 7). Missing data, when at random, were handled by multiple imputations [35]. All statistical analyses were carried out with SAS 9.4 (SAS Institute Inc., Cary, NC, USA). A *p*-value of 0.05 and lower was considered statistically significant.

#### **5. Conclusions**

Our results suggest that MR-proADM is a promising predictor of outcome in critically ill patients with severe SARS-CoV-2 infection, superior to procalcitonin or the SOFA score. Furthermore, serial measurements may help monitor disease progression. In the event of future COVID-19 waves, MR-proADM could be used both for risk stratification and triage of patients presenting with severe SARS-CoV-2 pneumonia and for monitoring of bacterial superinfection in COVID-19 ICU patients.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/antibiotics11091166/s1, Figure S1: Study flowchart; Figure S2: Survival curves according to a cut-point of 1 nmol/L of MR-proADM, on days 3 and 7; Figure S3: MR-proADM levels according to the occurrence of bloodstream infection or pneumonia, on days 1, 3, and 7; Table S1: Landmark analysis of the association of day-3 and day-7 MR-proADM with day-60 mortality, according to day-1 MR-proADM.

**Author Contributions:** Conceptualization, E.d.M., K.P., S.R., L.B., R.S. and J.-F.T.; methodology, E.d.M., K.P., S.R. and J.-F.T.; formal analysis, S.R.; investigation, E.d.M., M.M., P.-H.W., J.P. and P.J.; resources, K.P. and J.-F.T.; data curation, S.R.; writing—original draft preparation, E.d.M.; writing review and editing, E.d.M., K.P., M.M., S.R., P.-H.W., J.P., P.J., R.S., L.B. and J.-F.T.; visualization, E.d.M. and S.R.; supervision, J.-F.T.; project administration, S.R. and J.-F.T.; funding acquisition, K.P. and J.-F.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by a clinical research agreement from THERMO FISHER SCIEN-TIFIC (BRAHMS GmbH, Hennigsdorf, Germany).

**Institutional Review Board Statement:** The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board (or Ethics Committee) of the French society of intensive care (CE-SRLF 22-76); 12 September 2021.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

**Conflicts of Interest:** The authors received a research grant from Thermo Fisher to perform the study. Outside of the study, J.-F.T. declares a participation to the boards of Merck, Pfizer, BD, Gilead and Paratek, and lecture fees for Pfizer, Shionogi, BD and Merck.

#### **References**


## *Article* **Diagnostic Accuracy of Procalcitonin upon Emergency Department Admission during SARS-CoV-2 Pandemic**

**Stefano Malinverni 1,\* , Silvia Lazzaroni <sup>1</sup> , Maïa Nu ˇnez <sup>2</sup> , Thierry Preseau <sup>2</sup> , Frédéric Cotton <sup>3</sup> , Delphine Martiny <sup>4</sup> , Fatima Bouazza <sup>1</sup> , Vincent Collot <sup>1</sup> , Deborah Konopnicki <sup>5</sup> , Stéphane Alard <sup>6</sup> and Magali Bartiaux <sup>1</sup>**


## **Highlights:**


**Abstract:** Introduction: Procalcitonin is a marker for bacterial diseases and has been used to guide antibiotic prescription. Procalcitonin accuracy, measured at admission, in patients with communityacquired pneumonia (CAP), is unknown in the current severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic. Objectives: To evaluate the diagnostic accuracy of procalcitonin to assess the need for antibiotic treatment in patients with CAP presenting to the emergency department during the SARS-CoV-2 pandemic. Methods: We performed a real-world diagnostic retrospective accuracy study of procalcitonin in patients admitted to the emergency department. Measures of diagnostic accuracy were calculated based on procalcitonin results compared to the reference standard of combined microbiological and radiological analysis. Sensitivity, specificity, positive and negative predictive values, and area under (AUC) the receiver-operating characteristic (ROC) curve were calculated in two analyses: first assessing procalcitonin ability to differentiate microbiologically proven bacteria from viral CAP and then clinically diagnosed bacterial CAP from viral CAP. Results: When using a procalcitonin threshold of 0.5 ng/mL to identify bacterial etiology within patients with CAP, we observed sensitivity and specificity of 50% and 64.1%, and 43% and 82.6%, respectively, in the two analyses. The positive and negative predictive values of a procalcitonin threshold of 0.5 ng/mL to identify patients for whom antibiotics should be advised were 46.4% and 79.7%, and 48.9% and 79% in the two analyses, respectively. The AUC for the two analyses was 0.60 (95% confidence interval [CI] 0.52–0.68) and 0.62 (95% CI, 0.55–0.69). Conclusions: Procalcitonin measured upon admission during the SARS-CoV-2 pandemic should not guide antibiotic treatment in patients with CAP.

**Citation:** Malinverni, S.; Lazzaroni, S.; Nu ˇnez, M.; Preseau, T.; Cotton, F.; Martiny, D.; Bouazza, F.; Collot, V.; Konopnicki, D.; Alard, S.; et al. Diagnostic Accuracy of Procalcitonin upon Emergency Department Admission during SARS-CoV-2 Pandemic. *Antibiotics* **2022**, *11*, 1141. https://doi.org/10.3390/ antibiotics11091141

Academic Editors: Sabine Danielle Allard and Johan Van Laethem

Received: 7 July 2022 Accepted: 18 August 2022 Published: 23 August 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

**Keywords:** COVID-19; pandemics; procalcitonin; SARS virus; community-acquired infections; emergency service; hospital; pneumonia; viral; community-acquired pneumonia

#### **1. Introduction**

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection is associated with high rates of emergency attendance, hospitalization, and intensive care unit admission. Most cases present with mild symptoms, and a small proportion evolve to more severe presentations, such as oxygen-requiring pneumonia, acute respiratory distress syndrome, or fatal issues [1]. The most common symptoms are fever, fatigue, and dry cough. Less common symptoms include sputum production, anorexia, sore throat, chest pain, and nausea [1,2]. These symptoms are aspecific and are frequently observed in pneumonia caused by other viruses and bacteria. Differentiating between viral and bacterial pneumonia or bacterial coinfection of viral pneumonia is challenging because of the overlap in presentation between these entities [3,4]. The choice of administering broad-spectrum antibiotics to these patients is difficult, and overprescription of antibiotics has been reported in hospitalized patients with SARS-CoV-2 [5]. While a delay in antibiotic treatment of bacterial community-acquired pneumonia (CAP) is associated with increased mortality [6], systematic broad-spectrum antibiotic treatment of suspected bacterial CAP is associated with complications, side effects, and mortality [7].

Procalcitonin (PCT) is a prohormone produced by the thyroid gland. In response to bacterial infection, interleukin-6 (IL-6), tumor necrosis factor-α (TNF-α), and interleukin-1β (IL-β) induce PCT synthesis in extrathyroidal tissue [8] with a peak at 6 h from the onset of infection and a half-life of 24 h [9]. In most viral infections, increased interferon gamma production inhibits PCT synthesis, leading to bacterial specificity of PCT.

Studies have suggested that PCT is a useful serum biomarker that supports clinical decisions regarding antibiotic treatment in patients with CAP. Higher serum PCT levels are associated with a higher probability of bacterial disease [10]. Some clinical trials have reported that PCT can guide clinicians in decisions on empiric antibiotic coverage without incurring higher rates of adverse outcomes [11,12], while others have reported no differences in antibiotic use among patients with suspected lower respiratory infection when PCT values were integrated in the treatment decision [4].

