記住我
Asmarawati TP, Rosyid AN, Suryantoro SD et al. The clinical impact of bacterial co-infection among moderate, severe and critically ill COVID-19 patients in the second referral hospital in Surabaya. F1000Res 10113 Publ 2021 Feb 1510113 Publ 2021 Feb 15 2021;
,9The role of co-infections and secondary infections in patients with COVID-19.]. Current guidelines advocate using empirical antibiotics for COVID-19 patients under mechanical ventilation [10Alhazzani W, Møller MH, Arabi YM, et al. Surviving Sepsis Campaign: guidelines on the management of critically ill adults with Coronavirus Disease 2019 (COVID-19). 2020.
], which have led to substantial antimicrobial use as high as 80% [4Langford BJ So M Raybardhan S et al.Antibiotic prescribing in patients with COVID-19: rapid review and meta-analysis.], despite the low described incidence of co-infection. Identifying community-acquired bacterial co-infection among patients with COVID-19 remain a challenge; however, the broad administration of antibiotics is not justified since overuse and wrong consumption of antibiotics could result in fatal effects for the patient and an increase in antimicrobial resistance.Using biomarkers, could allow higher feasibility for the detection of bacterial co-infections. Procalcitonin (PCT) is a well-known biomarker used clinically and which can be potentially used to differentiate bacterial from viral or fungal infections considering bacterial infections typically show higher PCT serum concentration in recent evidence and serves as a valuable tool in guiding the initiation of antibiotic treatment [11Bouadma L Luyt CE Tubach F et al.Use of procalcitonin to reduce patients’ exposure to antibiotics in intensive care units (PRORATA trial): a multicentre randomised controlled trial.]. In influenza pneumonia, the role of PCT in the early recognition of bacterial co-infection has been widely demonstrated [12Rodríguez AH Avilés-Jurado FX Díaz E et al.Procalcitonin (PCT) levels for ruling-out bacterial coinfection in ICU patients with influenza: A CHAID decision-tree analysis.]. Conversely, in COVID-19 patients, several studies have shown the association of PCT with the severity of illness [13Procalcitonin in patients with severe coronavirus disease 2019 (COVID-19): A meta-analysis.,14Shen Y Cheng C Zheng X et al.Elevated procalcitonin is positively associated with the severity of covid-19: A meta-analysis based on 10 cohort studies.], however, limited data are available that have investigated the role of PCT or C-reactive protein (CRP) in identifying co-infection.Therefore, the main objectives of our study are to determine the clinical value of PCT or CRP at ICU admission to identify bacterial co-infection in patients with COVID-19 pneumonia. Secondarily, to evaluate its role as definite indicators of prognosis.
DiscussionThe main finding of this study was that initial PCT and CRP levels did not have a suitable ability to predict community-acquired respiratory co-infection among ICU patients with COVID-19 pneumonia. The main application of PCT could be its ability to exclude at admission a bacterial co-infection when a cut-off <0.30 ng/mL was applied, corresponding with NPV>90%. Furthermore, elevated PCT levels at admission was independently associated with higher mortality, as opposed to CRP values.
