Beruflich Dokumente
Kultur Dokumente
Key Words: Cervical neoplasia; Histology; Colposcopy biopsy; Sensitivity; Specificity; Positive predictive value
DOI: 10.1309/EWYGWFRRM8798U5P
Materials and Methods diagnoses were identical (perfect agreement). (2) The diag-
noses differed, but all were CIN 1 or less severe. (3) The diag-
The NTCC study enrolled about 95,000 women who noses differed, and all were CIN 2 or more severe. (4) One
were screened according to various schedules. Protocols have diagnosis was CIN 2 or worse, whereas the others were CIN 1
been detailed elsewhere.12,13 Briefly, in the first phase (about or better, or vice versa (major disagreement, to review further).
45,000 women), women in the control arm had a convention- (5) At least one diagnosis was carcinoma (to review further).
al Pap smear, and women in the experimental arm had a liq- Categories 4 and 5 were reviewed and discussed by all
uid-based sample taken and tested for cytologic features and pathologists involved, using a multiheaded microscope.
HPV. Women older than 35 years with atypical squamous Again, all available material was examined, and only 1 diag-
cells of undetermined significance (ASCUS)+ or HPV+ nosis per woman, corresponding to the worst pathologic find-
results were directly referred for colposcopy, whereas women ing, was formulated. When possible, the diagnosis was agreed
younger than 35 years with ASCUS+ results were directly on by consensus. Otherwise the majority diagnosis was cho-
referred for colposcopy independent of the HPV result and sen.
recalled after 1 year for cytologic examination and/or HPV
control if the HPV result was positive and the cytologic find- Statistical Analysis
ings negative. These latter results are not included in the pres- We considered original diagnoses of CIN 2 or worse
ent analysis. In the second phase (50,000 women), women in given by the pathology clinic after colposcopy-guided histol-
the control arm were still tested with conventional cytologic ogy but before treatment. We computed the proportion that
examination, whereas women in the experimental arm were not confirmed by the reviewers diagnosis as an estimate
received only the HPV test and positive results were referred of the probability of unneeded treatment. Indeed, the decision
directly for colposcopy, independent of age. to refer for treatment is based on colposcopy-guided histolog-
Colposcopic and histologic diagnoses were made without ic findings. We used as the gold standard the reviewers final
masking cytologic and/or HPV results. Histologic diagnoses diagnosis, based on all available material as described earlier,
were given by 10 pathology services. because we considered it to represent the best available assess-
The histologic samples diagnosed as CIN 2 or 3 or CIN 1 ment of whether a CIN 2 or worse lesion was present. The
were blindly reviewed. We retrieved all available histologic estimated proportion corresponds to 1 (the PPV of the pre-
samples taken within 1 year from referral for colposcopy: treatment histologic diagnosis).
punch biopsies, in some cases from more than 1 colposcopy, We also computed the sensitivity and specificity of the
and the surgical specimens when an ablative treatment was original pretreatment diagnosis vs the gold standard. In that
performed. Each case was randomly assigned to 1 or 2 pathol- analysis, biopsy specimens showing no CIN, which were not
ogists for review. Only H&E-stained slides were used. The reviewed, were considered true-negatives. We calculated 95%
only data available to reviewers were patient age and date the confidence intervals (CIs) for all proportions assuming a bino-
sample was obtained. Each reviewer was asked to formulate a mial distribution.
single diagnosis for each patient, corresponding to the worst All of these parameters were estimated separately accord-
pathologic finding among all the patients tests, according to ing to the cytologic and HPV results that preceded colposcopy
the following 5 categories: (1) negative or inflammation, (2) to evaluate the association between the PPV of the diagnostic
CIN 1 encompassing HPV condyloma, (3) CIN 2, (4) CIN 3, category and the probability of a false-positive histologic
or (5) carcinoma (encompassing squamous carcinoma and result.
adenocarcinoma).
In the second phase, 2 independent pathologists random-
ly selected from a pool of 9 (1 per center, excluding the cen-
Results
ter that made the first diagnosis) reviewed all cases. In phase
1, all diagnoses of CIN 2 or 3 and 42% of CIN 1 diagnoses There were 52 CIN cases unavailable for review, and
were reviewed. During phase 1 the remaining CIN 1 cases 1,128 cases were reviewed: 747 had an initial diagnosis of
were reviewed by only 1 pathologist who was randomly CIN 1, 349 were CIN 2 or worse, and 20 were CIN not other-
selected from the 3 most expert. This was done to reduce the wise specified colposcopy-guided biopsy diagnoses. The lat-
workload and seems not to have influenced the results. Indeed, ter 20 were considered CIN 1 or better in the analysis. We
the proportion of CIN 1 upgraded to CIN 2 or worse did not excluded 12 cases from this analysis because the colposcopy-
differ by method (P = .98). guided biopsy diagnosis was unclear or not available.
