A Simple Method for Differentiating Complicated Parapneumonic Effusion/Empyema from Parapneumonic Effusion Using the Split Pleura Sign and the Amount of Pleural Effusion on Thoracic CT

Background Pleural separation, the “split pleura” sign, has been reported in patients with empyema. However, the diagnostic yield of the split pleura sign for complicated parapneumonic effusion (CPPE)/empyema and its utility for differentiating CPPE/empyema from parapneumonic effusion (PPE) remains unclear. This differentiation is important because CPPE/empyema patients need thoracic drainage. In this regard, the aim of this study was to develop a simple method to distinguish CPPE/empyema from PPE using computed tomography (CT) focusing on the split pleura sign, fluid attenuation values (HU: Hounsfield units), and amount of fluid collection measured on thoracic CT prior to diagnostic thoracentesis. Methods A total of 83 consecutive patients who underwent chest CT and were diagnosed with CPPE (n=18)/empyema (n=18) or PPE (n=47) based on the diagnostic thoracentesis were retrospectively analyzed. Results On univariate analysis, the split pleura sign (odds ratio (OR), 12.1; p<0.001), total amount of pleural effusion (≥30 mm) (OR, 6.13; p<0.001), HU value≥10 (OR, 5.94; p=0.001), and the presence of septum (OR, 6.43; p=0.018), atelectasis (OR, 6.83; p=0.002), or air (OR, 9.90; p=0.002) in pleural fluid were significantly higher in the CPPE/empyema group than in the PPE group. On multivariate analysis, only the split pleura sign (hazard ratio (HR), 6.70; 95% confidence interval (CI), 1.91-23.5; p=0.003) and total amount of pleural effusion (≥30 mm) on thoracic CT (HR, 7.48; 95%CI, 1.76-31.8; p=0.006) were risk factors for empyema. Sensitivity, specificity, positive predictive value, and negative predictive value of the presence of both split pleura sign and total amount of pleural effusion (≥30 mm) on thoracic CT for CPPE/empyema were 79.4%, 80.9%, 75%, and 84.4%, respectively, with an area under the curve of 0.801 on receiver operating characteristic curve analysis. Conclusion This study showed a high diagnostic yield of the split pleura sign and total amount of pleural fluid (≥30 mm) on thoracic CT that is useful and simple for discriminating between CPPE/empyema and PPE prior to diagnostic thoracentesis.


Introduction
Before diagnostic thoracentesis, pleural infection should be suspected in all patients with pneumonia persistent fever, and elevation of serum inflammatory markers such as C-reactive protein and white blood cell count. However, those clinical findings do not always indicate complicated parapneumonic effusion (CPPE)/empyema rather than parapneumonic pleural effusion (PPE). Among patients with CPPE/empyema, the frequency of surgery ranges from 15% [1] to 68% [2] and the mortality rate in patients with empyema is 15-20% [3][4][5]. Rapid recognition of CPPE/empyema is thus crucial to successful treatment. In this regard, thoracic computed tomography (CT) could play a pivotal role in differentiating between CPPE/empyema and PPE. The split pleura sign has been considered a diagnostic sign for empyema. However, no reports have evaluated the use of the split pleura sign to differentiate CPPE/empyema from PPE. These two clinical entities are sequential conditions, which leads to difficulty in assessing the diagnostic yield of the split pleura sign. We therefore undertook a retrospective study to evaluate the utility of the split pleura sign and total amount of pleural effusion on thoracic CT for differentiating between CPPE/empyema and PPE.

