A comparative study of diagnostic accuracy in 3026 pleural biopsies and matched pleural effusion cytology with clinical correlation

Abstract Background Pleural effusion can be caused by a wide range of benign and malignant conditions. Pleural biopsy and effusion cytology represent two key methods of pathological diagnosis. To compare the performance these two methods, a large cohort of matched pleural biopsy and effusion cytology with clinical follow‐up was reviewed. Methods Pleural biopsies and effusion cytology specimens over a period of 18 years were retrieved. Cytology specimens collected within 7 days of pleural biopsy were matched. Reports were reviewed, and the cause for pleural effusion was determined by hospital disease coding and clinical data. Results Totally, 3026 cases were included. The leading cause of benign effusion was tuberculosis (n = 650). Malignant pleural effusion (MPE) was more common in older females (p < 0.001) and mostly due to lung cancer (n = 959), breast cancer (n = 64), and mesothelioma (n = 48). The inadequate/insufficient (B1/C1) rate of biopsy was higher than cytology (15.6% vs. 0.3%) but the rates for other diagnostic categories were similar. Biopsy and cytology showed a correlation coefficient of 0.315, improving to 0.449 when inadequate/insufficient (B1/C1) cases were excluded. The ROM for benign cytology (C2) was lower than biopsy (B2) (p < 0.001). Compared with biopsy, the diagnostic accuracy was higher in cytology overall and for metastatic carcinomas (p < 0.001) but lower for hematolymphoid malignancies (p = 0.014) and mesotheliomas (p = 0.002). Conclusions These results suggest that effusion cytology may be better for confirming benignity and diagnosing carcinomatous MPE. In these cases, pleural biopsy may be withheld to reduce procedural risks. However, for suspected hematolymphoid malignancies and mesothelioma, biopsy should be considered.


| INTRODUCTION
Pleural effusion is frequently encountered in hospital admissions. The underlying causes include a wide range of benign, reactive, infective, autoimmune, and neoplastic diseases. 1 Tissue diagnosis is crucial for the diagnosis and subsequent management of malignant pleural effusion (MPE), requiring either pleural fluid cytology or pleural biopsy. 2,3 While thoracentesis for pleural fluid cytology is minimally invasive and cost-efficient, tissue biopsy is generally considered more definitive and targeted for obtaining a larger volume of tissue. 4 However, in the context of pleural effusion, fluid cytology has the potential advantage of sampling exfoliated cells from the entire pleural cavity. 5 The diagnostic accuracy and yield of these two methods may vary according to the nature of pleural effusion, as small or mediastinal pleural lesions can be difficult to biopsy, 6 whereas tumor cells shed to pleural fluid may be scanty in sarcomas. 7 A direct comparison of the diagnostic accuracy and yield of these two methods, with respect to the underlying condition, is necessary for selecting the optimal mode of tissue acquisition. In this study, we reviewed a large cohort of matched pleural effusion cytology and biopsy, compared the diagnostic concordance of these two methods, and further analyzed their accuracy by reviewing hospital disease coding confirmed with clinical, radiological, microbiological, serological, and /or pathological data.

| METHODOLOGY
A computerized search was performed in the department archives for pleural biopsies and effusion cytology specimens collected from February 2001 to February 2019. The reports were reviewed for patient demographical data and histologic/cytologic diagnosis. The diagnoses were classified into five-tiered categories (B1-5/C1-5-1: insufficient/inadequate, 2: benign, 3: atypia, 4: suspicious, and 5: malignant; B-biopsy, C-cytologic aspirate) (Figure 1), and the microscopic description and diagnostic line of the reports were reviewed for diagnostic classifiers, such as lymphocytosis in effusion cytology or line of differentiation in neoplastic pleural biopsies. To establish the underlying cause of pleural effusion, hospital disease coding and relevant clinical data were retrieved. For malignant conditions, a corresponding hospital disease coding was acquired and in addition confirmed by clinical case notes, radiology reports, tumor markers, and/or previous pathological diagnoses; for benign causes, either an attributable clinical cause has to be documented (e.g., infection, renal failure) or a period of at least 1 year of unremarkable follow-up has to be recorded.
Pleural biopsies were paired with effusion cytology specimens by a hospital accession number unique to each patient in the region-wide clinical management system. Only effusion cytology specimens collected within 7 days before or after the date of pleural biopsy were considered a match. When multiple cytology specimens were matched, the most high-grade diagnosis was taken as the final diagnosis. This study was approved by the Joint Chinese University of Hong Kong-New Territories East Cluster Clinical Research Ethics Committee with an exemption of requirement of written consent (reference number: 2020.289).
Statistical analysis was performed with the statistics software R 4.10. Descriptive statistics and predictive values-sensitivity (Sn), specificity (Sp), positive predictive value (PPV), negative predictive value (NPV), accuracy, and risk of malignancy (ROM) were calculated. The t test and Chi-square test were used for comparison of continuous and categorical variables. κ statistic was used to measure the concordance between pleural biopsy and effusion cytology diagnoses. A p value of <0.05 was considered statistically significant.

