Low pericardial uid glucose concentration combined with computed tomography evaluation as a malignancy-related marker in symptomatic patients undergoing pericardiocentesis: a retrospective study

Background: We evaluated malignancy according to the characteristics of pericardial uid in symptomatic Japanese patients undergoing pericardiocentesis and computed tomography (CT). Methods: This was a retrospective, single-center, observational study of 125 symptomatic patients undergoing pericardiocentesis. The patients were classied into two groups: a malignancy group and a non-malignancy group, according to the primary disease and cytology of the pericardial effusion (PE). We compared the pericardial uid sample and CT measurements between both groups. Results: All patients were diagnosed as having exudative PE by Light’s criteria. PE with malignant cells was demonstrated in 76.8% of the malignancy group patients. A ratio of pericardial to serum lactate dehydrogenase (LDH) >0.6 as one of Light’s criteria, and lower serum brain natriuretic peptide (BNP: 126.9±89.8 pg/ml vs 409.2±97.7 pg/ml, malignancy vs non-malignancy groups, respectively; p = 0.037) were associated with malignancy (both, p < 0.05). A signicant difference was observed in pericardial uid glucose level between the malignancy and non-malignancy groups (pericardial uid glucose: 78.25±48.29 mg/dl vs 98.41±44.85, respectively; p < 0.05). Moreover, CT attenuation values (Hounseld units (HU)) tended to be higher in the malignancy group vs the non-malignancy group (22.7 [interquartile range (IQR), 17.4–26.0] vs 17.4 [IQR, 13.7–26.4], respectively; p = 0.08). Pericardial uid glucose level ≤ 70 mg/dl and CT attenuation values >20 HU were strongly associated with malignancy. Conclusions: Lower pericardial uid glucose level with higher CT attenuation values may suggest malignancy-related PE.


Background
The normal pericardial sac contains up to 50 ml of pericardial uid as a plasma ultra ltrate. Pericardial effusion (PE), which is classi ed according to composition, distribution, and volume, results from many diseases [1]. In patients with PE, the prevalence of malignancy or infection ranges from 15-50% [2,3], and malignant PE can be diagnosed by cytology. In addition to imaging and pericardial uid analysis, the clinical course of the underlying disease can contribute to the diagnosis of malignancy-related PE. In particular, large PE is a common manifestation in malignant disease, and patients with malignant PE have a poor prognosis because of advanced-disease stage [4,5]. In cases without pre-existing cancer, malignancy must be excluded because PE may be the rst manifestation of malignancy. Some studies reported PE attributed to unknown causes at a rate of 10-20% [5,6]. It may be di cult to diagnose malignant PE even if cytology or epicardial and pericardial biopsy are performed [7]. A previous study of pericardial uid analysis showed that diseases with exudative PE had lower pericardial uid glucose levels compared with diseases with transudative PE [8]. However, the relationship between malignancyrelated PE and non-malignant PE evaluated using computed tomography (CT) to analyze the pericardial uid is unknown. Therefore, we examined malignancy according to the characteristics of the pericardial uid and CT measurements in symptomatic patients undergoing pericardiocentesis.

Patients
We enrolled 125 consecutive symptomatic patients undergoing pericardiocentesis between April 2010 and April 2020 as a retrospective observational study in a single institution.
Twenty-eight patients without pericardial uid sampling and CT data were excluded because of out-ofhospital cardiopulmonary arrest, Stanford type A aortic dissection, and blow-out rupture in acute myocardial infarction as an intra-procedural complication. Therefore, 97 patients were nally enrolled in this study. The patients were classi ed into two groups: a malignancy group and a non-malignancy group. The malignancy group constituted patients with known existing malignancy or documented malignant cells in drained PE.

Blood samples and pericardial uid samples
Blood samples were collected approximately 24 hours after drainage. We evaluated the following in the drained pericardial uid: pH, total protein, lactate dehydrogenase (LDH), albumin, adenosine deaminase (ADA), hyaluronic acid, hemoglobin (Hb), and hematocrit. Collected pericardial uid was cytologically diagnosed with malignant cells or non-malignant cells by two pathologists. Pericardial uid testing was performed in 88 patients, and cytological analysis was performed in 94 patients. Patients were diagnosed with exudative PE according to Light's criteria, which constitute a pericardial uid-to-serum total protein ratio > 0.5 and a pericardial uid-to-serum LDH ratio > 0.6 or two-thirds of the upper limit of the normal (UNL) serum LDH value.
CT Protocol and measurement CT scans before drainage were performed within 24 hours of admission using a 64-multidetctor CT scanner (Discovery HD 750; GE Healthcare, Milwaukee, WI, USA). Eighty-one patients were evaluated, with 5-mm slice thickness, gantry rotation speed of 350 ms, 64 × 0.625-mm collimation, tube voltage of 120 kV, and effective tube current of 325-750 mA. Axial images were transferred to an available workstation and analyzed using SYNAPSE VINCENT software (Fuji lm™ Co., Tokyo, Japan). The CT attenuation values in the PE were calculated and analyzed by SYNAPSE VINCENT [9]. In addition to a previous method that was determined by the region of interest (ROI), CT attenuation values and pericardial area were measured by plotting around the PE using a two-dimensional (2D) image analysis system in three axial view slices as follows: 1) upper slice: bifurcation of the pulmonary trunk, 2) middle slice: fourchamber heart view, and 3) lower slice: upper edge of the liver (Fig. 1) [10]. The average CT attenuation value from the three slices was calculated as Houns eld units (HU). The measurements were reanalyzed by the same interpreters to evaluate intraobserver agreement.

