Article Text

Download PDFPDF
Invasive carcinoma versus pseudoinvasion: interobserver variability in the assessment of left-sided colorectal polypectomies
  1. Michael Lee1,
  2. Satoru Kudose1,
  3. Armando Del Portillo1,
  4. Huaibin Mabel Ko2,
  5. Hwajeong Lee3,
  6. Meredith E Pittman4,
  7. Marcela A Salomao5,
  8. Antonia R Sepulveda1,
  9. Stephen M Lagana1
  1. 1Department of Pathology and Cell Biology, Columbia University Medical Center, New York City, New York, USA
  2. 2Department of Pathology and Laboratory Medicine, Icahn School of Medicine, New York City, New York, USA
  3. 3Department of Pathology and Laboratory Medicine, Albany Medical Center, Albany, New York, USA
  4. 4Department of Pathology and Laboratory Medicine, Weill Cornell Medical Center, New York City, New York, USA
  5. 5Department of Laboratory Medicine and Pathology, Mayo Clinic, Phoenix, Arizona, USA
  1. Correspondence to Dr Michael Lee, Department of Pathology and Cell Biology, Columbia University Medical Center, New York, NY 10032, USA; mjl2197{at}cumc.columbia.edu

Abstract

Objectives Misplaced epithelium in adenomas can occasionally be difficult to distinguish from invasive adenocarcinoma. We evaluated interobserver variability in the assessment of left-sided colon polypectomies for pseudoinvasion versus invasive adenocarcinoma and further investigated relevant histological findings.

Methods 28 consecutive left-sided colon polyps with the keywords “pseudoinvasion”, “epithelial misplacement”, “herniation”, “prolapse” or “invasive adenocarcinoma” were collected from 28 patients and reviewed by eight gastrointestinal pathologists. Participants assessed stromal hemosiderin, lamina propria/eosinophils surrounding glands, desmoplasia, high grade dysplasia/intramucosal adenocarcinoma and margin status and rendered a diagnosis of pseudoinvasion, invasive adenocarcinoma, or both.

Results Agreement among pathologists was substantial for desmoplasia (κ=0.70), high grade dysplasia/intramucosal adenocarcinoma (κ=0.66), invasive adenocarcinoma (κ=0.63) and adenocarcinoma at the margin (κ=0.65). There was moderate agreement for hemosiderin in stroma (κ=0.53) and prolapse/pseudoinvasion (κ=0.50). Agreement was low for lamina propria/eosinophils around glands (κ=0.12). For invasive adenocarcinoma, seven or more pathologists agreed in 24 of 28 cases (86%), and there was perfect agreement in 19/28 cases (68%). For pseudoinvasion, seven or more pathologists agreed in 19 of 28 cases (68%), and there was perfect agreement in 16/28 cases (57%).

Conclusion Moderate to substantial, though imperfect, agreement was achieved in the distinction of pseudoinvasion from invasive carcinoma.

  • carcinoma
  • colon
  • colorectal neoplasms
  • intestine
  • large

Data availability statement

All data relevant to the study are included in the article.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

Pseudoinvasion, or epithelial misplacement, in polyps occurs when adenomatous mucosa herniates into submucosal tissue, a process that occurs not infrequently in left-sided polyps. This can be mistaken for T1 colorectal cancer. Given that herniated adenoma is a benign lesion with no risk for metastasis, overinterpretation as adenocarcinoma can lead to unnecessary surgical resection (with its inherent morbidity and mortality risks), increased frequency of endoscopic surveillance and no trivial degree of anxiety.1–4 Pseudoinvasion occurs in approximately 2%–4% of adenomas, more often on the left colon with the rectosigmoid colon as the most common site.5 6 In a study of 256 colon polyps with pseudoinvasion, 78.9% were in the rectosigmoid colon. The left colon is a common location because of its anatomically narrow lumen and predilection for diverticulosis.7

There are several histological features that may aid in the identification of pseudoinvasion over invasive carcinoma including lobulated architecture, hemosiderin-laden macrophages, lamina propria/eosinophils surrounding the crypts and mucin pools. Desmoplasia and differing degree of dysplasia favour invasive carcinoma.5–9 However, this can be a challenging differential with significant treatment implications and microscopic assessment may vary between pathologists.10 This study evaluated the interobserver variability in the assessment of left-sided colon polypectomies for pseudoinvasion versus invasive carcinoma and further assessed specific histological features of interest. We were also interested in observer’s confidence in each case (ie, were we aware of which cases were difficult).

