lntraoperative Imprint Cytology: Its Significance as a Diagnostic Adjunct Kitai Kim, M.D., F.I.A.c., Edwin R. Phillips, and Michael Paolino. M.D.
M.D.,
Ph.D.,
We retrospectively evaluated 664 specimens submitted f o r intraoperative frozen-section analysis f o r which cytologic imprints or smears were also prepared; 238 (36%) were malignant neoplasms. These preparations were retrospectively evaluated independently by three reviewers of varied experience in the detection of malignancy. The number of false-positive and false-negative results were recorded, and various assessment parameters (sensitivity. specificity. efficiency. and predictive value) were calculated. The imprint was of chief value as an adjunct to the frozen section, particularly in avoiding falsepositive and, to a lesser extent, false-negative interpretations. Experience with the use of intraoperative cytology demonstrated the technique to be of value in providing abbreviated preparation time (3-5 min); supportive diagnostic information when frozen section was equivocal: diagnostic infor,wation when frozen-section evaluation could not be done (e.g.. excessively small sample); contributory informationf o r final diagnosis on dificult cases; and exceffent teaching material f o r cytopathology. Diagn Cytopathol 1990;6:304-307. Key Words: Cytology; Intraoperative diagnosis
Rapid intraoperative histologic diagnosis has been traditionally based on the frozen section. Cytologic preparations are well recognized to provide information that, in some cases, exceeds that of the frozen section. Cytologic detail in imprint preparations is better preserved than that in frozen sections. In addition, the arrangement of cells may provide information as to the epithelial or nonepithelial nature of a lesion, as well as revealing specific histogenetic features. Despite the introduction of cytologic evaluation of fresh tissue specimens in 1927 by Dudgeon and Patrick,’ this technique has not gained widespread acceptance. The reason for this lack of acceptance is not entirely clear, but it may be related to limited cytopathologic expertise on the part of the examiners, compounded by apprehension over medical practice liability. Received June 15, 1989. Accepted October 18, 1989. From the Department of Pathology, Medical College of Ohio, Toledo, OH. Address reprint requests to Kitai Kim, M.D., Professor of Pathology, Director of Surgical Pathology and Cytology, Department of Pathology, Medical College of Ohio, PO Box 10008, Toledo, OH 43699.
304
Diagnostic Cytopathology,Vol6. No 5
A number of reports on the value of cytologic support or supplementation of frozen-section diagnoses have appeared in recent The classic report of Blaustein and Silverberg brought the value of the technique to attention and suggested that it could even substitute for frozen section in certain cases.’ Such emphasis on its value may gradually bring the technique into routine use in intraoperative d i a g n ~ s i s . ~ ” At this institution, cytologic imprints or smears are prepared from most of the fresh surgical specimens submitted for intraoperative frozen-section diagnosis. Those cases in which both imprints and frozen sections were obtained over a 2-yr period were evaluated retrospectively, resulting in 664 cases. Of these, 238 (36%) were malignant neoplasms. Although imprint cytologic preparations frequently offer more specific information than simple discrimination between benign and malignant tissue, evaluation principally focused on this issue, and the purpose of this study was to evaluate the sensitivity and specificity of imprint cytology with reference to intraoperative diagnosis.
Materials and Methods Over a 2-yr period, imprint smears were prepared from 664 fresh surgical specimens received for frozen-section diagnosis. Of these, 426 specimens were benign neoplasms or nonneoplastic tissues and 238 were malignant neoplasms. The specimens were classified with respect to anatomic site or organ system (Table I). Cytologic specimens were prepared from a selected freshly cut surface of the tissue by one or more of the following methods8-10: (1) lightly touching a glass slide with the specimen; (2) smearing the surface of the specimen against the slide; or (3) “rolling” small specimens on the slide. Occasionally, small particles received in saline solution were collected on a glass slide and “crushed” with a second slide. On fibrous or sclerotic lesions, the surface could be scraped with the edge of a glass slide or scalpel blade and the collected material smeared on a slide. All of these smears were fixed while a 1990 WILEY-LISS, INC.
