Cytologic-Histologic Correlation Practices for Nongynecologic Cytology Specimens: A Survey by the College of American Pathologists Cytopathology Committee Open Access

This survey was conducted in essentially an identical manner to our previously published study for gynecologic cytologic histologic correlation.⁴  The description of the methods is repeated here. A subset of the CAP Cytopathology Committee members and CAP technical staff, including a staff biostatistician, developed a survey and reviewed it for question clarity and validity. The final survey included 22 questions. Participants were not required to answer all questions in the survey. The survey was delivered in the 2018-D Nongynecologic Cytopathology (NGC) Education Program mailing to a total of 2146 laboratories. Data validity adjustments were applied to address skip sequence direction and missingness. The missingness check was based on the expected number of responses, which varied by laboratory because it accounted for nonresponse due to skip sequence reporting. In brief, multiple sections begin by asking if the institution performs a specific action. If they do not, then that section is “skipped,” and the lack of responses are not considered missing answers. For those sections not skipped, the percentage of missing answers (missingness) was assessed. Surveys with high missingness were excluded.

A 2-level institution location variable was created to define North American and international laboratories. North American laboratories included respondents from the United States, Canada, and Mexico. Laboratory institution type was extracted from the NGC Demographics Questionnaire results. Multivariate logistic regression models were used to test for practice differences and were fit with 2 factors: institution type and location. Institution type was defined as a 5-level factor (nonacademic hospital, clinic or regional/local independent laboratory, Veterans Administration/Department of Defense [VA/DOD] hospital, university hospital/academic medical center, and national/corporate laboratory). A significance level of .05 was used for this analysis, and all analyses were performed with SAS v9.3 and v9.4 (SAS Institute, Cary, North Carolina).

The survey was distributed in the 2018-D NGC Education Program mailing to 2146 laboratories, and 879 returned the survey, for a 41% response rate. There were 802 surveys that qualified for analysis; 77 surveys were excluded owing to duplication (1) and multiple missing responses (76).

Of the participants surveyed, 710 of 802 laboratories (88.5%) reported that CHC was performed on nongynecologic and/or FNA specimens (Figure 1). There was a statistically significant practice difference (P < .001) between institution types. More than 90% (29 of 30) of Veterans/DOD institutions, university hospital/academic medical centers (47 of 49), and nonacademic hospitals (435 of 477) performed CHC on nongynecologic and/or FNA specimens, whereas 78.0% of national/corporate laboratories (39 of 50) and 78.7% (85 of 108) of clinic or regional/local independent laboratories reported performing this CHC practice (Figure 2).

A high percentage of laboratories (91.9%, 647 of 704) had a written policy for performing CHC on nongynecologic specimens. Access to cytologic and surgical material was also available in a high percentage of 708 institutions: cytology slides, 94.4% (668); surgical slides, 93.4% (661); cytology final report, 92.9% (658); surgical final report, 92.8% (657); surgical final diagnosis, 89.5% (634); and cytology final interpretation, 89.4% (633) (Table 1). Access to final reports is preferable compared to cytology final interpretations because other components of cytology report adequacy, ancillary studies, and comments may not be conveyed in the interpretation. CHC was routinely performed by using correlation of reports with review of both cytology and surgical slides in 75.8% (530 of 699) of institutions. Other types of cytology correlations were performed in certain tissues, with the 2 most common being flow cytometry for correlation of lymph node diagnoses (67.1%, 245 of 365) and Afirma-Veracyte/Thyroseq for correlation of thyroid diagnoses (49.3%, 180 of 365).

CHC was most commonly performed in real time/concurrently: when the corresponding surgical specimen was reviewed (74.4%, 525 of 706) or when the cytology was reviewed (54.2%, 383 of 706) (multiple responses were allowed) (Table 2). Although at lower rates, retrospective CHC was also done when comparing final diagnosis for each specimen from a report (47.6%, 336 of 706), when compiling statistics (45.9%, 324 of 706), or during clinical patient conferences (24.2%, 171 of 706). Most often, cases were selected by the interpreting pathologist (70.1%, 386 of 551) for real-time concurrent CHC (multiple responses allowed). Laboratory-specified selection criteria were used in a far lower percentage of cases, with 18.1% (100 of 551) of laboratories using criteria for specimen type, 15.8% (87) using criteria for body site, and 15.2% (84) using diagnosis-specific criteria.