Materials and methods

This study was a collaborative effort between the surgical pathology departments at Columbia University Irving Medical Center (CUIMC), Icahn School of Medicine, Albany Medical Center, Weill Cornell Medical Center and Mayo Clinic. Following institutional research board approval, the pathology archive of CUIMC was searched for left-sided polypectomies with the keywords “pseudoinvasion”, “epithelial misplacement”, “herniation” or “prolapse” in the diagnosis or microscopic description from January 2015 to April 2020, and 19 cases were identified. A cohort of nine left-sided polypectomies diagnosed as ‘invasive adenocarcinoma’ in the diagnostic line were also included. All cases were consecutive. Following the entirety of the review, patient charts were queried for recurrence/metastasis.

These slides were anonymised, digitally scanned and reviewed by eight experienced gastrointestinal pathologists across multiple medical institutions. The definition of ‘gastrointestinal pathologist’ was a practising pathologist who had done a gastrointestinal pathology fellowship and spends the majority of their clinical time on gastrointestinal specimens. For each case, we examined and recorded the presence or absence of hemosiderin in the stroma, lamina propria/eosinophils surrounding the glands, desmoplasia, high grade dysplasia/intramucosal carcinoma, margin status and whether these findings represented pseudoinvasion/epithelial misplacement or invasive carcinoma in their assessment. The margin status was recorded as 0, 1 or NA with 0=negative for invasive tumour cells at the margin, 1=positive for invasive tumour cells at the resection margin and NA (no invasive adenocarcinoma in the polypectomy). Each pathologist was also asked to record their degree of confidence in the diagnostic assessment of pseudoinvasion or adenocarcinoma as 0, 1 or 2 (0=not at all confident, 1=fairly confident and 2=extremely confident).

Statistics: Fleiss’ κ statistics were computed to evaluate interobserver reliability among the eight gastrointestinal pathologists. All statistical analysis was performed using R (V.3.6.1). κ statistics were interpreted according to the Landis and Koch scale (κ values: <0: poor agreement, 0.01–0.20: slight agreement: 0.21–0.40: fair agreement, 0.41–0.60: moderate agreement, 0.61–0.80: substantial agreement and >0.80: nearly perfect agreement). κ scores range from −1 (complete disagreement) to 1 (perfect agreement).11 Spearman’s correlation (ρ) between histopathological features were computed. Statistical significance was assumed at p<0.05.

Results

In total, 19 patients with left-sided polypectomies diagnosed as either tubular or tubulovillous adenomas with pseudoinvasion, epithelial misplacement, herniation or prolapse type changes were identified with available H&E stained slides as were nine consecutive left-sided polypectomies diagnosed as invasive adenocarcinoma, meaning the entire test cohort was comprised of 28 consecutive cases. Figure 1 highlights the microscopic features and criteria we evaluated to diagnose pseudoinvasion (figure 1A,B) versus adenocarcinoma (figure 1C,D). Table 1 demonstrates that agreement among pathologists was substantial for desmoplasia (κ=0.70), high grade dysplasia/intramucosal adenocarcinoma (κ=0.66), invasive adenocarcinoma (κ=0.63) and invasive adenocarcinoma at the resection margin (κ=0.65). There was moderate agreement for hemosiderin in stroma (κ=0.53) and prolapse/pseudoinvasion (κ=0.50). Agreement was low for lamina propria/eosinophils around glands with slight agreement (κ=0.12) (figure 2).

Figure 1

Tubular glands within the lamina propria are surrounded by hemosiderin laden macrophages and stroma composed of eosinophils and lymphocytes consistent with pseudoinvasion (A and B, 4× and 10× magnification, H&E stain). Tubular glands with a cribriform architecture and higher grade cytologic atypia are surrounded by a desmoplastic stroma composed of dense connective tissue. These features are consistent with an invasive adenocarcinoma (C and D, 4× and 10× magnification, H&E stain).