In our cohort, the incidence of bacterial co-infection was 7.6%. Previous studies have reported similar low incidence [4Langford BJ So M Raybardhan S et al.Antibiotic prescribing in patients with COVID-19: rapid review and meta-analysis.,22Rawson TM Moore LSP Zhu N et al.Bacterial and Fungal Coinfection in Individuals with Coronavirus: A Rapid Review to Support COVID-19 Antimicrobial Prescribing.,23Garcia-Vidal C Sanjuan G Moreno-García E et al.Incidence of co-infections and superinfections in hospitalized patients with COVID-19: a retrospective cohort study.]. Moreover, we did not find an association between bacterial co-infection and mortality, although some studies have shown different results in this regard. Mussuuza et al. [6Musuuza JS Watson L Parmasad V Putman-Buehler N Christensen L Safdar N. Prevalence and outcomes of co-infection and superinfection with SARS-CoV-2 and other pathogens: A systematic review and metaanalysis.] carried out a systematic review of 118 studies, finding that co-infection was associated with an almost three times higher risk of death (OR = 2.84; 95% CI 1.42–5.66). However, in this meta-analysis, most of the studies were conducted in a case-mixed setting (ward and ICU) with few critical patients involved and including up to 27% of the pediatric population, making it difficult to compare with our results observed in a more homogeneous cohort. Furthermore, some studies did not clearly state differences between bacterial co-infection and superinfection when reporting the results [9The role of co-infections and secondary infections in patients with COVID-19.,24Nasir N Rehman F Omair SF. Risk factors for bacterial infections in patients with moderate to severe COVID-19: A case-control study.]. Although we did not find a significant association of co-infection and mortality, the presence of bacterial co-infection led to more prolonged ICU and hospital length of stay along with longer duration of mechanical ventilation, yielding higher use of healthcare resources, as seen consistently reported in previous data [8Asmarawati TP, Rosyid AN, Suryantoro SD et al. The clinical impact of bacterial co-infection among moderate, severe and critically ill COVID-19 patients in the second referral hospital in Surabaya. F1000Res 10113 Publ 2021 Feb 1510113 Publ 2021 Feb 15 2021;
,9The role of co-infections and secondary infections in patients with COVID-19.].The most frequent isolated pathogen causing co-infection was the methicillin-sensitive S. aureus. It is noteworthy to note that the high incidence of P. aeruginosa has been shown to be representative of a causative microorganism of co-infection in these patients. These findings coincide with the results observed by some investigators [25Relph KA, Russell CD, Fairfield CJ, et al. Procalcitonin Is Not a Reliable Biomarker of Bacterial Coinfection in People With Coronavirus Disease 2019 Undergoing Microbiological Investigation at the Time of Hospital Admission. 2019; :1–6.
]. In fact, this high incidence of P. aeruginosa co-infection has also been reported among critically ill patients with other viral infections such as severe influenza [26Loeches IM Schultz MJ Vincent JL et al.Increased incidence of co ‑ infection in critically ill patients with influenza.]. However, it is uncertain if the broadly administration of empirical antibiotics, corticosteroids and other immunomodulatory agents used against the COVID-19, could partly facilitate the respiratory co-infection of P aeruginosa among critically ill patients. The explanation of this high incidence of P. aeruginosa in patients with community acquired respiratory co-infection requires further investigation.Despite the low incidence of co-infection, the consumption of antibiotics on admission was as high as 80.7%. Since the pandemic began, this broad use of empirical antibiotics amongst COVID-19 hospitalized patients has been common practice [4Langford BJ So M Raybardhan S et al.Antibiotic prescribing in patients with COVID-19: rapid review and meta-analysis.]. Nonetheless, the indiscriminate use of antibiotics is not justified, even in a pandemic. Numerous efforts have been made to predict the presence of bacterial co-infection in patients with SARS-CoV-2 pneumonia at hospital admission in order to better guide the initial treatment with antibiotics and also to predict the evolution of the disease and how it leads to worsening outcomes. Consequently, it has been suggested that some biomarkers may be helpful for this purpose. Our study investigated the most common biomarkers used in routine clinical care. PCT and CRP levels were higher in the bacterial co-infection group, although these biomarkers were not independently associated with co-infection; hence it can be argued that PCT and CRP could not adequately predict bacterial co-infection. Moreover, the low ability of such biomarkers to accurately predict co-infection was confirmed through the ROC curve analysis, obtaining poor AUC results (0.56 for PCT and 0.54 for CRP). Additionally, a non-linear model performed using the CHAID analysis showed that neither PCT nor CRP levels were related to co-infection, which reinforced our findings.A recent multicenter cohort study conducted in United Kingdom including 1040 hospitalized adults with SARS-CoV-2 infection [25Relph KA, Russell CD, Fairfield CJ, et al. Procalcitonin Is Not a Reliable Biomarker of Bacterial Coinfection in People With Coronavirus Disease 2019 Undergoing Microbiological Investigation at the Time of Hospital Admission. 2019; :1–6.