Specific questions regarding practices for discordances showed a few responses that were overwhelmingly similar. The most common reason for discordance was sampling (inadequate sampling or no sampling of target lesion) (93.8%, 645 of 688) (Table 3). Most frequently, participants communicated discordances with the patient care provider directly (84.4%, 591 of 700), but other mechanisms were also common, including in the surgical pathology report (70.1%, 491 of 700) and in the cytology report (56.4%, 395 of 700). Most discordances were addressed at the time of the surgical pathology sign-out (69.5%, 491 of 706), after compilation of cases to generate correlation statistics (56.9%, 402 of 706), or at the time of cytology interpretation/sign-out (47.6%, 336 of 706). The threshold for discordance between cytology and surgical specimens varied among institutions, with no definition used by a majority. The most common was a 2-step difference (41.2%, 283 of 687). The methods used to investigate discordance also varied, with no definition used by a majority. Single-pathologist review of all discrepant surgical and cytology slides (48.2%, 340 of 706) was most frequent, followed closely by review of surgical/cytology specimens by the original pathologist (45.3%, 320 of 706). The person primarily responsible for final discordance resolution was the pathologist responsible for the corresponding surgical specimen (33.5%, 231 of 690), the designated pathologist (24.8%, 171 of 690), or the cytopathology section director (23.6%, 163 of 690).

Laboratories were asked about statistical metrics from nongynecologic CHC. Most laboratories monitored the percentage of cytology tests that correlate with subsequent surgical specimens of the same site (68.5%, 406 of 593) (Table 4). More than 30% monitored sampling discordance rate and interpretative error rate by pathologist (both 34.6%, 205 of 593). Statistical metrics were summarized either monthly (38.4%, 220 of 573) or quarterly (35.3%, 202 of 573) and distributed mostly to pathologists (73.9%, 425 of 575), and less commonly to a laboratory quality committee (54.1%, 311 of 575), hospital quality committee (40.2%, 231 of 575), or cytotechnologists (32.2%, 185 of 575). The data obtained from the metrics were used to improve patient care mostly by summarizing the results and sharing them as a quality improvement document (64.6%, 451 of 698), with individual results shared with the pathologist on record (64.2%, 448 of 698).

Remediation/retraining occurred most commonly in response to error trends: 48.8% (303 of 621) for pathologists and 44.8% (278 of 621) for cytotechnologists. Every major discordance was also a common criterion for remediation/retraining: 42.2% (262 of 621) for pathologists and 30.0% (186 of 621) for cytotechnologists.

A number of statistically significant differences were identified when practices between North American and international laboratories were compared (Table 5). Significantly more commonly, North American laboratories used a laboratory information system to generate reports to use for CHC (74.8% [481 of 643] versus international, 59.1% [39 of 66]; P = .02) and had more access to cytology final interpretations (91.6% [588 of 642] versus international, 68.2% [45 of 66]; P < .001), surgical pathology final diagnoses (91.4% [587 of 642] versus international, 71.2% [47 of 66]; P < .001), cytology slides (95.0% [610 of 642] versus international, 87.9%, [58 of 66]; P = .007), and surgical pathology slides (94.2% [605 of 642] versus international, 84.8% [56 of 66]; P = .03). Interestingly, international laboratories were more likely to routinely perform correlation of reports with review of both cytology and surgical pathology slides (89.4% [59 of 66] versus North American, 74.4% [471 of 633]; P = .04) and real-time/concurrent CHC at the time a corresponding surgical specimen was reviewed (84.8% [56 of 66] versus North American, 73.3% [469 of 640]; P = .03). For discordance investigation, North American laboratories were more likely to have a single pathologist review all discrepant surgical slides and cytology slides (49.8% [319 of 640] versus international, 31.8% [21 of 66]; P = .02), whereas international laboratories more frequently had a second pathologist review surgical slides to investigate discordance (40.9% [27 of 66] versus North American, 28.8% [184 of 640]; P = .01). International laboratories were also more likely to address discordances in the pathology report when cytology specimens did not reflect the tumor/malignant surgical specimen diagnosis (62.1% [41 of 66] versus North American, 48.5% [306 of 631]; P = .03). Positive predictive value (PPV) of a positive cytology finding was rarely monitored in North American laboratories (8.2%, 44 of 536), while in international laboratories it was monitored significantly more frequently (26.3%, 15 of 57; P < .001).