Figure 2

Two cases with a high level of interobserver variability and low agreement. While there is a lobulated architecture with surrounding lamina propria and eosinophils, there is also severe cytological atypia and a glandular profile with retraction (A and B, 4× and 10× magnification, H&E stain). Case 2 shows a lobulated group of glands at the cauterised, inked resection margin of a polypectomy (C, 4× magnification, H&E stain). Lobulated architecture suggests benignity. On higher power, there is arguable desmoplasia and a focus of small, angulated glands concerning for carcinoma; however, these are both crushed and cauterised (D, black arrow, 20× magnification, H&E stain). Case 2 was classified as pseudoinvasion by two pathologists; however, the patient developed subsequent metastases to the liver.

Table 1

Interobserver variability among eight gastrointestinal pathologists for 28 cases

Table 2 summarises the number of gastrointestinal pathologists who agreed with each histopathological finding. For invasive adenocarcinoma, five or more pathologists agreed in 27 of 28 cases (96%), six or more pathologists agreed in 25 of 28 cases (89%) and seven or more pathologists agreed in 24 of 28 cases (86%), and there was perfect agreement among all 8 gastrointestinal (GI) pathologists in 19/28 cases (68%). Thus, 9/28 cases (32%) failed to achieve perfect agreement when assessing for invasive adenocarcinoma. For pseudoinvasion, five or more pathologists agreed in 28 of 28 cases (100%), six or more pathologists agreed in 24 of 28 cases (86%) and seven or more pathologists agreed in 19 of 28 cases (68%), and there was perfect agreement in 16/28 cases (57%). Conversely, 12/28 cases (43%) failed to achieve perfect agreement when assessing for pseudoinvasion.

Table 2

Per cent agreement for histopathological features among eight pathologists for 28 cases

We analysed the relationship between a diagnosis of invasion and individual histologic features for each pathologist (table 3). The presence of desmoplasia was most strongly correlated with the diagnosis of invasion for every observer (Spearman’s ρ: 0.68–1, median 0.95; all p<0.05). Histological findings that were negatively associated with a diagnosis of invasion included pseudoinvasion, hemosiderin and eosinophils, though only the presence of pseudoinvasion reached statistical significance (Spearmans’ ρ: −0.52 to −1; median −0.68; all p<0.05). The presence of hemosiderin and adjacent eosinophils demonstrated a tendency towards negative correlation with a diagnosis of invasion for most pathologists (Spearman’s ρ: median −0.305 and −0.225, respectively), but these associations were statistically significant for only two pathologists. Table 4 demonstrates the correlation between a diagnosis of pseudoinvasion and individual histologic features.

Table 3

Spearman’s ρ between the diagnosis of invasion and other microscopic features

Table 4

Spearman’s ρ between the presence of pseudoinvasion and other features

A greater degree of agreement for the presence of invasion was seen in cases with high confidence rating. The degree of confidence was positively associated with agreement for the presence of invasion (Spearman’s ρ: 0.76, p<0.05, figure 3). Furthermore, the agreement for the presence of invasion was near perfect (κ=0.90, 95% CI 0.92 to 0.99) in 19 cases that were rated as at least fairly confident by all pathologists. In contrast, the degree of confidence only showed a tendency towards positive association with agreement for the presence of pseudoinvasion (Spearman’s ρ: 0.32, p=0.09, figure 4). In 19 cases rated as at least fairly confident by all pathologists, only moderate to substantial agreement (κ=0.62, 95% CI 0.53 to 0.70) for the presence of pseudoinvasion was achieved.

Figure 3

Scatterplot of mean confidence and per cent overall agreement among eight pathologists for the presence or absence of invasive adenocarcinoma for each case (n=28).

Figure 4

Scatterplot of mean confidence and per cent overall agreement among eight pathologists for the presence or absence of pseudoinvasion for each case (n=28).

There is no definitive gold standard, as only the presence of metastatic disease renders an opinion incontrovertible (its absence does not). Nonetheless, we searched the medical records for each patient, and 1 of 28 patients developed liver metastases. For this case, six of eight pathologists diagnosed it as invasive adenocarcinoma and two of eight pathologists thought it was pseudoinvasion (figure 2). The other 27 patients either had unremarkable follow-up screening colonoscopies, partial colectomy without nodal disease or additional clinical information was unavailable.

Discussion

The histological features and criteria pathologists evaluate in the consideration of pseudoinvasion versus adenocarcinoma have been well characterised in the literature.9 12 13 However, we observed that this issue is relatively common in consultation material and that there can be disagreements among participants in intradepartmental consensus meetings. Such discussions can become spirited as this distinction is pivotal in deciding whether a patient needs surgical intervention.