INTRAOPERATIVE IMPRINT CYTOLOGY Table I. Organ Systems Evaluated for Imprint and Frozen-Section Diagnostic Accuracy Number of specimens Svstem Digestive Genitourinary Gynecologic Respiratory Bone/soft tissue Lymph nodes Breast Thyroid/parathyroid NS Body surface Total
Positive
Nezative
Total
106 10 43 13 9 17 20 6 8 4 236
157 39 43 37 23 24 78 17 2 8 428
263 49 86 50 32 41 98 23
10 12 664
wet in 95% ethanol, and one or more of the slides were stained rapidly with hematoxylin-eosin for immediate evaluation. The collection of 664 specimens was reviewed retrospectively by three participants. Each examined the imprint smears and corresponding frozen-section specimens separately and recorded diagnostic impressions before examining the paraffin sections. Only the anatomic site of origin was known at the time of evaluation. Reviewer 1 was a third-year resident in pathology who had completed a 1-mo rotation in cytopathology. Reviewer 2 was a surgical pathologist with only an ancillary interest in cytopathology who does not routinely interpret cytologic specimens. Reviewer 3 was a surgical pathologist with a strong interest and special expertise in cytopathology. “Positive” diagnoses were recorded as (1) malignant neoplasm, not otherwise specified; (2) carcinoma or sarcoma; ( 3 ) adenocarcinoma, squamous carcinoma, or lymphoma; or (4) atypical cells, inconclusive for diagnosis. Evaluation of cytologic preparations for the purposes of this article was focused on the simple discrimination between benign and malignant disease. Each reviewer evaluated all specimens to distinguish malignant neoplasms from nonneoplastic tissues or benign neoplasms. The sensitivity, specificity, positive predic-
tive value, negative predictive value, and efficiency for each organ system were determined. Calculations were made according to formulas described elsewhere.“ Calculations were applied to imprint results, frozen-section results, and combined imprint-frozen-section results.
Results Efficiency (EFF) reflects the relative accuracy of a test system, i.e., the relative number of times the correct result is achieved. Table I1 shows the efficiency of imprint and frozen-section evaluations as a function of the organ system and the reviewer’s experience. Although absolute concordance between the two methods and between observers did not result, certain trends emerged. Imprint examination of some systems and organs, such as central nervous system (CNS) and lymph nodes, tended to be of low efficiency; for others, such as body surface and breast lesions, it was of high efficiency. Efficiency was consistently lower for imprint evaluations than for frozen sections. As would be expected, relative efficiency also correlated favorably with reviewer experience. To demonstrate the value of the adjunctive use of imprint cytology, false-positive and false-negative results are provided in Table 111. They are separated according to imprint assessment, frozen-section assessment, and combined imprint and frozen-section assessment, when the imprint and frozen-section results were in accord. Results are also expressed as a function of reviewer experience. Not surprisingly, false-negative results were more frequent on the imprint evaluations. Moreover, combined imprint and frozen-section results improved only slightly on the false-negative results for frozen section alone. However, in the cases of false-positive results, the number of incorrect readings was substantially lower than for false-negatives, and the use of combined imprint and frozen section results markedly reduced the error rate. This clearly underscores the improved accuracy of combined frozen sections and imprints. Specificity, sensitivity, and efficiency (Table IV) and predictive values (Table V) were expressed as a function
Table 11. Relative Diagnostic Efficiency in Frozen-Section and Imprint Evaluation (Percentages) Reviewer 2
Reviewer 1 System
CNS Lymph node Thyroid/parathyroid Bone/soft tissue Gastrointestinal Gynecologic Genitourinary Surface tissue Respiratory Breast
Reviewer 3
Imprint
FS
Imprint
FS
Imprint
FS
20 69 75 78 69 62 80 67 75 89
50 71 87 94 83 85 84 75 92 94
70
100 73 78 94 95 95 96 100 96 100
60 76 79 81 89 88 90 92 94 97
90 90 91 91 97 99 100 100 100 99
71 79 81 84 87 88 92 84 92
FS, frozen section. Diagnostic Cytopathology, Vol6, N o 5
305
KIM ET AL Table 111. Number of False-Negatives and False-Positives False-negatives Reviewer 1 2 3
False-positives
Imprint
FS
Imprint FS
172 98 71
83 38 17
76 33 15
+
Imprint
FS
15 6 5
20
Imprint + FS
1
10 0
6
1
FS, frozen section
of imprint evaluation; frozen-section evaluation; combined imprint and frozen-section evaluation; and reviewer experience. These results also reflect the lower sensitivity of the imprint method and the very similar specificities of imprints and frozen sections. Specificity is improved slightly with the use of combined imprints and frozen section than with frozen sections alone; efficiency also tended to improve in combinations, although only slightly. In these studies, equivocal results were counted as negative, since an uncertain or suspicious diagnosis would not provide a suitable basis for further surgical therapy. These results included instances of nondiagnostic frozen section when the smear was conclusive, and vice versa. The study only included those cases in which both frozen section and imprint were available. Some specimens, excluded from the enumeration, had only cytology; tiny parathyroid samples, as well as bone biopsies, belong in this group. In such cases, cytologic evaluation offered the only intraoperative assessment.