Finally, we reviewed statistically different practices among the various institution types. There were statistically significant practice differences for when CHC is performed (P < .001) (Table 6). Nonacademic hospitals performed retrospective correlation—at a later date when comparing final diagnosis from each specimen from a report (41.8%, 180 of 431), but less often than any other institution type. More than 55% (55.3%–62.1% [47 of 85 to 18 of 29]) performed this retrospective correlation for every other institution type. Discordance investigation occurred at the time of surgical pathology sign-out at a frequency greater than 69% for all institution types except for university hospital/academic medical centers (69.2%–72.4% [27 of 39 to 21 of 29] versus 48.9% [23 of 47]; P = .02), which had the highest frequency of investigating discordances after compilation of cases to generate correlation statistics (78.7%, 37 of 47; P = .003). Where discordances were addressed also varied among the institution types (P = .04).

The CLIA-mandated CHC in gynecologic cytopathology set a precedent that has for many years developed the discipline to be more comprehensive and thorough, with well-established guidelines.^5,6  In contrast, nongynecologic cytopathology, the focus of this article, is more varied.⁷  A much broader array of tissues may be sampled, and a much larger number of techniques used. This article reports the results of a survey included with the CAP’s NGC Education Program mailing to a total of 2146 laboratories.

Perhaps the most interesting finding from these survey results is that, overall, institutions perform CHC on nongynecologic specimens with the same degree of rigor as they perform gynecologic CHC, and with more attention to guidelines than previously.⁸  CHC on nongynecologic cases was performed at 88.5% (710 of 802) of the reporting institutions. This is much greater than in a 2014 CAP survey of nongynecologic cytology practice patterns that found a rate of 71.6% (722 of 1008). Overall, all metrics were comparable with the findings from the previous gynecologic CHC survey: rates of CHC performance, laboratory information system use, access to CHC components (availability to obtain slides, reports, and diagnoses for correlation), timing of CHC, selection of cases, time interval used for selecting correlating specimens, timing of discordance adjudication, personnel involved in CHC discordance investigation, the person responsible for the final CHC discordance resolution, most common reasons for specimen discordance, communication mechanisms when a discordance changes current patient care, monitored CHC statistical metrics, and the frequency with which CHC results are summarized for statistical analysis. These results are reassuring, as laboratories appear to adopt practices that improve patient safety, quality assurance, and error reduction.^9,10 

CHC is not performed with equal rigor across all institution types, however. More than 90% of hospitals, university hospitals/academic medical centers, and Veterans/DOD institutions performed CHC on nongynecologic cytology cases. Significantly fewer clinic or regional/local independent laboratories and national/corporate laboratories performed CHC on these cases. While the reason for this deficit is not clear from this survey, there may be structural reasons, such as limited access to biopsy specimens or lack of a standardized correlation method, why fewer of these institutions perform CHC on nongynecologic specimens.

The breakdown of responses by institution type revealed additional surprising differences. Most striking was the degree to which university hospitals/academic medical centers differed from the other institution types. University hospitals/academic medical centers most often reported a time point for discordance investigation after the compilation of cases to generate correlation statistics and, conversely, least often reported a time point for discordance investigation at the time of surgical pathology sign-out. It is unclear why university hospitals and academic medical centers differ in their responses from the other institutions, and studies to inquire about specifics may be worthwhile in the future. Importantly, the absolute number of responses for most institution types was relatively low. Although 477 institutions in the nonacademic hospital category responded to this survey, the numbers were far lower for all other institution types: 108 clinic or regional/local independent laboratories, 50 national/corporate laboratories, 49 university hospital/academic medical centers, and 30 Veterans/DOD medical centers. These low numbers may limit the generalizability of the findings.

There were subtle differences in the selection of cases for real-time concurrent CHC between gynecologic and nongynecologic cases. For real-time/concurrent correlations, the most common correlations, CHC review was selected by the interpreting pathologist in 60.2% (240 of 399) of gynecologic cytology cases compared to 70.1% (386 of 551) of nongynecologic cytology cases. In unpublished data from our previous gynecologic cytology paper, for gynecologic cytology cases, laboratory-specific selection criteria were more commonly used: 32.8% (131 of 399) for criteria based on diagnosis and 29.8% (119 of 399) for criteria based on specimen type. For nongynecologic cytology cases, criteria based on diagnosis were used for only 15.2% (84 of 551), and criteria based on specimen type were used for only 18.1% (100 of 551). The sheer diversity of tissues and techniques in nongynecologic cytopathology may hinder the adoption of uniform laboratory-based selection criteria for real-time concurrent CHC.

