Context.—

Neoplastic cellularity assessment has become an essential component of molecular oncology testing; however, there are currently no best practice recommendations or guidelines for this potentially variable step in the testing process.

Objective.—

To describe the domestic and international practices of neoplastic cellularity assessment and to determine how variations in laboratory practices affect neoplastic cellularity assessment accuracy.

Design.—

Data were derived from 57 US and international laboratories that participated in the 2019 College of American Pathologists Neoplastic Cellularity Proficiency Testing Survey (NEO-B 2019). NEO-B 2019 included 29 laboratory practice questions and 5 images exhibiting challenging histologic features. Participants assessed the neoplastic cellularity of hematoxylin-eosin–stained digital images, and results were compared to a criterion standard derived from a manual cell count.

Results.—

The survey responses showed variations in the laboratory practices for the assessment of neoplastic cellularity, including the definition of neoplastic cellularity, assessment methodology, counting practices, and quality assurance practices. In some instances, variation in laboratory practice affected neoplastic cellularity assessment performance.

Conclusions.—

The results highlight the need for a consensus definition and improved standardization of the assessment of neoplastic cellularity. We put forth an initial set of best practice recommendations to begin the process of standardizing neoplastic cellularity assessment.

In the era of precision medicine, molecular oncology testing has become critical to identifying clinically relevant mutations in a variety of cancer genes that are used in diagnosis, prognosis, and/or therapeutic management. Assessment of tissue adequacy and determination of neoplastic cellularity are part of a multistep process of successful molecular analysis. Inherent to all biopsy and surgical specimens are tissue heterogeneity and admixtures of both neoplastic and nonneoplastic cells. The ability to accurately assess tumor cellularity within a specimen is essential to both the analytic sensitivity and the interpretation of a molecular test's results.1 

Previous studies have demonstrated that the estimation of the percentage of neoplastic cellularity often exhibits interobserver variability and lacks precision.24  Contributors to this variation include differing definitions of neoplastic cellularity, multiple methodologies used for assessment, subjectivity in assessment of tissue content and tumor delineation, and lack of training and/or feedback. Although molecular testing guidelines state that the neoplastic cell content of each specimen should be assessed,5  no best practice recommendations or guidelines have been established for how to perform this assessment. However, given that neoplastic cellularity assessment plays such a critical role in molecular oncology testing and is a College of American Pathologists (CAP) Accreditation Program requirement (CAP checklist MOL.32395), standardizing this practice is of paramount importance.6 

The first goal of this study was to evaluate the current practices of neoplastic cellularity assessment as part of the multi-institutional CAP Neoplastic Cellularity (NEO) Proficiency Testing Survey. The second goal was to assess how variations in practices may affect performance on a series of performance challenge specimens. We suggest initial best practice recommendations for neoplastic cellularity assessment, based on our findings, with the intent of improving consistency in the performance of this determinative step in the molecular oncology testing process.

Data were derived from a single 2019 mailing of the CAP Neoplastic Cellularity Proficiency Testing Survey (NEO-B 2019). The survey included a total of 29 questions about laboratory practices (Supplemental Table 1, see the supplemental digital content containing 4 tables at https://meridian.allenpress.com/aplm in the September 2022 table of contents). In addition, five ×20 magnification digital pathology images were used in an educational performance challenge. In the performance challenge, participating laboratories were asked to estimate the neoplastic cellularity using their routine method (excluding necrotic cells) and to report that number as a percentage. The instructions did not specify if in situ carcinoma should be included in the assessment. Laboratories were also asked to determine whether neoplastic cellularity was sufficient for molecular testing based on the provided lower limit of detection of an assay. The participants' results were compared to the reference or criterion standard, which was derived by the manual counting of neoplastic versus nonneoplastic cells in each image, as previously described in Viray et al.2  Data for 12 repeated practice questions from the 2015–2019 (“A” mailing) NEO proficiency testing surveys were also analyzed in this study to determine whether these practices have changed over time.

