Leukocyte Morphological Parameters as Prognostic Markers in CAP

Community-acquired pneumonia (CAP) has a high incidence of 2-8 cases per 1000 inhabitants per year and a mortality rate of around 5%. Mortality in admitted patients is10-25%, and is much higher in those requiring admission to intensive care (ICU). Most patients hospitalized for CAP respond satisfactorily to treatment, but 10-15% experience therapeutic failure and 6% may develop a rapid and progressive deterioration that can be life-threatening. Mortality from CAP occurs mainly in patients with therapeutic failure. The prognostic factors of mortality have been related to the germ-host binomial. Despite the emergence of antibiotic resistance, there are studies that show that mortality is more associated with patient-dependent factors than with germ resistance. For the management of CAP, it is essential to improve the microbiological diagnosis and the assessment of its severity, which will allow the choice of antimicrobial agents, establish the need for hospital admission, monitoring and care during admission, the appropriate time for hospital discharge, as well as post-discharge management. Determining the etiologic agent of CAP remains problematic, due to failure to detect the microorganism with the usual diagnostic methods. To establish the severity of CAP, severity scales have been created that allow predicting the patient's evolution. The best known are: PSI, CURB-65, SCAP score, and ATS / IDSA. The objectives of stratifying patients with CAP are multiple: defining which patients can be managed out-of-hospital, establishing the patients who require greater monitoring in intermediate respiratory care units (ICUs) or ICUs, establishing severity models to select patients for whom perform new diagnostic tests or therapeutic trials. The prognostic scales measure the physiological effect of the infection on the host but not the inflammatory response mechanisms against the microorganism.

A better understanding of the early inflammatory response may have clinical significance determined by greater therapeutic efficacy in patients with more severe CAP. Microbial invasion of lung tissue causes an inflammatory response aimed at limiting the progression of the infection and destroying the microorganism. The objective of this response is to facilitate the arrival of leukocytes and other inflammatory biomarkers to exercise their defense function. Among the biomarkers we can highlight C-reactive Protein (CRP), studied for the diagnosis and monitoring of inflammatory processes, and with which a certain relationship has been established with the severity of CAP. PCT, referred to as a sensitive marker of severity in bacterial infection and sepsis, and as a guide to adjust antibiotic treatment in patients with CAP. Pro-adrenomodulin (ProADM) has been associated with a prognostic marker in patients with sepsis and as a useful marker in the risk stratification of patients with CAP. Plasma levels of inflammatory mediators appear to correlate with the severity of sepsis or pneumonia.

The correct diagnosis of infections by the clinic, biochemical and microbiological markers can be expensive and time consuming. It is important to look for new and low-cost alternatives to evaluate this condition, with the aim of making an early and timely diagnosis and evaluation and instituting the best therapeutic strategy and follow-up. Among these new alternatives, the "Cellular Popular Data" (CPD) stand out, which are morphological parameters of different types of leukocytes. The CPDs of the XN analyzers (Sysmex Corporation, Kobe, Japan) report quantitative information on the morphological and functional characteristics of leukocytes. They are morphological parameters that characterize neutrophils, lymphocytes, and monocytes and classify them according to their volume, shape, granularity, and their nucleic acid content. The composition of activated cell membranes is different from that of resting cells, due to the expression of receptors and signaling molecules on their surface, in response to activation. This membrane is more sensitive to analyzer reagents, and more fluorescent dye can penetrate the activated cell, and bind to the cytoplasmic organelles and nucleic acids. The optical signals are different, which makes it possible to distinguish the morphological changes produced and that are directly related to the functionality of the cell. Activated neutrophils and monocytes are characterized by increased "deformability", mobility, and their ability to adhere, granulate, and release cytokines.

The CPD values reflect the morphological and functional transformation of these activated cells, offering very valuable information on the state of the cell and the patient at the time of obtaining the sample. Recent studies have shown that these parameters are valuable for the detection and control of infections and inflammation. Neutrophil structural parameters NE-SSC, (NEUT GI granularity index) and NE-SFL (NEUT RI reactivity index) could predict the appearance of later-stage infection markers, such as the presence of immature granulocytes, suggesting that they can be used to detect bacterial infections very early. It has been shown to be useful in acute bacterial infection, particularly in the differentiation of bacterial infection and early detection of sepsis. The mean volume of the neutrophil and its variability are more sensitive indicators of bacteremia than the leukocyte count and the percentage of neutrophils. Neutrophils in sepsis are larger and their volumes more heterogeneous than in the healthy population; the same happens to monocytes, larger and more heterogeneous than in localized infections, and in the ROC analysis they had the highest sensitivity for detection of sepsis. Lymphocyte CPDs show specific changes in viral infection, providing potential for differential diagnosis between viral and bacterial infection. Together, the CPDs support the differentiation between viral and bacterial infections, or between acute or evolving infections, and if there is an inflammatory condition without infection, with better diagnostic performance, especially in postsurgical bacterial infection, than the conventional parameters.