Materials and Methods
This retrospective study was approved by the Ethics Board of Kyorin University (number: H26-032) (Mitaka, Tokyo, Japan). All patients were referred to our respiratory department in outpatient or inpatient settings in Mitaka City, Tokyo, Japan, between May 2006 and May 2014. No informed consent was required for this study, but patient records and information were anonymized and de-identified prior to analysis. The definition of CPPE/empyema was based on Light's criteria (including only classes 6 and 7) [6] and category 3 or 4 of the American College of Chest Physicians consensus [7]. To be enrolled in the study, patients had to be older than 15 years and show pleural effusion on thoracic CT. This single-institution study retrospectively assessed patients who satisfied at least one of the criteria for CPPE or empyema mentioned above. PPE was defined as a case with clinical and radiological improvement after initiation of antibiotic therapy regardless of the presence of pneumonia and which satisfied none of the criteria for CPPE/empyema. Furthermore, patients with other etiologies for exudative pleural effusion, such as cytologically confirmed malignant pleural effusion (mesothelioma, lung cancer, and other metastatic cancers) and collagen vascular diseases or drugassociated pleural effusion or effusions of unknown etiology were not enrolled in this study. All laboratory data including thoracentesis and thoracic CT were obtained within 48 h after admission or on the same day in outpatient settings. Three respiratory physicians and one radiologist with 10 years of experience, all of whom were blinded to the clinical findings of patients, reviewed high-resolution CT (HRCT) findings independently and reached decisions by consensus.

Measurement of each finding on thoracic CT
The amount of pleural effusion was semiquantitated as the distance between layers of parietal and visceral pleura by drawing a vertical line (Fig 1). Thickening of the visceral or parietal pleura was defined as pleura visible on thoracic CT. If thickening of either the visceral or parietal pleura was noted on thoracic CT, this was called the "hemi-split pleura sign" (Fig 2A and  2B), and if both layers of pleura were thickened and separated by effusion, this was defined as the "split pleura sign" (Fig 2C and 2D). Both pleura signs were used and evaluated even with non-enhanced CT. Hounsfield unit (HU) values of pleural fluid were assessed at three slices not adjacent to ribs, lung parenchyma, or areas of pleural thickening. Pleural effusion showing septal walls was called septated pleural effusion, and the presence of multiple septa was considered multiloculated effusion.

Laboratory discrimination between CPPE/empyema and PPE
Correlations between serum and pleural inflammatory markers were analyzed together with correlations among pleural parameters. Predictive parameters for CPPE/empyema were extracted, and a receiver operating characteristic curve was constructed.

Statistical analysis
Numeric data were evaluated for normal distribution and for equal variance using the Kolmogorov-Smirnov test and Levene's median test, respectively. Categorical data are presented as percentages of the total or numerically, as appropriate. Statistical comparisons of nonparametric data were performed using the Mann-Whitney test. Comparisons of categorical data were made with Pearson's chi-square test. Logistic regression modelling was used for uni-and multivariate analyses to identify risk factors for CPPE/empyema. Receiver-operator characteristic (ROC) curves defining the sensitivity and specificity for diagnosing empyema were constructed for both parameters (split-pleura sign and amount of pleural effusion) on chest CT. All tests were two-sided. Significance was indicated by values of p<0.05. Data were analyzed using SPSS version 19.0 software for Windows.

Clinical characteristics of the CPPE/empyema and PPE groups
A total of 36 patients had CPPE/empyema (male, n = 30; female, n = 6) and 47 had PPE (male, n = 40; female, n = 7). Characteristics of the two groups are shown in Table 1. Mean (±standard deviation (SD)) age was 71.6±10.6 years (range, 49-94 years) for the CPPE/empyema group and 68.0±12.3 years (range, 41-88 years) for the PPE group. Groups were similar in age, maleto-female ratio, and proportion of patients who had underlying diseases or symptoms. Overall 8 patients had COPD, but only two were taking corticosteroids. In both groups, main symptoms on admission to hospital were pyrexia, chest pain, productive cough, and dyspnea, showing no significant differences between groups. The proportions of smokers and alcohol drinkers were comparable between groups, and time from initial onset to visiting our hospital did not differ significantly between groups (CPPE/empyema group, 13.0±15.3 days; PPE group, 8.6±8.4 days) ( Table 1).

