logo SBA


Digital archive of theses discussed at the University of Pisa


Thesis etd-11152015-002416

Thesis type
Tesi di dottorato di ricerca
Thesis title
Risk stratification in patients with acute pulmonary embolism: the role of arterial blood gas exchange
Academic discipline
Course of study
tutor Prof. Palla, Antonio
  • gas exchange
  • prognosis
  • pulmonary embolism
Graduation session start date
Release date
Introduction - Currently, acute pulmonary embolism (PE) remains one of the leading causes of morbidity and mortality in the cardiovascular setting. PE is a great health problem and may present as a cardiovascular emergency. PE management and treatment are customized by mortality risk. Indeed, the European Society of Cardiology/European Respiratory Society (ESC/ERS) 2014 guidelines (GL) divide PE patients, initially, based on the presence of clinical parameters, such as systemic arterial hypotension or cardiogenic shock, subsequently, using first prognostic compositum score (pulmonary embolism severity index (PESI)), then right ventricular (RV) dysfunction signs, detected by an imaging test and cardiac laboratory biomarkers1. PESI is the most extensively validated prognostic score1, 2–5; however, owing to the complexity of the original PESI, that includes eleven variables, a simplified version known as simplified PESI (sPESI) has been developed and validated6,7.
According to current GL1 on the diagnosis and management of PE, prognostic assessment is required, concurrently with diagnosis, to define risk stratification and therapeutic decision-making. Proposed risk stratification, using these tools, divides PE patients in high, intermediate-high, intermediate-low and low early mortality risk, considering early death risk as PE correlated in-hospital or 30-day mortality rate GL1. Clinical symptoms and signs of acute RV failure, such as persistent arterial hypotension and cardiogenic shock, indicate a high risk of early death and those patients need a thrombolytic therapy1, if it is not contraindicated. Subsequently, the acute RV dysfunction findings, demonstrated by an imaging exams, are determinant factors of outcome in acute PE, as well as RV dysfunction or myocardial injury biomarkers1. Many studies have shown RV dysfunction, as assessed by trans-thoracic echocardiography (TTE), as one of the strongest predictors of early mortality in PE patients8-10: TTE demonstrates RV dilatation, akinesia/hypokinesia or pressure overload. However, the test is operator-dependent and not necessarily available around the clock in all institutions, moreover, TTE criteria of RV dysfunction are not definitely an universally standardized1. Tomographic pulmonary angiography (CTPA) is currently considered the diagnostic gold standard for PE1. In addition to its diagnostic role, the test has also the ability to highlight RV dilation and other radiological signs with prognostic significance1,11-13. Finally, brain natriuretic peptide (BNP) or N-terminal pro-brain natriuretic peptide (NT-proBNP) increased release reflects a RV pressure overload indirected sign and the severity of haemodynamic impairment after an acute PE. Their rise correlate with worse prognosis, such as elevated plasma troponin concentrations that is associated with high mortality both in unselected patients and in haemodynamically stable patients1,14–17.
Nowadays, arterial blood gas analysis (BGA) remains a first-level test in patients with suspected PE for the evaluation of acide-base status and gas exchange. BGA is usually used to describe the clinical presentation of patients with suspected PE or to explain pathophysiological mechanisms, but it does not play a major role as prognostic factor1. Only arterial hypoxaemia and low arterial oxygen saturation are contained in some prognostic score such as the Geneva and PESI ones, in association with many other markers1,5,18. In one study, hypoxemia was found to be an independent predictor of three-month all causes mortality in PE patients19. Currently, despite importance advances in diagnosis and treatment, assessment of risk and appropriate management of patients with PE remains a difficult task in clinical practice. This is particularly true in haemodynamically stable intermediate risk patients who, in clinical practice, are the most of the cases. Besides, there is an increasing interest in risk stratification using standardized blood tests, that do not require advanced skills and can be easily standardized.
Aim - Primary end-point of the study is to investigate whether a new parameter based on arterial blood gas exchange may predict in-hospital clinical deterioration. Clinical deterioration is defined as rescue thrombolysis, need for positive inotropic support, ventilation support or endotracheal intubation, cardiopulmonary resuscitation or need of recovery in intensive care units. This, in fact, might help risk stratification, mostly in the intermediate class of risk proposed by current GL. Secondary end-points consists in evaluating the arterial blood gas exchange data to predict in-hospital and one-month all-cause mortality.
Materials and methods. - In our study we evaluate consecutive patients with a CTPA confirmed diagnosis of pulmonary embolism, hospitalized in a tertiary and a secondary care units (Cisanello Hospital, Pisa and Lotti Hospital, Pontedera), coming from the emergency department, from April 2013 to August 2015. Patients were eligible if their PE were objectively confirmed according to current GL by the presence of filling defects within the pulmonary arterial bed on CTPA study1. In every patient the clinical assessment of the haemodynamic status was performed. Moreover sPESI, RV dysfunction signs detected by TTE, cardiac laboratory biomarkers (BNP and/or NT-proBNP, high-sensibility troponin (TrHS) were determined. Besides, we collected the values of room air or oxygen support arterial BGA parameters, performed in the first hour after emergency room admission and before performing the CTPA study. We followed our patients in their in-hospital iter, while, after demission, the patients have been followed by their family physicians for at least one month; in case of death all relevant data have been obtained from family physicians or hospital files or civil registry.