The available literature has focused on the potential usefulness of CPD in diagnosis, but we do not have data on prognostic value or its applicability to pneumonia. The classical inflammatory biomarkers are expensive and often not accessible in clinical practice, while the evaluation of new leukocyte markers through hematimetry analysis, cheaper and more accessible in clinical practice, can help to monitor the inflammatory response and recognize to patients who may have poor evolution. No study, to date, has related all these parameters (CPDs) together with the severe evolution of pneumonia or mortality, nor have clear cut-off points been established for each of them. We propose an observational study, in which these markers are related to the severity and prognosis of CAP, and a comparison between them, in addition to incorporating these biomarkers into the prognostic rules currently in use and seeing how their predictive capacity is modified.

Objectives

1. To evaluate de predictive ability of new biomarkers such as leukocyte morphology parameters (CPDs) and their short-term change over time, with poor evolution (defined by therapeutic failure, and/or the need for admission to high-monitoring units such as ICU or Intermediate Respiratory Care Units and/or mortality in 30 days) in patients admitted for community-acquired pneumonia.
2. To analyze the relationship between these biomarkers (CPDs) and the etiology of pneumonia.
3. To compare the predictive capacity at baseline and during evolution of these markers of leucocyte morphology (CPDs) with other frequently used biomarkers (CRP, PCT and pro-ADM) in the cohort of patients admitted for community acquired pneumonia (CAP).
4. To incorporate these new biomarkers (CPDs) to the clinical prognostic scores such as PSI, CURB-65, SCAP score and ATS/IDSA score with the aim of complementing them and better identifying patients at high risk of poor evolution (who require monitoring and more aggressive therapies) and identifying patients with less severe disease that can be managed at the outpatient level, creating a new predictive rule with higher sensitivity and specificity.

Design

• Study design: Multicenter prospective observational study with longitudinal follow-up up to 25 months (24 months of inclusion and follow-up up to 30 days) of patients who attended the emergency services of the participating hospitals for community-acquired pneumonia.
• Scope of study: multicenter study to be carried out in 3 hospitals of the public network, Galdakao-Usansolo Hospital (Galdakao), Basurto Hospital (Bilbao) and San Pedro de Logroño Hospital. The first is a general acute teaching hospital that serves a population of 300,000 inhabitants. The second, Basurto, is a general university acute hospital that serves a population of 450,000 inhabitants. The third, San Pedro de Logroño, is a general university acute care hospital with a reference population of around 320,000 inhabitants.
• Sample size: We estimate, from studies carried out by our groups in previous years, to include about 1000-1200 useful patients with pneumonia who will require hospital admission in the 24-month recruitment period. Studies on the development of predictive models establish that it is necessary to have at least 10 events of the dependent variable of interest (in our case the dependent variables would be those included in the poor evolution: therapeutic failure, admission to the ICU or to monitoring units such as Intermediate Respiratory Care Units and in-hospital mortality/30 days; and on the other hand the etiological diagnosis) for each independent variable included in the multivariate logistic regression model. Given that our intention is to initially include a limited but comprehensive number of variables in the multivariate logistic regression models (predictably, not less than 2-3 but not more than 5), we estimate that it will be necessary for 50 to 100 of these to occur. Events of the dependent variable in the sample from which we will derive the most complex prediction rule to ensure that the logistic regression model converges properly. Data previously collected tell us that the number of events of our dependent variable would be a 5-6% mortality at 30 days for hospitalized patients plus 7% therapeutic failure and 4-5% of patients admitted to discharge units monitoring, which makes an expected number of events of the primary variable to be about 100 events in the bypass sample. With the recruitment time shown in this protocol and with data from 100 useful patients, we believe that it is sufficient to meet the main objectives set forth and develop and validate the predictive models.
• Sampling: consecutive sampling where all new cases of patients diagnosed with CAP will be collected consecutively in the 3 participating hospitals during a period of 24 months, who meet the selection criteria and sign the informed consent until the indicated sample size is achieved.
• Ethical aspects: All participants will sign the informed consent after having discussed with the investigators the objectives, risks and potential benefits of the study. The rights of patients will at all times be protected by the Declaration of Helsinki. This project will have the approval of the Ethics and Clinical Research Committee. The terms relating to the protection of personal data will be updated in the subject information sheet (HIP / CI) in relation to Regulation (EU) No. 2016/679 of the European Parliament and of the Council of April 27, 2016 on the Protection of Data (RGPD) when the data protection agency incorporates it.
• Study variables: .Independent variables:

1. Variables related to the patient's condition (socio-demographic, comorbidities, physical examination, laboratory tests), duration of symptoms at the time of diagnosis.
2. Variables related to severity at the time of admission. The variables nedeed will be collected to calculate the risk class established by the Pneumonia Severity Index (PSI) scale, by the CURB-65 scale (Confusion, Urea nitrogen, Respiratory rate, Blood pressure, age> 65) by the SCAP scale and by the ATS / IDSA scale collected during the first 8 hours of diagnosis. Biomarker analysis (CPDs, PCR, Procalcitonin) will be performed at the time of diagnosis in all patients and 72 hours after starting treatment. For the pro-ADM analysis, a plasma extraction will be performed upon admission, 72 hours, which will be frozen at -70º, for later centralized analysis. The PCR will be measured by immunoturbidimetry on a Roche Modular platform (CRPLX, reference no. 3002039). Procalcitonin and Pro-adrenomedullin, by immunolumonometric analysis (Time Resolved amplified crytate Emission, Brahms Diagnostica, Germany). The PDCs markers will be measured using the Sysmex XN analyzer that reports as research parameters those related to leukocyte morphology (CPD), 6 numerical values for each subpopulation, which describe each cell type according to size (volume), complexity (cytoplasmic granules) and activation (nucleic acid content) as described below: A.For Neutrophils:

