Fluids—Good and Bad Actors: Observations From the College of American Pathologists Interlaboratory Comparison Program in Nongynecologic Cytology

Body cavity fluid specimens are a common diagnostic dilemma in daily practice. Difficulties arise in distinguishing “reactive” or negative samples from malignant disorders involving the serous-lined body cavities. We attempted to subjectively assess the characteristics of specimens that performed well and compare them to poorly performing body cavity fluids used in the College of American Pathologists (CAP) Interlaboratory Comparison Program in Nongynecologic Cytology (NGC).

The CAP NGC program began in 1997. During the following 4 years, it grew to include 1121 laboratories, with 3914 pathologists and 3267 cytotechnologist subscribers. This purely educational program consists of a quarterly glass slide mailing with 5 slides and a brief history. Individual participants use the CAP NGC as an educational tool and a measure of their performance compared with other laboratories. However, since slides circulate extensively, most slides have a unique accrued slide profile, allowing evaluation of the characteristics of each slide in the program.

The members of the CAP Cytopathology Resource Committee submit cases used for the CAP NGC program. There are a broad variety of cases in the program, including washing and brushing specimens from the respiratory tract, gastrointestinal tract, and urinary tract. Cerebrospinal fluid and fine needle biopsy specimens of various body sites are also included. Body cavity fluids are a large component of the specimens in the program, including pleural, peritoneal, and pericardial sites.

Cases are selected for the NGC program by consensus at screening sessions. Two members, who agree on a general and specific interpretation, review all slides. The 2 members also agree that the slide is a good representation of the entity and that it is technically adequate for assessment. After circulation in the field, performance characteristics of each slide are available. Those slides that perform poorly are categorized as “least concordant,” and 2 committee members review the slides and decide to maintain the slides in the program or eliminate them if technically inadequate or a misidentification had occurred.

Using data from 4 years of the CAP NGC program, subjective assessment of “least concordant” and “good performer” body cavity fluids resulted in a summary of characteristics for specimens that performed well. Conversely, characteristics of “poor performers” were identified.

Cumulative histories of all responses for 1997–2001 were obtained through the CAP Scores computer system. Body cavity fluids selected included pleural, pericardial, and peritoneal fluids. The cases were divided into reference diagnostic categories for comparison within category. Positive categories were adenocarcinoma, squamous cell carcinoma, small cell carcinoma, mesothelioma, melanoma, and lymphoma. Negative categories combined normal or reactive mesothelial cells and inflammation.

Good performers were slides that had a cumulative maximum agreement by laboratory response to the reference diagnosis. Poor performance cases were identified by minimum cumulative agreement by laboratory response to the reference diagnosis. Concordance was defined as the number of laboratory responses identical to the reference diagnosis, expressed as a percentage of total laboratory responses for that slide during a 4-year interval.

Four members of the Cytopathology Resource Committee simultaneously reviewed most concordant and least concordant cases for each diagnostic category. They first confirmed the diagnosis and then evaluated each case for cellularity, numbers of abnormal cells, and specific diagnostic criteria, depending on the reference diagnosis. After consensus review, features describing good performers and observations noted to cause poor performance in the NGC program were identified.

There were 10 396 laboratory responses for slides representing pleural, peritoneal, and body cavity fluids during a 4-year period, 1997–2001. Pathologists responded to 18 350 challenges and cytotechnologist responses totaled 12 640 (Table 1). Diagnosis specific responses are seen in Table 2.

Good performers for the reference diagnosis of adenocarcinoma demonstrated 100% cumulative agreement as a laboratory response. These cases demonstrated increased cellularity with 3-dimensional clusters with smooth borders (Figure 1). The cells demonstrated cytomegaly, eccentric nuclei with cytoplasmic vacuoles, and indentation of the nuclei. Nuclear irregularity and coarse chromatin were present. Nucleoli were inconstant as a feature of good performers. There were at least 2 distinct populations of cells in good performance adenocarcinoma; adenocarcinoma cells were distinct from relatively abundant mesothelial cells.

Poorly performing adenocarcinoma cases ranged from 0% to 41% concordance with the reference diagnosis. Several patterns emerged from review: hypocellular adenocarcinoma, single cell pattern of adenocarcinoma, and hypercellular adenocarcinoma without a distinct second population of cells. Hypocellular adenocarcinoma, in which there were few malignant cells identified, demonstrated a dual population, but the malignant single cells were presumably undetected by participants in the CAP NGC. Poor performers with a single cell pattern demonstrated a monotonous cell population with few mesothelial cells for comparison. Single cell pattern adenocarcinomas demonstrated classic criteria of malignancy, but the changes were less obvious than hypocellular adenocarcinoma, with a background of benign mesothelial cells. Cells of single cell pattern adenocarcinoma had irregular nuclei, increased nuclear-cytoplasmic ratios, and eccentric nuclei (Figure 2, A). Chromatin was often pale. Occasional signet ring cells with indented nuclei were present (Figure 2, B). Hypercellular adenocarcinoma was highly cellular, with 3- dimensional cell clusters (Figure 2, C). The cells were relatively uniform, with scalloped borders, but without the intercellular windows usually seen in mesothelial clusters (Figure 2, D). Nucleoli were obvious. Failure to compare the dominant malignant population to the scattered background mesothelial cells was likely the cause of poor performance.

Good performers in the squamous carcinoma group demonstrated a “dirty” or necrotic background with atypical keratinizing cells (Figure 3, A). Poor performers (22% concordance) had no keratinizing cells and fewer malignant cells (Figure 3, B).

In contrast to good performers, poorly performing small cell undifferentiated carcinoma had few malignant cells (11%–41% concordance). However, nuclear molding, coarse chromatin, and rare linear cellular arrays were present. Failure to detect rare malignant cells contributed to poor performance (Figure 4).