Multivariate logistic regression models with exact methods were used to test for practice characteristic differences with institution location and type. Institution location was defined as a 2-level factor with domestic (United States) and international classifications. Institution type was defined as a 3-level factor for independent/commercial reference laboratory, academic (university or teaching) hospital/medical center laboratory, and other, which included nonacademic hospital/medical center laboratory and nonhospital site. For multivariate models that did not meet the model convergence criterion, the model was adjusted to only include the location factor. For models with significant institution type differences, pairwise testing was performed between independent/commercial reference laboratories and academic hospital/medical center laboratories. Wilcoxon rank sum tests were used to test for distribution differences between neoplastic cellularity results by the specific practice characteristic relevant to the case-specific histologic feature. To test for 5-year practice trends, generalized linear models were fit with a repeated measures covariance structure to account for the correlated results for laboratories participating in multiple mailings. Analyses were performed with SAS 9.4 and R 3.6.2. A significance level of .05 was used for the analyses.

Neoplastic Cellularity Assessment Practice and Interpretation

A total of 57 laboratories participated in the NEO-B 2019, which assessed current practices and performance. Of these laboratories, 37 (64.9%) were located in the United States and 20 (35.1%) were international (Table 1). The survey results encompassed the practices of different laboratory settings: 34 (59.6%) independent/commercial, 17 (29.8%) academic hospital, 3 (5.3%) nonacademic hospital, and 3 (5.3%) nonhospital sites that included 2 research laboratories.

Table 1

Laboratory Characteristics

Laboratory Characteristics
Laboratory Characteristics

Practices used for determining neoplastic cellularity varied across laboratories. In most laboratories, neoplastic cellularity was routinely determined by a pathologist who was board certified in anatomic pathology (AP; 84.2%; 48 of 57); however, in some laboratories, the assessment was completed by a clinical pathology (CP)–only board-certified pathologist (10.5%; 6 of 57) or a non–board-certified pathology trainee (5.3%; 3 of 57; Table 2). Neoplastic cellularity was typically estimated from glass slides (91.2%; 52 of 57), with estimations infrequently made from digital slides (15.8%; 9 of 57) or static images (1.8%; 1 of 57).

Table 2

Neoplastic Cellularity Assessment Practices

Neoplastic Cellularity Assessment Practices
Neoplastic Cellularity Assessment Practices

The definition of neoplastic cellularity varied among laboratories: 40 of the 57 laboratories (70.2%) defined neoplastic cellularity as the ratio of neoplastic cells or nuclei to the total number of cells or nuclei, whereas 16 laboratories (28.1%) defined it as the ratio of the area of neoplastic cells to the total area of cells. Overall, there was a greater tendency to estimate the neoplastic cellularity rather than perform a manual count (82.5% [47 of 57] versus 17.5% [10 of 57]). No laboratories reported using an automated counting method, such as an artificial intelligence–based system. Interestingly, the methods used by US and international laboratories differed significantly (P = .01). Estimation was by far the predominant method used by the 37 US laboratories (94.6%; 35 of 37), with only 2 laboratories (5.4%) using a manual counting strategy; however, the 20 international laboratories reported a more even distribution between those performing an estimation (60.0%; 12 of 20) versus a manual count (40.0%; 8 of 20). Overall, the neoplastic cellularity estimations were typically performed in less than 10 minutes (93.0%; 53 or 57), with only 4 laboratories reporting estimation times exceeding 10 minutes (7.0%).

A number of histologic features can potentially lead to interobserver variability in neoplastic cellularity assessments. The survey revealed that although most laboratories excluded apoptosis (77.2%; 44 of 57), necrosis (87.7%; 50 of 57), and the in situ component of an invasive tumor (73.7%; 42 of 57) in their neoplastic cellularity assessment, several laboratories included or were unsure how to assess such features (Table 3). Extracellular mucin is another feature that could potentially affect neoplastic cellularity assessment, particularly for those laboratories that define neoplastic cellularity based on tumor area. Of the 16 laboratories that determined neoplastic cellularity by area, 75.0% (12) excluded extracellular mucin, whereas 25.0% either included mucin (2) or were unsure whether mucin was included in their assessment (2).