Within the context of the ongoing pandemic, PCT has been associated with the severity and mortality of SARS-CoV-2 infections [13–15]. Meanwhile, evidence on the role of PCT in guiding antibiotic prescriptions remains insufficient [16,17].

This two-center case–control study aimed to evaluate the role of PCT in differentiating CAP with an indication for antibiotic treatment from other entities associated with a new radiological infiltrate and lower respiratory tract infection (LRTI) signs and symptoms.

#### **2. Material and Methods**

#### *2.1. Design*

This retrospective two-center case–control observational study assessed the diagnostic accuracy of PCT for antibiotic prescription guidance in patients with CAP admitted to the Emergency Department of Saint-Pierre and Brugmann University hospitals between 1 March 2020 and 31 October 2020. The ethics committee of each hospital (OM 007 and OM 026) approved the study protocol and waived the need for signed informed consent due to the retrospective design of the study (CE20-12-11 and CE 2022/132).

#### *2.2. Setting and Participants*

A sample size of 225 was calculated as sufficient to detect specificity and sensitivity of 0.8 with a two-sided type I error of 0.05 and a power of 80%, assuming a prevalence of 20% and H<sup>0</sup> of 0.6 for both sensitivity and specificity. For eligibility, we assessed a convenience sample of consecutive patient consultations and expected to achieve the previously calculated sample size within the study timeframe. Both centers routinely measure PCT levels of patients with suspected CAP upon admission. Patients with CAP having a serum PCT measurement performed within 24 h from ED admission and at least one viral and one bacterial investigation (a pair of hemocultures, sputum, or bronchoalveolar lavage) performed within 48 h from admission were included in this study. Minors, pregnant women, patients already on antibiotic treatment at the time of ED admission, and patients with an extrapulmonary site of infection diagnosed during their initial evaluation were excluded.

All enrolled patients had signs of acute infection (temperature of >38 ◦C, chills, altered mental status, and leukocyte count of >10,000/µL or <4000/µL). Moreover, all patients had at least one symptom of acute respiratory illness (cough, dyspnea, sputum production, tachypnea, pleuritic chest pain, ambient air oxygen saturation (SatO2) of <94%, or a loss of ≥4 SatO<sup>2</sup> points following a 1 min walking test) or at least one finding during auscultation (crackles and rales). All included patients had a new infiltrate on radiological imaging performed within 48 h from admission. Missing data were treated as missing in the analysis, and no imputation was performed.

#### *2.3. Outcome Measure and Analysis*

PCT concentrations were measured by technicians blinded to the clinical information using Lumipulse G B•R•A•H•M•S PCT immunoreaction cartridges on a Lumipulse G600II instrument (Fujirebio, Ghent, Belgium).

Bacteriological analysis included culture from a respiratory tract specimen (sputum or bronchoalveolar lavage) or a pair of blood cultures. Diagnosis of SARS-CoV-2 infection was based on the COVID-19 Ag Respi-Strip® (Coris Bioconcept, Gembloux, Belgium), followed by qRT-PCR in case of a negative result (RealStar® SARS-CoV-2 RT PCR Kit, Altona Diagnostics), both performed on a nasopharyngeal swab [18]. Additional microbiological tests were performed according to clinical presentation. Additional bacterial testing included the *Legionella pneumophila* urinary antigen test and serological tests for *Chlamydophila pneumoniae* and *Mycoplasma pneumonia.* Additional viral testing included immunochromatographic techniques for influenza and adenovirus, direct fluorescence antibody tests for parainfluenza viruses, and inoculation of three cell cultures. Upon request, for 13 patients, respiratory tract specimens were analyzed using a multiplex PCR system to detect an additional fourteen viral and three bacterial targets (Biofire™ Filmarray™, bioMérieux, Marcy l'Etoile, France).

Chest radiographs were classified as showing radiographic evidence of pneumonia whenever a new infiltrate was mentioned in the X-ray report. Thoracic CT scans were considered to have radiographic evidence of pneumonia when the summary report mentioned probable bacterial or viral pneumonia. Radiological assessment was performed by a radiologist blinded to the PCT results. During the study period, thoracic low-dose CT scans for suspected CAP were additionally classified through a simplified classification as having typical features of bacterial CAP, typical features of SARS-CoV-2 CAP, intermediate features of both viral and bacterial CAP, or features evoking a diagnosis different from CAP.

CAP was defined as a new infiltrate on chest radiological study in a patient presenting with LRTI signs and symptoms [19]. Patients were classified according to laboratory test results and interpretation of radiological images. CAP for which antibiotic treatment was recommended was defined as cases with a microbiological analysis positive for pathogenic bacteria or with a chest CT lung infiltrate typical of bacterial pneumonia. CAP for which antibiotic treatment was recommended included bacterial CAP and viral CAP cases with a documented bacterial coinfection. CAP cases in which antibiotics were discouraged were defined as cases with a typical viral pneumonia infiltrate on radiological studies, with a microbiological analysis positive for a pathogenic virus and a lack of any microbiological analysis positive for a bacterial pathogen or imaging test suggesting a possible bacterial coinfection.

Sixty-eight (19%) patients could not be classified using this method. These cases were classified, in addition to the previously categorized patients, by two independent

specialists blinded to the PCT results in a secondary analysis. These two specialists classified ambiguous cases according to clinical, microbiological, and radiographic results, antibiotic administration, and clinical evolution as CAP cases for which antibiotic treatment was recommended. In cases of disagreement, a third independent specialist provided a definitive classification. recommended. In cases of disagreement, a third independent specialist provided a definitive classification. Therefore, we retrospectively studied the accuracy of PCT in identifying CAP cases in which antibiotic therapy was recommended in two separate analyses, using the two aforementioned definitions for recommended or discouraged antibiotic treatment.

ambiguous cases according to clinical, microbiological, and radiographic results, antibiotic administration, and clinical evolution as CAP cases for which antibiotic treatment was

Therefore, we retrospectively studied the accuracy of PCT in identifying CAP cases in which antibiotic therapy was recommended in two separate analyses, using the two aforementioned definitions for recommended or discouraged antibiotic treatment. **3. Analysis**  *3.1. Procalcitonin among Groups* 

#### **3. Analysis** We compared the PCT distribution within the two aforementioned analyses using the Wilcoxon rank-sum test. PCT cutoff values described in the literature as thresholds for

#### *3.1. Procalcitonin among Groups* identifying bacterial infections and guiding antibiotic therapy [11,12] were used to cate-

We compared the PCT distribution within the two aforementioned analyses using the Wilcoxon rank-sum test. PCT cutoff values described in the literature as thresholds for identifying bacterial infections and guiding antibiotic therapy [11,12] were used to categorize CAP according to PCT values in four strata: <0.1 ng/mL, 0.1–0.249 ng/mL, 0.25–0.499 ng/mL, and >0.5 ng/mL. gorize CAP according to PCT values in four strata: <0.1 ng/mL, 0.1–0.249 ng/mL, 0.25– 0.499 ng/mL, and >0.5 ng/mL. *3.2. Accuracy of Procalcitonin for Identifying Antibiotic-Requiring CAP*  We calculated a nonparametric receiver-operating characteristic (ROC) curve for the