Discussion Since its advent, the frozen section has been the mainstay of rapid intraoperative diagnosis and will likely remain so unless a more precise technique is developed. At our institution, the use of cytologic preparations as a companion to the frozen section was begun in 1980. It has been believed to be a consistent aid in the assessment of difficult frozen sections, but this opinion largely stemmed from anecdotal experience. To demonstrate the value of diagnostic intraoperative cytology and, more specifically, of combined frozen-section and cytologic evaluations, we examined specimens accrued over a 2-yr period in which both types of preparations were available. The evaluation of these specimens, although retrospective, provided a comparison of specificity, sensitivity, and efficiency as a function of the results of cytologic evaluation alone, Table IV.
frozen section evaluation alone, and combined frozen section and cytology. In addition, the results were also examined as a function of reviewer experience. This includes experience in anatomic pathology in general but particularly in cytopathology. This experience is important because confidence in cytopathologic interpretation may contribute to the acceptance of this method. It should be mentioned that in this review, false readings may be anomalously high because of (at least) two sources of bias: ( 1 ) lack of clinical information at the time of retrospective reading; and (2) a possible tendency toward more aggressive interpretation because of the lack of any adverse consequence of error under the conditions of this study. Moreover, cases judged as “suspicious” were classed as negative for the purpose of this study. In none of the evaluations did the accuracy of imprint evaluation alone surpass that of frozen section alone. That false-negativeswere consistently higher in cytologic assessment than in frozen section assessment is largely a result of improper sampling of specimens (e.g., air-dried specimens, sclerotic tissues) and is not surprising. The falsenegative rate in sclerotic tissue can be reduced by scraping the tissue using a scalpel or wooden spatula.’ In addition, some specimens were particularly prone to give false-negative results (e.g., microfocal carcinoma in lymph nodes or bone) or interpretations (e.g., certain lymphomas). On the other hand, instances were encountered in which imprint cytologic evaluation provided the only morphologic evidence of malignancy. The low efficiency of imprint evaluation, e.g., in CNS specimens, is in contrast to that reported by Marshall et al.” However, in this study, the number of encounters with CNS material was small and their imprint samples were frequently not appropriate. Among the false-positive imprint interpretations, the leader appeared to be adenomatous polyps of the colon,
Percentages of Specificity, Sensitivity, and Efficiency
-
specificity
Sensitivity Imprint
Reviewer
Imprint
FS
+ FS
Imprint
FS
+ FS
Imprint
FS
Imprint f FS
96.5 98.6 98.8
95.3 99.8 98.6
91.6 100.0 99.8
21.1 58.5 69.9
64.8 83.9 92.8
42.9 80.1 91.6
71.8 84.3 88.6
84.5 94.1 96.5
84.2 94.4 91.3
. . ~ ~
1 2 3
FS, frozen section.
