AJCP / Original Article
Impact of p16INK4a Immunohistochemistry Staining on Interobserver Agreement on the Diagnosis of Cervical Intraepithelial Neoplasia Alberto Pacchiarotti, MD,1 Simona Galeotti, MSc,2 Paola Bellardini, MD,3 Francesco Chini, MSc,4 Guido Collina, MD,5 Paolo Dalla Palma, MD,6 Bruno Ghiringhello, MD,7 Vincenzo Maccallini, MD,8 Fabio Musolino, MSc,1 Giovanni Negri, MD,9 Roberto Pisa, MD,10 Ilaria Sabatucci, MD,11 and Paolo Giorgi Rossi, PhD2 From 1Lega Italiana per la Lotta contro i Tumori, Sezione Provinciale di Latina, Latina, Italy; 2Servizio Interaziendale di Epidemiologia, AUSL Reggio Emilia, Reggio Emilia, Italy; 3Screening, Latina Local Health Authority ASL di Latina, Latina, Italy; 4Laziosanità, Agenzia di Sanità Pubblica Regione Lazio, Rome, Italy; 5UO Anatomia Patologica, AUSL di Bologna, Bologna, Italy; UO Anatomia Patologica, Ospedale Maggiore, AUSL di Bologna, Bologna, Italy; 6Anatomia Patologica, Ospedale S. Chiara, Trento, Italy; 7Anatomia Patologica, OIRM S. Anna, Turin, Italy; 8UO Aziendale di Screening, ASL 1 Abruzzo, Pescina (AQ), Italy; 9Dipartimento di Patologia, Central Hospital Bolzano, Bolzano, Italy; 10Servizio Anatomia Patologica, Azienda Ospedaliera S. Camillo-Forlanini, Rome, Italy; and 11Department of Gynecologic–Obstetrical and Urologic Sciences, University of Rome, Rome, Italy. Key Words: p16 protein/p16INK4a (cyclin-dependent kinase inhibitor); Cervical cancer; Mass screening; Immunohistochemistry; Cervical intraepithelial neoplasia; Reproducibility DOI: 10.1309/AJCPCYWVL61SVKFU
ABSTRACT Objectives: This study aimed to compare the interobserver Cohen κ on H&E staining and on H&E plus p16INK4a staining of all cervical biopsy specimens in a populationbased screening program. Methods: All the colposcopy-guided biopsies generated by the routine screening of 23,258 women aged 25 to 64 years were stained with H&E and H&E plus p16. Biopsy specimens were reviewed by six external experts. Results: The four diagnoses were available in 441 cases. The interobserver κ values were 0.52 (95% confidence interval [CI], 0.45-0.58) and 0.48 (95% CI, 0.42-0.56) with H&E and H&E + p16, respectively, when using a five-group classification (normal, CIN 1, CIN 2, CIN 3, and cancer); adopting a two-group classification (≤CIN 1 and ≥CIN 2), the values were 0.75 (95% CI, 0.66-0.82) and 0.70 (95% CI, 0.61-0.79), respectively. Conclusions: The use of p16 on all cervical biopsy specimens in a screening program showed virtually no effect on reproducibility of the histologic diagnosis.
© American Society for Clinical Pathology
Cervical cancer is a well-controlled disease in industrialized countries because of the widespread use of the Papanicolaou (Pap) test and the implementation of screening programs.1 With decreasing prevalence of the disease, it is important to pay more attention to the specificity of the screening algorithm and to reducing overdiagnosis and overtreatment of regressive lesions. This will be even more critical in human papillomavirus (HPV)–based screening.2 In this scenario, the use of diagnostic tools to better assess high-grade histologic lesions is relevant because these lesions are the targets of surgical treatment.3 Thus far, the most promising biomarkers that improve the diagnostic accuracy in histology have been those linked to E6/E7 overexpression.4 Specifically, several studies have tested the cross-sectional accuracy of p16,5-7 a cyclin-dependent kinase inhibitor that is involved in the genesis of different tumors. In cervical cancer, the overexpression of p16 is linked to the oncogenic transformation caused by persistent high-risk (HR) HPV infection. Testing for HR-HPV cannot distinguish between the transient infections that make up the majority of cases and transforming infections. However, p16 overexpression is correlated with the start of the oncogenic transformation process, which is necessary for the development of cervical cancer and its precursor lesions. Histology is usually considered the gold standard, and it is always challenging to measure the sensitivity and specificity of a test to improve the gold standard itself. Moreover, in cervical cancer precursors, it is impossible to measure the validity of the histologic diagnosis through clinical follow-up of the lesions because most of the lesions targeted for surgical
367
Am J Clin Pathol 2014;141:367-373 DOI: 10.1309/AJCPCYWVL61SVKFU
367 367
Pacchiarotti et al / Impact of p16INK4a on Interobserver Agreement
treatment are actually regressive, and recurrence after treatment is a rare event.8 As an indirect measure of diagnostic accuracy, some authors have proposed9,10 using the interobserver Cohen κ11 if the κ values of concordance between two expert observers on the same p16-assessed samples are higher than those assessed with H&E. This would imply that the diagnostic criteria are less subjective and more reproducible and thus that the final diagnosis resulting from these criteria is probably more accurate. The usefulness of p16 in the differential diagnosis of cervical intraepithelial neoplasia (CIN) when resolving uncertain morphologies is well recognized.6,12-14 However, it is not clear when it should be used, and only recently has there been an effort in systematically studying the cases in which p16 staining is recommended.15 Furthermore, to our knowledge, no population-based studies have been conducted on the use of p16 as a routine auxiliary method for all histologic specimens in a cervical cancer screening program. The objective of this study was to evaluate whether p16 in histology is a valid biomarker in improving the accuracy of diagnosis. Because of the lack of a gold standard, we aimed to determine whether the diagnostic concordance of different pathologists increases when using p16 and whether the classification of doubtful cases decreases.
Materials and Methods Population and Study Design The study was conducted in the province of Latina in central Italy, with a population of 550,000 in 2010; 160,000 women were in the target age group of the screening program (25-64 years). The province-wide organized screening program, which actively invites all the target population for Pap tests every 3 years, has been in place since 1999. The population base of the study consisted of 23,258 women who underwent a Pap test for a new screening round between May 2008 and September 2010; all the biopsy samples collected during the first colposcopy following the Pap test were considered for the inclusion criteria. Women in posttreatment or postcolposcopy follow-up care were excluded. All women with positive Pap tests with a diagnosis of atypical squamous cells of undetermined significance (ASC-US) or more severe according to the 2001 Bethesda System were invited to undergo a colposcopy. Almost all biopsy specimens (590/595) were stained with routine methods and with p16. All biopsy specimens were analyzed as part of the screening in Latina using H&E plus p16 staining. Slides were rendered anonymous by independent operators and reassessed after a 12-month period of 368 368
Am J Clin Pathol 2014;141:367-373 DOI: 10.1309/AJCPCYWVL61SVKFU
washout on the basis of H&E staining only by the pathologist from Latina, who made the first diagnosis. The 590 cases were then randomly assigned to six experts in cervicovaginal pathology. These pathologists analyzed the slides with H&E and, after at least 3 months of washout, reanalyzed the cases using both H&E and p16 slides. Immunohistochemistry Tissue blocks were cut, stained, and interpreted at the Lega Italiana per la Lotta contro i Tumori pathology laboratory in Latina. Serial slices of the same punch biopsy specimen were alternately placed on two slides, one of which was processed using routine procedures for H&E staining and the other for immunostaining for p16INK4a using the CINtec Histology Kit (Roche MTM Laboratories, Heidelberg, Germany) according to the manufacturer’s instructions: 10 minutes in a water bath at 95°C to 99°C to retrieve the antigen; 5-minute wash in a peroxidase-blocking reagent; 30-minute incubation and wash with the p16INK4a antibody (clone E6H4); 30-minute wash in a secondary antibody reagent (polymer-based goatantimouse antibody fragment conjugated to horseradish peroxidase); and finally a chromogenic visualization step using the 3,3′-diaminobenzidine chromogen. Slides were then counterstained with hematoxylin and coverslipped. Microscopic evaluation was performed according to the CIN classification (CIN 1, CIN 2, and CIN 3). Data Analysis and Statistical Methods A gold standard is not available for histology because histology itself is the final and most accurate diagnosis we can obtain for most cervical lesions. Without a gold standard, however, it is not possible to directly measure test validity. As an indirect measure of diagnostic accuracy, some authors propose using the Cohen κ: if the agreement between two different pathologists on the same sample is higher using p16 than with H&E alone, this means that diagnoses are more reproducible and, probably, more accurate. A contingency table and the Cohen κ have been calculated to analyze the concordance between observers and to determine whether p16 improves the diagnostic reproducibility. We present the unweighted κ statistics for a five-category classification (normal, including metaplasia; CIN 1; CIN 2; CIN 3, including adenocarcinoma in situ; and invasive cancer) and a binary classification (≥CIN 2 and ≤CIN 1); 95% confidence intervals of κ statistics were calculated with a bootstrap method (Stata 11.0, StataCorp, College Station, TX).16 The intraobserver concordance between H&E and H&E plus p16 was analyzed to identify which cases were reclassified and to verify whether p16 staining reduces inconclusive diagnoses (inadequate and CIN not otherwise specified [NOS]) and low reproducible diagnoses (CIN 2).9,17,18 © American Society for Clinical Pathology
AJCP / Original Article
The study was evaluated by the Direzione Sanitaria of the Latina Local Health Unit, which asked for a slightly modified informed consent for colposcopy-guided biopsies explaining the review process and, given the analogy with routine quality control procedures, waived the study from ethical review board submission.