#### *3.2. Accuracy of Procalcitonin for Identifying Antibiotic-Requiring CAP* two aforementioned dichotomous analyses to study the diagnostic accuracy of PCT in

*Antibiotics* **2022**, *11*, x FOR PEER REVIEW 4 of 11

We calculated a nonparametric receiver-operating characteristic (ROC) curve for the two aforementioned dichotomous analyses to study the diagnostic accuracy of PCT in identifying antibiotic-requiring CAP. Sensitivity, specificity, negative predictive values, and positive predictive values were calculated using PCT cutoff values of 0.1 ng/mL, 0.25 ng/mL and 0.5 ng/mL. All statistical analyses were performed using Stata software version 16 (StataCorp, College Station, TX, USA). identifying antibiotic-requiring CAP. Sensitivity, specificity, negative predictive values, and positive predictive values were calculated using PCT cutoff values of 0.1 ng/mL, 0.25 ng/mL and 0.5 ng/mL. All statistical analyses were performed using Stata software version 16 (StataCorp, College Station, TX, USA). **4. Results**  During the study period, 476 patients presented with complaints related to CAP. Af-

#### **4. Results** ter application of inclusion and exclusion criteria, 359 (75.4%) patients were included in

During the study period, 476 patients presented with complaints related to CAP. After application of inclusion and exclusion criteria, 359 (75.4%) patients were included in the current analysis (Figure 1). the current analysis (Figure 1). Patient characteristics at inclusion are illustrated in Table 1. Table 2 illustrates the etiological pathogens identified.

**Figure 1.** Study-inclusion flowchart.

Patient characteristics at inclusion are illustrated in Table 1.

**Table 1.** Demographic and clinical characteristics at study inclusion.


Abbreviations: No., number; IQR, interquartile range; COPD, chronic obstructive pulmonary disease; PaO2/FiO2, arterial oxygen partial pressure to fractional inspired oxygen ratio.

Table 2 illustrates the etiological pathogens identified.

**Table 2.** Etiological diagnosis using microbiological and molecular methods.


Abbreviations: No., number.

In total, 77 cases (26.5%) were classified as having CAP with an indication for antibiotic treatment in the primary analysis and 100 (27.9%) in the secondary analysis. All 214 (100%) patients with microbiologically documented viral CAP had SARS-CoV-2-related pneumonia.

PCT concentrations were higher in the CAP group where antibiotics were recommended (0.22 ng/mL; IQR, 0.11–2.22 ng/mL) than the CAP group for which antibiotic were discouraged (0.19 ng/mL; IQR, 0.1–0.39 ng/mL; *p* = 0.01) (Supplementary Digital Content; Figure S1).

Antibiotic-requiring CAP was more prevalent in higher PCT strata. The prevalence of antibiotic-requiring CAP was 24% among patients with PCT <0.1 ng/mL and increased to 46% among patients with PCT ≥ 0.5 ng/mL (Supplementary Digital Content Table S1). Results were ambiguous in the intermediate strata.

*4.1. Accuracy of PCT for Identifying CAP Cases in Which Antibiotic Therapy Was Recommended* 4.1.1. Nested Cohort without Cases Classified According to Specialist Opinion

PCT performed poorly in identifying cases for which antibiotics were recommended (area under the curve [AUC], 0.60; 95% confidence interval [CI], 0.52–0.68) in the nested cohort, excluding patients classified according to specialist opinion (Figure 2).

Procalcitonin [confirmed and clinical bacterial vs.

*4.1. Accuracy of PCT for Identifying CAP Cases in Which Antibiotic Therapy Was Recommended* 

PCT performed poorly in identifying cases for which antibiotics were recommended (area under the curve [AUC], 0.60; 95% confidence interval [CI], 0.52–0.68) in the nested

4.1.1. Nested Cohort without Cases Classified According to Specialist Opinion

cohort, excluding patients classified according to specialist opinion (Figure 2).

**Figure 2.** Receiver-operating characteristic curve of serum procalcitonin for the diagnosis of microbiologically proven bacterial pneumonia. **Figure 2.** Receiver-operating characteristic curve of serum procalcitonin for the diagnosis of microbiologically proven bacterial pneumonia.

A PCT threshold of ≥0.25 ng/mL to identify CAP for which antibiotics were recommended resulted in sensitivity of 48.1% (95% CI, 36.5–59.7%) and specificity of 61.7% (95% CI, 54.8–68.2%) (Table 3). A PCT threshold of ≥0.25 ng/mL to identify CAP for which antibiotics were recommended resulted in sensitivity of 48.1% (95% CI, 36.5–59.7%) and specificity of 61.7% (95% CI, 54.8–68.2%) (Table 3).


Abbreviations: PPV, positive predictive value; NPV, negative predictive value.

≥0.5 41.6 (30.4–53.4) 82.7 (77.0–87.5) 32 (11.0) 45 (15.5) 37 (12.7) 177 (60.8) (34.3– 58.8) (73.8– 84.8) 29.0 75.9 A threshold of ≥0.5 ng/mL to identify CAP in which antibiotics were recommended resulted in sensitivity of 41.6% (95% CI: 30.4–53.4%) and specificity of 82.7% (95% CI: 77.0–87.5%).

46.4

79.7

>0.1 79.0 (69.7–86.5) 25.5 (20.3–31.2) 79 (22.0) 21 (5.9) <sup>193</sup> (53.8) 66 (18.3) (23.7– 34.8) (65.5– 84.4) 4.1.2. Complete Cohort including Patients Classified According to Specialist Opinion

PCT performed poorly in identifying CAP cases for whom antibiotics were recommended [AUC: 0.62 (95% CI: 0.55–0.69)] in the entire cohort (Figure 3).

87.5%).

confirmed and clinical viral CAP] n = 359

>0.25 50.0 (39.8–60.2) 64.1 (57.9–69.9) 50 (13.9) 50 (13.9) 93 (25.9) 166 (46.2)

≥0.5 43.0 (33.1–53.3) 82.6 (77.5–87.0) 43 (12.0) 57 (15.9) 45 (12.5) 214 (59.6)

Abbreviations: PPV, positive predictive value; NPV, negative predictive value.

mended [AUC: 0.62 (95% CI: 0.55–0.69)] in the entire cohort (Figure 3).

A threshold of ≥0.5 ng/mL to identify CAP in which antibiotics were recommended resulted in sensitivity of 41.6% (95% CI: 30.4–53.4%) and specificity of 82.7% (95% CI: 77.0–

PCT performed poorly in identifying CAP cases for whom antibiotics were recom-

4.1.2. Complete Cohort including Patients Classified According to Specialist Opinion

35.0 (27.2– 43.4)

48.9 (38.1– 59.8)

76.9 (70.6– 82.3)

79.0 (73.6– 83.7)

**Figure 3.** Receiver-operating characteristic curve of serum procalcitonin for the diagnosis of clinical bacterial pneumonia. **Figure 3.** Receiver-operating characteristic curve of serum procalcitonin for the diagnosis of clinical bacterial pneumonia.