• NE-SSC, mean value for cytoplasmic granularity; NE-WX dispersion of NE-SSC values
• NE-SFL, mean value RNA / DNA content; NE-WY dispersion of NE-SFL values
• NE-FSC, mean cell volume; NE-WZ dispersion of NE-FSC values

1. For Lymphocytes:

• LY-X, mean value for cytoplasmic granularity; LY-WX dispersion of LY X values
• LY-Y mean value RNA / DNA content; LY-WY dispersion of LY-Y values
• LY-Z, mean cell volume; LY-WZ dispersion of LY-Z values C.For Monocytes
• MO-X, mean value for cytoplasmic granularity; MO-WX dispersion of MO-X values
• MO-Y, mean value of RNA / DNA content; MO-WY dispersion of MO-Y values
• MO-Z, mean cell volume; MO-WZ dispersion of MO-Z values

1. Variables related to evolution (variables that can be analyzed as independent in some cases and as dependent in others). .Early therapeutic failure (first 72 hours of treatment): when the clinical situation deteriorates and is accompanied by hemodynamic instability, the appearance or worsening of respiratory failure, the need for mechanical ventilation, radiological progression or the appearance of a new infectious focus. .Late therapeutic failure (after the first 72 hours of treatment): Admission to the Intensive Care Unit and / or admission to the Intermediate Respiratory Care Unit (ICU). .Complications established during its evolution: shock, respiratory failure (Po2 / Fio2 <250), renal failure (plasma creatinine> 2 mg), pleural effusion.
2. Variables related to the treatment administered. .Antibiotic administration prior to diagnosis and days of treatment .Class of antibiotic used at the time of diagnosis. .Adherence of antibiotic treatment to SEPAR regulations (categorical variable). .Time to go from intravenous to oral medication. .Use of invasive mechanical ventilation and time with this treatment. .Use of non-invasive mechanical ventilation and time with this treatment.
3. Variables related to bacteriological diagnosis: At the time of diagnosis, all patients will undergo a nasopharyngeal smear to perform an RT-PCR. In addition, the bacteriological diagnosis will include 2 blood cultures, the determination of urinary antigens of pneumococcus and legionella in the acute phase (BinaxNOW) and the Serological tests for atypical bacteria and viruses both during the acute phase and in remission or convalescence.

• Main dependent variables: <!-- -->

1. Main dependent variable: 30-day mortality, presence of therapeutic failure and need for ICU and / or ICU.
2. Microbiological diagnosis (bacterial / atypical / viral ..)
3. Mortality will be initially determined by means of a consultation established in that period of time and in the absence of a telephone interview 30 days after diagnosis. Deaths and their corresponding dates will be confirmed through hospital computer support and by public records of death certificates.

• Statistical analysis:

The data processing procedure of this project will be established by following the following steps:

1. A descriptive analysis of the recruited sample will be carried out.
2. To create predictive models: a. The collected sample will be divided into two subsamples: Derivation Group 1: The total sample will be divided into 60% for the derivation of the predictive models for each of the results studied; Group 2 of validation of the predictive rules: the models will be validated in this sample (40% of the sample). b. In Group 1, you will identify the risk factors of risk factors and create predictive models. The unit of study will be the patient (each patient can be included only once). A bivariate analysis will be carried out to study which variables, of the possible predictors, are related to each outcome parameter. Those variables with a p-value <0.20 and using variable selection techniques (LARS, LASSO, shrinkage), the potential predictors to be introduced in a multivariate logistic regression model will be identified. Those variables that are statistically significant will be chosen for the final scale. C. In the same way, statistical techniques of machine learning will be applied (random forest, neural networks, machine support vectors, nearest neighbor classification methods). d. The outcome variable will also be taken as temporary dependent variables (time until the event) and Cox regression models, competitive risk models and joint models will be used following the same process as that described above.
3. Goodness of fit and comparison of the developed predictive models. On the one hand, in the case of dichotomous dependent variables, the area under the ROC curve (AUC, discriminative capacity) will be calculated, considering a value> 0.80 as a robust predictive model. In addition to the AUC, the calibration of the model will be estimated through the Hosmer-Lemeshow test (good calibration for a p-value ≥0.05). Finally, the different proposed models will be contrasted by constructing ROC curves and comparing the respective ROC curves.
4. Internal validation of predictive models:

The validation of the predictive model will be carried out in the validation group (Group 2). The predictive model and the scale will be validated in the second subsample, making use of the predicted values obtained in the derivation sample.