Results Of the 590 biopsies performed, 53 were excluded because it was not possible to recover the slides for external review, 90 were excluded because one of the external reviewers decided not to participate in the study after slide transmission, and six were excluded because it was impossible to link the slide to the case. Thus, the final data set consisted of 441 women with a mean age of 37 years and a range of 25 to 64 years.
These women underwent colposcopy because their Pap test findings were positive. In more than 80% of the cases, a low-grade squamous intraepithelial lesion (L-SIL) or ASCUS was observed ❚Table 1❚. The overall positive predictive value of colposcopy-guided biopsy was 18.7%, with 8.8% for biopsies in women referred for ASC-US and L-SIL and 68.2% for those referred for abnormal squamous cell suggesting high grade and high-grade squamous intraepithelial lesion. Cohen κ comparison between the local pathologist and the external reviewers for the diagnoses made with only H&E staining is 0.517 and 0.753 for the five- and two-category classifications, respectively ❚Table 2❚ and ❚Table 3❚. The κ values for diagnoses made using H&E plus p16 immunostaining are extremely similar: 0.483 and 0.700 for the five- and two-category classifications, respectively (Tables 2 and Table 3). The results do not show differences between the reviewers ❚Table 4❚.
❚Table 1❚ Distribution of the Cytologic and Histologic Diagnoses According to the Original Diagnosis of the Reviewed Cases Histology Cytology
Negative
CIN 1
CIN 2
CIN 3
Invasive Cancer Inadequate
ASC-US 82 13 8 2 1 L-SIL 170 66 17 4 0 AGC 4 0 1 0 0 ASC-H 6 0 2 1 0 H-SIL 17 4 25 18 1 Carcinoma 0 0 0 2 0 Adenocarcinoma 1 0 0 0 1 Inadequate 2 0 0 0 0 Total 282 83 53 27 3 % 62.8 18.5 11.8 6.0 0.7
0 0 0 0 1 0 0 0 1 0.2
Total No. (%)
% PPV for CIN 2+
106 (23.6) 10.4 257 (57.2) 8.2 5 (1.1) 20.0 9 (2.0) 33.3 66 (14.7) 68.2 2 (0.4) 100.0 2 (0.4) 50.0 2 (0.4) 0.0 449 (100) 18.7 100
AGC, abnormal glandular cells; ASC-H, abnormal squamous cell suggesting high grade; ASC-US, atypical squamous cells of undetermined significance; CIN, cervical intraepithelial neoplasia; CIN 2+, cervical intraepithelial neoplasia grade 2 or more severe; H-SIL, high-grade squamous intraepithelial lesion; L-SIL, low-grade squamous intraepithelial lesion; PPV, positive predictive value.
❚Table 2❚ Contingency Table of Interobserver Agreement on Diagnosis Made With H&E and H&E + p16 Staining Based on a Five-Category Classification Diagnosis Made by External Pathologists Diagnosis Made by Local Pathologist
Negative
CIN 1
CIN 2
CIN 3
Invasive Cancer
H&E staining Negative 220 38 1 1 0 CIN 1 16 48 13 3 0 CIN 2 4 8 17 20 1 CIN 3 2 0 9 13 1 Invasive cancer 0 0 0 0 3 Total 242 94 40 37 5 H&E + p16 staining Negative 193 30 1 0 0 CIN 1 37 58 22 2 0 CIN 2 2 7 13 17 0 CIN 3 2 3 7 20 3 Invasive cancer 0 0 0 0 1 Total 234 98 43 39 4
Total 260 80 50 25 3 418 224 119 39 35 1 418
CIN, cervical intraepithelial neoplasia.
© American Society for Clinical Pathology
369
Am J Clin Pathol 2014;141:367-373 DOI: 10.1309/AJCPCYWVL61SVKFU
369 369
Pacchiarotti et al / Impact of p16INK4a on Interobserver Agreement
❚Table 3❚ Contingency Table of Interobserver Agreement on Diagnosis Made With H&E and H&E + p16 Staining Based on a Binary Classification
Diagnosis Made by External Pathologists
Diagnosis Made by Local Pathologist CIN 1–
CIN 2+
H&E staining CIN 1– 322 18 CIN 2+ 14 64 Total 336 82 H&E + p16 staining CIN 1– 318 25 CIN 2+ 14 61 Total 332 86
Total 340 78 418 343 75 418
reclassified as negative (among these there could be atypical immature metaplasia), 13 as CIN 1, and 16 as CIN 3. With p16, two cases of CIN NOS changed to negative, but one negative case and two CIN 1 cases changed to CIN NOS. Finally, if diagnoses are grouped into two classes (≤CIN 1 and ≥CIN 2), the two different methods lead to the same conclusion regarding the necessity to treat in 95% of the cases: with p16, 19 cases are reclassified as more severe and 18 as less severe. The intraobserver κ of the two readings was 0.71 and 0.86 for the five- and two-category classifications, respectively ❚Table 6❚ and ❚Table 7❚, with individual values ranging from 0.53 to 0.95 and from 0.69 to 1, respectively ❚Table 8❚.
CIN, cervical intraepithelial neoplasia; CIN 1–, cervical intraepithelial neoplasia grade 1 or less severe; CIN 2+, cervical intraepithelial neoplasia grade 2 or more severe.
The contingency table of the diagnoses made using H&E and H&E plus p16 by the same pathologist illustrates that p16 staining does not reduce inconclusive diagnoses (inadequate and CIN NOS) and low reproducible diagnoses (CIN 2) ❚Table 5❚. Of the 94 cases classified as CIN 2 with H&E, 61 maintained this classification, four were
Discussion In this population-based study, we did not find any difference in concordance among pathologists when diagnoses were made using H&E alone or H&E plus p16. Furthermore, we did not observe any reduction in inconclusive diagnoses (inadequate and CIN NOS) and low reproducible diagnoses (CIN 2).