This study shows that diagnostic agreement as to whether there is or is not invasive carcinoma in a polypectomy is substantial. The fact that this is a multi-institutional study suggests that the suggested published criteria are generally broadly followed. There remains, however, room for improvement. Studies have looked at the utility of MMP-1, p53, E-cadherin, Ki-67 and collagen IV immunohistochemical stains to elucidate difficult cases. Invasive adenocarcinoma shows increased nuclear p53 reactivity and MMP1 stromal staining, with a decrease in e-cadherin and collagen IV. However, routine use of such ancillary testing is rare to our knowledge.14 15 In our experience, evaluation of deeper levels and review with colleagues prior to sign out is the more typical ‘work up’ of such cases.

The selection of consecutive cases was a strength of the study, as we did not want to select a cohort of particularly hard cases, in which suboptimal reproducibility would be the expected finding. The result that all eight pathologists agreed on the diagnosis of invasive adenocarcinoma in only 68% of cases was surprising, although seven of eight agreed in 86%, which is somewhat reassuring. These results are likely to reflect a ‘worst-case (though still “real world”) scenario’ since the ability to section deeper into the block and seek consensus was not available to study participants. In fact, greater degree of agreement for the presence of invasion in cases with high confidence rating (figures 3 and 4) suggests that pathologists were aware which cases were the most diagnostically challenging. Flagging these cases accordingly would allow for the ‘typical steps’ described above. However, pseudoinvasion appears to be a more challenging diagnosis to make. Two key histological features for pseudoinvasion, lamina propria/eosinophils surrounding glands and hemosiderin-laden macrophages showed lower levels of agreement among pathologists, and the degree of confidence only showed a tendency towards positive association with agreement. Given that most experts consider desmoplasia as being pathognomonic for invasion in colonic adenocarcinoma and that for some observers the correlation between desmoplasia and invasion was less than 1, it is likely some participants were unsure in certain instances as to whether focal fibrosis did or did not represent true desmoplasia. It is not surprising that most pathologists did not regard pseudoinvasion and true invasion as mutually exclusive since whole slides were evaluated. Though any particular glandular focus which appears to be in the submucosa should be classifiable as either invasive or pseudoinvasive, there could be both foci of prolapsed/pseudoinvasive glands and distinct foci of true invasion. Other potential challenges include, tangential orientation, cautery artefact, uncertainty regarding lymphovascular invasion versus retraction artefact, arguable desmoplasia, and pseudoinvasion and invasive glands within the same case. Some of these diagnostic pitfalls are highlighted in figure 2.

We have shown that among experienced gastrointestinal pathologists, evaluation of left-sided colon polyps for pseudoinvasion versus invasive carcinoma is relatively reproducible but that room for improvement remains. The distinction between pseudoinvasion and invasion ultimately may have a significant clinical impact, as the decision to perform a major surgery may hinge on this assessment. Unless and until there can be a more readily applicable gold standard than the rare event of metastatic disease in the setting of a superficially invasive polyp, we suggest sampling deeply into the block and soliciting additional opinions. Pathologists seem to recognise which cases are likely to be controversial.

Take home messages

  • Pseudoinvasion in adenomas can occasionally be difficult to distinguish from invasive adenocarcinoma. We evaluated interobserver variability in the assessment of left-sided colon polypectomies for pseudoinvasion versus invasive adenocarcinoma.

  • Agreement among pathologists was substantial for desmoplasia (κ=0.70) and invasive adenocarcinoma (κ=0.63). There was moderate agreement for prolapse/pseudoinvasion (κ=0.50).

  • Among experienced gastrointestinal pathologists, evaluation of left-sided colon polyps for pseudoinvasion versus invasive carcinoma is relatively reproducible, but room for improvement remains.

  • Until a more readily applicable gold standard than the rare event of metastatic disease in the setting of a superficially invasive polyp, we suggest sampling deeply into the block and soliciting additional opinions.

Data availability statement

All data relevant to the study are included in the article.

Ethics statements

Ethics approval

This study was approved by the appropriate institutional research boards of each institution.

References

Footnotes

  • Handling editor Runjan Chetty.

  • Contributors All authors except for SK reviewed the digital slides, provided diagnostic assessments and contributed to writing the manuscript. SK performed the statistical analysis and created tables and scatterplots.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.