Table 3

Neoplastic Cellularity Counting Practices

Neoplastic Cellularity Counting Practices
Neoplastic Cellularity Counting Practices

To enhance the neoplastic cellularity of a specimen and decrease the risk of a false-negative result, tumor enrichment methods prior to molecular testing are commonplace in laboratory practice. Of the 57 laboratory participants, 63.2% (36) performed enrichment on select cases, 31.6% (18) performed enrichment in all cases, and 5.3% (3) did not perform enrichment (Table 4). Tumor enrichment was reportedly done before performing a variety of molecular assays, including next-generation sequencing (85.2%; 46 of 54), single-gene polymerase chain reaction (64.8%; 35 of 54), Sanger sequencing (35.2%; 19 of 54), pyrosequencing (13.0%; 7 of 54), and other assays (ie, microsatellite instability testing, loss of heterozygosity testing, microarray testing, RNA-based gene expression analysis, and primer extension-based analysis; 9.3%; 5 of 54).

Table 4

Tumor Enrichment Practices

Tumor Enrichment Practices
Tumor Enrichment Practices

Laboratories reported using multiple methods for enrichment. The most common enrichment methodologies included manual macrodissection from a slide (90.7%; 49 of 54), manual macrodissection by coring (18.5%; 10 of 54), manual microdissection with a dissecting microscope (16.7%; 9 of 54), and laser-capture microdissection (3.7%; 2 of 54). Following tumor enrichment, 44.4% (24 of 54) of the laboratories reported routinely confirming the appropriate area was used for tumor enrichment. In laboratories that confirmed enrichment of the appropriate area, this step was typically performed by an AP board-certified pathologist (73.9%; 17 of 23); however, less frequently, laboratory technologists (17.4%; 4 of 23), CP board-certified pathologists (8.7%; 2 of 23), or a non–board-certified AP trainee (8.7%; 2 of 23) performed this step as well (Table 4).

More than half of the laboratories (54.4%; 31 of 57) refrained from testing when the neoplastic cellularity was below the determined minimum percentage cutoff or limit of detection (Table 5). Although some laboratories (38.6%; 22 of 57) proceeded with testing when the neoplastic cellularity was below this cutoff, the threshold for testing varied by laboratory and likely depended on the molecular assay used. A total of 4 of the 57 laboratories (7.0%) claimed to have no minimum cutoff and would perform testing regardless of the neoplastic cellularity. More than 80.0% of the laboratories used neoplastic cellularity either consistently (26 of 57; 45.6%) or occasionally (21 of 57; 36.8%) in the postanalytic interpretation of the results. When no variants were detected, 64.3% (36 of 56) of laboratories routinely or situationally re-reviewed the neoplastic cellularity. A subset of laboratories (61.4%; 35 of 57) documented the neoplastic cellularity percentage in the final report, most commonly as a numeric estimate (88.6%; 31 of 35). The CAP Molecular Oncology Committee has annually assessed the percentage of laboratories that document the neoplastic cellularity percentage in their final report. Although only 61.4% of laboratories in the 2019 survey reported documenting neoplastic cellularity in the final report, this percentage has significantly increased, from 14.3% (4 of 28) in 2015 (P < .001; Supplemental Table 2).

Table 5

Preanalytic and Postanalytic Neoplastic Cellularity Assessment Practices

Preanalytic and Postanalytic Neoplastic Cellularity Assessment Practices
Preanalytic and Postanalytic Neoplastic Cellularity Assessment Practices

Notably, some of the postanalytical practices of US and international laboratories, as well as those of academic and nonacademic laboratories, also differed significantly. For instance, 90.0% (18 of 20) of international laboratories consistently or occasionally used the neoplastic cellularity percentage in result interpretation compared with 78.4% (29 of 37) of US laboratories (P = .007). In addition, 77.8% (14 of 18) of the international laboratories confirmed that the appropriate area was used for tumor enrichment compared with 27.8% (10 of 36) of the US laboratories (P = .003). More international laboratories (80.0%; 16 of 20) also reported documenting the neoplastic cellularity of a sample in the final report than US laboratories (51.4%; 19 of 37; P = .04). Further, significantly more academic medical laboratories (94.1%; 16 of 17) used neoplastic cellularity (sometimes or always) in result interpretation compared with independent or commercial laboratories (76.5%; 26 of 34; P = .03; Supplemental Table 3).