❚Table 4❚ Interobserver Agreement and Cohen k Values by Each External Reviewer vs Local Pathologist
Five Categories
H&E
κ
Two Categories
H&E + p16
95% Confidence Interval κ
Overall 0.517 0.453-0.584 Reviewer 1 0.599 0.454-0.735 Reviewer 2 0.593 0.447-0.738 Reviewer 3 0.413 0.258-0.590 Reviewer 4 0.489 0.371-0.626 Reviewer 5 0.450 0.284-0.601
H&E
95% Confidence Interval κ
0.483 0.420-0.557 0.495 0.346-0.635 0.418 0.277-0.563 0.488 0.339-0.652 0.538 0.395-0.668 0.455 0.299-0.611
H&E + p16
95% Confidence Interval κ
0.753 0.663-0.821 0.795 0.551-0.954 0.819 0.588-0.959 0.690 0.445-0.881 0.691 0.487-0.881 0.764 0.553-0.907
95% Confidence Interval
0.700 0.610-0.786 0.680 0.452-0.861 0.639 0.352-0.847 0.805 0.544-0.953 0.731 0.551-0.887 0.641 0.433-0.816
❚Table 5❚ Contingency Table of Intraobserver Agreement of H&E vs H&E + p16 Staining Based on Descriptive Classification H&E
H&E + p16 Inadequate Negative
Metaplasia CIN 1
Inadequate 14 1 0 Negative 7 407 4 Metaplasia 0 28 2 CIN 1 0 24 0 CIN 2 0 4 0 CIN 3 0 0 0 AIS 0 0 0 CIN NOS 0 2 0 SCC 0 0 0 AK 0 0 0 Total 21 466 6
CIN 2
CIN 3
AIS
CIN NOS SCC
AK
Total
0 0 0 0 0 0 0 15 62 6 3 0 1 0 0 490 2 1 0 0 0 0 0 33 141 7 2 0 2 0 0 176 13 61 16 0 0 0 0 94 3 8 50 1 0 0 0 62 0 0 0 0 0 0 0 0 0 1 0 0 1 0 0 4 0 0 3 0 0 4 1 8 0 0 0 0 0 0 0 0 221 84 74 1 4 4 1 882
AIS, adenocarcinoma in situ; AK, adenocarcinoma; CIN, cervical intraepithelial neoplasia; NOS, not otherwise specified; SCC, squamous cell carcinoma.
370 370
Am J Clin Pathol 2014;141:367-373 DOI: 10.1309/AJCPCYWVL61SVKFU
© American Society for Clinical Pathology
AJCP / Original Article
❚Table 6❚ Intraobserver Concordance Between H&E and H&E + p16 Diagnoses Based on Five-Category Classification
Diagnosis Based on H&E + p16 Staining
Diagnosis Based on H&E Staining Negative
CIN 1
CIN 2
CIN 3
Invasive Cancer
Total
Negative 431 61 7 3 0 502 CIN 1 24 141 7 2 0 174 CIN 2 3 12 60 15 0 90 CIN 3 0 3 8 51 0 62 Invasive cancer 0 0 0 3 5 8 Total 458 217 82 74 5 836 CIN, cervical intraepithelial neoplasia.
This conclusion is in contrast with those of previous studies.10,19-25 Bergeron and colleagues19 noted a significant 13% increase in sensitivity for CIN 2+ when p16INK4a-immunostained slides were added and interpreted together with the H&E slides; they also found that the κ values significantly improved from 0.566 to 0.749. Gurrola-Díaz and colleagues21 observed an increase in κ values among three pathologists when using a binary classification from 0.36, 0.12, and 0.18 for each pair of observers using H&E to 0.59, 0.70, and 0.59 using p16INK4a stains. Dijkstra and colleagues20 found a similar increase in interobserver agreement among three pathologists from a weighted κ ranging from 0.44 to 0.66 in H&E slides to 0.80 with H&E plus p16. Horn and colleagues22 found an increase in κ values for punch biopsies—from 0.49 to 0.64—when H&E-stained slides were read together with p16INK4a-stained slides. Finally, Sari Aslani et al23 found that p16 increased the interobserver agreement as well as consensus. Our results are in line with the recent College of American Pathologists and American Society for Colposcopy and Cervical Pathology working group recommendations “against the use of p16 IHC as a routine adjunct to histological assessment of biopsy specimens with morphologic interpretations of negative, -IN1, and -IN3” (wherein IN can be interpreted as CIN for our purposes).15 The difference with previous studies could be the result of two factors: (1) κ in this study is high in the H&E analysis
❚Table 7❚ Intraobserver Concordance Between H&E and H&E + p16 Diagnoses Based on Two-Category Classification H&E
H&E + p16 CIN 1–
CIN 2+
Total
CIN 1– 657 19 676 CIN 2+ 18 142 160 Total 675 161 836 CIN 1–, cervical intraepithelial neoplasia grade 1 or less severe; CIN 2+, cervical intraepithelial neoplasia grade 2 or more severe
as well, and (2) the study population had a low frequency of cases with uncertain interpretation. In fact, the κ for the binary classification was higher than 0.7, which is considered as having very good accordance and standing in the upper part of the distribution of published studies. Dalla Palma and colleagues9 found an overall κ value of 0.65 for a binary classification of CIN 2+, Stoler and Schiffman26 found a value of 0.68, and Carreon and colleagues18 found a 0.70 κ value between two reviewers from the same center and values from 0.42 to 0.46 between reviewers from different centers. Only Malpica and colleagues27 and Crum and colleagues28 reported κ values higher than 0.75. This study is population based; therefore, it considers all women screened and thus makes it possible to foresee the real impact that using p16 could have. On the other hand, this
❚Table 8❚ Intraobserver Concordance Between H&E and H&E + p16 Diagnosesa
Five Categories
Two Categories
κ
95% Confidence Interval
κ
95% Confidence Interval
Overall Local pathologist Reviewer 1 Reviewer 2 Reviewer 3 Reviewer 4 Reviewer 5
0.705 0.694 0.626 0.527 0.952 0.828 0.615
0.663-0.745 0.627-0.743 0.483-0.784 0.346-0.675 0.871-1.000 0.712-0.924 0.462-0.744
0.857 0.849 0.920 0.694 1 0.910 0.777
0.806-0.899 0.773-0.910 0.762-1.000 0.451-0.888
a
0.791-1.000 0.603-0.929
Cohen κ values shown are those obtained by each pathologist for five- and two-category classifications.
© American Society for Clinical Pathology
371
Am J Clin Pathol 2014;141:367-373 DOI: 10.1309/AJCPCYWVL61SVKFU
371 371
Pacchiarotti et al / Impact of p16INK4a on Interobserver Agreement
approach makes it difficult to observe rare morphologies for which p16 immunostaining could be more useful. In fact, p16 has been demonstrated to be useful in differential diagnosis of some ambiguous morphologies (ie, glands with immature atypical squamous metaplasia,12-14 which in H&E analysis can be erroneously interpreted as CIN 2). In the current study, four cases were classified as CIN 2 using H&E which, after p16 analysis, were found to be negative, including the atypical immature metaplasia.6 The main limitation of our study was that the pathologists did not interpret the slides under the same circumstances: the local pathologist made the interpretation with H&E plus p16 under the psychological pressure of a clinical diagnosis and used H&E only as a quality control and for study purposes. The external reviewers, on the other hand, performed both interpretations for study purposes, without any psychological pressure and in the opposite time order: first H&E and then H&E plus p16. Although the reviewers came from centers throughout Italy, most had participated in several quality control and concordance studies, thereby increasing the baseline agreement in all conditions. Finally, this population-based study included about 1 year’s workload of a medium-sized screening program without any selection. Unfortunately, we can only report the results from five of the six reviewers who performed the external review of the slides because the data of the sixth reviewer were lost, and it was impossible to re-evaluate the slides. However, because the slides were assigned to the reviewers according to a random sampling, the final estimates were not biased by the incomplete nature of the data. Although the usefulness of p16 in resolving uncertain morphologies is well known, the use of p16 on all cervical biopsy specimens in a screening program showed virtually no effect on reproducibility of the histologic diagnosis. The study was supported by the Lega Italiana per la Lotta contro i Tumori, Sezione Provinciale di Latina, and the Latina Local Health Authority (ASL). Address reprint requests to Dr Giorgi Rossi: Servizio Interaziendale di Epidemiologia, AUSL Reggio Emilia, Via Amendola 2, 42122, Reggio Emilia, Italy; [email protected]. Dr Rossi is the principal investigator of a trial funded by the Italian Ministry of Health for which reagents were obtained at a reduced price or for free from Roche, Hologic-GenProbe, and Qiagen.