Page 4/13
Categorical variables were compared using the χ 2 test for proportions and the unpaired t-test or analysis of variance for continuous variables, as appropriate. The linearity of the relationship between two variables was assessed by linear regression analysis. Correlation analysis was performed with Pearson's correlation coe cient. Parameters with non-normal distributions were analyzed using the Mann-Whitney U test. p < 0.05 was considered signi cant, and results are expressed as the mean ± standard deviation (SD). All analyses were performed using JMP version 11.0.

Baseline clinical characteristics
The baseline characteristics of the 97 subjects are shown in Table 1; the majority were male and > 65 years of age. Approximately 10% of the patients had previously taken antiplatelet agents, but few patients had received prior anticoagulation therapy. The most frequent etiology in patients undergoing pericardiocentesis was malignant disease (58.8%), and among these patients, 59.6% had lung cancer ( Table 2). Pericardial uid with malignant cells: positive-cytology was demonstrated in 76.8% of the patients in the malignancy group.   Blood samples (Table 3) No signi cant differences in serum C-reactive protein (CRP), total protein, LDH, albumin, glucose, and Hb were observed between the malignancy and non-malignancy groups. In contrast, estimated glomerular ltration rate (eGFR) was signi cantly higher in patients with vs without malignancy (60.56 ± 31.42 ml/min/1.73m 2 vs 45.97 ± 29.08 ml/min/1.73m 2 , respectively; p = 0.024). Serum brain natriuretic peptide (BNP) levels were signi cantly lower in patients with vs without malignancy (126.89 ± 183.35 pg/ml vs 409.18 ± 868.67 pg/ml, respectively; p = 0.037). Values are expressed as mean ± standard deviation CRP C-reactive protein, eGFR estimated glomerular ltration rate, LDH lactate dehydrogenase, BNP brain natriuretic protein, ADA adenosine deaminase, UNL upper limit of normal Pericardial uid samples, pericardial/serum ratio, Light's criteria, and cytology (Table 3).
No signi cant differences in pericardial uid pH levels, total protein, LDH, albumin, ADA, and hyaluronic acid were observed between the two groups. Pericardial uid glucose levels were signi cantly lower in patients with vs without malignancy (78.24 ± 48.29 vs 98.41 ± 44.85, respectively; p = 0.048) (Fig. 2).
Malignant cells in the pericardial uid was demonstrated in 76.8% of the patients in the malignancy group. All patients were diagnosed as having exudative PE by Light's criteria. Of the three items in Light's criteria, LDH ratio > 0.6 and pericardial LDH > 2/3 of the UNL were associated with malignancy (p = 0.017 and p = 0.036, respectively). The pericardial uid-to-serum LDH ratios were signi cantly higher in patients with malignancy than in those without malignancy (5.23 ± 7.69 vs 2.59 ± 2.71, respectively; p = 0.047). The pericardial uid-to-serum glucose ratios were signi cantly lower in patients with vs without malignancy (0.57 ± 0.34 vs 0.83 ± 0.45, respectively; p = 0.005) (Fig. 2).

CT measurements
No signi cant differences in CT parameters were observed between the malignancy and non-malignancy groups. However, CT attenuation values (HU) tended to be higher in the malignancy group compared with the non-malignancy group ( The relationship between pericardial uid samples and CT measurements The malignancy group had a higher median CT attenuation value of 20.6 HU (p < 0.05). Moreover, overall, CT attenuation values > 20 HU were associated with malignancy (p < 0.05) (Fig. 3). The relationship between pericardial uid measurements and CT attenuation values is shown in Fig. 4. CT attenuation values were strongly correlated with pericardial uid hematocrit (r = 0.908). CT attenuation values were also correlated with pericardial uid glucose level (r = − 0.425), protein level (r = 0.541), glucose ratio (r = − 0.461), and protein ratio (r = 0.583). Pericardial uid glucose ≤ 70 mg/dl and CT attenuation values > 20 HU were cutoff values associated with malignancy (Fig. 5). Additionally, pericardial uid glucose ≤ 70 mg/dl and CT attenuation values > 20 HU were also cutoff values associated with positive-cytology patients (see Supplementary Fig. 1, Additional File 1).