Laboratories implemented several strategies to standardize and improve the quality of neoplastic cellularity assessment (Table 6). More than half (59.6%; 34 of 57) of the laboratories reported having implemented some form of training (eg, teaching slide set, online modules, peer training, or lecture), and three-fourths (42 of 56) had established a tumor cellularity procedure or protocol. Of the 22 institutions with a pathology residency training program, 54.5% (12) included neoplastic cellularity in the training curriculum. Approximately half of the laboratories (50.9%; 29 of 57) had performed an interobserver variability assessment between 2 different readers as a quality assurance measure. Finally, 63.6% (35 of 55) of the laboratories communicated, either routinely or occasionally, the variant allele fraction and percent of neoplastic cellularity back to the individual(s) who assessed the neoplastic cellularity.

Table 6

Standardization and Quality Assurance Practices

Standardization and Quality Assurance Practices
Standardization and Quality Assurance Practices

Neoplastic Cellularity Performance Challenge

In addition to the neoplastic cellularity practice assessment, NEO-B 2019 contained an educational performance challenge, with 5 images exhibiting features that had the potential to elicit discrepancies in neoplastic cellularity estimations. The 5 images consisted of breast carcinomas (2), colorectal carcinomas (2), and lung carcinoma (1). These images were used to evaluate stated laboratory practices in relation to performance. Specifically, the 5 images were selected to evaluate the participants' performance in the presence of in situ carcinoma, apoptosis, abundant lymphocytes, necrosis, and extracellular mucin (Figure 1; Table 7). Fifty-four participants provided assessments for all 5 images (Figure 2). Supplemental Table 4 contains a summary of the educational challenge neoplastic cellularity assessments by laboratory practice questions.

Figure 1

The 5 hematoxylin-eosin–stained ×20 magnification images distributed as part of the educational challenge in the College of American Pathologists Neoplastic Cellularity Proficiency Testing Survey (NEO-B 2019). A, NEO-95 ROI-3. B, NEO-96 ROI-2. C, NEO-97 ROI-3. D, NEO-98 ROI-3. E, NEO-99 ROI-1.

Figure 1

The 5 hematoxylin-eosin–stained ×20 magnification images distributed as part of the educational challenge in the College of American Pathologists Neoplastic Cellularity Proficiency Testing Survey (NEO-B 2019). A, NEO-95 ROI-3. B, NEO-96 ROI-2. C, NEO-97 ROI-3. D, NEO-98 ROI-3. E, NEO-99 ROI-1.

Close modal
Table 7

Summary of Performance Challenge Images and Participant Performance

Summary of Performance Challenge Images and Participant Performance
Summary of Performance Challenge Images and Participant Performance
Figure 2

Box-and-whisker plots demonstrating differences in the neoplastic cellularity estimates for the 5 performance challenge cases. Box plots extend to the minimum and maximum results.

Figure 2

Box-and-whisker plots demonstrating differences in the neoplastic cellularity estimates for the 5 performance challenge cases. Box plots extend to the minimum and maximum results.

Close modal

Image NEO-95 ROI-3 contained breast tissue with predominantly ductal carcinoma in situ and microscopic foci of invasive ductal carcinoma (Figure 1, A). The criterion standard was 3.1%. The overall mean response for this image was 27.8% with a standard deviation of 23.6%. The 10 participants that reported including in situ carcinoma in their estimation of neoplastic cellularity had significantly higher tumor cellularity estimations for this image (median, 45.0%) compared with the 40 participants that did not include in situ carcinoma (median, 10.0%; P = .01; Figure 3, A). Importantly, all 10 of the participants (100.0%) that included the in situ component in the assessment reported this image as adequate for molecular testing. Conversely, only 16 of the 40 participants (40.0%) that did not include in situ carcinoma reported the specimen as adequate. In addition, the percentage of all the performance challenge participants that classified the image as adequate was greater for participants without AP board certification (80.0%; 8 of 10) compared with AP board-certified participants (43.2%; 19 of 44), although the difference was not significant (P = .08).

Figure 3

Box-and-whisker plots demonstrating differences in the neoplastic cellularity estimates for performance challenge cases. A, Case NEO-95 ROI-3 results for participants who did and did not include in situ carcinoma in their neoplastic cellularity estimates. B, Case NEO-97 ROI-3 results for participants who estimated neoplastic cellularity by cell number versus tumor area.

Figure 3

Box-and-whisker plots demonstrating differences in the neoplastic cellularity estimates for performance challenge cases. A, Case NEO-95 ROI-3 results for participants who did and did not include in situ carcinoma in their neoplastic cellularity estimates. B, Case NEO-97 ROI-3 results for participants who estimated neoplastic cellularity by cell number versus tumor area.

