In a prospective cohort study, we approached consecutive eligible patients for enrollment between January 1, 2017, and May 31, 2023.

We screened outpatients presenting with symptoms of acute PE at the Emergency Department of Ramón y Cajal Hospital, Madrid, Spain (Fig. 1). Eligible patients were required to have symptomatic PE confirmed by objective testing. We excluded patients with a diagnosis of COVID-associated PE. Patients with haemodynamic instability (defined as cardiogenic shock, systolic blood pressure <90mmHg, or need of inotropic support), those in whom thrombolytic therapy was indicated by decision of attending physician, and those in whom CRP levels could not be determined within 24h from the diagnosis of PE, were also excluded. The study was approved by the Institutional Review Board of Ramón y Cajal Hospital.

Fig. 1.

Patient flow diagram. Abbreviations: PE, pulmonary embolism; CRP, C-reactive protein.

(0.16MB).

For the diagnosis of PE, the study required the finding of an intraluminal filling defect on pulmonary embolism protocol contrast-enhanced helical chest computed tomography,14 a high probability ventilation–perfusion scan according to the criteria of the Prospective Investigation of Pulmonary Embolism Diagnosis,15 or an indeterminate ventilation–perfusion lung scan associated with a confirmed lower limb (proximal or distal) deep vein thrombosis on venous compression ultrasonography.16

From blood drawn in a serum gel tube, we analyzed serum for CRP. We employed a conventional particle-enhanced nephelometry to determine CRP levels (Dimension Vista, Siemens, Berlin, Germany). This system measured concentrations from 2.9 to 290mg/L.17 For higher concentrations, we diluted samples 2-fold and reanalyzed them.

Patients underwent transthoracic echocardiography within 24h after diagnosis of PE. Trained and certified local cardiologists, blinded to the patient's clinical data, interpreted each echocardiogram based on guideline recommendations.18 The study defined echocardiographic RV dysfunction as the presence of at least two of the following criteria: (i) RV diastolic diameter at the midlevel (within the chamber) >30mm in the parasternal window; (ii) RV diameter at the midlevel greater than that of the left ventricle in the apical or subcostal space; (iii) RV free wall hypokinesis; and (iv) estimated pulmonary artery systolic pressure >30mmHg.19

Prognostic test results did not mandate treatment options. In the absence of an absolute contraindication to anticoagulation, clinicians-initiated treatment with parenteral anticoagulation combined with oral vitamin-K antagonist therapy, or with direct oral anticoagulants. In patients whose clinical status deteriorated after enrollment, clinicians administered reperfusion treatment and/or inotropic support as deemed appropriate.

We used transthoracic echocardiographic RV dysfunction within 24h after diagnosis of PE as the study's primary outcome. For 30 days after the diagnosis of PE, we also evaluated patients for all-cause and PE-related death. Two independent experts (authors C.R. and D.J.), blinded to the CRP results, classified the cause of death as PE-related if1 autopsy deemed PE as the cause of death, or2 death was most likely due to a PE-related deterioration or an objectively confirmed recurrent event, and most likely not due to another cause.

Data are presented as mean±SD or as median (IQR), as appropriate, for continuous variables and frequencies for categorical variables. Patients were divided into quartiles of CRP, with the lowest quartile (i.e., CRP<11.0mg/L) as the reference category. Quartiles were chosen to ensure a similar number of patients in each CRP category. We first conducted bivariable analysis comparing the demographic and clinical characteristics of patients across the quartiles of CRP with analysis of variance or Kruskal–Wallis test for continuous variables, and the chi-square test for categorical variables.

Association between CRP quartiles and the presence of RV dysfunction were determined using the Kendall's tau-b correlation coefficient. To determine CRP's diagnostic accuracy for echocardiographic RV dysfunction, we used receiver operating characteristic curves and analyzed area under the curves (AUC).

To assess the relationship between CRP and the outcomes of interest, we constructed multivariable logistic regression models for the overall cohort. We addressed potential confounding by controlling for the severity of illness and additional variables related to the outcome. The following models were generated sequentially to determine the successive influence of potential confounders on the relationship between CRP levels and mortality1: unadjusted2; adjusted only for age (as a continuous variable) and sex; and3 adjusted for age (as a continuous variable), sex, and the severity of PE (assessed by the simplified Pulmonary Embolism Severity Index [sPESI]). All these models were built with CRP as a continuous variable, and as a 4-category predictor variable.

We assessed the sensitivity of our findings by repeating the primary analysis under varying assumptions. Sensitivity analyses comprised comparing outcomes between those with normal and those with elevated CRP levels, using cut-off levels of <10mg/L and <5mg/L for a normal CRP.20 Since active cancer may produce an inflammatory response that elevated the CRP level,21 we also repeated analyses after excluding patients who had an active cancer (n=90).

