Refined pancreatobiliary UroVysion criteria and an approach for further optimization

Abstract

Pancreatobiliary strictures are a common source of false negatives for malignancy detection. UroVysion is more sensitive than any other method but remains underutilized because of conflicting sensitivities and specificities due to a lack of standardized cutoff criteria and confusion in interpreting results in the context of primary sclerosing cholangitis. We set out to determine the sensitivities and specificities of UroVysion, brushing cytology, forceps biopsies, and fine needle aspiration (FNAs) for pancreatobiliary stricture malignancy detection. A retrospective review was performed of all biopsied pancreatobiliary strictures at our institution over 5 years. UroVysion was unquestionably the most sensitive method and all methods were highly specific. Sensitivity was highest while maintaining specificity when a malignant interpretation was limited to cases with 5+ cells with the same polysomic signal pattern and/or loss of one or both 9p21 signals. Only UroVysion detected the metastases and a neuroendocrine tumor. In reviewing and analyzing the signal patterns, we noticed trends according to location and diagnosis. Herein we describe our method for analyzing signal patterns and propose cutoff criteria based upon observations gleaned from such analysis.

Keywords: FISH; cytogenetics; cytopathology; pancreatobiliary; stricture.

FIGURE 1

This is an abridged representation of an actual score sheet from one of the pancreatic ductal adenocarcinoma (PDAC) cases. The signal patterns could mistakenly be attributed to independent gains of 3, 7, and 17. This pattern is instead the result of hemizygous loss of 9p21 followed by whole‐genome doubling. As shown in Table 1, in PDACs loss of 9 is more common than the gain of 3, 7, or 17

FIGURE 2

This is an abridged representation of an actual score sheet from one of the pancreatic ductal adenocarcinoma (PDAC) cases. The signal patterns can be seen to be due to cytogenetic instability in a tetrasomic cell population instead of from gains in a diploid population. Based on the frequencies with which each aneusomy is seen in PDAC (listed in Table 1) as compared to the frequency of whole‐genome doubling (WGD) in solid tumors, it is far more likely that this many chromosomes with exactly four signals would be due to WGD than independent gains. Furthermore, when a number other than four signals is present, the number is either one less or one more than four. When three signals are seen the possibilities could be that the cell gained a copy of that chromosome or that there were four copies and one was lost. When three of five cells contain three signals for a given probe and the other cells contain four signals for said probe it is usually more likely that copies were lost, especially when the probe in question is CEP17 and the lesion is a PDAC. Loss of 17, loss of 9p21, and gain of 3 are the three most common whole chromosomal aneusomies affecting UFISH probes and occur in a 3:1:1 as illustrated by the ideogram from Kowalski et al. When the observed signal patterns could not be explained by the expected aneusomic frequencies it was denoted in the inferred cytogenetic sequence by RLAG (random losses and gains)

FIGURE 3

This is an abridged representation of an actual score sheet from a benign case. If one were using polysomy in four cells as the threshold for malignancy this would be interpreted as positive since Cells 1–4 would be called polysomic. Cell 2 may simply represent a tetrasomic cell that randomly lost one CEP17 signal since there are other tetrasomic cells and chromosome 17 is frequently lost. In cases with multiple tetrasomic cells, it is common to see one or more that have lost a copy of CEP17 or another probe. This case illustrates why we believe that positivity should require five instead of four polysomic cells, that the polysomic cells have the same signal pattern, and that cases with tetrasomic or near‐tetrasomic cells be interpreted with caution

FIGURE 4

This is an abridged representation of an actual score sheet from a pancreatic ductal adenocarcinoma (PDAC) case. This example demonstrates how one can deduce the order of events from the different populations of cells. Looking at Cell 1 one may wonder whether 3 or 7 was gained first and may not consider that 17 and 9p21 were lost. Cell 4 having only one CEP17 signal and one 9p21 signal reveals that at least some cells lost these signals at some point. After noticing this, it becomes evident that Cells 1–3 represent cells such as Cell 4 that have undergone whole‐genome doubling (WGD). This tells you that the losses of both 17 and 9p21 occurred prior to WGD. The presence of cells such as Cells 6 and 7 combined with the absence of cells showing loss of only 9p21 suggests that loss of 17 preceded loss of 9p21. When the signal patterns of the different cells did not allow for the determination of one event preceding or following another it was denoted in the inferred cytogenetic sequence by OCBD (order cannot be determined)