Close modal

Image NEO-96 ROI-2 was from a breast carcinoma specimen notable for numerous apoptotic bodies, and it had a criterion standard of 22.0% (Figure 1, B). The mean response for this image was 40.1% with a standard deviation of 12.4%. Although there were no significant differences in estimation between laboratories that did and did not claim to count apoptotic cells (P = .59), 27.8% (15 of 54) of laboratories incorrectly reported this image as adequate or meeting the 50% threshold required for testing.

Image NEO-97 ROI-3 contained enlarged lung adenocarcinoma cells with numerous small lymphocytes and had a criterion standard of 9.3% (Figure 1, C). The overall mean response for this image was 22.3% with a standard deviation of 10.3%. For this image, there was an association between neoplastic cellularity practice and performance. Laboratories that estimated neoplastic cellularity by area had a significantly higher estimation (median, 30.0%; minimum–maximum, 15.0%–65.0%) than those estimating by neoplastic cell or nucleus number (median, 20.0%; minimum–maximum, 5.0%–40.0%; P < .001; Figure 3, B). In addition, 20.4% (11 of 54) of participants incorrectly reported that the neoplastic cellularity exceeded the 30% threshold required for testing. Only 3 of the 37 laboratories (8.1%) that estimated neoplastic cellularity by cell or nucleus number reported this image as adequate. This is considerably lower than the 53.3% (8 of 15) of laboratories that used the neoplastic area approach and indicated the image was adequate.

Image NEO-98 ROI-3 was from a colonic adenocarcinoma specimen that was notable for abundant necrosis and had a criterion standard of 35.3% (Figure 1, D). The mean response for this image was 24.0% with a standard deviation of 11.0%. Image NEO-99 ROI-1 was from a colonic adenocarcinoma specimen with abundant extracellular mucin (Figure 1, E). This image had a criterion standard of 71.6%. The mean response for this image was 53.1% with a standard deviation of 10.7%. For both NEO-98 ROI-3 and NEO-99 ROI-1, the number of participants that included necrosis and extracellular mucin in their counting practices, respectively, was small and precluded a meaningful comparison to laboratories that do not include these features. A total of 100% (54 of 54) and 90.7% (49 of 54) of participants correctly classified images NEO-98 ROI-3 and NEO-99 ROI-1 as adequate for molecular testing in relation to the threshold standards of 10% and 40%, respectively.

Neoplastic cellularity assessment has become routine practice for molecular oncology laboratories and is a critical component of test performance; however, no best practice recommendations or guidelines currently exist. Therefore, understanding the current state of practice of neoplastic cellularity assessment could potentially contribute to optimal performance of molecular testing. In this study, we report the results of a CAP neoplastic cellularity practice survey and performance challenge. The results identify variations in the practices used for neoplastic cellularity assessment, which in some instances affected laboratory performance. Overall, the results highlight a need for standardization of neoplastic cellularity assessment.

To more accurately assess neoplastic cellularity, a consensus definition is required.4,7  Although most laboratories defined neoplastic cellularity as the ratio of the number of neoplastic cells or nuclei to the overall number of cells or nuclei, many laboratories used a ratio of the area of neoplastic cells to the area of the total cells. Neoplastic cellularity defined by area is prone to error given that there can be variations in cell size.7  This pitfall was illustrated in the NEO-97 ROI-3 image containing enlarged lung adenocarcinoma cells and numerous smaller lymphocytes. Specifically, those participants that assessed tumor cellularity by area overestimated the neoplastic cellularity compared with those that used the ratio of the number of neoplastic cells or nuclei to the overall number of cells or nuclei. Overestimation becomes particularly problematic when the true neoplastic cellularity is close to the threshold of detection for an assay because of the potential for a false-negative result.8  Defining neoplastic cellularity by cells or nuclei is considered to be more accurate given that the number of cells or nuclei directly correlates with DNA content.9  Hence, laboratories should define neoplastic cellularity as the number of neoplastic cells or nuclei compared to the overall or total number of cells or nuclei.