306
Eficiency
Diagnostic Cytopathology, Vol6, N o 5
Imprint
INTRAOPERATIVE IMPRINT CYTOLOGY Table V. Percentages of Positive and Negative Predictive Value PV-
PV+ Reviewer 1
2 3
Imprint
FS
Imprint + FS
81.0 95.8 97.1
88.4 99.5 92.3
85.1 100.0 99.4
Imprint
FS
Imprint + FS
70.6 81.2 85.6
83.1 91.8 96.1
84.1 92.7 96.5
PV+, positive predictive value; PV -, negative predictive value, FS, frozen section.
which provided four, three, and two false-positives for reviewers 1, 2, and 3, respectively. Diagnosis based on cytologic evaluation of adenomatous polyps should be avoided, and adjunctive use at frozen section should be approached with caution. Of central importance to this study was the effect of using cytopathologic evaluation as an adjunct to frozen section. In the expression of results in terms of sensitivity, specificity, and efficiency (Table IV), it appears that the sum of cytologic and cryohistologic assessment does not improve the sensitivity of frozen section, nor does it significantly improve efficiency. This is principally because of the high rate of false-negatives on cytology evaluations, as discussed above. However, sensitivity and efficiency of imprint cytologic evaluation can be significantly improved by better sampling technique. Specificity, which relates to the correctness of diagnosis of positive specimens, is very similar between imprints and frozen sections and appears to improve little numerically by the summation of cytologic and frozen section interpretations. However, this apparently slight difference in percentage, attributable to a reduction of falsepositives, translated into an impressive gap in the small number of specimens involved (Table 111). Here, the use of agreement between frozen section and cytology significantly reduces false-positives, regardless of the level of observer experience or expertise. At surgery, falsepositive diagnoses are potentially the most damaging diagnostic error. This strongly emphasizes the value of imprint evaluation in rapid intraoperative diagnosis as an adjunct to the frozen section. In addition to supportive information, useful in discriminating between benign conditions and cancer, imprint cytology provides the following benefits: (1) evaluation of small specimens, which risk being lost during preparation for frozen section; (2) preservation of material for special studies such as electron microscopy; (3) evaluation of specimens in which frozen section is not feasible for technical reasons (e.g., bone); (4) exhibition of histogenetic features aiding in evaluation of the specimen (e.g.,
parathyroid tissue, cell of origin of poorly differentiated neoplasms, and cytologic type of lymphoma); and (5) promotion of the teaching of cytologic interpretation and histologic and cytologic correlation. In spite of numerous previous studies describing the value of imprint cytology as an intraoperative diagnostic method, it has not become a routine practice in many institutions. This study provides a quantitative assessment of the use of imprint cytologic method in conjunction with frozen section and emphasizes its adjunctive value for practitioners over a wide range of experience and expertise.
Acknowledgrnent The authors are grateful to Dr. Bernard Naylor for his assistance with the manuscript and to Ms. Marilyn Cline for her excellent typing services.
References 1. Dudgeon LS, Patrick CV. A new method for the rapid microscopic diagnosis, with an account of 200 cases so examined. Br J Surg 1927;15:250+. 2. Schnicker KT, Hermanek P. Intraoperative histology or cytology? Virchows Arch [A] 1974;362:247-58. 3. Westlake PT. An aspiration and touch prep cytology smear as a teaching tool. Cytotechnol Bull 1982;19:26. 4. Kern WH. Presidential address. Acta Cytol 1982;26:755-8. 5. Blaustein PA, Silverberg SG. Rapid cytological examination of surgical specimens. Path Annu 1977;12:251-78. 6. Owings RM. Rapid cytologic examination of surgical specimens: a valuable technique in the surgical pathology laboratory. Hum Pathol 1984;15:605-14. 7. Wilkinson JA, Bonnin JM. Intraoperative cytology; an adjunct to frozen sections. New York: Igaku-Shoin, 1987:9-11. 8. Lee TK. The value of imprint cytology in tumor diagnosis: a retrospective study of 522 cases in Northern China. Acta Cytol 1983;26: 169-7 1. 9. Studham VB, Dravid MV, Grovers, Kher A. Role of scrape cytology in rapid intraoperative diagnosis. Acta Cytol 1984;28:477-82. 10. Galen RS, Gambino SR. Beyond normality: the predictive value and efficiency of medical diagnosis. New York: Wiley & Sons, 1975. 11. Marshall LF, A d a m H, Doyle D, Graham DI. The histological accuracy of the cancer technique for neurosurgical biopsies. J Neurosurg 1973;39:82-8.
