This is a subgroup analysis of the SLICE multicenter clinical trial. The protocol and main analysis of SLICE have been published elsewhere.10,11 In brief, SLICE randomized consecutive patients requiring hospitalization for exacerbation of their COPD to proactive screening for PE or conventional management based on local protocols and clinical practice guidelines. Screening for PE initially consisted of performing a D-dimer test. A negative D-dimer (according to the cut-off point used in each participating center) excluded PE. Patients with a positive D-dimer test underwent chest CT angiography within 12h of randomization.10 A negative CT angiography ruled out PE, while a positive CT angiography was an indication for anticoagulant treatment.

In this subgroup analysis, we included all patients in the intervention group of the SLICE trial that tested negative for PE on chest CT angiography (i.e., normal CT or CT finding of an alternative diagnosis to PE).

Demographic, clinical, and laboratory data were collected at the time of randomization. These included, among others: age, sex, weight, height, comorbidities (cancer [non-lung], ischemic heart disease, heart failure, sleep apnea), risk of VTE (previous surgery, immobilization, history of VTE), severity of COPD, number of exacerbations in the preceding year, treatment for the stable phase of COPD, clinical signs and symptoms (heart rate, blood pressure, blood oxygen) and laboratory test results at the time of admission (hemoglobin, leukocytes, platelets, serum creatinine).

Bronchial infection on chest CT angiography was defined as bronchial wall thickening, centrilobular nodules, and a tree-in-bud sign. Pneumonia was diagnosed in patients with a lobar or segmental alveolar pattern who showed good clinical and radiological response to broad-spectrum antibiotic therapy. Heart failure was diagnosed in patients with interstitial involvement (Kerley lines), an alveolar or ground-glass pattern predominantly in dependent regions, and bilateral pleural effusion. Histological confirmation of pulmonary nodules and masses was required for a diagnosis of cancer.

The primary outcome variable was the composite of readmission for COPD or all-cause mortality at 90 days. Secondary variables were each of the individual components of the primary variable, new onset or recurrent of non-fatal thrombotic events, and length of hospital stay.

Continuous data are shown as mean±standard deviation or median (interquartile range), and categorical data are shown as absolute values and proportions (%). Differences between groups were tested using the Student's t-test or the Wilcoxon rank-sum test for continuous variables and the χ2 test or Fisher's exact test for categorical variables.

A multivariate logistic regression model was constructed to determine whether an alternative diagnosis on CT (versus normal CT) was independently correlated with the primary outcome variable.13 The model was constructed using a maximum of 10 variables that showed a univariate correlation with the outcome (P<0.05) or that were considered clinically relevant based on expert opinion and published experience.14 These variables were: (a) age; (b) sex; (c) COPD severity; (e) congestive heart failure (CHF); (f) cancer; (g) an alternative diagnosis on CT; (h) immobilization; (i) heart rate; (j) systolic blood pressure; and (k) blood oxygen levels. Starting with the full model, we used the backward elimination technique to eliminate any variables that did not modify the coefficient of the variable (i.e., presence of an alternative diagnosis on CT) by 10% or more. Odds ratios (OR) were calculated with their 95% confidence intervals (CI). We then repeated the analyses to assess the prognostic significance of the presence (versus absence) of any diagnosis, including PE, on chest CT. As the model was not adjusted for multiplicity, the analyses of the secondary variables should be viewed as merely exploratory.

A two-sided P<0.05 was considered indicative of statistical significance in all cases. All analyses were performed on Stata, version 13.1 (StataCorp LLC; College Station, Texas, US).

Chest CT findings were suggestive of an alternative diagnosis to PE in 133 patients (76.0%; 95% CI, 69.0–82.1%). Of these, 84 (48.0%) were diagnosed with acute bronchial infection, 13 (7.4%) with lung cancer, 10 (5.7%) with heart failure, 8 (4.6%) with pneumonia, 18 (10.3%) with other diagnoses (5 with bronchiectasis, 4 with fibrosis, 3 with pleural effusion, 2 with pulmonary hypertension, 2 with pericardial effusion, 1 with hypersensitivity pneumonitis, and 1 with organizing pneumonia). Chest CT was normal in 42 (24.0%) patients. Diagnosis of these comorbidities led to a change in the treatment of 31 patients (31 of 133 patients; 23.3%; 95% CI, 16.4–31.4%): diuretics in 10 patients, antibiotics in 9, systemic corticosteroids in 7, and treatment for lung cancer in 5 patients.

Demographic, clinical and laboratory characteristics are shown in Table 1. There were no significant differences between the group of patients with an alternative diagnosis to PE on CT and the group of patients with normal CT. Purulent sputum during exacerbation was found more frequently in patients with an alternative diagnosis to PE on CT (8.3% versus 0%; P=0.07).

The primary event occurred in 55 (31.4%; 95% CI, 24.6–38.9%) of the 175 patients during the first 90 days of follow-up (Table 2). Forty-five patients (25.7%) were readmitted due to exacerbation of their COPD, and 16 patients (9.1%) died, of whom 6 had previously been readmitted. The primary event occurred in 40 of 133 patients (30.1%) with an alternative diagnosis to PE on chest CT, and in 15 of 42 patients (35.7%) with normal CT (difference 5.6%, 95% CI, −10.6 to 23.6%) (Fig. 1). Table 2 shows the distribution of events according to the presence or absence of an alternative diagnosis to PE on chest CT.

Fig. 1.

Cumulative percentage of events based on the presence of an alternative diagnosis to PE on chest CT or a normal chest CT.

(0.14MB).

In the univariate analysis, age>75 years (OR: 1.88; 95% CI: 0.99–3.59; P=0.06), cancer (OR: 4.78; 95% CI: 1.15–19.87; P=0.03), and immobilization for medical reasons (OR: 1.99; 95% CI: 0.96–4.16; P=0.07) significantly increased the risk of the composite endpoint occurring within 90 days of randomization (Table 3). In the multivariate analysis, the finding of an alternative diagnosis to PET on chest CT (versus normal CT) was not independently associated with the composite endpoint in the first 90 days of follow-up (ORadjusted: 0.64; 95% CI: 0.30–1.38; P=0.26) (Table 3).

After adjusting for age>75 years, cancer, and medical immobilization, an alternative diagnosis to PE on chest CT (versus normal CT) was not independently associated with all-cause mortality (difference 12.0%; 95% CI, 0.6–19.1%; P=.06) or readmission for COPD (ORadjusted: 0.48; 95% CI, 0.22–1.03; P=0.06) within 90 days of randomization.

We then analyzed the prognostic significance of the presence (versus absence [i.e., normal CT]) of any diagnosis, including PE, on chest CT. A specific CT finding was identified in 150 patients (150 of 192 patients; 78.1%; 95% CI, 71.6–83.8%). The finding of a specific diagnosis on chest CT was not independently associated with the composite endpoint (ORadjusted: 0.61; 95% CI: 0.29–1.30; P=0.20), all-cause death (difference 11.3%; 95% CI, 0.0–17.8%; P=0.07), or readmission for COPD (ORadjusted: 0.48; 95% CI: 0.22–1.03; P=0.06) within 90 days of randomization.