Although defining neoplastic cellularity is seemingly straightforward, certain histologic features can be sources of variability in interpretation. For example, the survey identified a lack of consensus about whether to include in situ carcinoma. Although most participants excluded it in their calculations, individuals should recognize that in situ and invasive components of a tumor can be molecularly distinct. For example, ERBB2 (HER2) amplified ductal carcinoma in situ (DCIS) of the breast can be associated with ERBB2 (HER2) nonamplified invasive ductal carcinoma.10  In this example, including DCIS in the neoplastic fraction can result in an overestimation of the tumor cellularity, which may lead to an inaccurate assessment of the mutational status of the invasive tumor. Conversely, some genes (eg, PIK3CA) have mutations that are typically conserved between DCIS and invasive carcinoma, making the inclusion of DCIS less problematic.11  In the performance challenge specimen containing DCIS (NEO-95 ROI-3), the participants that reported including DCIS in their estimation had significantly higher estimates of neoplastic cellularity than those that did not. In addition, 100% of participants that reported including in situ carcinoma designated the specimen as adequate for molecular testing compared with only 40% of laboratories that did not. Depending on the molecular assay or the biomarkers being evaluated, laboratories should be cautious about including in situ carcinoma in their neoplastic cellularity assessment, particularly when its inclusion results in an otherwise inadequate specimen being designated as adequate for molecular testing.

Other histologic features about which consensus was lacking in the assessment of neoplastic cellularity included necrosis, apoptosis, and extracellular mucin. Although most participants excluded these areas, several laboratories included them or reported being unsure about these practices. For the images containing abundant apoptotic cells (NEO-96 ROI-2), no significant differences in performance between laboratories that did or did not include apoptotic cells in their estimations of neoplastic cellularity were observed. For the assessment of tumor cellularity in the presence of apoptosis, the mean and median of the laboratories' neoplastic cellularity estimates tended to be higher than the criterion standard (Table 7). Conversely, in the presence of necrosis, the mean and median neoplastic cellularity estimates tended to be lower than the criterion standard. These findings raise the possibility that apoptotic cells and necrosis may be problematic histologic features for neoplastic cellularity assessments. The DNA quality and quantity in regions of apoptosis and necrosis still require more rigorous examination; however, prior studies have shown that DNA extracted from areas including necrosis generates genotyping results comparable to those of viable tumor by multiple methods, including next-generation sequencing.8,12  In addition, whether nonviable tumor in a posttreatment setting contains DNA is another area in need of better understanding in order to draw a consensus. Numerous other histologic features encountered in daily practice, including desmoplasia, reactive changes, and melanophages, can complicate neoplastic cellularity assessment and potentially represent additional sources of interobserver variability. Until the quality of nucleic acids from areas including necrosis and/or apoptosis is more extensively studied and better understood, laboratories should be cautious about including these areas in their assessments.

In most laboratories, an AP board-certified pathologist performs the neoplastic cellularity assessment; however, other individuals with different training backgrounds may also participate in this assessment. Assuming the practice questionnaire participants were also the individuals performing the educational image challenge, we did not observe any statistically significant differences in performance between AP board-certified pathologists and non–AP board-certified individuals as a group (eg, CP-only board-certified pathologist, pathology trainee, and technologist); however, we noted that individuals without AP board certification may be misclassifying DCIS more frequently. Given the limitations of the survey, further investigation is required to better understand these findings and determine the differences between AP board-certified pathologists and non–AP board-certified individuals in assessing tumor cellularity. For instance, if non-AP board-certified individuals are including DCIS in the tumor cellularity assessment, it would be important to know if they are unable to recognize the morphologic features of DCIS and/or are unaware of the different biology of DCIS versus invasive cancer for some genes. A study with more participants, more images, images with more complex histologic features, and more detailed survey questions about the individual reviewers may identify such differences among individuals of different training backgrounds. It is a CAP accreditation requirement for a qualified pathologist to perform the histologic assessment of samples for neoplastic cell content (CAP checklist MOL.32395).6 