We conducted analyses using SPSS software version 29.0.1.0 and considered a two-sided P value of <0.05 significant.

In the Emergency Department, we screened 791 consecutive patients diagnosed with non-COVID-19-associated acute symptomatic PE. We excluded 28 (3.5%) patients because of haemodynamic instability and an additional 130 patients (16.4%) because of CRP level not being determined within 24h from the diagnosis of PE. We enrolled the remaining eligible 633 patients (297 men and 336 women) (Fig. 1).

The cohort had a median CRP concentration of 34.0mg/L (IQR 11.0–84.2mg/L; distribution shown in Supplemental Fig. 1). Of the 633 enrolled patients, 477 (75.4%) had an elevated CRP level (i.e., >10mg/L) (CRP-positive group) and the remaining 156 (24.6%) patients had a normal serum CRP level (i.e., <10mg/L) (CRP-negative group).

Table 1 describes baseline characteristics stratified by quartiles of CRP. The CRP quartiles had similar mean age and vital signs (i.e., heart rate, oxygen saturation, and systolic blood pressure), and similar frequencies of immobilization, history of previous VTE, and comorbidities (i.e., chronic obstructive disease and congestive heart failure) at presentation. Patients in the higher quartiles were more likely to have male sex, history of cancer, previous surgery, and recent bleeding than patients in the lower CRP quartiles. Dyspnoea occurred more frequently in the lower CRP quartiles while chest pain occurred more frequently in the higher quartiles. The CRP quartiles did not have significantly different rates of treatment with a thrombolytic agent or insertion of an inferior vena cava filter during follow-up (Table 1).

One hundred and twenty-four (19.6%) patients met criteria for echocardiographic RV dysfunction (within 24h of PE diagnosis). Patients without RV dysfunction had significantly lower CRP levels compared with patients with RV dysfunction (31.7 [10.0–76.4]mg/L vs 45.4 [16.0–111.4]mg/L; P=0.018) (Fig. 2a). CRP quartile was significantly associated with RV dysfunction (Kendall's tau-b correlation coefficient, 0.095; P<0.01) with increasing quartiles showing a greater frequency of RV dysfunction (Fig. 2b). CRP (as a continuous variable) marginally discriminated between patients with and those without RV dysfunction (AUC, 0.58).

Fig. 2.

(a) Box plots with serum concentrations of C-reactive protein according to the presence or absence of right ventricle dysfunction. (b) Distribution of patients with right ventricle dysfunction among C-reactive protein quartiles in patients with pulmonary embolism (Kendall's tau-b correlation coefficient, 0.095; P<0.01). Abbreviations: CRP, C-reactive protein; RV, right ventricle.

(0.11MB).

The study had complete clinical outcome information for all patients at the end of the 30-day follow-up. The entire cohort had a 30-day all-cause mortality of 6.2% (39 of 633 patients), and a 30-day PE-related mortality rate of 2.8% (18 of 633 patients) (Table 2). Death unrelated to PE was caused by cancer in 10 (1.6%) patients, major bleeding in 3 (0.5%) patients, and unknown or other causes in 8 (1.3%) patients. When examined as a continuous variable (between 0.2 and 479.4mg/L), CRP level was not associated with 30-day all-cause mortality (unadjusted odds ratio [OR], 1.00; 95% CI, 1.00–1.01; P=0.095), but it was associated with 1% higher odds of PE-related death per 1mg/L increase (unadjusted OR, 1.01; 95% CI, 1.00–1.01; P=0.026). On univariable analysis, as compared with patients in the lowest quartile (quartile 1) of CRP, patients in quartiles 2, 3, and 4 had an increase in the odds of 30-day all-cause death of 2.58 (P=0.115), 3.16 (P=0.050), and 3.45 (P=0.034), respectively (Table 3).

Adjusted 30-day all-cause mortality was 2.5% for patients with CRP levels in the lowest quartile and 8.2% for patients with levels in the highest quartile. Similar findings were observed for 30-day PE-related mortality. The adjusted risk of 30-day PE-related death was 0% for patients in the lowest quartile of CRP levels and 5.1% for patients in the highest quartile of CRP levels.

To explore the sensitivity of our findings, we repeated the analysis with varying assumptions (Table 4). The use of a cut-off level of 10mg/L or 5mg/L for a normal CRP did not significantly change the main findings. After excluding patients with cancer, we found similar results, with a significant association between CRP quartile and the presence of RV dysfunction (Kendall's tau-b correlation coefficient, 0.105; P<0.01).