References 1. IARC Working Group on the Evaluation of Cancer Preventive Strategies. Cervix Cancer Screening: IARC Handbooks of Cancer Prevention No.10. Lyon, France: IARC Press; 2005. 2. Cuzick J, Clavel C, Petry CH, et al. Overview of the European and North American studies on HPV testing in primary cervical screening. Int J Cancer. 2006;119:1095-1101. 372 372
Am J Clin Pathol 2014;141:367-373 DOI: 10.1309/AJCPCYWVL61SVKFU
3. Arbyn M, Anttila A, Jordan J, et al, eds. European Guidelines for Quality Assurance on Cervical Cancer Screening. 2nd ed. Luxembourg, Luxembourg: Office for Official Publications of the European Communities; 2008. 4. Ghittoni R, Accardi R, Hasan U, et al. The biological properties of E6 and E7 oncoproteins from human papillomaviruses. Virus Genes. 2010;40:1-13. 5. Liang J, Mittal KR, Wei JJ, et al. Utility of p16INK4a, CEA, Ki67, P53 and ER/PR in the differential diagnosis of benign, premalignant, and malignant glandular lesions of the uterine cervix and their relationship with Silverberg scoring system for endocervical glandular lesions. Int J Gynecol Pathol. 2007;26:71-75. 6. Negri G, Egarter-Vigl E, Kasal A, et al. p16INK4a is a useful marker for the diagnosis of adenocarcinoma of the cervix uteri and its precursors: an immunohistochemical study with immunocytochemical correlations. Am J Surg Pathol. 2003;27:187-193. 7. Tringler B, Gup CJ, Singh M, et al. Evaluation of p16INK4a and pRb expression in cervical squamous and glandular neoplasia. Hum Pathol. 2004;35:689-696. 8. Schiffman M, Castle PE, Jeronimo J, et al. Human papilloma virus and cervical cancer. Lancet. 2007;370:890-907. 9. Dalla Palma P, Giorgi Rossi P, Collina G, et al. The reproducibility of CIN diagnoses among different pathologists: data from histology reviews from a multicenter randomized study. Am J Clin Pathol. 2009;132:125-132. 10. Klaes R, Benner A, Friedrich T, et al. p16INK4a immunohistochemistry improves interobserver agreement in the diagnosis of cervical intraepithelial neoplasia. Am J Surg Pathol. 2002;26:1389-1399. 11. Cohen J. A coefficient of agreement for nominal scales. Educ Psychol Meas. 1960;20:213-220. 12. Kong CS, Balzer BL, Troxell ML, et al. p16INK4a immunohistochemistry is superior to HPV in situ hybridization for the detection of high-risk HPV in atypical squamous metaplasia. Am J Surg Pathol. 2007;31:33-43. 13. Pinto AP, Schlecht NF, Woo TY, et al. Biomarker (ProEx C, p16(INK4A), and MiB-1) distinction of high-grade squamous intraepithelial lesion from its mimics. Mod Pathol. 2008;21:1067-1074. 14. Regauer S, Reich O. CK17 and p16 expression patterns distinguish (atypical) immature squamous metaplasia from high-grade cervical intraepithelial neoplasia (CIN III). Histopathology. 2007;50:629-635. 15. Darragh TM, Colgan TJ, Thomas Cox J, et al. The Lower Anogenital Squamous Terminology Standardization Project for HPV-associated lesions: background and consensus recommendations from the College of American Pathologists and the American Society for Colposcopy and Cervical Pathology. Int J Gynecol Pathol. 2013;32:76-115. 16. StataCorp. Stata Statistical Software: Release 11. College Station, TX: StataCorp LP; 2009. 17. Cai B, Ronnett B, Stoler M, et al. Longitudinal evaluation of interobserver and intraobserver agreement of cervical intraepithelial neoplasia diagnosis among an experienced panel of gynecologic pathologists. Am J Surg Pathol. 2007;31:1854-1860. 18. Carreon JD, Sherman ME, Guillén D, et al. CIN2 is a much less reproducible and less valid diagnosis than CIN3: results from a histological review of population-based cervical samples. Int J Gynecol Pathol. 2007;26:441-446.
© American Society for Clinical Pathology
AJCP / Original Article
19. Bergeron C, Ordi J, Schmidt D, et al. Conjunctive p16INK4a testing significantly increases accuracy in diagnosing highgrade cervical intraepithelial neoplasia. Am J Clin Pathol. 2010;133:395-406. 20. Dijkstra MG, Heideman DA, de Roy SC, et al. p16INK4a immunostaining as an alternative to histology review for reliable grading of cervical intraepithelial lesions. J Clin Pathol. 2010;63:972-977. 21. Gurrola-Díaz CM, Suárez-Rincón AE, Vázquez-Camacho G, et al. P16INK4a immunohistochemistry improves the reproducibility of the histological diagnosis of cervical intraepithelial neoplasia in cone biopsies. Gynecol Oncol. 2008;111:120-124. 22. Horn LC, Reichert A, Oster A, et al. Immunostaining for p16INK4a used as a conjunctive tool improves interobserver agreement of the histologic diagnosis of cervical intraepithelial neoplasia. Am J Surg Pathol. 2008;32:502-512. 23. Sari Aslani F, Safaei A, Pourjabali M, et al. Evaluation of Ki67, p16 and CK17 markers in differentiating cervical intraepithelial neoplasia and benign lesions. Iran J Med Sci. 2013;38:15-21. 24. Vinyuvat S, Karalak A, Suthipintawong C, et al. Interobserver reproducibility in determining p16 overexpression in cervical lesions: use of a combined scoring method. Asian Pac J Cancer Prev. 2008;9:653-657.
© American Society for Clinical Pathology
25. Zhang Q, Kuhn L, Denny LA, et al. Impact of utilizing p16INK4a immunohistochemistry on estimated performance of three cervical cancer screening tests. Int J Cancer. 2007;120:351-356. 26. Stoler MH, Schiffman M, for the Atypical Squamous Cells of Undetermined Significance–Low-grade Squamous Intraepithelial Lesion Triage Study (ALTS) Group. Interobserver reproducibility of cervical cytologic and histologic interpretations: realistic estimates from the ASCUS-LSIL Triage Study. JAMA. 2001;285:1500-1505. 27. Malpica A, Matisic JP, Niekirk DV, et al. Kappa statistics to measure interrater and intrarater agreement for 1790 cervical biopsy specimens among twelve pathologists: qualitative histopathologic analysis and methodologic issues. Gynecol Oncol. 2005;99(suppl 1):S38-S52. 28. Crum CP, Beach KJ, Hedley ML, et al. Dynamics of human papillomavirus infection between biopsy and excision of cervical intraepithelial neoplasia: results from the ZYC101a protocol. J Infect Dis. 2004;189:1348-1354.