Poorly performing mesotheliomas (4%–33% concordance) demonstrated increased cellularity, papillary 3-dimensional clusters, and few single cells (Figure 5, A). Most poorly performing slides were confused with adenocarcinoma. Good performers were composed of mostly single cells in a monotonous population. Cytoplasm was dense with a “fluffy” rim, small cytoplasmic vacuoles, and cytoplasmic windows (Figure 5, B).

Epithelioid melanomas with abundant cellularity and a discohesive pattern (single cells), prominent nucleoli, cytoplasmic vacuoles, and nucleomegaly characterized good performers. The absence of melanin pigment was not associated with poor performance. In contrast, poor performers (0%–58% concordance) were sparsely cellular with rare binucleate cells with prominent nucleoli (Figure 6). Without supporting history or confirmatory ancillary stains, these cases were not identifiable as melanoma.

Good performing lymphoma cases demonstrated concordance greater than 80%. Most slides were highly cellular with a single cell discohesive pattern. Apoptosis was present. History was undoubtedly contributory to the good performance of the lymphoma cases; many cases were accompanied by a history, which pointed to a hematopoietic neoplasm (Figure 7). Poor performers (<38% concordance) consisted of nearly pure populations of atypical large, pleomorphic single cells. However, some cases demonstrated infrequent multilobated cells with giant nuclei and prominent nucleoli, reminiscent of Reed-Sternberg cells. The most common diagnosis of those cases was carcinoma or melanoma.

Normal or reactive fluids demonstrated good performance when a polymorphous cell population was present, including lymphocytes, rare plasma cells, neutrophils, and mesothelial cells (Figure 8). Poor performing negative cases were those in which an overwhelming lymphocytic population was present, suggesting a hematopoietic neoplasm, or reactive mesothelial changes were interpreted as adenocarcinoma.

Body cavity fluids are a common specimen in most pathology practices and present difficult diagnostic challenges.1–3 The problems in everyday practice are the same challenges facing participants in the CAP NGC. Using a 4-year CAP slide performance database allows us to compare those slides that most laboratories identified correctly (good performers) with those that performed poorly in their reference diagnosis category (bad performers).

Unlike daily practice, the slides of the CAP NGC are “optimized” slides. Two Cytopathology Resource Committee members critically evaluate slides before acceptance into the NGC program. They must confirm the diagnosis and agree that the slides are technically acceptable and that the slide is a good representation of the diagnosis. Also, unlike daily practice, the CAP NGC program uses a single slide for interlaboratory comparison. There are no cell blocks or ancillary studies. This limits participants to morphologic interpretations only. What causes slides to perform poorly? And what lessons can we learn from the CAP NGC bad actors that will help in daily practice?

Poor performance in the CAP NGC relates to location error and interpretive error. Failure to locate malignant cells occurs most frequently in poorly performing adenocarcinoma, squamous cell carcinoma, small cell undifferentiated carcinoma, melanoma, and hypocellular lymphoma with large multilobate nuclei. Despite demonstrating classic diagnostic malignant features, the malignant cells are rare and overlooked. Identifying hypocellular or “rare event” malignancy is a function of careful, methodical screening, whether the case is for CAP NGC or a patient specimen.

Interpretive error occurs in all diagnostic reference categories. Poor performance occurs when malignant cells are seen singly and, surprisingly, when malignant cells are abundant and in clusters. The single cell pattern of adenocarcinoma is a common feature contributing to poor performance. Despite abundant malignant cells on the slide, the cells are difficult to distinguish from mesothelial cells and require critical application of classic cytologic criteria.4 

Helpful criteria for identifying single cell adenocarcinoma include nuclear membrane abnormalities, eccentric nuclei, and indentation of the nuclei by vacuoles when cytoplasmic vacuoles are noted.5 

Highly cellular poor performers were noted in the adenocarcinoma and mesothelioma reference categories. In both reference diagnostic categories, the malignant cells formed abundant 3-dimensional configurations. In both categories, rare benign cells were available for comparison on the slides.5 Helpful features used by the Cytopathology Resource Committee members for identifying hypercellular adenocarcinoma included 3-dimensional clusters of monotonous cells, cells with scalloped cytoplasmic borders, lack of intercellular windows, and finely granular or clear cytoplasm. Malignant nuclear features included nuclei that extended to the edge of the cell, slight nuclear irregularities, and prominent nucleoli. The cases of mesothelioma that were poor performers were those that demonstrated a papillary pattern, without the monotonous population of single cells with dense cytoplasm with the “fluffy” rims, characteristic of classic mesothelioma.6 

Slides with the reference diagnosis of lymphoma or hematopoietic malignancy performed well if there was an accompanying history and a monotonous, hypercellular population of cells. Reactive or negative cases were often mistaken for lymphoma if lymphocytes predominated. Negative or reactive cases that performed well demonstrated a polymorphous population of cells, including mesothelial cells, lymphocytes, plasma cells, and neutrophils.

Using a 4-year cumulative slide history from the CAP NGC database, we compared slides with good cumulative performance characteristics to slides that performed poorly. This database included 10 396 laboratory responses of body cavity fluids from 1997 to 2001. After consensus confirmation of the diagnosis, the committee evaluated the characteristics of the best and the worst performing slides. Trends leading to poor performance include rare events, malignancy involving single cells, and hypercellular malignancy. Careful screening, application of classic criteria, and careful search for a dual population may help improve performance characteristics of fluid specimens in the CAP NGC and in daily practice. Lessons from bad actors in the CAP NGC program may also help identify areas for laboratory quality improvement.