One aspect of nongynecologic cytology that differs from gynecologic cytology is the availability of additional correlative studies for some tissue types. Only 1 such test, flow cytometric analysis for lymph node assessment, was used in more than two-thirds of cases (67.1%, 245 of 365). For laboratories performing other cytology correlations, molecular testing for thyroid cytology was used in nearly half of cases (49.3%, 180 of 365). Other tests were used much less often. As clinician familiarity with these tests increases, and their clinical utility improves and guidelines for standards of practice change, these tests may likely see increased use as adjuncts to cytologic analysis.

Our previous work on gynecologic CHC found a relationship between the number of cases received at an institution and the likelihood of performing CHC. Specifically, those institutions evaluating fewer than 1200 specimens had the lowest CHC rate at 43.4%.⁴  Given the significant variability in specimen types and locations in nongynecologic cytology, caseload is even more likely to alter the probability of routine CHC performance. Since nongynecologic cytology case numbers were not assessed for this study, however, the effect of caseload on the likelihood of CHC cannot be evaluated.

Similar to our findings in gynecologic cytology,⁴  international laboratories calculate nongynecologic cytology PPV more frequently than North American laboratories. Worldwide, the cumulative percentage of laboratories that calculate PPV is very low. As population prevalence of lesions change over time, as they inexorably do, the utility and efficacy of cytologic studies also change. Continuously evaluating the positive and negative predictive values provides a mechanism for monitoring the utility and efficacy of cytologic examinations. Calculating the PPV is now a recommended component of the CHC for gynecologic cytology,¹¹  and it should also be incorporated into the practice of nongynecologic cytology.

This study has several limitations in addition to those described above. As the report of survey data, the findings rely on the accuracy of the reporting by individual institutions, which cannot be independently verified. This study was designed as a survey and not a comprehensive analysis of all institutions that review cytology cases, and thus selection bias may limit the utility of the findings. Moreover, given the incomplete response to the survey request, with responses received from 41% (879 of 2146) of individual institutions, the degree to which these data are representative of all institutions cannot be assessed.

Another significant limitation derives from the absence of guidelines for performing nongynecologic CHC. Given this void, institutions have adopted varying methods for performing NGC CHC. This lack of uniformity complicates comparisons across institutions.

Some elements of these limitations may be addressed in future studies. As primary human papillomavirus testing is being implemented across many countries and changing the field of gynecologic cytology, we may see more of an emphasis toward nongynecologic cytology. Current barriers such as staffing shortages, workload limits, and time limitations may be alleviated for laboratories owing to decreased gynecologic volumes. Nongynecologic cytology is so diverse, given the breadth of anatomic sites and procedures, that histologic correlations with cytology may not be the only quality measure. For instance, this study has demonstrated specific cytologic correlations with molecular studies (thyroid) or flow cytometry (lymph node). Additional types of correlation may be on the horizon, such as cytology correlation with biomarkers.^12,13  Cytopathology reporting options to make reports more standardized and data more extractable are on the horizon.^14,15  Furthermore, with improved tissue collection methods in cytology, incorporation of small tissue biopsies into the practice of cytopathology, and use of ancillary tests such as specific immunohistochemical markers and molecular tests, surgical pathology is no longer the sole gold standard, and the practice of CHC is a quality metric to assess both surgical and cytopathologic diagnosis. Given the importance of CHC, changes such as allowing reimbursements or billing for CHC may help incentivize laboratories to meet good practice measures. In conclusion, the field of cytopathology is evolving and ever expanding, and CHC positively contributes to the field’s further development.

This study surveyed a large number of institutions worldwide regarding CHC practices for nongynecologic cytology specimens. Overall, there is a relatively high rate of CHC for nongynecologic specimens among laboratories, much improved compared to the past. On most CHC parameters, the results closely mirror the findings of the previous CAP survey of gynecologic CHC practices. In an analysis of CHC by institution type, clinic or regional/local independent laboratories and national/corporate laboratories performed CHC less frequently than hospitals, university hospitals/academic medical centers, and Veterans/DOD institutions. Institutions were not surveyed on case numbers, so the influence of case volume on CHC performance could not be assessed.