], observed that PCT was not a useful biomarker or which provided an added clinical value to predict bacterial co-infection, reporting an AUROC 0.56, which coincide with the findings we reported in our cohort. The investigators found higher incidence of bacterial co-infection compared with our study, maybe due to the fact that one of the inclusion criteria was the requirement of blood and respiratory cultures at hospital admission. Additionally, two third of patients were diagnosed of co-infection with respiratory cultures, but only one third of the cohort needed mechanical ventilation, which implied that most of respiratory cultures results were taken from the upper respiratory tract trough sputum, hence, with more probability of overdiagnosis of bacterial infection, especially when the quality of sputum is poor. They recognized that rates of recorded microbiological investigation were low and culture positivity was high, whereby there may be a bias for preferential recording of positive microbiology results in their database. Conversely, almost 90% of patients with bacterial co-infection in our cohort of critically ill patients were mechanically ventilated, in whom the respiratory cultures were taken from the lower respiratory tract by tracheal aspirate or bronchoalveolar lavage. Moreover, the authors could not provide NPV to avoid misleading conclusions due to its highly selected cohort. Our study adds more valuable information in reporting a high NPV with a PCTPCT production is stimulated by bacterial endotoxin and proinflammatory cytokines (IL-6, IL-1β, and TNF-α) but inhibited by interferon‐gamma cytokine produced during viral infections. Given that PCT has relentlessly demonstrated its ability to discriminate between viral and bacterial pneumonia [27Schuetz Philipp BM Mirjam Christ-Crain Biomarkers to improve diagnostic and prognostic accuracy in systemic infections.,28Self WH Balk RA Grijalva CG et al.Procalcitonin as a Marker of Etiology in Adults Hospitalized with Community-Acquired Pneumonia.], it was faithfully believed that it could be helpful to distinguish the presence of bacterial co-infection in COVID-19 pneumonia. Nonetheless, several observational data have addressed this topic, with unexpected results as our findings. Some small studies [29Vanhomwegen C Veliziotis I Malinverni S et al.Procalcitonin accurately predicts mortality but not bacterial infection in COVID-19 patients admitted to intensive care unit.,30Roy A Powers HR Craver EC Nazareno MD Yarrarapu SNS Sanghavi DK. Antibiotic stewardship: Early discontinuation of antibiotics based on procalcitonin level in COVID-19 pneumonia.] observed that PCT levels were not useful to predict bacterial co-infection. In a larger retrospective study [31May M Chang M Dietz D et al.Limited utility of procalcitonin in identifying community-associated bacterial infections in patients presenting with coronavirus disease 2019.] including 2443 non-critical patients, PCT cut-offs (0.25 or 0.50 ng/ml) did not reliable identify co-infection. Also, Dolci et al. [32Dolci A Robbiano C Aloisio E et al.Searching for a role of procalcitonin determination in COVID-19: A study on a selected cohort of hospitalized patients.] investigated PCT and CRP levels in a small cohort of 83 hospitalized COVID-19 patients. They found that both biomarkers had AUCs Case in point, CRP is synthesized in response to cytokines IL-6 and IL-1 as an early but unspecific acute-phase inflammatory biomarker. However, it is well known that it has low specificity to predict bacterial etiology in respiratory infections [27Schuetz Philipp BM Mirjam Christ-Crain Biomarkers to improve diagnostic and prognostic accuracy in systemic infections.]. This is also in alignment with our findings. CRP levels were not an accurate reliable tool to predict bacterial co-infection. To date, there is no data regarding the power of CRP by itself to predict bacterial co-infection in the COVID-19 scenario.The predictive mortality value of biomarkers has also been investigated during the pandemic. Our data showed that PCT and CRP levels at admission were higher among non-survivors. However, only values of PCT >0.5ng/ml were independently associated with higher mortality, whereas CRP levels were not a prognostic factor. Our study found that most of the patients had normal PCT levels at ICU admission, which is consistent with past reported measures by other authors [29Vanhomwegen C Veliziotis I Malinverni S et al.Procalcitonin accurately predicts mortality but not bacterial infection in COVID-19 patients admitted to intensive care unit.,33Huang C Wang Y Li X et al.Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China.]