373
Am J Clin Pathol 2014;141:367-373 DOI: 10.1309/AJCPCYWVL61SVKFU
373 373
Impact of p16INK4a Immunohistochemistry Staining on Interobserver Agreement on the Diagnosis of Cervical Intraepithelial Neoplasia Alberto Pacchiarotti, MD,1 Simona Galeotti, MSc,2 Paola Bellardini, MD,3 Francesco Chini, MSc,4 Guido Collina, MD,5 Paolo Dalla Palma, MD,6 Bruno Ghiringhello, MD,7 Vincenzo Maccallini, MD,8 Fabio Musolino, MSc,1 Giovanni Negri, MD,9 Roberto Pisa, MD,10 Ilaria Sabatucci, MD,11 and Paolo Giorgi Rossi, PhD2 From 1Lega Italiana per la Lotta contro i Tumori, Sezione Provinciale di Latina, Latina, Italy; 2Servizio Interaziendale di Epidemiologia, AUSL Reggio Emilia, Reggio Emilia, Italy; 3Screening, Latina Local Health Authority ASL di Latina, Latina, Italy; 4Laziosanità, Agenzia di Sanità Pubblica Regione Lazio, Rome, Italy; 5UO Anatomia Patologica, AUSL di Bologna, Bologna, Italy; UO Anatomia Patologica, Ospedale Maggiore, AUSL di Bologna, Bologna, Italy; 6Anatomia Patologica, Ospedale S. Chiara, Trento, Italy; 7Anatomia Patologica, OIRM S. Anna, Turin, Italy; 8UO Aziendale di Screening, ASL 1 Abruzzo, Pescina (AQ), Italy; 9Dipartimento di Patologia, Central Hospital Bolzano, Bolzano, Italy; 10Servizio Anatomia Patologica, Azienda Ospedaliera S. Camillo-Forlanini, Rome, Italy; and 11Department of Gynecologic–Obstetrical and Urologic Sciences, University of Rome, Rome, Italy. Key Words: p16 protein/p16INK4a (cyclin-dependent kinase inhibitor); Cervical cancer; Mass screening; Immunohistochemistry; Cervical intraepithelial neoplasia; Reproducibility DOI: 10.1309/AJCPCYWVL61SVKFU
ABSTRACT Objectives: This study aimed to compare the interobserver Cohen κ on H&E staining and on H&E plus p16INK4a staining of all cervical biopsy specimens in a populationbased screening program. Methods: All the colposcopy-guided biopsies generated by the routine screening of 23,258 women aged 25 to 64 years were stained with H&E and H&E plus p16. Biopsy specimens were reviewed by six external experts. Results: The four diagnoses were available in 441 cases. The interobserver κ values were 0.52 (95% confidence interval [CI], 0.45-0.58) and 0.48 (95% CI, 0.42-0.56) with H&E and H&E + p16, respectively, when using a five-group classification (normal, CIN 1, CIN 2, CIN 3, and cancer); adopting a two-group classification (≤CIN 1 and ≥CIN 2), the values were 0.75 (95% CI, 0.66-0.82) and 0.70 (95% CI, 0.61-0.79), respectively. Conclusions: The use of p16 on all cervical biopsy specimens in a screening program showed virtually no effect on reproducibility of the histologic diagnosis.
© American Society for Clinical Pathology
Cervical cancer is a well-controlled disease in industrialized countries because of the widespread use of the Papanicolaou (Pap) test and the implementation of screening programs.1 With decreasing prevalence of the disease, it is important to pay more attention to the specificity of the screening algorithm and to reducing overdiagnosis and overtreatment of regressive lesions. This will be even more critical in human papillomavirus (HPV)–based screening.2 In this scenario, the use of diagnostic tools to better assess high-grade histologic lesions is relevant because these lesions are the targets of surgical treatment.3 Thus far, the most promising biomarkers that improve the diagnostic accuracy in histology have been those linked to E6/E7 overexpression.4 Specifically, several studies have tested the cross-sectional accuracy of p16,5-7 a cyclin-dependent kinase inhibitor that is involved in the genesis of different tumors. In cervical cancer, the overexpression of p16 is linked to the oncogenic transformation caused by persistent high-risk (HR) HPV infection. Testing for HR-HPV cannot distinguish between the transient infections that make up the majority of cases and transforming infections. However, p16 overexpression is correlated with the start of the oncogenic transformation process, which is necessary for the development of cervical cancer and its precursor lesions. Histology is usually considered the gold standard, and it is always challenging to measure the sensitivity and specificity of a test to improve the gold standard itself. Moreover, in cervical cancer precursors, it is impossible to measure the validity of the histologic diagnosis through clinical follow-up of the lesions because most of the lesions targeted for surgical
367
Am J Clin Pathol 2014;141:367-373 DOI: 10.1309/AJCPCYWVL61SVKFU
367 367
Pacchiarotti et al / Impact of p16INK4a on Interobserver Agreement
treatment are actually regressive, and recurrence after treatment is a rare event.8 As an indirect measure of diagnostic accuracy, some authors have proposed9,10 using the interobserver Cohen κ11 if the κ values of concordance between two expert observers on the same p16-assessed samples are higher than those assessed with H&E. This would imply that the diagnostic criteria are less subjective and more reproducible and thus that the final diagnosis resulting from these criteria is probably more accurate. The usefulness of p16 in the differential diagnosis of cervical intraepithelial neoplasia (CIN) when resolving uncertain morphologies is well recognized.6,12-14 However, it is not clear when it should be used, and only recently has there been an effort in systematically studying the cases in which p16 staining is recommended.15 Furthermore, to our knowledge, no population-based studies have been conducted on the use of p16 as a routine auxiliary method for all histologic specimens in a cervical cancer screening program. The objective of this study was to evaluate whether p16 in histology is a valid biomarker in improving the accuracy of diagnosis. Because of the lack of a gold standard, we aimed to determine whether the diagnostic concordance of different pathologists increases when using p16 and whether the classification of doubtful cases decreases.
Materials and Methods Population and Study Design The study was conducted in the province of Latina in central Italy, with a population of 550,000 in 2010; 160,000 women were in the target age group of the screening program (25-64 years). The province-wide organized screening program, which actively invites all the target population for Pap tests every 3 years, has been in place since 1999. The population base of the study consisted of 23,258 women who underwent a Pap test for a new screening round between May 2008 and September 2010; all the biopsy samples collected during the first colposcopy following the Pap test were considered for the inclusion criteria. Women in posttreatment or postcolposcopy follow-up care were excluded. All women with positive Pap tests with a diagnosis of atypical squamous cells of undetermined significance (ASC-US) or more severe according to the 2001 Bethesda System were invited to undergo a colposcopy. Almost all biopsy specimens (590/595) were stained with routine methods and with p16. All biopsy specimens were analyzed as part of the screening in Latina using H&E plus p16 staining. Slides were rendered anonymous by independent operators and reassessed after a 12-month period of 368 368
Am J Clin Pathol 2014;141:367-373 DOI: 10.1309/AJCPCYWVL61SVKFU
washout on the basis of H&E staining only by the pathologist from Latina, who made the first diagnosis. The 590 cases were then randomly assigned to six experts in cervicovaginal pathology. These pathologists analyzed the slides with H&E and, after at least 3 months of washout, reanalyzed the cases using both H&E and p16 slides. Immunohistochemistry Tissue blocks were cut, stained, and interpreted at the Lega Italiana per la Lotta contro i Tumori pathology laboratory in Latina. Serial slices of the same punch biopsy specimen were alternately placed on two slides, one of which was processed using routine procedures for H&E staining and the other for immunostaining for p16INK4a using the CINtec Histology Kit (Roche MTM Laboratories, Heidelberg, Germany) according to the manufacturer’s instructions: 10 minutes in a water bath at 95°C to 99°C to retrieve the antigen; 5-minute wash in a peroxidase-blocking reagent; 30-minute incubation and wash with the p16INK4a antibody (clone E6H4); 30-minute wash in a secondary antibody reagent (polymer-based goatantimouse antibody fragment conjugated to horseradish peroxidase); and finally a chromogenic visualization step using the 3,3′-diaminobenzidine chromogen. Slides were then counterstained with hematoxylin and coverslipped. Microscopic evaluation was performed according to the CIN classification (CIN 1, CIN 2, and CIN 3). Data Analysis and Statistical Methods A gold standard is not available for histology because histology itself is the final and most accurate diagnosis we can obtain for most cervical lesions. Without a gold standard, however, it is not possible to directly measure test validity. As an indirect measure of diagnostic accuracy, some authors propose using the Cohen κ: if the agreement between two different pathologists on the same sample is higher using p16 than with H&E alone, this means that diagnoses are more reproducible and, probably, more accurate. A contingency table and the Cohen κ have been calculated to analyze the concordance between observers and to determine whether p16 improves the diagnostic reproducibility. We present the unweighted κ statistics for a five-category classification (normal, including metaplasia; CIN 1; CIN 2; CIN 3, including adenocarcinoma in situ; and invasive cancer) and a binary classification (≥CIN 2 and ≤CIN 1); 95% confidence intervals of κ statistics were calculated with a bootstrap method (Stata 11.0, StataCorp, College Station, TX).16 The intraobserver concordance between H&E and H&E plus p16 was analyzed to identify which cases were reclassified and to verify whether p16 staining reduces inconclusive diagnoses (inadequate and CIN not otherwise specified [NOS]) and low reproducible diagnoses (CIN 2).9,17,18 © American Society for Clinical Pathology
AJCP / Original Article
The study was evaluated by the Direzione Sanitaria of the Latina Local Health Unit, which asked for a slightly modified informed consent for colposcopy-guided biopsies explaining the review process and, given the analogy with routine quality control procedures, waived the study from ethical review board submission.