Correlations among radiological parameters
Among the parameters obtained from radiological findings, correlations were assessed using total amount of fluid (mm), thickness of visceral pleura (mm), and HU value (Fig 5). The correlation between total amount of fluid and thickness of the visceral pleura was significant in both the CPPE/empyema (r = 0.394; p = 0.019) and PPE groups (r = 0.318; p = 0.03). Similarly, the thickness of visceral pleura and HU level showed a significant moderately positive correlation in both groups (CPPE/empyema group: r = 0.454, p = 0.006; PPE group: r = 0.438, p = 0.002). A significant correlation was seen between total amount of pleural fluid and HU value in the PPE group, but not in the CPPE/empyema group.
Receiver operating characteristic curve using two factors ROC curve analysis was performed for the two parameters (split pleura sign and amount of pleural effusion 30 mm) identified by multivariate analysis. The split pleura sign was better than amount of pleural effusion 30 mm for diagnosing CPPE/empyema, with 80.6%   Fig 6).

Discussion
Discriminating CPPE/empyema from PPE is often quite difficult, especially before diagnostic thoracentesis, because of the clinical similarities of these two sequential conditions, as seen in the present study (Tables 1 and 2). Interestingly, the time from initial onset to the first visit to hospital and the proportions of underlying diseases did not differ significantly between the CPPE/empyema and PPE groups. Furthermore, no significant correlation between systemic inflammation (serum WBC or CRP levels) and local thoracic inflammation (pleural fluid TCC) was found in either group (Fig 3). Although serum LDH level was significantly higher in the PPE group than in the CPPE/empyema group, LDH levels in the pleural fluid were higher in CPPE/empyema group than that of PPE group. This indicates that pleural inflammation is not necessarily reflected in the serum. However, some studies that examined correlations between systemic inflammation (inflammatory markers) and local inflammation (pleural cavity) using interleukin-18 [8] or soluble Fas ligand [9] found no clear evidence of positive correlations with those markers. This presents a diagnostic dilemma for physicians. In this regard, radiological assessment for differentiating CPPE/empyema from PPE seemed to be pivotal in the diagnostic process. To the best of our knowledge, only six reports have been published regarding the split pleura sign [10][11][12][13][14][15]. Stark et al. reported that the split pleura sign was seen in 68% of pleural empyema cases [13] and is considered the most reliable CT sign helping to distinguish empyema from pulmonary abscess [13] or noninfectious pleural effusion [14]. However, no report has described the utility of the split pleura sign [15] in differentiating CPPE/empyema from PPE. From this perspective, the present series showed that the split pleura sign could be a useful marker for differentiating CPPE/empyema (HR, 6.70; 95%CI, 1.91-23.5; p = 0.003) from PPE. Furthermore, the amount of pleural fluid (30 mm) on thoracic CT was also found to be a suggestive factor for CPPE/empyema (HR, 7.48; 95%CI, 1.76-31.8; p = 0.006). Being aware of these two predictive factors for CPPE/empyema is a simple way for physicians to assess the probability of CPPE/empyema prior to performing diagnostic thoracentesis.
The present study has some limitations in that: 1) only 15 of 36 patients (41.7%) in the CPPE/empyema group and 8 of 47 patients (17.0%) in the PPE group underwent enhanced thoracic CT, which might have affected the results for the incidence of the split pleura sign; 2) thoracic CT might not be available in local hospitals or clinics; and 3) other pleural diseases such as malignant effusion, mesothelioma, and tuberculous pleuritis were outside of the scope of this study, and also need to be ruled out in general practice; 4) the study was retrospective. However, this study showed a strong correlation between pleural LDH and ADA levels (Fig 4) using the combined data from both CPPE/empyema and PPE groups, implying that these measurement might be useful for differentiation from the other pleural diseases described above. In conclusion, this is the first study to demonstrate the high diagnostic yield of both split pleura sign and large pleural effusion (30 mm) on thoracic CT for discriminating between CPPE/ empyema and PPE. Early recognition of CPPE/empyema using those signs could help decrease morbidity and mortality.