A PCT threshold of ≥0.25 ng/mL to identify CAP for which treatment with antibiotics was indicated resulted in a sensitivity of 50.0% (95% CI, 39.8–60.2%) and a specificity of 64.1% (95% CI, 57.9–69.9%) (Table 2). A PCT threshold of ≥0.5 ng/mL to identify CAP for which treatment with antibiotic was indicated resulted in sensitivity of 43% (95% CI, 33.1– 53.3%) and specificity of 82.6% (95% CI, 77.5–87.0%). A PCT threshold of ≥0.25 ng/mL to identify CAP for which treatment with antibiotics was indicated resulted in a sensitivity of 50.0% (95% CI, 39.8–60.2%) and a specificity of 64.1% (95% CI, 57.9–69.9%) (Table 2). A PCT threshold of ≥0.5 ng/mL to identify CAP for which treatment with antibiotic was indicated resulted in sensitivity of 43% (95% CI, 33.1–53.3%) and specificity of 82.6% (95% CI, 77.5–87.0%).

We performed a sensitivity analysis nested on patients with SARS-CoV-2 to study the accuracy of procalcitonin in identifying a bacterial coinfection within CAP patients with SARS-CoV-2. Results were consistent with previous analysis. A threshold of ≥0.5 ng/mL to identify bacterial coinfections within SARS-CoV-2 CAP resulted in sensitivity of 40% (95% CI: 16.3–67.7%) and specificity of 80.4% (95% CI: 74.9–85.1%). PCT performed poorly in identifying bacterial coinfection (area under the curve [AUC], 0.60; 95% confidence interval [CI], 0.52–0.68) within SARS-CoV-2 CAP patients [AUC: 0.59 (95% CI: 0.44– We performed a sensitivity analysis nested on patients with SARS-CoV-2 to study the accuracy of procalcitonin in identifying a bacterial coinfection within CAP patients with SARS-CoV-2. Results were consistent with previous analysis. A threshold of ≥0.5 ng/mL to identify bacterial coinfections within SARS-CoV-2 CAP resulted in sensitivity of 40% (95% CI: 16.3–67.7%) and specificity of 80.4% (95% CI: 74.9–85.1%). PCT performed poorly in identifying bacterial coinfection (area under the curve [AUC], 0.60; 95% confidence interval [CI], 0.52–0.68) within SARS-CoV-2 CAP patients [AUC: 0.59 (95% CI: 0.44–0.76)].

#### 0.76)]. **5. Discussion**

In this multicenter retrospective study of 359 adults with CAP admitted to the ED during the SARS-CoV-2 pandemic, including 298 with microbiologically documented pathogens, no PCT threshold identified CAP for which antibiotic treatment was recommended.

Trials and meta-analyses, partially performed in EDs, suggested that PCT could tailor antibiotic prescription in CAP without increasing adverse outcomes [11,12,20]. Guidelines based on previous results of trials provided graded recommendations based on four tiers of PCT levels and discouraged the use of antibiotics for patients with PCT values ≤0.1 ng/mL while strongly recommending antibiotics in patients with PCT values ≥0.5 ng/mL. Other studies have suggested thresholds of 0.2 ng/mL to differentiate CAP from bronchitis [21]. Studies performed during the H1N1 pandemic indicated that PCT could help distinguish patients with a bacterial etiology from those with viral pneumonia [22,23]. However, a recent trial cast doubt on the ability of PCT to reduce antibiotic exposure in ED-diagnosed CAP [4]. The results from our study, carried out during the SARS-CoV-2 pandemic, further challenged PCT-based recommendations to guide antibiotic administration in CAP. In this cohort, withholding antibiotic treatment in patients with PCT levels ≤0.1 ng/mL

would have resulted in undertreating 17 (24.6%) cases of all CAP with a microbiological indication for antibiotic treatment. Moreover, the routine administration of antibiotics in our cohort in patients with PCT ≥0.5 ng/mL would have resulted in the inappropriate overtreatment of 37 (53.6%) of 69 patients (Supplement Table S1). These high rates of over- and undertreatment, as well as the low AUC observed for PCT as a tool to identify antibiotic-requiring CAP from viral CAP, do not seem to support the role of PCT in guiding antibiotic prescription in CAP within the context of the SARS-CoV-2 pandemic.

Our results are in line with previous studies [4,10,24] and with recent American guidelines for the management of CAP that recommend against routine PCT measurements to determine the need for initial antibacterial therapy [25,26]. The first ROC analysis, restricted to patients for whom a causative microbiological pathogen was documented, reported an AUC of 0.60 without any clear cutoff point that could identify patients for whom antibiotic therapy should be recommended. PCT performed poorly in identifying cases for which antibiotics were recommended, suggesting that PCT alone is not sufficient in CAP to guide antibiotic prescription in the ED. While being methodologically sound to restrict the analysis of patients with a proven microbiological etiology or a typical focal, bacterial condensation on CT scans, results from this population are not directly generalizable to clinical practice, as pathogens may be detected in less than 40% of CAP cases. Moreover, thoracic CT is not routinely performed for the diagnosis of pneumonia [27].

The secondary analysis, including both CAP with a straightforward etiologic diagnosis and pneumonia classified according to an independent blinded review, showed a similar AUC of 0.62 for the ability of PCT to identify CAP with an indication for antibiotic treatment. This analysis yielded similar results in a more pragmatic clinical scenario where the clinician has to decide whether to initiate antibiotic treatment for CAP, irrespective of whether a pathogen will be documented from the microbiological analyses.

PCT is higher in SARS-CoV-2 CAP than in CAP associated with other viral etiologies [28]. In addition, PCT had extremely low (<10%) positive predictive values for bacterial pneumonia in a multicenter study investigating empiric antibacterial therapy for suspected SARS-CoV-2 CAP [16]. The lower discriminatory performance of PCT compared to that previously reported in the literature might be explained by the hyperinflammatory status and cytokine storm caused by SARS-CoV-2, resulting in higher PCT concentrations than in other viral CAP [14], thereby lowering the discriminatory power of PCT for bacterial infections. High levels of IL-1b and TNF-α, together with high IL-6, have been reported in SARS-CoV-2 patients [29] which might increase PCT plasma levels. Alternatively, PCT trajectories rather than absolute values have been suggested as possible markers of bacterial infection without solid evidence in favor of this practice [14,30]. This finding is relevant, as PCT-driven antibiotic prescriptions in the current pandemic context might drive antibiotic overconsumption and expose patients and hospitals to potential harmful effects. Moreover, the reported findings are in line with evidence suggesting that PCT might be a marker of severity in SARS-CoV-2 pneumonia, which would hinder its ability to be used as a diagnostic tool to withdraw or withhold antibiotic prescription [13].

#### **6. Limitations**

This study has both strengths and limitations. First, the retrospective design may have introduced a selection bias, as 24.6% of the patients with CAP were not included in the analysis according to the exclusion criteria (Figure 1). Second, the low proportion of CAP cases requiring antibiotic treatment might have reduced the diagnostic performance of procalcitonin. Third, while some studies recommend serial PCT measurements to guide antibiotic prescription in critically ill patients [30], we studied PCT at the time of ED admission. Fourth, the two-center design may limit the generalization, and fifth, the emergence of new SARS-CoV-2 variants that might elicit distinct procalcitonin responses. Finally, immunosuppression, which might alter cytokine expression in response to SARS-CoV-2 infection, could influence procalcitonin response to different CAP etiologies, and was not accounted for in this analysis.