The practices used to assess neoplastic cellularity assessment will likely continue to differ among laboratories and are subject to change over time. Currently, most laboratories assess neoplastic cellularity from glass slides using an estimation strategy. Far fewer laboratories assess neoplastic cellularity from a digital slide or image or perform manual counting. Although manual counting may yield an accurate result and was used to derive the criterion standard in this survey, it is time consuming and less practical in daily practice. Although there was no statistically significant difference in the accuracy of the neoplastic cellularity assessment between those laboratories that performed a manual count compared with an estimation, only 9 laboratories performed a manual count, and our study may be unable to detect any differences in accuracy between these methodologies. With the more widespread adoption of digital pathology, automated methods for determining neoplastic cellularity, including molecular methods and/or artificial intelligence, may emerge and have the potential to reduce variability and provide more accurate results. Nevertheless, image-based automated systems will still need to address the sources of variability identified in this study, including the definition of neoplastic cellularity and the identification of the components of a tumor to include in the estimate (ie, necrosis, apoptosis, and in situ carcinoma). Currently, estimation strategies for assessing tumor cellularity are prevalent and are likely to remain so until more accurate strategies become feasible and practical. Therefore, it is important to recognize the limitations of this method when making decisions about the cutoffs for accepted tumor cellularity for individual assays. Setting such cutoffs near the limit of analytic sensitivity of an assay may lead to an insufficient/diluted mutant allele fraction, and ultimately a false-negative result.8  Regardless of the method used, laboratories should consistently use the method(s) described in their standard operating procedure.

Tumor enrichment has the potential to improve detection and minimize false-negative risk and is currently a common practice, with 94.7% of laboratories surveyed performing enrichment on select or all cases. In laboratories performing enrichment on select cases, it is unclear whether the select cases included those with lower tumor cellularity, cases being tested with a less sensitive assay, or some combination of both. Tumor enrichment by manual macrodissection from a slide was the most commonly used method, followed by macrodissection by coring, microdissection using a dissecting microscope, and laser-capture microdissection. This survey did not address more specific details of tumor enrichment practices, including the size and features of the tissue selected. Optimal areas for enrichment include those with the highest and purest tumor content. Factors that dilute the tumor DNA, including tumor-infiltrating lymphocytes and desmoplastic stroma, or that reduce cellularity, including red blood cells and mucin, should be avoided when possible.4  When considering the areas selected for tumor enrichment, additional factors that may need to be considered include relative proximity of the selected tumor to dense inflammatory infiltrates or lymphoid aggregates/follicles. This is of particular importance when manual macrodissection is performed because it carries a greater risk of including tissue from unintended areas.

Molecular oncology testing results should always be interpreted in the context of the neoplastic cellularity. In this survey, 36.8% of participants only sometimes used and 17.5% of participants did not use the neoplastic cellularity in the result interpretation step of molecular testing. As emphasized in the CAP Accreditation Program Molecular Pathology Checklist (CAP checklists MOL.32395 and MOL.36108), an estimation and consideration of the percentage of tumor cells in conjunction with the lower limit of detection of a sequencing assay is essential for proper postanalytic interpretation of a negative test result.6  Although it may be clinically reasonable to accept marginal specimens for testing, the neoplastic cellularity should be re-reviewed when no variants are identified to ensure adequacy of the specimen for the molecular test being used. Neoplastic cellularity may also provide the context needed to verify mutant to nonmutant allelic ratios and possible intratumoral heterogeneity and to interpret aneuploidy, amplification, and loss of heterozygosity. In addition, all pathology reports should contain a disclaimer about the risk of a false-negative result when the tumor content is near or below the diagnostic threshold. In 2019, 61.4% of laboratories documented the neoplastic cellularity in the final molecular report, which increased from 14.3% documenting this in 2015.

There are several limitations of this study. First, the educational performance challenges consisted of digital images of discrete microscopic foci rather than an assessment of larger tissue areas on the slide. Although discrete foci permit manual counting to establish the criterion standard, examination of such foci does not exactly match the clinical practice in most laboratories. Second, this study contained a limited number of performance challenge images. Although statistically significant differences in performance were identified based on laboratory practices identified in the survey, more specimens are needed to further confirm performance differences based on reported practices and testing personnel.

Overall, the results of this practice assessment and performance challenge highlight the variations in neoplastic cellularity assessment and the associated variations in performance. Based on the findings of this survey, we propose initial best practice recommendations to begin the process of standardizing the practice of neoplastic cellularity assessment (Table 8). Although tissue-based neoplastic cellularity assessment was the focus of this study, other specimen types, including cytologic specimens, will also be important to consider in improving and broadening best practices in the future.