The diagnoses provided by the reviewers were compared Table 1 shows the distribution of the colposcopy-guid-
with the diagnoses originally given by the screening center. ed biopsy histologic diagnoses according to the final diagno-
The following concordance-categories were defined: (1) All sis after revision. The sensitivity of the pretreatment original
Table 1
Original Pretreatment Histologic Diagnosis by Reviewed Diagnosis Based on Pretreatment and Posttreatment Specimens
Unavailable 0 1 5 3 3 0 12
CIN NOS 1 7 7 5 0 0 20
CIN 1 0 44 651 48 4 0 747
CIN 2 0 6 42 107 44 1 200
CIN 3 0 0 4 32 112 1 149
Total 1 58 709 195 163 2 1,128
reason for the colposcopy referral, ie, ASCUS, low-grade CIN, cervical intraepithelial neoplasia.
* Positive predictive value, 85.1%; sensitivity, 83.9%.
squamous intraepithelial lesion (LSIL)+, HPV+ and negative
cytologic results or HPV+ and positive cytologic results
Table 3 and Figure 1. Sensitivity ranged between 75% and
approximately 90% (except for the HPV group that included conventional arm (P = .02) for women with ASCUS cytolog-
only 3 cases that were CIN 2 or worse), and there was no evi- ic findings (26.7%; 95% CI, 12%-46%) than for women who
dence of significant variation between groups. had LSIL or more severe cytologic results (8.6%; 95% CI,
On the other hand, the proportion of false-positive diag- 4%-16%). In the experimental arm, it was higher (P = .003)
noses of CIN 2 or worse varied according to cytologic and for HPV but cytologically abnormal (ASCUS+) cases
HPV test results (Table 3 and Figure 1). It was higher in the (66.7%; 95% CI, 22%-96%) than for HPV+ cases (10.6%;
Table 3
Number of Biopsies and CIN 2 or Worse Diagnoses After Revision, Sensitivity, Specificity, and Probability of Unnecessary
Treatment According to the HPV and Cytologic Results Used to Refer Women for Colposcopy
No. of No. of Cases No. of CIN 2 or Sensitivity, % Specificity, % Probability of False CIN 2 or
HPV/Cytologic Results Cases Reviewed Worse After Review (95% CI) (95% CI) Worse Diagnosis, % (95% CI)
No HPV test
Overall 694 335 122 87.7 (74-88) 97.1 (95-98) 15.0 (8-20)
ASCUS 302 133 29 75.9 (56-90) 97.0 (94-99) 26.7 (12-45)
LSIL+ 392 202 93 91.4 (84-97) 97.2 (95-99) 8.6 (4-16)
HPV+
No test 787 371 122 87.7 (80-93) 96.9 (95-98) 15.7 (10-23)
Overall HPV and cytology 635 331 112 82.3 (74-89) 97.9 (96-99) 10.6 (5-18)
Normal cytology 255 93 20 75.0 (51-91) 99.0 (97-100) 11.8 (1-36)
ASCUS 123 60 23 87.0 (66-97) 99.0 (94-100) 4.8 (0-24)
LSIL+ 257 178 69 85.3 (74-93) 95.7 (92-98) 12.1 (5-22)
HPV
Overall 319 81 3 66.7 () 98.7 (98-100) 66.7 (22-96)
ASCUS 182 43 1 100.0 () 99.4 (97-100) 50.0 ()
LSIL+ 137 38 2 50.0 () 97.8 (94-100) 80.0 ()
ASCUS, atypical squamous cells of undetermined significance; CI, confidence interval; CIN, cervical intraepithelial neoplasia; HPV, human papillomavirus; LSIL, low-grade
squamous intraepithelial lesion.