Diagnostic Cytopathology, Vol6, No 5
307
M.D.,
Ph.D.,
We retrospectively evaluated 664 specimens submitted f o r intraoperative frozen-section analysis f o r which cytologic imprints or smears were also prepared; 238 (36%) were malignant neoplasms. These preparations were retrospectively evaluated independently by three reviewers of varied experience in the detection of malignancy. The number of false-positive and false-negative results were recorded, and various assessment parameters (sensitivity. specificity. efficiency. and predictive value) were calculated. The imprint was of chief value as an adjunct to the frozen section, particularly in avoiding falsepositive and, to a lesser extent, false-negative interpretations. Experience with the use of intraoperative cytology demonstrated the technique to be of value in providing abbreviated preparation time (3-5 min); supportive diagnostic information when frozen section was equivocal: diagnostic infor,wation when frozen-section evaluation could not be done (e.g.. excessively small sample); contributory informationf o r final diagnosis on dificult cases; and exceffent teaching material f o r cytopathology. Diagn Cytopathol 1990;6:304-307. Key Words: Cytology; Intraoperative diagnosis
Rapid intraoperative histologic diagnosis has been traditionally based on the frozen section. Cytologic preparations are well recognized to provide information that, in some cases, exceeds that of the frozen section. Cytologic detail in imprint preparations is better preserved than that in frozen sections. In addition, the arrangement of cells may provide information as to the epithelial or nonepithelial nature of a lesion, as well as revealing specific histogenetic features. Despite the introduction of cytologic evaluation of fresh tissue specimens in 1927 by Dudgeon and Patrick,’ this technique has not gained widespread acceptance. The reason for this lack of acceptance is not entirely clear, but it may be related to limited cytopathologic expertise on the part of the examiners, compounded by apprehension over medical practice liability. Received June 15, 1989. Accepted October 18, 1989. From the Department of Pathology, Medical College of Ohio, Toledo, OH. Address reprint requests to Kitai Kim, M.D., Professor of Pathology, Director of Surgical Pathology and Cytology, Department of Pathology, Medical College of Ohio, PO Box 10008, Toledo, OH 43699.
304
Diagnostic Cytopathology,Vol6. No 5
A number of reports on the value of cytologic support or supplementation of frozen-section diagnoses have appeared in recent The classic report of Blaustein and Silverberg brought the value of the technique to attention and suggested that it could even substitute for frozen section in certain cases.’ Such emphasis on its value may gradually bring the technique into routine use in intraoperative d i a g n ~ s i s . ~ ” At this institution, cytologic imprints or smears are prepared from most of the fresh surgical specimens submitted for intraoperative frozen-section diagnosis. Those cases in which both imprints and frozen sections were obtained over a 2-yr period were evaluated retrospectively, resulting in 664 cases. Of these, 238 (36%) were malignant neoplasms. Although imprint cytologic preparations frequently offer more specific information than simple discrimination between benign and malignant tissue, evaluation principally focused on this issue, and the purpose of this study was to evaluate the sensitivity and specificity of imprint cytology with reference to intraoperative diagnosis.
Materials and Methods Over a 2-yr period, imprint smears were prepared from 664 fresh surgical specimens received for frozen-section diagnosis. Of these, 426 specimens were benign neoplasms or nonneoplastic tissues and 238 were malignant neoplasms. The specimens were classified with respect to anatomic site or organ system (Table I). Cytologic specimens were prepared from a selected freshly cut surface of the tissue by one or more of the following methods8-10: (1) lightly touching a glass slide with the specimen; (2) smearing the surface of the specimen against the slide; or (3) “rolling” small specimens on the slide. Occasionally, small particles received in saline solution were collected on a glass slide and “crushed” with a second slide. On fibrous or sclerotic lesions, the surface could be scraped with the edge of a glass slide or scalpel blade and the collected material smeared on a slide. All of these smears were fixed while a 1990 WILEY-LISS, INC.