. However, many studies coincided that high PCT levels upon admission are associated with worse clinical outcomes. The meta-analysis conducted by Lippi et al. showed that increased PCT values were associated with a higher risk of severe SARS-CoV-2 infection (OR, 4.76; 95% CI, 2.74–8.29). Su et al. [34Sex differences in clinical characteristics and risk factors for mortality among severe patients with COVID-19: a retrospective study.], in a retrospective study of 651 severe COVID-19 patients, found that PCT >0.1ng/mL on admission was independently associated with increased in-hospital death (OR 6.35, 95% CI 1.39-28.88, p = 0.017). Also, two recent metanalyses concluded that elevated PCT levels were significantly associated with increased mortality (24,30), suggesting that the best threshold may be PCT >0.5ng/ml. Likewise, all these findings are in line with our results, in which PCT ≥0.50 ng/mL was independently associated with ICU mortality (OR 1.5, 95% CI 1.21-1.88; p 29Vanhomwegen C Veliziotis I Malinverni S et al.Procalcitonin accurately predicts mortality but not bacterial infection in COVID-19 patients admitted to intensive care unit.,36Zeng Z Yu H Chen H et al.Longitudinal changes of inflammatory parameters and their correlation with disease severity and outcomes in patients with COVID-19 from Wuhan.,37Hematological and Inflammatory Parameters to Predict the Prognosis in COVID-19.]. Despite our findings suggested that PCT had a low discriminatory performance to distinguish bacterial co-infection, it might still aid in predicting poor prognosis among COVID-19 patients admitted to the ICU. This could be explained by the hyperinflammatory state induced by the SARS-CoV-2 infection with the consequent and extensive release of cytokines [38Profiling serum cytokines in COVID-19 patients reveals IL-6 and IL-10 are disease severity predictors.], resulting in greater production of PCT as a marker of a more severe viral infection, regardless of the presence of bacterial co-infection. While CRP values seemed to be associated with higher disease severity [39Yitbarek GY Walle Ayehu G Asnakew S et al.The role of C-reactive protein in predicting the severity of COVID-19 disease: A systematic review.,40Ahnach M Zbiri S Nejjari S Ousti F Elkettani C. C-reactive protein as an early predictor of COVID-19 severity.] and the development of acute respiratory distress syndrome [41Risk Factors Associated with Acute Respiratory Distress Syndrome and Death in Patients with Coronavirus Disease 2019 Pneumonia in Wuhan, China.], several studies have reported that CRP levels were not independently associated with death [35Huang I Pranata R Lim M.A Oehadian A Alisjahbana B. C-reactive protein, procalcitonin, D-dimer, and ferritin in severe coronavirus disease-2019: a meta-analysis.,37Hematological and Inflammatory Parameters to Predict the Prognosis in COVID-19.,41Risk Factors Associated with Acute Respiratory Distress Syndrome and Death in Patients with Coronavirus Disease 2019 Pneumonia in Wuhan, China.], in accordance with our research.Although PCT thresholds could be helpful to exclude bacterial co-infection regardless of its ability to predict prognosis, further research exploring the predictive ability of bacterial co-infection with combination of CRP and PCT may provide additional results to guide clinical decisions when there is clinical suspicion of bacterial co-infection in COVID-19 patients. There is recent data that suggests that the combination of PCT and CRP levels showed greater accuracy to predict bacterial co-infection in children with influenza H1N1 [42Li Z, Hou Q. Combination of procalcitonin and C- reactive protein levels in the early diagnosis of bacterial co- infections in children with H1N1 influenza. 2019; :184–190.
]. Furthermore, not only CRP and PCT levels need to be investigated in additional studies but also some new biomarkers such as inflammatory mediators in the bronchoalveolar lavage fluid, the detection of bacterial genetic code by molecular amplification techniques by polymerase chain reaction and omics which could provide promising results and useful information for clinical decision making in the diagnoses of bacterial co-infection among patients with suspected sepsis [43Póvoa P, Coelho L. Which Biomarkers Can Be Used as Diagnostic Tools for Infection in Suspected Sepsis ? 2021; :662–671.
].To the best of our knowledge, this is one of the largest studies evaluating the diagnostic accuracy of PCT and CRP to predict bacterial co-infection and its prognostic value among ICU patients with COVID-19 pneumonia. The main strengths of this analysis were the inclusion of a sizeable homogeneous cohort of critical COVID-19, the multicentric design across ten countries in Latin America and Europe, and the use of distinct statistical methods to evaluate and confirm our primary outcome. However, we acknowledge some limitations present in the analysis. First, the retrospective design nature may have introduced a selection bias that frequently arises in observational studies. Nevertheless, the study was conducted in a homogeneous cohort of critically ill patients, and potential confounding factors have been considered. Second, the exclusion criteria were based on PCT and CRP's missing data, which could result in higher inclusion of patients with bacterial co-infection. Missing data may introduce bias, primarily if such data are not appropriately handled. However, considering the pandemic context and the large dataset of the study, we could assume that the underlying mechanism might be due to missing data entirely at random. Therefore, in such a case, the results can be said to carry less bias and only affected by a reduced statistical power.