Cytology
Biopsy Biopsy
302 392
No review Reviewed No review Reviewed
169 (2 CIN 2) 133 188 (14 CIN 2) 204
None
After Review
After Review
CIN 1 CIN 2+ Total CIN 1 CIN 2+ Total
CIN 1 96 8 104 CIN 1 101 8 109
CIN 2+ 7 22 29 CIN 2+ 8 85 93
Total 103 30 133 Total 109 93 202
Positive
After Review
After Review
After Review
CIN 1 CIN 2+ Total CIN 1 CIN 2+ Total CIN 1 CIN 2+ Total CIN 1 CIN 2+ Total
CIN 1 70 2 72 CIN 1 36 1 37 CIN 1 101 8 109 CIN 1 229 20 249
CIN 2+ 5 15 20 CIN 2+ 3 20 23 CIN 2+ 10 58 68 CIN 2+ 15 107 122
Total 75 17 92 Total 39 21 80 Total 111 66 177 Total 244 127 371
Biopsy Biopsy
182 137
No review Reviewed No review Reviewed
Negative
After Review
Figure 1 Flow chart of the women enrolled in the trial according to the results of the first-level screening test (cytology in
columns and human papillomavirus [HPV] in rows) and results of histologic review. For the original diagnosis, we considered only
the colposcopy-guided biopsy specimen, whereas the review also included the postsurgical histologic findings, when present.
ASCUS, atypical squamous cells of undetermined significance; CIN, cervical intraepithelial neoplasia; CIN 1, CIN or better; CIN
2+, CIN 2 or worse; LSIL, low-grade squamous intraepithelial lesion.
95% CI, 5%-18%), without significant variation or trend probability that a histologic diagnosis of CIN 2 or worse from
according to the severity of cytologic findings. Finally, the a colposcopy-guided biopsy specimen was false: the r2 for lin-
lesions found in HPV+ women without a Pap test, examined ear correlation between log(PPV) and log(proportion false-
during the second phase of the trial, had a probability of being positive histologic diagnoses) was 0.66.
falsely classified as CIN 2 or worse of 15.7% (95% CI, 10%-
23%). This proportion was not statistically different from that
observed overall among women with conventional cytologic
Discussion
diagnoses (P = .4) or among HPV+ women recruited during
phase 1 who had simultaneous cytologic examination and The main goal of cervical screening is to identify women
HPV testing (P = .2). with high-grade intraepithelial lesions (CIN 2 or more
The highest probability of a false-positive histologic diag- severe). These women require treatment, whereas women
nosis of CIN 2 or worse (66.7%) was observed in the group with low-grade lesions (CIN 1 or less severe), which are fre-
with abnormal cytologic findings (ASCUS or worse) but neg- quently regressive, should be referred for cytologic or colpo-
ative HPV results that also had the lowest PPV among all scopic control.14
screening test combinations (0.4%). There was a statistically In our study, 7.4% of women with CIN 1 or less severe pre-
significant (P = .015) inverse correlation between the PPV of treatment histologic findings were judged to have CIN 2 or
the screening test combination that led to referral and the worse. The clinical consequences of this upgrade were minimal
because every woman with a lesion was monitored and the worse missed by colposcopy is not negligible.5,17 Because
lesion controlled with strict follow-up. biopsy specimens from women with negative diagnoses were
On the other hand, about 15% of cases diagnosed as CIN not reviewed, we assumed they were all negative for CIN 2 or
2 or worse before treatment were downgraded to CIN 1 or worse. It is possible that a small number of true cases of CIN
less. This proportion changed according to the combination of 2 or worse are present in this group. This would result in a
screening test results that preceded the colposcopy. This phe- slight overestimate of sensitivity and a negligible overestimate
nomenon is the consequence of the fact that sensitivity and of specificity.16
specificity were constant in these groups, whereas the preva- Remarkably, sensitivity and specificity were stable over
lence of CIN 2 or worse among women who had a biopsy var- the different groups, which suggests that knowing why the
ied. Because the prevalence of CIN 2 or worse among women women were referred for colposcopy had little influence on
with a biopsy is correlated to the prevalence in the entire the interpretation of histologic findings. The similar values of
group, the PPV of histologic diagnoses was correlated with specificity and sensitivity among groups with such a different
the PPV of the initial screening test combination. prevalence of lesions give consistency to our results.