Results Of the 590 biopsies performed, 53 were excluded because it was not possible to recover the slides for external review, 90 were excluded because one of the external reviewers decided not to participate in the study after slide transmission, and six were excluded because it was impossible to link the slide to the case. Thus, the final data set consisted of 441 women with a mean age of 37 years and a range of 25 to 64 years.
These women underwent colposcopy because their Pap test findings were positive. In more than 80% of the cases, a low-grade squamous intraepithelial lesion (L-SIL) or ASCUS was observed ❚Table 1❚. The overall positive predictive value of colposcopy-guided biopsy was 18.7%, with 8.8% for biopsies in women referred for ASC-US and L-SIL and 68.2% for those referred for abnormal squamous cell suggesting high grade and high-grade squamous intraepithelial lesion. Cohen κ comparison between the local pathologist and the external reviewers for the diagnoses made with only H&E staining is 0.517 and 0.753 for the five- and two-category classifications, respectively ❚Table 2❚ and ❚Table 3❚. The κ values for diagnoses made using H&E plus p16 immunostaining are extremely similar: 0.483 and 0.700 for the five- and two-category classifications, respectively (Tables 2 and Table 3). The results do not show differences between the reviewers ❚Table 4❚.
❚Table 1❚ Distribution of the Cytologic and Histologic Diagnoses According to the Original Diagnosis of the Reviewed Cases Histology Cytology
Negative
CIN 1
CIN 2
CIN 3
Invasive Cancer Inadequate
ASC-US 82 13 8 2 1 L-SIL 170 66 17 4 0 AGC 4 0 1 0 0 ASC-H 6 0 2 1 0 H-SIL 17 4 25 18 1 Carcinoma 0 0 0 2 0 Adenocarcinoma 1 0 0 0 1 Inadequate 2 0 0 0 0 Total 282 83 53 27 3 % 62.8 18.5 11.8 6.0 0.7
0 0 0 0 1 0 0 0 1 0.2
Total No. (%)
% PPV for CIN 2+
106 (23.6) 10.4 257 (57.2) 8.2 5 (1.1) 20.0 9 (2.0) 33.3 66 (14.7) 68.2 2 (0.4) 100.0 2 (0.4) 50.0 2 (0.4) 0.0 449 (100) 18.7 100
AGC, abnormal glandular cells; ASC-H, abnormal squamous cell suggesting high grade; ASC-US, atypical squamous cells of undetermined significance; CIN, cervical intraepithelial neoplasia; CIN 2+, cervical intraepithelial neoplasia grade 2 or more severe; H-SIL, high-grade squamous intraepithelial lesion; L-SIL, low-grade squamous intraepithelial lesion; PPV, positive predictive value.
❚Table 2❚ Contingency Table of Interobserver Agreement on Diagnosis Made With H&E and H&E + p16 Staining Based on a Five-Category Classification Diagnosis Made by External Pathologists Diagnosis Made by Local Pathologist
Negative
CIN 1
CIN 2
CIN 3
Invasive Cancer
H&E staining Negative 220 38 1 1 0 CIN 1 16 48 13 3 0 CIN 2 4 8 17 20 1 CIN 3 2 0 9 13 1 Invasive cancer 0 0 0 0 3 Total 242 94 40 37 5 H&E + p16 staining Negative 193 30 1 0 0 CIN 1 37 58 22 2 0 CIN 2 2 7 13 17 0 CIN 3 2 3 7 20 3 Invasive cancer 0 0 0 0 1 Total 234 98 43 39 4
Total 260 80 50 25 3 418 224 119 39 35 1 418
CIN, cervical intraepithelial neoplasia.
© American Society for Clinical Pathology
369
Am J Clin Pathol 2014;141:367-373 DOI: 10.1309/AJCPCYWVL61SVKFU
369 369
Pacchiarotti et al / Impact of p16INK4a on Interobserver Agreement
❚Table 3❚ Contingency Table of Interobserver Agreement on Diagnosis Made With H&E and H&E + p16 Staining Based on a Binary Classification
Diagnosis Made by External Pathologists
Diagnosis Made by Local Pathologist CIN 1–
CIN 2+
H&E staining CIN 1– 322 18 CIN 2+ 14 64 Total 336 82 H&E + p16 staining CIN 1– 318 25 CIN 2+ 14 61 Total 332 86
Total 340 78 418 343 75 418
reclassified as negative (among these there could be atypical immature metaplasia), 13 as CIN 1, and 16 as CIN 3. With p16, two cases of CIN NOS changed to negative, but one negative case and two CIN 1 cases changed to CIN NOS. Finally, if diagnoses are grouped into two classes (≤CIN 1 and ≥CIN 2), the two different methods lead to the same conclusion regarding the necessity to treat in 95% of the cases: with p16, 19 cases are reclassified as more severe and 18 as less severe. The intraobserver κ of the two readings was 0.71 and 0.86 for the five- and two-category classifications, respectively ❚Table 6❚ and ❚Table 7❚, with individual values ranging from 0.53 to 0.95 and from 0.69 to 1, respectively ❚Table 8❚.
CIN, cervical intraepithelial neoplasia; CIN 1–, cervical intraepithelial neoplasia grade 1 or less severe; CIN 2+, cervical intraepithelial neoplasia grade 2 or more severe.
The contingency table of the diagnoses made using H&E and H&E plus p16 by the same pathologist illustrates that p16 staining does not reduce inconclusive diagnoses (inadequate and CIN NOS) and low reproducible diagnoses (CIN 2) ❚Table 5❚. Of the 94 cases classified as CIN 2 with H&E, 61 maintained this classification, four were
Discussion In this population-based study, we did not find any difference in concordance among pathologists when diagnoses were made using H&E alone or H&E plus p16. Furthermore, we did not observe any reduction in inconclusive diagnoses (inadequate and CIN NOS) and low reproducible diagnoses (CIN 2).