Results - A total of 371 PE patients were included during the study. We divided population study in 4 prognostic classes, according to GL. The normality of all prognostic markers was observed in only the 6.2% of study population (low-risk class), the most part of patients (89.8%) belonged to the intermediate risk class, mostly to the low-intermediate one, only 4.0% were hemodynamically instable (high-risk class). All study patients had a complete echocardiographic examination and more over the 50% of the cases showed at least a RV dysfunction sign. Overall, 319 patients (86%) presented a sPESI score ≥ 1. About biomarker alteration, Tr HS was elevated ( ≥ 60 pg * mL-1) in 184 patients (49.6%), and BNP ( ≥ 75 pg * mL-1) in 185 (49.9%). NT – proBNP (≥ 600 pg * mL-1) was measured only in 252 patients and it resulted altered in the 44.8% of the cases. About BGA, we measured all parameters, but we chose PaO2 st./FiO2 and PaO2 /FiO2 as possible prognostic markers, because of their simplicity and immediacy of calculation. We fixed a cut-off value of 300 mmHg, because a relative cut-off in PE patients is not available in the literature. This threshold, generally used to define acute lung injury (ALI), indicates a condition of severe respiratory failure. In overall population, PaO2/FiO2 was altered in 43% of the cases, while PaO2st./FiO2 is low in 220 patients (59.3%). PaO2/FiO2 and PaO2 st./FiO2 were significantly associated with all other considered clinical or blood markers or RVD detected by TTE. During the follow-up (one month), 34 patients died (9.2%), of these 23 (6.2% of study population, 67.7% total death), during hospital stay. Clinical deterioration presented in 84 patients (22.6%), within hospital stay. Univariate cox regression analysis showed a significantly elevated risk (HR = 4.3; p < 0.001) in patients with altered troponin values considering the compositum score, but this was also true correlating PaO2 st./FiO2 (HR = 3.3; p = 0.007). Moreover, the risk based on BNP and NT pro-BNP (respectively HR = 3.9 , p = 0.001 and HR = 3.7, p = 0.001) is high. Concerning echocardiogram signs, the contemporary of RV free wall akinesia or hypokinesia, RV enlargement and pulmonary hypertension findings increased more than 3 times the risk of in-hospital mortality (HR = 3.4, p < 0.001). sPESI class also elevated risk (HR = 1.5, p < 0.001). However, in multivariate logistic regression analisys, including the described variables and age quartiles, no parameters, including troponin and other biomarkers, was resulted independently significant in our population, considering the compositum score. Univariate cox regression analysis showed similar results considering in-hospital or one-month mortality, while in multivariate logistic regression showed only sPESI was independent predictor of in-hospital mortality (HR = 2.1, p = 0.002). A significant difference in pH values only between high-risk and not high risk populations was observable. Besides, high-risk prognostic class was characterized by a significant increased fraction of inspired oxygen and a greater arterial lactate values, compared to other prognostic classes. PaO2/FiO2 and PaO2 standard/FiO2 values demonstrated a decreasing trend, considering increased risk of adverse events. Elaborating Kaplan – Meier curve by PaO2/FiO2 and PaO2 standard/FiO2 values, altered BGA values correlated with worse prognosis patients. Besides, considering only intermediate-risk patients, analysis showed that PaO2/FiO2 and PaO2 standard/FiO2 alteration were significantly different in better and worse prognosis patients (p < 0,05), both in low-intermediate and in high-intermediate risk classes. Finally, calculating positive predictive value (PPV) for primary and secondary end-points, it was similar to that obtained in other studies109, 110, 111 concerning “classical” PE prognostic markers, such as troponin, BNP and NT-pro BNP: PO2/FiO2 PPV is 37.3% for primary end-point and 11.2% and 13% for secondary ones, while PO2 st./FiO2 PPV is 32.3% for compositum score and respectively 10% and 12.7% for in-hospital and one-month mortality. However, it is more interesting negative predictive value (NPV). PO2/FiO2 and PO2 st./FiO2 NPV are 88.6% and 91.4% for compositum score, while for in-hospital mortality they are respectively 97.6% and 99.3%, they are 93.8% and 96% for one-month mortality. It means that if PO2/FiO2 or PO2 st./FiO2 are normal, patients probably will not have a bad clinical course.