The study's main strength is its pragmatic design and the few exclusion criteria allowing us to interpret our results on PCT accuracy in identifying bacterial CAP in real-life ED conditions. The second strength is the high proportion of patients (83.0%) with a microbiologically documented etiology of CAP, which is higher than that found in previous PCT trials.

#### **7. Conclusions**

Procalcitonin measured at ED admission performed poorly as a guide for antibiotic prescription in CAP during the SARS-CoV-2 pandemic.

**Supplementary Materials:** The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/antibiotics11091141/s1, Figure S1: Box plot of serum procalcitonin comparing patients with viral and antibiotic-requiring CAP; Table S1: Prevalence of CAP class by procalcitonin strata.

**Author Contributions:** Conceptualization, S.M., S.L., M.N., T.P., F.C., D.M., F.B., V.C., D.K., S.A. and M.B.; methodology, S.M., S.L., M.N., T.P., F.B., V.C. and M.B.; software, F.C.; validation, S.M., S.L., M.N., T.P., F.C., D.M., F.B., V.C., D.K., S.A. and M.B.; formal analysis, S.M., S.L., M.N., F.B. and V.C.; investigation, S.M., S.L., M.N., T.P., F.C., D.M., F.B., V.C., D.K., S.A. and M.B.; resources, S.M., T.P., F.C., D.M., S.A. and M.B.; data curation, S.M., S.L., M.N. and F.B.; writing—original draft preparation, S.M. and S.L.; writing—review and editing, S.M., S.L., M.N., T.P., F.C., D.M., F.B., V.C., D.K., S.A. and M.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** The study was conducted in accordance with the Declaration of Helsinki, and approved by the Ethics Committee) of CHU Saint Pierre and CHU Brugmann (protocol code CE20-12-11 and CE 2022/132).

**Informed Consent Statement:** The two Ethics Committee waived the need for signed informed consent due to the retrospective design of the study.

**Data Availability Statement:** The complete dataset of the study will be available in the figsharer depository with the following DOI: https://doi.org/10.6084/m9.figshare.19283672.v1.

**Acknowledgments:** We acknowledge Genderini, Krebs, and Gabrowska for independently classifying ambiguous CAP cases. We acknowledge IFCSM and Amis du SIAMU for their unconditional support of this research.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Review* **Azithromycin through the Lens of the COVID-19 Treatment**

**Georgia G. Kournoutou \* and George Dinos \***

Department of Biochemistry, School of Medicine, University of Patras, 26504 Patras, Greece

**\*** Correspondence: gkurnutu@upatras.gr (G.G.K.); dinosg@upatras.gr (G.D.); Tel.: +30-2610996259 (G.D.)

**Abstract:** Azithromycin has become famous in the last two years, not for its main antimicrobial effect, but for its potential use as a therapeutic agent for COVID-19 infection. Initially, there were some promising results that supported its use, but it has become clear that scientific results are insufficient to support such a positive assessment. In this review we will present all the literature data concerning the activity of azithromycin as an antimicrobial, an anti-inflammatory, or an antivirus agent. Our aim is to conclude whether its selection should remain as a valuable antivirus agent or if its use simply has an indirect therapeutic contribution due to its antimicrobial and/or immunomodulatory activity, and therefore, if its further use for COVID-19 treatment should be interrupted. This halt will prevent further antibiotic resistance expansion and will keep azithromycin as a valuable anti-infective therapeutic agent.

**Keywords:** macrolides; azithromycin; virus; coronavirus; COVID-19; immunolides; antivirus

## **1. Introduction**

Azithromycin (Azi) belongs to the large family of macrolide antibiotics, an important class of first-line antimicrobial agents [1]. Azi belongs to the second generation of macrolides, as a semisynthetic derivative of erythromycin with a modified macrolactone ring with 15 members instead of 14 members as in erythromycin (Figure 1). Although Azi did not exhibit improved activity against Gram-positive bacteria compared to the mother compound erythromycin [2–4], it was selected for further development due to its enhanced pharmacokinetic profiles. In particular, it was selected for its high half-life time and the ability to accumulate at high levels within lung tissue [5–9]. Clarithromycin (Figure 1) is another key second-generation 14-membered macrolide with similar features and structure, and while it was initially included in a few trial schemes as a potential therapeutic drug for COVID-19, it was rapidly discontinued [10,11]. Azi, like most of the other macrolides, is not only known for its antimicrobial activity, but it also has additional actions as either antiinflammatory or antivirus agents. In this review we will present a summary of the existing literature data concerning azithromycin and will explain why it was initially hypothesized to have activity for COVID-19 treatment. Additionally, all studies on the use or non-use of azithromycin in the treatment of COVID-19 will be presented. Finally, we will discuss why Azi is not included anymore in therapeutic protocols and why its use must be interrupted to avoid increasing Azi pathogen resistance thereby maintaining the antibiotic as a useful therapeutic weapon for a longer time.

**Citation:** Kournoutou, G.G.; Dinos, G. Azithromycin through the Lens of the COVID-19 Treatment. *Antibiotics* **2022**, *11*, 1063. https://doi.org/ 10.3390/antibiotics11081063

Academic Editors: Sabine Danielle Allard and Johan Van Laethem

Received: 29 June 2022 Accepted: 1 August 2022 Published: 5 August 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

*Antibiotics* **2022**, *11*, x FOR PEER REVIEW 2 of 12

**Figure 1.** Molecular structure of the mother macrolide molecule erythromycin and its semisynthetic derivatives azithromycin (15-membered) and clarithromycin (14-membered). Blue and red colors in the structures represent modifications of the mother molecule (black). **Figure 1.** Molecular structure of the mother macrolide molecule erythromycin and its semisynthetic derivatives azithromycin (15-membered) and clarithromycin (14-membered). Blue and red colors in the structures represent modifications of the mother molecule (black). derivatives azithromycin (15-membered) and clarithromycin (14-membered). Blue and red colors in the structures represent modifications of the mother molecule (black).

#### **2. Methods 2. Methods**  The authors searched the PubMed and the Scopus database using the terms "Coro-

The authors searched the PubMed and the Scopus database using the terms "Coronavirus", "COVID-19", and "Azithromycin". The same search terms were used for searching the Clinical Trials database. The abstracts were screened and only the relevant articles were considered in the review. Articles from the Clinical Trials database that were limited by a small sample size or other criteria were declared not relevant, and therefore, were not considered. Articles published as recently as 15 April 2022 were included. The authors searched the PubMed and the Scopus database using the terms "Coronavirus", "COVID-19", and "Azithromycin". The same search terms were used for searching the Clinical Trials database. The abstracts were screened and only the relevant articles were considered in the review. Articles from the Clinical Trials database that were limited by a small sample size or other criteria were declared not relevant, and therefore, were not considered. Articles published as recently as 15 April 2022 were included. navirus", "COVID-19", and "Azithromycin". The same search terms were used for searching the Clinical Trials database. The abstracts were screened and only the relevant articles were considered in the review. Articles from the Clinical Trials database that were limited by a small sample size or other criteria were declared not relevant, and therefore, were not considered. Articles published as recently as 15 April 2022 were included.