INTRAOPERATIVE IMPRINT CYTOLOGY Table I. Organ Systems Evaluated for Imprint and Frozen-Section Diagnostic Accuracy Number of specimens Svstem Digestive Genitourinary Gynecologic Respiratory Bone/soft tissue Lymph nodes Breast Thyroid/parathyroid NS Body surface Total
Positive
Nezative
Total
106 10 43 13 9 17 20 6 8 4 236
157 39 43 37 23 24 78 17 2 8 428
263 49 86 50 32 41 98 23
10 12 664
wet in 95% ethanol, and one or more of the slides were stained rapidly with hematoxylin-eosin for immediate evaluation. The collection of 664 specimens was reviewed retrospectively by three participants. Each examined the imprint smears and corresponding frozen-section specimens separately and recorded diagnostic impressions before examining the paraffin sections. Only the anatomic site of origin was known at the time of evaluation. Reviewer 1 was a third-year resident in pathology who had completed a 1-mo rotation in cytopathology. Reviewer 2 was a surgical pathologist with only an ancillary interest in cytopathology who does not routinely interpret cytologic specimens. Reviewer 3 was a surgical pathologist with a strong interest and special expertise in cytopathology. “Positive” diagnoses were recorded as (1) malignant neoplasm, not otherwise specified; (2) carcinoma or sarcoma; ( 3 ) adenocarcinoma, squamous carcinoma, or lymphoma; or (4) atypical cells, inconclusive for diagnosis. Evaluation of cytologic preparations for the purposes of this article was focused on the simple discrimination between benign and malignant disease. Each reviewer evaluated all specimens to distinguish malignant neoplasms from nonneoplastic tissues or benign neoplasms. The sensitivity, specificity, positive predic-
tive value, negative predictive value, and efficiency for each organ system were determined. Calculations were made according to formulas described elsewhere.“ Calculations were applied to imprint results, frozen-section results, and combined imprint-frozen-section results.
Results Efficiency (EFF) reflects the relative accuracy of a test system, i.e., the relative number of times the correct result is achieved. Table I1 shows the efficiency of imprint and frozen-section evaluations as a function of the organ system and the reviewer’s experience. Although absolute concordance between the two methods and between observers did not result, certain trends emerged. Imprint examination of some systems and organs, such as central nervous system (CNS) and lymph nodes, tended to be of low efficiency; for others, such as body surface and breast lesions, it was of high efficiency. Efficiency was consistently lower for imprint evaluations than for frozen sections. As would be expected, relative efficiency also correlated favorably with reviewer experience. To demonstrate the value of the adjunctive use of imprint cytology, false-positive and false-negative results are provided in Table 111. They are separated according to imprint assessment, frozen-section assessment, and combined imprint and frozen-section assessment, when the imprint and frozen-section results were in accord. Results are also expressed as a function of reviewer experience. Not surprisingly, false-negative results were more frequent on the imprint evaluations. Moreover, combined imprint and frozen-section results improved only slightly on the false-negative results for frozen section alone. However, in the cases of false-positive results, the number of incorrect readings was substantially lower than for false-negatives, and the use of combined imprint and frozen section results markedly reduced the error rate. This clearly underscores the improved accuracy of combined frozen sections and imprints. Specificity, sensitivity, and efficiency (Table IV) and predictive values (Table V) were expressed as a function
Table 11. Relative Diagnostic Efficiency in Frozen-Section and Imprint Evaluation (Percentages) Reviewer 2
Reviewer 1 System
CNS Lymph node Thyroid/parathyroid Bone/soft tissue Gastrointestinal Gynecologic Genitourinary Surface tissue Respiratory Breast
Reviewer 3
Imprint
FS
Imprint
FS
Imprint
FS
20 69 75 78 69 62 80 67 75 89
50 71 87 94 83 85 84 75 92 94
70
100 73 78 94 95 95 96 100 96 100
60 76 79 81 89 88 90 92 94 97
90 90 91 91 97 99 100 100 100 99
71 79 81 84 87 88 92 84 92
FS, frozen section. Diagnostic Cytopathology, Vol6, N o 5
305
KIM ET AL Table 111. Number of False-Negatives and False-Positives False-negatives Reviewer 1 2 3
False-positives
Imprint
FS
Imprint FS
172 98 71
83 38 17
76 33 15
+
Imprint
FS
15 6 5
20
Imprint + FS
1
10 0
6
1
FS, frozen section
of imprint evaluation; frozen-section evaluation; combined imprint and frozen-section evaluation; and reviewer experience. These results also reflect the lower sensitivity of the imprint method and the very similar specificities of imprints and frozen sections. Specificity is improved slightly with the use of combined imprints and frozen section than with frozen sections alone; efficiency also tended to improve in combinations, although only slightly. In these studies, equivocal results were counted as negative, since an uncertain or suspicious diagnosis would not provide a suitable basis for further surgical therapy. These results included instances of nondiagnostic frozen section when the smear was conclusive, and vice versa. The study only included those cases in which both frozen section and imprint were available. Some specimens, excluded from the enumeration, had only cytology; tiny parathyroid samples, as well as bone biopsies, belong in this group. In such cases, cytologic evaluation offered the only intraoperative assessment.