This observation has important clinical implications. Our results also stress the need for quality assurance in
Histologic diagnoses determine the decision to treat. cervical histologic diagnosis. Indeed, even if the diagnosis of
Therefore, the proportion of cases of CIN 2 or worse down- CIN is believed to be simple and reproducible by most pathol-
graded after review can be assumed to represent the probabil- ogists, in the historic study by Siegler,18 10 of 20 cases of car-
ity of a treatment for CIN 2 or worse to be unnecessary. Our cinoma in situ (actually CIN 3 or high-grade SIL [HSIL])
data show that this probability can be relevant in some groups were missed by a group of 25 pathologists. More recent work
of women if they are directly referred for colposcopy, such as found higher agreement: Robertson and collaborators10 found
women with negative HPV results but abnormal cytologic a Cohen (the statistic that measures agreement not due to
results, especially ASCUS, which supports the use of HPV for chance19) of 0.46 for HSIL vs less severe than HSIL; Stoler et
cytologic triage. In general, our results stress the need for ini- al5 found a of 0.69; and Park et al7 and Parker et al4 found a
tial tests with a high PPV. In screening, a low PPV of the diag- of about 0.8. Indeed biopsy diagnoses are always considered
nostic tests that directly refer to colposcopy results not only in the gold standard in any study of diagnostic accuracy, but the
increased workload and costs for colposcopy units, but also subjective interpretation of histologic classification is a big
unnecessary treatments. obstacle in many interinstitutional studies.5
In the study by Hopman and colleagues,15 20% of low- In screening, a low PPV of the diagnostic test that direct-
grade lesions were upgraded and 26% of high-grade lesions ly precedes colposcopy not only has consequences on the col-
were downgraded after review. In the analysis by Stoler et al,5 poscopy workload and costs but also leads to unnecessary
reviewers reclassified 27% of CIN 1 cases to CIN 2 or worse treatment. This consideration is true independent of the type
and 24% of CIN 2 or worse cases to CIN 1. of first-level test, cytologic examination or HPV testing.
Our high negative predictive value is the result of a low
prevalence of true CIN 2 or worse in our series. This also led From the 1Pathology Unit, S. Chiara Hospital of Trento, Trento,
to a relatively low PPV despite high specificity, 98% overall. Italy; 2Agency for Public Health, Lazio Region, Rome; 3Pathology
Section, University of Bologna, Bologna; 4Department of Human
It must be kept in mind that this last figure estimates the speci- Pathology and Oncology, University of Firenze, Firenze;
ficity of pretreatment biopsy and, therefore, represents only 5Pathology Unit, S. Anna Hospital of Torino, Torino; 6Department
women who actually had a biopsy. The specificity of the entire of Pathology, University of Verona, Verona; 7Section of
second level diagnostic phase, including colposcopy and Cytopathology, Hospital of Padova, Padova; 8Pathology Unit,
biopsy, is probably even higher because many women did not Hospital of Imola, Bologna; 9Pathology Unit, Maggiore Hospital
of Bologna, Bologna; 10Pathology Unit, Hospital of Faenza,
have a biopsy and most of them plausibly represent true-neg-
Ravenna; 11Pathology Unit, Belcolle Hospital of Viterbo,
ative results for CIN 2 or worse. Viterbo; and 12Centro per la Prevenzione Oncologica, Torino.
Because we compared pretreatment diagnosis with
The NTCC trial was financially supported by the European
reviewed posttreatment diagnoses, false-negatives were
Union (Europe against Cancer contracts SI.2.327046 and
indicative not only of errors of local pathologists in interpret- SPC.2002475), by the Italian Ministry of Health (Progetto
ing histologic features but also of the different specimens Speciale Valutazione di nuove tecnologie per lo screening del
available before and after treatment. The true sensitivity of the cervicocarcinoma and project ex L 138/2004), Regione Piemonte,
whole diagnostic process of colposcopy plus biopsy is even Torino, Regione Toscana, Firenze, Regione Veneto, Venezia,
lower because biopsies were not done for all women, and Regione Emilia-Romagna, Bologna, Provincia Autonoma di
Trento, Agenzia di Sanit Pubblica, Regione Lazio, by the Special
some of the biopsy specimens may not have been taken from Project Oncology, Compagnia di S. Paolo FIRMS, Torino.
the most severe lesion.16 This results in an overestimate of Address reprint requests to Dr Giorgi Rossi: Agency for Public
sensitivity. It has been shown that the portion of CIN 2 or Health, Lazio Region, via di S. Costanza 53, 00198 Rome, Italy.
* The following are contributing members of the NTCC 6. Grenko RT, Abendroth CS, Frauenhoffer EE, et al. Variance
Pathology Group: G.L. Taddei, F. Castiglione, and A.M. in the interpretation of cervical biopsy specimens obtained
Buccolero, Unit of Pathology, University, Florence; P. Dalla for atypical squamous cells of undetermined significance.
Palma and D. Aldovini, Unit of Pathology, Ospedale di Trento; B. Am J Clin Pathol. 2000;114:735-740.
Ghiringhello and S. Privitera, Unit of Pathology, OIRM, S. Anna, 7. Park JJ, Genest DR, Sun D, et al. Atypical immature
Turin; G. Collina, Unit Section of Anatomic Pathology M. metaplastic-like proliferations of the cervix: diagnostic
Malpighi, Bellaria Hospital, Bologna University, Bologna; G.P. reproducibility and viral (HPV) correlates. Hum Pathol.