#### **3. Azithromycin as an Antimicrobial Agent 3. Azithromycin as an Antimicrobial Agent 3. Azithromycin as an Antimicrobial Agent**

**2. Methods** 

The antimicrobial activity of azithromycin results from its binding with high affinity to the entrance of the ribosomal exit tunnel of prokaryotic 70S ribosomes and strongly inhibiting the bacterial protein synthesis [12–14]. According to the crystal structure data, it binds to the entrance of the nascent peptide exit tunnel and partially occludes it (Figure 2). The antimicrobial activity of azithromycin results from its binding with high affinity to the entrance of the ribosomal exit tunnel of prokaryotic 70S ribosomes and strongly inhibiting the bacterial protein synthesis [12–14]. According to the crystal structure data, it binds to the entrance of the nascent peptide exit tunnel and partially occludes it (Figure 2). The antimicrobial activity of azithromycin results from its binding with high affinity to the entrance of the ribosomal exit tunnel of prokaryotic 70S ribosomes and strongly inhibiting the bacterial protein synthesis [12–14]. According to the crystal structure data, it binds to the entrance of the nascent peptide exit tunnel and partially occludes it (Figure 2).

**Figure 2.** Structure of azithromycin in complex with the 70S ribosome carrying A-, P-, and E-site tRNAs. (**A**,**B**) Location of the ribosome-bound azithromycin (yellow) in the macrolide binding pocket at the entrance to the nascent peptide exit tunnel (NPET) of the 70S ribosome relative to tRNAs viewed as cross-cut sections through the ribosome. The 30S subunit is shown in light yellow, the 50S subunit is in light blue, the mRNA is in magenta, and the A-, P-, and E-site tRNAs are colored green, dark blue, and orange, respectively. The phenylalanyl and formyl-methionyl moieties of the A- and P-site tRNAs are shown as spheres [15]. **Figure 2.** Structure of azithromycin in complex with the 70S ribosome carrying A-, P-, and E-site tRNAs. (**A**,**B**) Location of the ribosome-bound azithromycin (yellow) in the macrolide binding pocket at the entrance to the nascent peptide exit tunnel (NPET) of the 70S ribosome relative to tRNAs viewed as cross-cut sections through the ribosome. The 30S subunit is shown in light yellow, the 50S subunit is in light blue, the mRNA is in magenta, and the A-, P-, and E-site tRNAs are colored green, dark blue, and orange, respectively. The phenylalanyl and formyl-methionyl moieties of the A- and P-site tRNAs are shown as spheres [15]. **Figure 2.** Structure of azithromycin in complex with the 70S ribosome carrying A-, P-, and E-site tRNAs. (**A**,**B**) Location of the ribosome-bound azithromycin (yellow) in the macrolide binding pocket at the entrance to the nascent peptide exit tunnel (NPET) of the 70S ribosome relative to tRNAs viewed as cross-cut sections through the ribosome. The 30S subunit is shown in light yellow, the 50S subunit is in light blue, the mRNA is in magenta, and the A-, P-, and E-site tRNAs are colored green, dark blue, and orange, respectively. The phenylalanyl and formyl-methionyl moieties of the A- and P-site tRNAs are shown as spheres [15].

Thus, Azi was considered as 'tunnel plugs' that inhibit the synthesis of every protein entering the exit tunnel [1,16]. However, more recent evidence demonstrates that Thus, Azi was considered as 'tunnel plugs' that inhibit the synthesis of every protein entering the exit tunnel [1,16]. However, more recent evidence demonstrates that Thus, Azi was considered as 'tunnel plugs' that inhibit the synthesis of every protein entering the exit tunnel [1,16]. However, more recent evidence demonstrates that macrolides selectively inhibit the translation of a subset of cellular proteins and that their action crucially depends on the nascent protein sequence and the antibiotic structure [17].

Recent studies have shown that the translation of many genes was arrested at a few distinct sites through the length of the gene after treatment with macrolide antibiotics. Analysis of the sites of the stops revealed the existence of specific sequence signatures that induce pronounced drug-induced translation arrest and lead to specific regulation of protein synthesis. Ribo-seq and toeprinting experiments have revealed leader ORFs of macrolide resistance genes carrying the +x+ motif, where + stands for positively charged amino acids lysine or arginine, and x stands for any amino acid [17,18].

Therefore, Azi emerges as a modulator of translation rather than as a global inhibitor of protein synthesis. In general, macrolide antibiotics are active mainly against Gram-positive bacteria and have a lower activity against Gram-negative bacteria [19,20]. Macrolides are very active against Gram-positive bacteria *Staphylococcus*, *Streptococcus*, and *Diplococcus*; among Gram-negative cocci, *Neisseria gonorrhea*, *Haemophilus influenzae*, *Bordetella pertussis,* and *Neisseria meningitis*; and are extremely active against various Mycoplasmas. Since its discovery Azi has been extensively used in the treatment of bacterial and mycobacterial infections of the respiratory, gastrointestinal, genitourinary, and cutaneous systems [21–23]. Azi is a member of the WHO list of essential medications [24] and is available in large quantities worldwide. Despite some mild side effects, including mainly diarrhea and QT prolongation, Azi is proven to be safe and cheap, and therefore, easily available to humans worldwide [21,22].

#### **4. Azithromycin as an Anti-Inflammatory Agent**

Beyond the antibacterial activity of azithromycin, and broadly most macrolides, their anti-inflammatory effects have been established and some of them have been used in chronic inflammatory diseases such as chronic rhinosinusitis, bronchial asthma, bronchiectasis, chronic obstructive pulmonary disease, cystic fibrosis, etc. [23–27]. Probably, the most striking example of their immunomodulation comes from diffuse panbronchiolitis, an idiopathic inflammation and progressively destructive disease of the bronchioles which can be converted from a lethal to a treatable disease with daily low-dose erythromycin or Azi [23,28,29]. This has been accredited to the ability of Azi to normalize the upregulated activities of IL-1β, IL-2, TNF, and GM-CSF [30]. Azi is rapidly absorbed after oral administration with a half-life time of approximately 3 days, leading to a high and constant tissue concentration [23,31]. As a result, Azi accumulates in human cells, including epithelial cells, and most notably in phagocytes where it has been concentrated hundreds to thousands of times with a focus on phagocyte lysosomes [9,31]. Its anti-inflammatory or immunomodulatory activity reported in several studies includes the most frequent effects on neutrophils, monocytes, and lymphocytes [27,29,32]. Among the usually measured immunological modified markers are the number of decreased neutrophils; the concentrations of neutrophil elastase; cytokines release; surface-expressed molecules (mainly Toll-like receptors); superoxide production; and cell homeostasis, mainly apoptosis and phagocytosis (Figure 3) [27,29,32,33]. Neutrophil function inhibition has been reported more frequently than eosinophil function inhibition. Azi stimulates neutrophil degranulation and phagocytosis-associated oxidative burst, mediated via modulation of ERK 1/2 signaling [19]. These initial stimulatory effects are followed by modulation of transcription factors activator protein (AP)-1, nuclear factor kappa B (NFκB), inflammatory cytokines, and mucin release, with overall anti-inflammatory effects [34].