Discussion Since its advent, the frozen section has been the mainstay of rapid intraoperative diagnosis and will likely remain so unless a more precise technique is developed. At our institution, the use of cytologic preparations as a companion to the frozen section was begun in 1980. It has been believed to be a consistent aid in the assessment of difficult frozen sections, but this opinion largely stemmed from anecdotal experience. To demonstrate the value of diagnostic intraoperative cytology and, more specifically, of combined frozen-section and cytologic evaluations, we examined specimens accrued over a 2-yr period in which both types of preparations were available. The evaluation of these specimens, although retrospective, provided a comparison of specificity, sensitivity, and efficiency as a function of the results of cytologic evaluation alone, Table IV.
frozen section evaluation alone, and combined frozen section and cytology. In addition, the results were also examined as a function of reviewer experience. This includes experience in anatomic pathology in general but particularly in cytopathology. This experience is important because confidence in cytopathologic interpretation may contribute to the acceptance of this method. It should be mentioned that in this review, false readings may be anomalously high because of (at least) two sources of bias: ( 1 ) lack of clinical information at the time of retrospective reading; and (2) a possible tendency toward more aggressive interpretation because of the lack of any adverse consequence of error under the conditions of this study. Moreover, cases judged as “suspicious” were classed as negative for the purpose of this study. In none of the evaluations did the accuracy of imprint evaluation alone surpass that of frozen section alone. That false-negativeswere consistently higher in cytologic assessment than in frozen section assessment is largely a result of improper sampling of specimens (e.g., air-dried specimens, sclerotic tissues) and is not surprising. The falsenegative rate in sclerotic tissue can be reduced by scraping the tissue using a scalpel or wooden spatula.’ In addition, some specimens were particularly prone to give false-negative results (e.g., microfocal carcinoma in lymph nodes or bone) or interpretations (e.g., certain lymphomas). On the other hand, instances were encountered in which imprint cytologic evaluation provided the only morphologic evidence of malignancy. The low efficiency of imprint evaluation, e.g., in CNS specimens, is in contrast to that reported by Marshall et al.” However, in this study, the number of encounters with CNS material was small and their imprint samples were frequently not appropriate. Among the false-positive imprint interpretations, the leader appeared to be adenomatous polyps of the colon,
Percentages of Specificity, Sensitivity, and Efficiency
-
specificity
Sensitivity Imprint
Reviewer
Imprint
FS
+ FS
Imprint
FS
+ FS
Imprint
FS
Imprint f FS
96.5 98.6 98.8
95.3 99.8 98.6
91.6 100.0 99.8
21.1 58.5 69.9
64.8 83.9 92.8
42.9 80.1 91.6
71.8 84.3 88.6
84.5 94.1 96.5
84.2 94.4 91.3
. . ~ ~
1 2 3
FS, frozen section.