Discussion
In the current study, we investigated the characteristics of malignant pericardial uid in symptomatic patients undergoing pericardiocentesis and CT; lower pericardial uid glucose level and higher CT attenuation values were related to malignancy. Strobbe et al. evaluated the etiologies of PE and distinguished radiation therapy for malignancy as a cause of PE [11]. Moreover, they reported that bone marrow transplantation was an etiology in the idiopathic group [11]. Additionally, Sullivan et al. reported that chronic graft-versus-host disease (GVHD) presented as an immune reaction after bone marrow transplantation, similar to the mechanism in autoimmune diseases [12]. Therefore, non-recurrent cancer patients with prior radiation therapy or chronic GVHD were included in our non-malignancy group. In the current study, the number of patients in the malignancy group was higher than the number in the nonmalignancy group, and approximately 60% of the patients in the malignancy group had lung cancer.
One study in Western Asia and another in Africa reported PE attributed to unknown causes at rates of approximately 10-15% [6,13]. In contrast, a study in Western Europe and another in North America reported that approximately 80-90% of patients with PE were classi ed as having an idiopathic etiology [14,15]. Distinguishing malignant PE from non-malignant PE has important implications regarding prognosis. However, two previous studies reported that 44.9-92% of malignant patients often demonstrated malignant cells in PE [8,16]. As a result of cytology by repeat drainage, the current study showed that 76.8% of patients in the malignancy group demonstrated malignant cells in the pericardial uid. Standard pericardial cytology does not necessarily exclude malignant PE; therefore it is necessary to detect neoplasia using blood samples, pericardial uid analysis, and imaging modalities, such as positron-emission tomography-CT (PET-CT). Cytology-negative patients in the malignancy group in this study were de ned as patients who had untreatable recurrent or metastatic carcinoma.
Although pericardial uid analysis is performed routinely, the results of these analysis are not effectively utilized owing to a lack of speci city. Meyers et al. reported that pericardial uid glucose level was lower in patients with exudates than in those with transudates, and there was a signi cant overlap in the etiology of PE between patients with transudates vs exudates [8]. All cases in our study, in both the malignancy and non-malignancy groups, were diagnosed as having exudative PE by Light's criteria. No signi cant differences in pericardial uid LDH levels were observed between the groups, while LDH ratio > 0.6 and pericardial LDH > 2/3 of the UNL were associated with malignancy. Ben-Horin et al. demonstrated pericardial uid LDH level > 2.4 times the serum level and mean pericardial uid protein level 0.6 times the serum level in patients undergoing elective open surgery without a history of pericardial diseases [17]. Despite these ndings, a recent study showed that malignant patients had lower PE LDH levels than nonmalignant patients, and another study showed that malignant patients had lower PE glucose levels compared with patients with heart failure-related effusion [16,18]. Furthermore, Ben-Horin et al. demonstrated a mean pericardial uid glucose level of 133 mg/dl and mean pericardial uid-to-serum glucose ratio of 1.0 in patients without a history of pericardial diseases [17]. In the current study, pericardial uid glucose levels and pericardial uid-to-serum glucose ratio were signi cantly lower in the malignancy group than in the non-malignancy group. In an additional analysis, pericardial uid glucose levels were signi cantly lower in positive-cytology patients than in negative-cytology patients. However, neither pericardial uid glucose levels nor the pericardial uid-to-serum glucose ratio in our study was helpful for differentiating the malignancy group from the non-malignancy group because the areas under the receiver operating characteristic (ROC) curve were < 0.7. Pericardial uid glucose level may decrease in diseases such as tuberculosis, rheumatism, and in ammatory diseases in non-malignancy patients, and serum glucose level can change secondary to various factors. In our study, no signi cant differences in the rates of diabetes mellitus and hypoglycemic agents were observed between the groups regarding pericardial uid glucose levels.
CT attenuation values may be useful in determining the etiology of PE. We measured CT attenuation values in PE areas lager than the area in the ROI and used the former as the CT attenuation value. Previous studies reported that CT attenuation values were positively correlated with pleural and pericardial total protein levels and the pleural and pericardial/serum protein ratio [10,19]. Moreover, Rifkin et al. demonstrated that CT attenuation values were strongly correlated with pericardial uid hematocrit [20]. Our study showed that CT attenuation values were strongly correlated with pericardial uid hematocrit, and that these values also had a good correlation with pericardial uid glucose and protein levels, and glucose ratio and protein ratio. CT attenuation values from 20 HU to 60 HU indicated exudate in hypothyroidism, purulent pericarditis, or malignancy [21]. The median CT attenuation values in both groups was > 20 HU for exudates, while the cutoff value was determined to be 20 HU according to a previous report indicating that patients with CT attenuation values > 20 HU had proteinaceous or hemorrhagic PE [22]. As a result, CT attenuation values > 20 HU were associated with malignancy. Accordingly, the combination of pericardial uid glucose ≤ 70 mg/dl and CT attenuation values > 20 HU might be associated with malignancy in PE. In addition, this study showed that this combination was associated with positive-cytology patients.

Study limitations
There are several limitations to our study. First, the size of the non-malignancy group was small because this was a single-institution study. Second, the malignancy group had higher proportions of patients with lung cancer.