Azi inhibits lipopolysaccharide-induced pro-inflammatory cytokines; increases phagocytosis by inhibiting AP-1 [35]; improves lysosomal resistance to oxidant challenge [36]; and promotes M2 polarization of macrophages (a process in which macrophages produce distinct functional phenotypes in response to specific microenvironmental stimuli and signals) [37–39]. Azi can also increase the phagocytosis of apoptotic epithelial cells [40] and neutrophils by macrophages [41] further supporting its anti-inflammatory activity. Studies have shown that part of the immunomodulatory effects of macrolides could be attributed to the impairment of TLR signaling by reducing the release of PAMPs (Pathogen-Associated

Molecular Patterns) and inhibiting TLR expression, either of dendritic cells or macrophages, thereby regulating the immune response [33,42]. *Antibiotics* **2022**, *11*, x FOR PEER REVIEW 4 of 12

**Figure 3.** Immunomodulatory effects of azithromycin. **Figure 3.** Immunomodulatory effects of azithromycin.

Azi inhibits lipopolysaccharide-induced pro-inflammatory cytokines; increases phagocytosis by inhibiting AP-1 [35]; improves lysosomal resistance to oxidant challenge [36]; and promotes M2 polarization of macrophages (a process in which macrophages produce distinct functional phenotypes in response to specific microenvironmental stimuli and signals) [37–39]. Azi can also increase the phagocytosis of apoptotic epithelial cells [40] and neutrophils by macrophages [41] further supporting its anti-inflammatory activity. Studies have shown that part of the immunomodulatory effects of macrolides could be attributed to the impairment of TLR signaling by reducing the release of PAMPs (Pathogen-Associated Molecular Patterns) and inhibiting TLR expression, either of dendritic cells or macrophages, thereby regulating the immune response [33,42]. Immunomodulatory effects, although similar to most therapeutic macrolides, are Immunomodulatory effects, although similar to most therapeutic macrolides, are likely to differ among them. Few studies have examined the anti-inflammatory effects of macrolides on more than one macrolide, and none of the human trials have explicitly compared different macrolides. Furthermore, the majority of these trials were conducted on healthy volunteers and/or Azi was administered in varying doses at a time [27,43–45]. Clinical investigations in CF patients, on the other hand, revealed that Azi, but not Clarithromycin, improves respiratory function and reduces pulmonary exacerbations [46,47]. Additionally, another study showed that Azi, but not clarithromycin or roxithromycin, inhibits IL-1alpha and IL-1beta production [48]. In general, azithromycin inhibits the synthesis of pro-inflammatory cytokines by both innate and adaptive immune cells, as well as the accumulation, adhesion, and death of pulmonary neutrophils [32].

likely to differ among them. Few studies have examined the anti-inflammatory effects of macrolides on more than one macrolide, and none of the human trials have explicitly compared different macrolides. Furthermore, the majority of these trials were conducted on healthy volunteers and/or Azi was administered in varying doses at a time [27,43–45]. Clinical investigations in CF patients, on the other hand, revealed that Azi, but not Clar-Azithromycin, like other macrolides, has very low activity against eukaryotes due to their low affinity for binding to eukaryotic ribosomes [1]. There are specific differences between eukaryotic and bacterial ribosomes (differences between rRNA bases or ribosomal proteins) that mediate the selectivity and toxicity of ribosomal drugs, as established by rRNA sequencing studies and X-ray crystallography [49].

#### ithromycin, improves respiratory function and reduces pulmonary exacerbations [46,47]. **5. Azithromycin as an Antivirus Agent**

Additionally, another study showed that Azi, but not clarithromycin or roxithromycin, inhibits IL-1alpha and IL-1beta production [48]. In general, azithromycin inhibits the synthesis of pro-inflammatory cytokines by both innate and adaptive immune cells, as well as the accumulation, adhesion, and death of pulmonary neutrophils [32]. Azithromycin, like other macrolides, has very low activity against eukaryotes due to their low affinity for binding to eukaryotic ribosomes [1]. There are specific differences between eukaryotic and bacterial ribosomes (differences between rRNA bases or ribosomal proteins) that mediate the selectivity and toxicity of ribosomal drugs, as established by rRNA sequencing studies and X-ray crystallography [49]. **5. Azithromycin as an Antivirus Agent**  Azi's antiviral effects have been demonstrated in vitro, albeit not all examples have been confirmed in vivo [32]. Since Azi mediated exacerbations in airway diseases, particularly in asthma [25,50], its effects were studied against viruses that cause such airway Azi's antiviral effects have been demonstrated in vitro, albeit not all examples have been confirmed in vivo [32]. Since Azi mediated exacerbations in airway diseases, particularly in asthma [25,50], its effects were studied against viruses that cause such airway infections such as rhinoviruses (RV). Azi inhibits RV replication and releases in primary human bronchial epithelial cells in vitro [51]. The AMAZES research, the largest clinical trial of a long-term macrolide on airway diseases, found that Azi reduced asthma exacerbations by 40% in vivo [25]. The mechanism is not known, but metagenomic analysis suggested that it could be related to an antibacterial effect versus *Haemophilus influenzae* and possibly its abundance in inhaled air [52–54]. Pre-treatment with azithromycin inhibits RV replication in CF bronchial epithelial cells, probably by amplifying the antiviral response mediated by the IFN pathway [55]. Additionally, Azi showed a reduction in H1N1 viral replication in A549 cells with IC<sup>50</sup> <sup>58</sup> <sup>µ</sup>M interfering with the internalization of viruses [56].In experiments with the Zika virus, within glial cell lines and human astrocytes, there was a reduction in viral growth and virus-induced cytotoxicity [57]. Equally, Azi inhibited Ebola

infections such as rhinoviruses (RV). Azi inhibits RV replication and releases in primary human bronchial epithelial cells in vitro [51]. The AMAZES research, the largest clinical trial of a long-term macrolide on airway diseases, found that Azi reduced asthma exacerreplication with EC<sup>50</sup> 5.1 µM and low toxicity; however, it did not boost survival in mice or guinea pigs when tested in vivo in a mouse model [58].

The precise mechanism of the antiviral activity of Azi remains unclear. Given that Azi is a weak base it can accumulate in acidic intracellular organelles such as endosomal vesicles and lysosomes [59]. In keeping with lysosomal accumulation, azithromycin causes lysosomal pH change [23]. This modified acidic environment caused by accumulation of Azi could also be responsible for uncoating enveloped viruses such as influenza and maybe coronavirus [59]. Data also suggest that the antiviral activity of Azi could be attributed to its ability to increase the expression of the epithelial interferon genes, leading to a reduction in viral replication [60].

Recently, Azi and spiramycin (a natural 16-membered ring macrolide) provided significant in vivo protection against enterovirus-A71 infection in mice [61]. Spiramycin was found to interfere with EV-A71 viral RNA synthesis, and it is likely that spiramycin and Azi function in concert after the viral entrance; thereby, inhibiting viral RNA synthesis either directly or indirectly.