306
Eficiency
Diagnostic Cytopathology, Vol6, N o 5
Imprint
INTRAOPERATIVE IMPRINT CYTOLOGY Table V. Percentages of Positive and Negative Predictive Value PV-
PV+ Reviewer 1
2 3
Imprint
FS
Imprint + FS
81.0 95.8 97.1
88.4 99.5 92.3
85.1 100.0 99.4
Imprint
FS
Imprint + FS
70.6 81.2 85.6
83.1 91.8 96.1
84.1 92.7 96.5
PV+, positive predictive value; PV -, negative predictive value, FS, frozen section.
which provided four, three, and two false-positives for reviewers 1, 2, and 3, respectively. Diagnosis based on cytologic evaluation of adenomatous polyps should be avoided, and adjunctive use at frozen section should be approached with caution. Of central importance to this study was the effect of using cytopathologic evaluation as an adjunct to frozen section. In the expression of results in terms of sensitivity, specificity, and efficiency (Table IV), it appears that the sum of cytologic and cryohistologic assessment does not improve the sensitivity of frozen section, nor does it significantly improve efficiency. This is principally because of the high rate of false-negatives on cytology evaluations, as discussed above. However, sensitivity and efficiency of imprint cytologic evaluation can be significantly improved by better sampling technique. Specificity, which relates to the correctness of diagnosis of positive specimens, is very similar between imprints and frozen sections and appears to improve little numerically by the summation of cytologic and frozen section interpretations. However, this apparently slight difference in percentage, attributable to a reduction of falsepositives, translated into an impressive gap in the small number of specimens involved (Table 111). Here, the use of agreement between frozen section and cytology significantly reduces false-positives, regardless of the level of observer experience or expertise. At surgery, falsepositive diagnoses are potentially the most damaging diagnostic error. This strongly emphasizes the value of imprint evaluation in rapid intraoperative diagnosis as an adjunct to the frozen section. In addition to supportive information, useful in discriminating between benign conditions and cancer, imprint cytology provides the following benefits: (1) evaluation of small specimens, which risk being lost during preparation for frozen section; (2) preservation of material for special studies such as electron microscopy; (3) evaluation of specimens in which frozen section is not feasible for technical reasons (e.g., bone); (4) exhibition of histogenetic features aiding in evaluation of the specimen (e.g.,
parathyroid tissue, cell of origin of poorly differentiated neoplasms, and cytologic type of lymphoma); and (5) promotion of the teaching of cytologic interpretation and histologic and cytologic correlation. In spite of numerous previous studies describing the value of imprint cytology as an intraoperative diagnostic method, it has not become a routine practice in many institutions. This study provides a quantitative assessment of the use of imprint cytologic method in conjunction with frozen section and emphasizes its adjunctive value for practitioners over a wide range of experience and expertise.
Acknowledgrnent The authors are grateful to Dr. Bernard Naylor for his assistance with the manuscript and to Ms. Marilyn Cline for her excellent typing services.
References 1. Dudgeon LS, Patrick CV. A new method for the rapid microscopic diagnosis, with an account of 200 cases so examined. Br J Surg 1927;15:250+. 2. Schnicker KT, Hermanek P. Intraoperative histology or cytology? Virchows Arch [A] 1974;362:247-58. 3. Westlake PT. An aspiration and touch prep cytology smear as a teaching tool. Cytotechnol Bull 1982;19:26. 4. Kern WH. Presidential address. Acta Cytol 1982;26:755-8. 5. Blaustein PA, Silverberg SG. Rapid cytological examination of surgical specimens. Path Annu 1977;12:251-78. 6. Owings RM. Rapid cytologic examination of surgical specimens: a valuable technique in the surgical pathology laboratory. Hum Pathol 1984;15:605-14. 7. Wilkinson JA, Bonnin JM. Intraoperative cytology; an adjunct to frozen sections. New York: Igaku-Shoin, 1987:9-11. 8. Lee TK. The value of imprint cytology in tumor diagnosis: a retrospective study of 522 cases in Northern China. Acta Cytol 1983;26: 169-7 1. 9. Studham VB, Dravid MV, Grovers, Kher A. Role of scrape cytology in rapid intraoperative diagnosis. Acta Cytol 1984;28:477-82. 10. Galen RS, Gambino SR. Beyond normality: the predictive value and efficiency of medical diagnosis. New York: Wiley & Sons, 1975. 11. Marshall LF, A d a m H, Doyle D, Graham DI. The histological accuracy of the cancer technique for neurosurgical biopsies. J Neurosurg 1973;39:82-8.
Diagnostic Cytopathology, Vol6, No 5
307