❚Table 4❚ Interobserver Agreement and Cohen k Values by Each External Reviewer vs Local Pathologist
Five Categories
H&E
κ
Two Categories
H&E + p16
95% Confidence Interval κ
Overall 0.517 0.453-0.584 Reviewer 1 0.599 0.454-0.735 Reviewer 2 0.593 0.447-0.738 Reviewer 3 0.413 0.258-0.590 Reviewer 4 0.489 0.371-0.626 Reviewer 5 0.450 0.284-0.601
H&E
95% Confidence Interval κ
0.483 0.420-0.557 0.495 0.346-0.635 0.418 0.277-0.563 0.488 0.339-0.652 0.538 0.395-0.668 0.455 0.299-0.611
H&E + p16
95% Confidence Interval κ
0.753 0.663-0.821 0.795 0.551-0.954 0.819 0.588-0.959 0.690 0.445-0.881 0.691 0.487-0.881 0.764 0.553-0.907
95% Confidence Interval
0.700 0.610-0.786 0.680 0.452-0.861 0.639 0.352-0.847 0.805 0.544-0.953 0.731 0.551-0.887 0.641 0.433-0.816
❚Table 5❚ Contingency Table of Intraobserver Agreement of H&E vs H&E + p16 Staining Based on Descriptive Classification H&E
H&E + p16 Inadequate Negative
Metaplasia CIN 1
Inadequate 14 1 0 Negative 7 407 4 Metaplasia 0 28 2 CIN 1 0 24 0 CIN 2 0 4 0 CIN 3 0 0 0 AIS 0 0 0 CIN NOS 0 2 0 SCC 0 0 0 AK 0 0 0 Total 21 466 6
CIN 2
CIN 3
AIS
CIN NOS SCC
AK
Total
0 0 0 0 0 0 0 15 62 6 3 0 1 0 0 490 2 1 0 0 0 0 0 33 141 7 2 0 2 0 0 176 13 61 16 0 0 0 0 94 3 8 50 1 0 0 0 62 0 0 0 0 0 0 0 0 0 1 0 0 1 0 0 4 0 0 3 0 0 4 1 8 0 0 0 0 0 0 0 0 221 84 74 1 4 4 1 882
AIS, adenocarcinoma in situ; AK, adenocarcinoma; CIN, cervical intraepithelial neoplasia; NOS, not otherwise specified; SCC, squamous cell carcinoma.
370 370
Am J Clin Pathol 2014;141:367-373 DOI: 10.1309/AJCPCYWVL61SVKFU
© American Society for Clinical Pathology
AJCP / Original Article
❚Table 6❚ Intraobserver Concordance Between H&E and H&E + p16 Diagnoses Based on Five-Category Classification
Diagnosis Based on H&E + p16 Staining
Diagnosis Based on H&E Staining Negative
CIN 1
CIN 2
CIN 3
Invasive Cancer
Total
Negative 431 61 7 3 0 502 CIN 1 24 141 7 2 0 174 CIN 2 3 12 60 15 0 90 CIN 3 0 3 8 51 0 62 Invasive cancer 0 0 0 3 5 8 Total 458 217 82 74 5 836 CIN, cervical intraepithelial neoplasia.
This conclusion is in contrast with those of previous studies.10,19-25 Bergeron and colleagues19 noted a significant 13% increase in sensitivity for CIN 2+ when p16INK4a-immunostained slides were added and interpreted together with the H&E slides; they also found that the κ values significantly improved from 0.566 to 0.749. Gurrola-Díaz and colleagues21 observed an increase in κ values among three pathologists when using a binary classification from 0.36, 0.12, and 0.18 for each pair of observers using H&E to 0.59, 0.70, and 0.59 using p16INK4a stains. Dijkstra and colleagues20 found a similar increase in interobserver agreement among three pathologists from a weighted κ ranging from 0.44 to 0.66 in H&E slides to 0.80 with H&E plus p16. Horn and colleagues22 found an increase in κ values for punch biopsies—from 0.49 to 0.64—when H&E-stained slides were read together with p16INK4a-stained slides. Finally, Sari Aslani et al23 found that p16 increased the interobserver agreement as well as consensus. Our results are in line with the recent College of American Pathologists and American Society for Colposcopy and Cervical Pathology working group recommendations “against the use of p16 IHC as a routine adjunct to histological assessment of biopsy specimens with morphologic interpretations of negative, -IN1, and -IN3” (wherein IN can be interpreted as CIN for our purposes).15 The difference with previous studies could be the result of two factors: (1) κ in this study is high in the H&E analysis
❚Table 7❚ Intraobserver Concordance Between H&E and H&E + p16 Diagnoses Based on Two-Category Classification H&E
H&E + p16 CIN 1–
CIN 2+
Total
CIN 1– 657 19 676 CIN 2+ 18 142 160 Total 675 161 836 CIN 1–, cervical intraepithelial neoplasia grade 1 or less severe; CIN 2+, cervical intraepithelial neoplasia grade 2 or more severe
as well, and (2) the study population had a low frequency of cases with uncertain interpretation. In fact, the κ for the binary classification was higher than 0.7, which is considered as having very good accordance and standing in the upper part of the distribution of published studies. Dalla Palma and colleagues9 found an overall κ value of 0.65 for a binary classification of CIN 2+, Stoler and Schiffman26 found a value of 0.68, and Carreon and colleagues18 found a 0.70 κ value between two reviewers from the same center and values from 0.42 to 0.46 between reviewers from different centers. Only Malpica and colleagues27 and Crum and colleagues28 reported κ values higher than 0.75. This study is population based; therefore, it considers all women screened and thus makes it possible to foresee the real impact that using p16 could have. On the other hand, this
❚Table 8❚ Intraobserver Concordance Between H&E and H&E + p16 Diagnosesa
Five Categories
Two Categories
κ
95% Confidence Interval
κ
95% Confidence Interval
Overall Local pathologist Reviewer 1 Reviewer 2 Reviewer 3 Reviewer 4 Reviewer 5
0.705 0.694 0.626 0.527 0.952 0.828 0.615
0.663-0.745 0.627-0.743 0.483-0.784 0.346-0.675 0.871-1.000 0.712-0.924 0.462-0.744
0.857 0.849 0.920 0.694 1 0.910 0.777
0.806-0.899 0.773-0.910 0.762-1.000 0.451-0.888
a
0.791-1.000 0.603-0.929
Cohen κ values shown are those obtained by each pathologist for five- and two-category classifications.
© American Society for Clinical Pathology
371
Am J Clin Pathol 2014;141:367-373 DOI: 10.1309/AJCPCYWVL61SVKFU
371 371
Pacchiarotti et al / Impact of p16INK4a on Interobserver Agreement
approach makes it difficult to observe rare morphologies for which p16 immunostaining could be more useful. In fact, p16 has been demonstrated to be useful in differential diagnosis of some ambiguous morphologies (ie, glands with immature atypical squamous metaplasia,12-14 which in H&E analysis can be erroneously interpreted as CIN 2). In the current study, four cases were classified as CIN 2 using H&E which, after p16 analysis, were found to be negative, including the atypical immature metaplasia.6 The main limitation of our study was that the pathologists did not interpret the slides under the same circumstances: the local pathologist made the interpretation with H&E plus p16 under the psychological pressure of a clinical diagnosis and used H&E only as a quality control and for study purposes. The external reviewers, on the other hand, performed both interpretations for study purposes, without any psychological pressure and in the opposite time order: first H&E and then H&E plus p16. Although the reviewers came from centers throughout Italy, most had participated in several quality control and concordance studies, thereby increasing the baseline agreement in all conditions. Finally, this population-based study included about 1 year’s workload of a medium-sized screening program without any selection. Unfortunately, we can only report the results from five of the six reviewers who performed the external review of the slides because the data of the sixth reviewer were lost, and it was impossible to re-evaluate the slides. However, because the slides were assigned to the reviewers according to a random sampling, the final estimates were not biased by the incomplete nature of the data. Although the usefulness of p16 in resolving uncertain morphologies is well known, the use of p16 on all cervical biopsy specimens in a screening program showed virtually no effect on reproducibility of the histologic diagnosis. The study was supported by the Lega Italiana per la Lotta contro i Tumori, Sezione Provinciale di Latina, and the Latina Local Health Authority (ASL). Address reprint requests to Dr Giorgi Rossi: Servizio Interaziendale di Epidemiologia, AUSL Reggio Emilia, Via Amendola 2, 42122, Reggio Emilia, Italy; [email protected]. Dr Rossi is the principal investigator of a trial funded by the Italian Ministry of Health for which reagents were obtained at a reduced price or for free from Roche, Hologic-GenProbe, and Qiagen.