Discussion - Concurrently, with the diagnosis of PE, prognostic assessment is required in order to stratify patients for risk of early death and, consequently, chose the best available therapeutic decision-making. Prognostic markers actually proposed by GL are the following: hemodynamic stability, prediction rules, cardiac biomarkers and RVD findings detected by an imaging-test1. Hemodynamic instability detects early high-risk mortality patients, who need a specific diagnostic-therapeutic pathway. On the contrary, patients with hemodynamic stability can be divided in early-death-risk classes, evaluating others prognostic markers1. sPESI < 1 provides an accurate identification of low death risk patients, while cardiac biomarkers and echocardiography further classify patients with sPESI > 1 in low-intermediate and high-intermediate risk classes. This prognostic classification is based on many studies which had the power to correlate each marker to adverse events. The main strength of sPESI lies in the reliable identification of patients at low-risk for one-month mortality and, also in our study, we obtained the same result. Besides, RV dilatation, akinesia/hypokinesia or pressure overload collected by TTE have been identified as independent predictors of an adverse outcome92, even if this type of skill is operator-dependent, not always available any time in any hospital and diagnostic criteria are not always evaluated in a standard way. In the population we studied RVD was significantly more frequent both in patients dying in acute phase or within one month. In particular, all patients dead during in-hospital stay showed RVD signs. In the literature TTE proved a low positive predictive value 94,95, even if TTE would be a very sensible skill. Besides, the increased release of some cardiac biomarkers, such as BNP or NT-proBNP, correlate with worse prognosis15, 97, 110, 111. At the same time, elevated plasma troponin concentrations are associated with high mortality both in unselected patients and in haemodynamically stable patients1,14–17. In the patients we studied we obtained this type of correlation, when considering the compositum score (including clinical deterioration and in-hospital death) and in-hospital or one-month mortality. A meta-analysis, covering a total of 2000 patients, showed elevated cardiac troponin I or T concentrations in about 50% of the patients with acute PE99, however, other reports have suggested a limited prognostic value of elevated troponins in normotensive patients100. The reported positive predictive value of troponin elevation for PE-related early mortality ranges from 12 – 44%, while the negative predictive value is high. As it happens in the case of troponin, BNP and NT-proBNP, the positive predicted value obtained in the case of PaO2/FiO2 and PaO2 st./FiO2 is almost low, while the negative predicted value is high, both for adverse and mortality events. In our study, the univariate analysis reflects all these correlations and highlights same results even in main BGA parameters (PaO2/FiO2 e PaO2 st./FiO2). The distribution of altered PaO2/FiO2 and PaO2 st./FiO2 values in PE patients with adverse or benign outcomes, is similar to the distribution of “classical” prognostic markers, considering primary end-point (clinical deterioration and in-hospital death) and secondary ones (in-hospital and one-month death). However, a multivariate regression analysis, including all the markers and quartiles for patients age, showed that sPESI was the only independent variable which predicts in-hospital mortality risk. This suggests a possible interaction towards classical markers and PaO2/FiO2 and PaO2 st./FiO2 values: Correlating PaO2/FiO2 and PaO2 st./FiO2 values with the commonly utilized markers, we obtained significant statistic concordance. Inside the intermediate class, the PaO2/FiO2 and PaO2 st./FiO2 alterations increase considerably the risk of adverse events and it allows speculate speculate a possible new stratification of the low-intermediate and high-intermediate classes. Moreover gas exchange data show significant differences among the four patients prognostic classes: the high risk class is characterized by increased fraction of inspired oxygen, greater arterial lactate values (Table 17). At the same time PaO2/FiO2 and PaO2 st./FiO2 show a decreasing trend from low to high risk class; these data could reflect a greater gas exchange inefficiency in the high risk class. The new prognostic parameter (PaO2 st./FiO2) that we introduced reflects the efficiency of blood exchanges, it is commonly calculated in clinical practice since it is really simple to measure. Perhaps, it has not employed insofar, since the classically utilized prognostic markers derive from a “cardiologic” culture, that makes use mostly of echocardiogram and biochemical markers originally investigated in the left heart pathologies. However, a good correlation is present between both PaO2/FiO2 and PaO2 st./FiO2 findings, as well as troponin and echocardiography, with clinical deterioration and PE-related deaths.
Limits of the study
The main limit of this study is that multivariate regression analysis revealed PaO2/FiO2 and PaO2 st./FiO2 were not independent predictor of a worse clinical course, however we also need to consider that this result could be the expression of classical markers and gas exchange data interaction. In fact, cardiac laboratory biomarkers and gas exchange data could be different expressions of the same pathofisiological process.
On the other hand, it is possible that the sPESI prognostic strength is so sobvious that the other markers are not statistically significant.
Conclusion - The use of the new parameter, PaO2/FiO2 and/or PaO2 st./FiO2 values, may help stratify patients with PE for the risk of early death. Indeed, when this ratio is < 300 mmHg patients are at high risk and their in-hospital prognosis is worse. On the contrary, a ratio of PaO2/FiO2 and PaO2 st./FiO2 > 300 mmHg seems a sign of better clinical course. Such patients, therefore, could be candidate for a short-term hospitalization or for an outpatient treatment of PE. In the clinical practice, the following flow chart could be used. Once the patients are included in a prognostic GL classe by using sPESI and the other markers, PaO2/FiO2 and PaO2 st./FiO2 values could provide a higher probability of worse or better clinical course. Further data are however necessary to assess whether PaO2/FiO2 and/or PaO2 st./FiO2 values, alone or in combination with clinical, blood or imaging findings of RVD, can be used to manage and treat PE patients.