Even so, after the exclusion criteria, the sample size was considerably large and the final dataset for the analysis had a negligible rate of missing values (less than 5%) in most variables. Third, antibiotic administration before the ICU admission was not collected, which could have influenced a misclassification of some patients with negative respiratory cultures, hence the classification in the non-co-infection group. However, the observed bacterial co-infection incidence was similar to that reported in previous data. Several international guidelines during the pandemic, recommended to initiate empirical antibiotics at admission for severe COVID-19, which led to most patients receiving empirical antimicrobial treatment at hospital admission. In this context, the true incidence of co-infection could be difficult to know. Nevertheless, our study provides real data of clinical practices from an international cohort of critically ill patients and, therefore, the observed incidence could be considered very close to the real incidence of co-infection, globally. Additionally, prior antibiotics could influence the levels of biomarkers. Nonetheless, co-infected patients only took one day between the hospital arrival and ICU admission, therefore, the levels of biomarkers were taken early in the acute phase of the illness, which makes the results as valid due to a proper sample timing.
Fourth, biomarker levels could be influenced by immunomodulatory therapies such as tocilizumab and corticosteroids. Nevertheless, it has been observed that the decreased of CRP after tocilizumab and corticosteroid treatment is progressive and this reduction appears to be significant after the 72 hours of treatment initiation, whereas the effects on PCT might be less pronounced [44Surana PD, Nayak R, Sheikh A, Haldankar P, Kale J. Western India. 2022; :123–132.
, 45Raess N, Schuetz PP, Cesana-nigro N, et al. Running Title : Prednisone , effect on inflammatory markers.
, 46Cui Z, Merritt Z, Assa A, et al. Early and Signi cant Reduction in C-Reactive Protein Levels After Corticosteroid Therapy Is Associated With Reduced Mortality in Patients With COVID-19. 2021; :1–7.
]. In our study, biomarker levels were taken within the first 48 hours of hospital admission, hence, the influence of these treatments on biomarkers is low, yielding a reliable analysis. Fifth, attrition bias may occur when large number of participants drop out from a study, which might have an influence on the results. Most of excluded patients, dropped out due to missing data on biomarker levels, our variable of interest. Nevertheless, excluded patients had similar characteristics in terms of demographics, illness severity, time course of disease and outcomes, without relevant clinical differences that could influence the values of baseline biomarkers. Moreover, the incidence of bacterial co-infection in the excluded group was low (3.8%), hence, the exclusion of such subjects probably would not affect the results and the diagnostic performance of biomarkers to predict bacterial co-infection. Sixth, we recognize moderate imbalanced data on the incidence of bacterial co-infection leading to class imbalance when performing logistic regression analysis, as predictive models tend to be more biased towards majority classes. Notwithstanding, recent research suggests that logistic regression model in our analysis is reliable as imbalance ratio effect is limited with larger sample sizes, whereby the estimates get closer to the true value when the sample size increases [47Aryani H, Rahman A, Wah YB, Huat OS. Predictive Performance of Logistic Regression for Imbalanced Data with Categorical Covariate. 2021; 29:181–197.
,48Binti S, Lai S, Huda N, Binti N, Mohamad MB. Comparing the Performance of AdaBoost , XGBoost , and Logistic Regression for Imbalanced Data. 2021; 9:379–385.
].The latter suggests, that procalcitonin and C-reactive protein did not successfully identify bacterial co-infection among ICU patients with COVID-19 pneumonia. However, a threshold of procalcitonin <0.3 ng/ml may be helpful to rule out bacterial co-infection. Therefore, it allows clinicians to carry out the proper use of antibiotics and limit its overuse. Finally, elevated procalcitonin concentration on admission may be a helpful biomarker to predict prognosis. Further research should be conducted to verify our findings.
留言 (0)