| Distribution and concordance of diagnoses
The diagnoses of biopsy and cytology showed an overall concordance of 0.626 with a κ statistic of 0.315, which improved to a concordance of 0.743 and a κ statistic of 0.449 when all B1 and/or C1 cases were excluded.

| DISCUSSION
In this cohort of 3026 paired pleural biopsies and effusion cytology specimens, there were more male than female patients (p < 0.001) and benign than malignant diagnoses (p < 0.001). These demographic differences were also reported in cohorts of pleural effusion without paired biopsy and cytology, [8][9][10] and may be contributed delayed health-seeking behaviors and higher smoking prevalence in the local male population. 11,12 The leading cause of benign effusion was tuberculosis, corresponding to the high incidence in the locality 13 and similar to tuberculosis endemic areas. 14,15 It was followed by parapneumonic effusion and heart failure. Other causes included autoimmune, infective, and inflammatory reactive conditions and organ failure ( Table 2). For MPEs, the common underlying malignancies were lung cancer, breast cancer, and mesothelioma. Lung and breast cancers are known for their high propensity of developing pleural metastasis and MPE, 16,17 and the majority of patients with mesothelioma suffer from pleural effusion on presentation. 18 In line with the literature, MPEs more commonly presented in older and female patients than benign effusions (p < 0.001), 8,10 and mesotheliomas were predominantly observed in male patients (p < 0.001). 19 The female genital tract (2.2%, n = 27/1212) followed the lung (79.1%, n = 959/1212) and the breast (4.1%, n = 50/1212) as the third most common primary site contributing to MPEs caused by metastatic carcinomas. The distribution of primary tumors was similar to previous studies, including a cohort of Asian demographic. 15,20,21 Despite that metastatic carcinoma of lung primary was more common in male patients (52.2%, n = 501/959), breast and gynecological cancers in the cohort resulted in an overall predisposition to female patients for MPEs caused by metastatic carcinoma (50.8%, n = 616/1212) ( Table 1).
There was a higher rate of inadequacy/insufficiency in pleural biopsy compared with effusion cytology (15.6% vs. 0.3%). As the cohort was collected prior to the publication of International System for serous fluid cytopathology, fluid volume was not considered for the diagnosis of insufficiency. Specimens were considered inadequate/insufficient when no mesothelial or lesional cells were included. The rate of inadequacy for pleural biopsy reported in the concordance between biopsy and cytology improved when B1 and/or C1 cases were excluded. However, it should be noted that there were 27 C4 diagnoses and 47 C5 diagnoses matched to B1 pleural biopsies. Out of these 74 pairs, 72 had an underlying malignant condition. On the contrary, only two B5 and no B4 biopsies were matched to C1 diagnoses, and only one (n = 1/2, 50%) was malignant on the follow-up (Table 4).
Comparing the ROMs of each diagnostic category between biopsy and cytology, only the B2/C2 and B3/ C3 categories showed significant difference. The lower ROM of the C3 in cytology may stem from the diagnostic uncertainty in cytology specimens, 28 contributed by the interpretative difficulties and qualitative issues such as degeneration and preparation artifacts. 29 Although cytology appears to show superior diagnostic accuracy in the B2/ C2 category and all categories combined (Table 5), in subgroup analysis, the difference is only reproduced in cases of metastatic carcinoma (Table 6). Biopsy outperforms cytology for diagnosing hematolymphoid malignancies (p = 0.014) and mesothelioma (p = 0.002) ( Table 6). The majority of MPEs in this cohort were due to metastatic carcinomas, and this would increase the accuracies in the combined and B2/C2 categories for cytology. Inability of demonstrating invasion and overlapping cytomorphology with reactive mesothelial cells are some of the difficulties in cytologic diagnosis of mesothelioma. 30,31 As for hematolymphoid malignancies, aspiration disrupts the lymph node architecture and limits the diagnosis of low-grade lymphomas. 32 However, in carcinomatous MPEs, previous studies have shown that blinded pleural biopsy only provides limited additional diagnostic information, reflecting the irregular distribution of malignant deposits across the pleura. 33,34 In contrast, effusion fluid usually immerses the entire pleural cavity, and in this situation effusion cytology may be more representative than pleural biopsy. Further review of biopsy and cytology reports of the hematolymphoid malignancy and mesothelioma subgroups from Table 6, immunocytochemistry was performed in 56.8% (n = 21/37) and 76.5% (n = 26/34) of biopsy cases included for analysis in hematolymphoid malignancy and mesothelioma subgroups. This is higher than the proportion of cytology specimens with immunocytochemistry performed (hematolymphoid malignancy subgroup: 7.3%, n = 3/41; mesothelioma subgroup: 4.9%, n = 2/41). Immunocytochemistry and other ancillary tests provide additional information for cytologic interpretation, but still have limitations in differentiating reactive lymphoid hyperplasia from low-grade lymphomas, 35 and can give contradicting or noncontributory results. 