References 1. IARC Working Group on the Evaluation of Cancer Preventive Strategies. Cervix Cancer Screening: IARC Handbooks of Cancer Prevention No.10. Lyon, France: IARC Press; 2005. 2. Cuzick J, Clavel C, Petry CH, et al. Overview of the European and North American studies on HPV testing in primary cervical screening. Int J Cancer. 2006;119:1095-1101. 372 372
Am J Clin Pathol 2014;141:367-373 DOI: 10.1309/AJCPCYWVL61SVKFU
3. Arbyn M, Anttila A, Jordan J, et al, eds. European Guidelines for Quality Assurance on Cervical Cancer Screening. 2nd ed. Luxembourg, Luxembourg: Office for Official Publications of the European Communities; 2008. 4. Ghittoni R, Accardi R, Hasan U, et al. The biological properties of E6 and E7 oncoproteins from human papillomaviruses. Virus Genes. 2010;40:1-13. 5. Liang J, Mittal KR, Wei JJ, et al. Utility of p16INK4a, CEA, Ki67, P53 and ER/PR in the differential diagnosis of benign, premalignant, and malignant glandular lesions of the uterine cervix and their relationship with Silverberg scoring system for endocervical glandular lesions. Int J Gynecol Pathol. 2007;26:71-75. 6. Negri G, Egarter-Vigl E, Kasal A, et al. p16INK4a is a useful marker for the diagnosis of adenocarcinoma of the cervix uteri and its precursors: an immunohistochemical study with immunocytochemical correlations. Am J Surg Pathol. 2003;27:187-193. 7. Tringler B, Gup CJ, Singh M, et al. Evaluation of p16INK4a and pRb expression in cervical squamous and glandular neoplasia. Hum Pathol. 2004;35:689-696. 8. Schiffman M, Castle PE, Jeronimo J, et al. Human papilloma virus and cervical cancer. Lancet. 2007;370:890-907. 9. Dalla Palma P, Giorgi Rossi P, Collina G, et al. The reproducibility of CIN diagnoses among different pathologists: data from histology reviews from a multicenter randomized study. Am J Clin Pathol. 2009;132:125-132. 10. Klaes R, Benner A, Friedrich T, et al. p16INK4a immunohistochemistry improves interobserver agreement in the diagnosis of cervical intraepithelial neoplasia. Am J Surg Pathol. 2002;26:1389-1399. 11. Cohen J. A coefficient of agreement for nominal scales. Educ Psychol Meas. 1960;20:213-220. 12. Kong CS, Balzer BL, Troxell ML, et al. p16INK4a immunohistochemistry is superior to HPV in situ hybridization for the detection of high-risk HPV in atypical squamous metaplasia. Am J Surg Pathol. 2007;31:33-43. 13. Pinto AP, Schlecht NF, Woo TY, et al. Biomarker (ProEx C, p16(INK4A), and MiB-1) distinction of high-grade squamous intraepithelial lesion from its mimics. Mod Pathol. 2008;21:1067-1074. 14. Regauer S, Reich O. CK17 and p16 expression patterns distinguish (atypical) immature squamous metaplasia from high-grade cervical intraepithelial neoplasia (CIN III). Histopathology. 2007;50:629-635. 15. Darragh TM, Colgan TJ, Thomas Cox J, et al. The Lower Anogenital Squamous Terminology Standardization Project for HPV-associated lesions: background and consensus recommendations from the College of American Pathologists and the American Society for Colposcopy and Cervical Pathology. Int J Gynecol Pathol. 2013;32:76-115. 16. StataCorp. Stata Statistical Software: Release 11. College Station, TX: StataCorp LP; 2009. 17. Cai B, Ronnett B, Stoler M, et al. Longitudinal evaluation of interobserver and intraobserver agreement of cervical intraepithelial neoplasia diagnosis among an experienced panel of gynecologic pathologists. Am J Surg Pathol. 2007;31:1854-1860. 18. Carreon JD, Sherman ME, Guillén D, et al. CIN2 is a much less reproducible and less valid diagnosis than CIN3: results from a histological review of population-based cervical samples. Int J Gynecol Pathol. 2007;26:441-446.
© American Society for Clinical Pathology
AJCP / Original Article
19. Bergeron C, Ordi J, Schmidt D, et al. Conjunctive p16INK4a testing significantly increases accuracy in diagnosing highgrade cervical intraepithelial neoplasia. Am J Clin Pathol. 2010;133:395-406. 20. Dijkstra MG, Heideman DA, de Roy SC, et al. p16INK4a immunostaining as an alternative to histology review for reliable grading of cervical intraepithelial lesions. J Clin Pathol. 2010;63:972-977. 21. Gurrola-Díaz CM, Suárez-Rincón AE, Vázquez-Camacho G, et al. P16INK4a immunohistochemistry improves the reproducibility of the histological diagnosis of cervical intraepithelial neoplasia in cone biopsies. Gynecol Oncol. 2008;111:120-124. 22. Horn LC, Reichert A, Oster A, et al. Immunostaining for p16INK4a used as a conjunctive tool improves interobserver agreement of the histologic diagnosis of cervical intraepithelial neoplasia. Am J Surg Pathol. 2008;32:502-512. 23. Sari Aslani F, Safaei A, Pourjabali M, et al. Evaluation of Ki67, p16 and CK17 markers in differentiating cervical intraepithelial neoplasia and benign lesions. Iran J Med Sci. 2013;38:15-21. 24. Vinyuvat S, Karalak A, Suthipintawong C, et al. Interobserver reproducibility in determining p16 overexpression in cervical lesions: use of a combined scoring method. Asian Pac J Cancer Prev. 2008;9:653-657.
© American Society for Clinical Pathology
25. Zhang Q, Kuhn L, Denny LA, et al. Impact of utilizing p16INK4a immunohistochemistry on estimated performance of three cervical cancer screening tests. Int J Cancer. 2007;120:351-356. 26. Stoler MH, Schiffman M, for the Atypical Squamous Cells of Undetermined Significance–Low-grade Squamous Intraepithelial Lesion Triage Study (ALTS) Group. Interobserver reproducibility of cervical cytologic and histologic interpretations: realistic estimates from the ASCUS-LSIL Triage Study. JAMA. 2001;285:1500-1505. 27. Malpica A, Matisic JP, Niekirk DV, et al. Kappa statistics to measure interrater and intrarater agreement for 1790 cervical biopsy specimens among twelve pathologists: qualitative histopathologic analysis and methodologic issues. Gynecol Oncol. 2005;99(suppl 1):S38-S52. 28. Crum CP, Beach KJ, Hedley ML, et al. Dynamics of human papillomavirus infection between biopsy and excision of cervical intraepithelial neoplasia: results from the ZYC101a protocol. J Infect Dis. 2004;189:1348-1354.
373
Am J Clin Pathol 2014;141:367-373 DOI: 10.1309/AJCPCYWVL61SVKFU
373 373