Table 8

Initial Recommendations for Assessing Neoplastic Cellularity

Initial Recommendations for Assessing Neoplastic Cellularity
Initial Recommendations for Assessing Neoplastic Cellularity

The authors thank David Rimm, MD, PhD, for creating the criterion standards for the images used for this study. They also thank Ellen Lazarus, MD, for providing medical editing support.

1.
Committee on Policy Issues in the Clinical Development and Use of Biomarkers for Molecularly Targeted Therapies
;
Board on Health Care Services; Institute of Medicine; The National Academies of Science, Engineering, and Medicine;
LA
Graig,
Phillips
JK,
Moses
HL,
eds.
Biomarker Tests for Molecularly Targeted Therapies: Key to Unlocking Precision Medicine
.
Washington, DC
:
The National Academies Press;
2016
.
2.
Viray
H,
Li
K,
Long
TA,
et al
A prospective, multi-institutional diagnostic trial to determine pathologist accuracy in estimation of percentage of malignant cells
.
Arch Pathol Lab Med
.
2013
;
137
(11)
:
1545
1549
.
3.
Smits
AJ,
Kummer
JA,
de Bruin
PC,
et al
The estimation of tumor cell percentage for molecular testing by pathologists is not accurate
.
Mod Pathol
.
2014
;
27
(2)
:
168
174
.
4.
Lhermitte
B,
Egele
C,
Weingertner
N,
et al
Adequately defining tumor cell proportion in tissue samples for molecular testing improves interobserver reproducibility of its assessment
.
Virchows Arch
.
2017
;
470
(1)
:
21
27
.
5.
Lindeman
NI,
Cagle
PT,
Beasley
MB,
et al
Molecular testing guideline for selection of lung cancer patients for EGFR and ALK tyrosine kinase inhibitors: guideline from the College of American Pathologists, International Association for the Study of Lung Cancer, and Association for Molecular Pathology
.
Arch Pathol Lab Med
.
2013
;
137
(6)
:
828
860
.
6.
Commission on Laboratory Accreditation, ed.
Molecular Pathology Checklist
.
Northfield, IL: College of American Pathologists;
2020
.
7.
Dufraing
K,
De Hertogh
G,
Tack
V,
et al
External quality assessment identifies training needs to determine the neoplastic cell content for biomarker testing
.
J Mol Diagn
.
2018
;
20
(4)
:
455
464
.
8.
Weichert
W,
Schewe
C,
Lehmann
A,
et al
KRAS genotyping of paraffin-embedded colorectal cancer tissue in routine diagnostics: comparison of methods and impact of histology
.
J Mol Diagn
.
2010
;
12
(1)
:
35
42
.
9.
Dudley
JC,
Gurda
GT,
Tseng
LH,
et al
Tumor cellularity as a quality assurance measure for accurate clinical detection of BRAF mutations in melanoma
.
Mol Diagn Ther
.
2014
;
18
(4)
:
409
418
.
10.
Lambein
K,
Van Bockstal
M,
Vandemaele
L,
et al
Comparison of HER2 amplification status among breast cancer subgroups offers new insights in pathways of breast cancer progression
.
Virchows Arch
.
2017
;
471
(5)
:
575
587
.
11.
Trinh
A,
Gil del Alcazar
CR,
Sachet
Shukla,
et al
Genomic alterations during the in situ to invasive ductal breast carcinoma transition shaped by the immune system
.
Mol Cancer Res
.
2021
;
19
(4)
:
623
635
.
12.
Conroy
JM,
Pabla
S,
Glenn
ST,
et al
Analytical validation of a next-generation sequencing assay to monitor immune responses in solid tumors
.
J Mol Diagn
.
2018
;
20
(1)
:
95
109
.

Author notes

Supplemental digital content is available for this article at https://meridian.allenpress.com/aplm in the September 2022 table of contents.

The authors have no relevant financial interest in the products or companies described in this article.

Competing Interests

The views expressed in this article are those of the authors and do not reflect the official policy of the Departments of the Army/Navy/Air Force, the Department of Defense, or the US government.

All authors are current or past members of the College of American Pathologists Molecular Oncology Committee. Souers and Vasalos are employees of the College of American Pathologists.

Supplementary data