36 For the diagnosis of mesothelioma, equivocal invasion on solid tissue remains the most decisive diagnostic feature, which outperforms immunocytochemical and serologic markers in terms of specificity and reproducibility. 30 A lower proportion of cases with immunocytochemistry performed may have contributed to the difference in diagnostic performance between biopsy and cytology. However, barriers to immunocytochemistry exclusive to cytology preparations should also be considered, including insufficient material for conversion to cell block and technical difficulties in antibody staining for alcohol-fixed specimens. 37 Cytomorphology remains the crux of cytologic interpretation with ancillary tests adding value only within specific contexts. 38 For the accuracy in typing of malignant lesions (carcinoma vs. mesothelioma vs. hematolymphoid), analysis was limited by the small number of cases. In addition, a significant number of reports lacked diagnostic classifiers, only issuing a diagnosis of "malignancy" or "malignant cells." When mesothelial lineage was specified in the report, both cytology and biopsy were very accurate (specific ROMs = 1) ( Table 7). The diagnosis of malignant hematolymphoid lesion on biopsy (B5) was correct in 93% of cases (n = 13/14), greater than that of cytology (C4: n = 4/5, 80%, C5: 3/4, 75%).
Considerations in performing concurrent pleural biopsy at the time of thoracentesis is crucial in the management of pleural effusion. Factors include the availability of resource, likelihood of underlying disease determined by clinical impression and previous workup results, and the added risk of complication for pleural biopsy. For example, bedside blinded pleural biopsy is more commonly performed alongside diagnostic thoracentesis in tuberculosis endemic area for the exclusion of tuberculous pleuritis. 33,39 In the latest Society of Interventional Radiology consensus guidelines, 40 pleural biopsy is regarded as deep organ biopsy and should only be performed under strict monitoring of platelet count and clotting profile. Whereas thoracocentesis is considered as a procedure with low bleeding risk, thus does not require withholding T A B L E 6 Comparison of diagnostic accuracy between pleural biopsy and effusion cytology in clinically confirmed a) hematolymphoid malignancy, b) mesothelioma, and c) metastatic carcinoma  anticoagulation or antiplatelet medication beforehand. Therefore, if there is a high pretest probability of carcinomatous MPE, an early diagnostic thoracentesis right after clinical consultation with real-time ultrasound guidance may be preferable, if the patient has low bleeding tendency or is even on antiplatelet agents. The techniques in pleural biopsy have undergone significant improvements over the period in which data were collected for this cohort. The performance of pleural biopsy has been improved by prebiopsy ultrasound localization of pleural abnormality and real-time image guidance. 41 The sensitivity and accuracy of pleural biopsy in this cohort may have been lowered by the inclusion of blinded biopsy, variation in operator experience, and usage of less recent methods. Another limitation was that the cohort assembled before the introduction of the International system for reporting of serous fluid. 3 The diagnostic criteria and reproducibility may be negatively impacted. Volume criteria from the International system were not incorporated in this cohort, but its effect may be rather insubstantial as the insufficiency rate was low (0.3%).

| CONCLUSION
This large cohort of paired pleural biopsy and effusion cytology specimens demonstrated that the inadequate/ insufficient (B1/C1) rate was higher in biopsy than cytology, while that of the other diagnostic categories were similar. Biopsy and cytology showed fair to moderate concordance depending on whether inadequate/insufficient (B1/C1) specimens were included. The malignant (B5/ C5) and suspicious (B4/C4) categories had similar ROMs. The ROM for atypia in biopsy (B3) was higher than that of cytology (C3). Cytology showed better performance in terms of ROM for the benign category (B2/C2) and overall diagnostic accuracy. Subgroup analysis showed that cytology was more accurate in carcinomatous MPEs but not for hematolymphoid malignancies and mesotheliomas. These results suggest that for patients with a high pretest probability of MPE, particularly when metastatic carcinoma is suspected, diagnostic thoracentesis alone may be sufficient as the initial step. However, for workup of hematolymphoid malignancies and mesothelioma, biopsy should be considered.

ETHICS STATEMENT
This study was approved by the Joint Chinese University of Hong Kong -New Territories East Cluster Clinical Research Ethics Committee (reference number: 2020.289).