#### **6. Azithromycin and Betacoronovirus**

From the beginning of the current SARS-CoV-2 pandemic, several drug screens were conducted in a rapid, urgent manner to evaluate potential candidate medications against this pathogen. The requirements were: to be approved, to be inexpensive, to be safe, and to be available as quickly as was feasible worldwide. Previous screens had recognized more than 90 drugs that inhibited SARS-CoV-2 viral replication with EC<sup>50</sup> nearly to 10 µM [62]. The tested drugs included protease inhibitors, ATPase proton pump inhibitors, viral protease inhibitors, compounds targeting the angiotensin pathway, and antibiotics. Azi tested in Vero E6 cells had an EC<sup>50</sup> of 2.12 µM and EC<sup>90</sup> equal to 8.65 µM, and selectivity index >19 [63], which is very comparable to the control antiviral-compound remdesivir (EC<sup>50</sup> = 1.65, EC<sup>90</sup> = 2.52), the first antiviral agent with proven clinical efficacy against SARS-CoV-2 in all clinical trials [64–66]. Azi was also discovered as a target in a bioinformatic screening investigation of potentially relevant pathways that may be turned into pharmaceutically acceptable forms. An initial study focused on two of the previous candidate molecules, hydroxychloroquine (HQL) and Azi, suggested a synergistic inhibition of SARS-CoV-2 replication in Vero cells at 5 and 10 µM concentrations, respectively [67,68]. This synergy was presented as a way to make hydroxychloroquine more effective at less hazardous concentrations. It was the first observational study suggesting that HQL, especially when combined with Azi, improved virological clearance [69]. However, because the data with Azi came from only six patients and the study was open-label and nonrandomized, no acceptable conclusions could be derived statistically [70]. This Azi-HQL combination was also investigated in nonhuman primates, but no substantial antiviral effect was observed in the five macaques given Azi in addition to hydroxychloroquine [71]. Furthermore, this initial favorable finding led to the immediate start of interventional trials to assess the efficacy of the COVID-19 therapy combination, as well as the efficacy of Azi with HQL. Hundreds of trials with Azi are listed on clinical trials.gov. Initially, Azi was prescribed as an adjunct to hydroxychloroquine, but later HQL was largely abandoned and Azi was used alone. From the beginning of 2020, decades of publications were released either favoring or discouraging the use of Azi, both with or without HQL [70]. According to them, Azi was initially favored with or without HQL [72–77] but at the same time more observations did not favor its use [78–86]. Since most of them were retrospective studies, it was clear that randomized control trials (RCTs) were necessary to clarify the previous controversial data. All these RCTs were integrated during the previous year and are presented in Table 1. The table gives an overview of the most currently published, up-to-date, peer-reviewed studies in the literature, in which the effect of Azi is evaluated. Although these RCTs in Table 1 differ in their outcomes and whether or not hospitalized patients are included, all of them suggested that azithromycin does not reduce hospital admissions, respiratory failure, or death when compared to conventional therapy, and therefore, Azi should no longer be

used to treat COVID-19. In a few words, all of them showed that, in hospitalized patients with COVID-19, azithromycin did not reduce the time to sustained clinical improvement or discharge. There is clearly no efficacy in terms of clinical status or mortality at the fixed time points used in all scientifically acceptable large trials.

**Table 1.** Published RCTs assessing the effect of Azithromycin on COVID-19 treatment.


Furthermore, according to Oldenburg et al., there was no significant difference in self-reported symptom absence 14 days after enrollment among patients assigned to azithromycin versus a placebo in their randomized controlled trial of single-dose oral azithromycin for outpatient COVID-19 [11]. This last finding supports earlier randomized clinical trials of azithromycin for COVID-19 in both outpatient and inpatient settings, none of which found azithromycin to be effective in treating COVID-19.

Given that azithromycin consumption during the pandemic was increased up to 3 times compared to the pre COVID period [91–93], it is important to reduce useless consumption, as it is an extremely dangerous practice, to avoid increasing antimicrobial resistance (AMR). Antimicrobial resistance (AMR) develops when bacteria, fungi, or viruses are exposed to antibiotics, antifungals, or antivirals leading to the development of a resistance to one or more antimicrobial drugs. As a result, the antimicrobials become ineffective and infections may persist. AMR is considered a serious and persistent therapeutic problem today being an economic and health burden. It is conservatively estimated that, in the US and Europe, 2.5 million people are affected by such infections each year and approximately 50,000 people die because of these infections [94]. The discovery of novel antibiotics has nearly halted over the past 30 years leading to the exhaustion of the pipeline reserve. The resistance of pathogens to antibiotics can be addressed with a rapid development of new effective and safe antibiotics [1,95]. Several studies have revealed a significant increase in drug resistance to azithromycin in some strains of *gonococci* [96]. Drug resistance to azithromycin is also increasing in *E. coli* [95].

Identifying strategies that can work to reduce the burden of bacterial AMR—either across a wide range of settings or those that are specifically tailored to the resources available and leading pathogen–drug combinations in a particular setting—is an urgent priority [97]. Since the prevalence of bacterial superinfection in COVID-19 is low [98], and unlike influenza [99–101], there is no preventive benefit against postviral pneumococcal and atypical pneumonia [98], it will be extremely helpful to avoid the useless consumption of any antibiotic prescription, specifically azithromycin, in COVID-19 treatment..

The outbreak of a pandemic led to a massive disruption of healthcare systems which overshadowed the misuse and incorrect prescription of some antibiotics [102]. Many COVID-19 patients received empirical antibiotic therapy for COVID 19 treatment in the early stages of pandemic since it was considered the safer option due to clinical uncertainty [93]. Control policies should be administrated in clinical practice regarding the use of drugs in the treatment of COVID-19. As AMR will be a major clinical problem, stewardship activities are necessary in the coming years to face the new-pandemic [102].

#### **7. Closing Remarks and Perspectives**

During the pandemic of coronavirus, antibiotics prescription was elevated without justification, partly because the medical community was unprepared for this burst and secondly because the clinical situation of patients changed dramatically each day after the initial day of infection. The administration of known antibiotics was considered to be the correct way of combating coronavirus but it soon became clear that there was no justification for the overuse of antibiotics as they did not decrease the risk of mortality in the patients who had no reason to receive this treatment [103].

To summarize, there is no scientific justification for the use of azithromycin in the treatment of COVID-19 up to now, and the only way to keep this antibiotic relevant in the future as a useful tool for combating pathogenic infections is to use it wisely, only after careful consideration and high expectations.

**Author Contributions:** G.G.K.: writing—review and editing and writing—original draft; G.D.: writing review and editing. All authors have read and agreed to the published version of the manuscript.

**Funding:** The publication of this article has been financed by the Research Committee of the University of Patras. G.G.K. is a Research Fellow supported by the Hellenic State Scholarships Foundation (IKΥ) through the Operational Program "Human Resources Development, Education and Lifelong Learning" co-financed by Greece and the European Union (European Social Fund, ESF).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** All relevant data are presented in the article.

**Acknowledgments:** The authors want to thank Yuri Polikanov for the kind preparation of Figure 2.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Abbreviations**

Azi: azithromycin; IL-1β: interleukin 1β; IL-2: interleukin 2; TNF: tissue necrosis factor; GM-CSF: granulocyte macrophage colony stimulating factor; TLR: Toll-like receptor; AP-1: activator protein-1; EV-A71: enterovirus A71; RV: rhinovirus; CF; cystic fibrosis; IFN: interferon; HQL: hydroxychloroquine; ERK1/2: extracellular signal related regulated kinase 1/2.

#### **References**


## *Article*