1999;30:1161-1165.
Casadei and P. Pierotti, Unit of Pathology, Ospedale Maggiore di
Bologna, Bologna; M. Aldi and C. Sintoni, Unit of Pathology, 8. de Vet HC, Knipschild PG, Schouten HJ, et al. Interobserver
variation in histopathological grading of cervical dysplasia.
Presidio Ospedaliero di Ravenna, Ravenna; G. Galanti, Unit of
J Clin Epidemiol. 1990;43:1395-1398.
Pathology, Presidio Ospedaliero di Imola, Bologna; V. Gomes and
G. Verrico, U.O.C. di Anatomia e Istologia Patologica, Ospedale 9. Ismail SM, Colclough AB, Dinnen JS, et al. Observer
variation in histopathological diagnosis and grading of cervical
Belcolle, Viterbo; M. Lestani, Unit of Pathology, University of
intraepithelial neoplasia. BMJ. 1989;298:707-710.
Verona; E. Bragantini, Unit of Pathology, Ospedale Borgo Trento,
Verona; and G.L. Onnis and D. Minucci, Unit of Pathology, 10. Robertson AJ, Anderson JM, Beck JS, et al. Observer
variability in histopathological reporting of cervical biopsy
University of Padova. specimens. J Clin Pathol. 1989;42:231-238.
The following are contributors to the article: P. Dalla Palma
11. Kalof AN, Cooper K. Our approach to squamous
was the coordinator of the NTCC Pathologist Working Group and
intraepithelial lesions of the uterine cervix. J Clin Pathol.
designed and conducted the review; P. Dalla Palma, G. Collina, 2007;60:449-455.
A.M. Buccoliero, B. Ghiringhello, M. Lestani, G. Onnis, D.
12. Ronco G, Giorgi-Rossi P, Carozzi F, et al. Human
Aldovini, G. Galanti, G. Casadei, M. Aldi, and V. Gomes reviewed
papillomavirus testing and liquid-based cytology in primary
the histologic specimens; G. Ronco was the NTCC project leader screening among younger women: results at recruitment
and designed the study. P. Giorgi Rossi with P. Dalla Palma and G. from the NTCC randomised controlled trial. Lancet Oncol.
Ronco drafted the manuscript. P. Giorgi Rossi, G. Ronco, and P. 2006;7:547-555.
Giubilato were responsible for data analysis. All authors critically 13. Ronco G, Segnan N, Giorgi Rossi P, et al, for the New
revised the manuscript. All members of the NTCC Pathologist Technologies for Cervical Cancer Working Group. Human
Working Group reviewed the histologic specimens. papillomavirus testing and liquid-based cytology: results at
Acknowledgments: We thank all staff who assisted in running recruitment from the New Technologies for Cervical Cancer
the study, Margaret Becker for editing the text, and the thousands randomized controlled trial. J Natl Cancer Inst. 2006;98:765-
of women who participated in the study. 774.
14. Advisory Committee on Cancer Prevention.
Recommendation on cancer screening in the European
References Union. Eur J Cancer. 2000;36:1473-1478.
15. Hopman EH, Kenemans P, Helmerhorst TJ. Positive
1. Joste NE, Crum CP, Cibas ES. Cytologic/histologic correlation predictive rate of colposcopic examination of the cervix uteri:
for quality control in cervicovaginal cytology; experience with an overview of literature. Obstet Gynecol Surv. 1998;53:97-
1,582 paired cases. Am J Clin Pathol. 1995;103:32-34. 106.
2. Tritz DM, Weeks JA, Spires SE, et al. Etiologies for non- 16. Adams AL, Eltoum I, Roberson J, et al. Negative colposcopic
correlating cervical cytologies and biopsies. Am J Clin Pathol. biopsy after positive human papillomavirus (HPV) DNA
1995;103:594-597. testing: false-positive HPV results or false-negative histologic
3. Malpica A, Matisic JP, Van Niekirk D, et al. Kappa statistics to findings? Am J Clin Pathol. 2006;125:413-418.
measure interrater and intrarater agreement for 1790 cervical 17. Pretorius RG, Peterson P, Azizi F, et al. Subsequent risk and
biopsy specimens among twelve pathologists: qualitative presentation of cervical intraepithelial neoplasia (CIN) 3 or
histopathologic analysis and methodology issues. Gynecol cancer after a colposcopic diagnosis of CIN 1 or less. Am J
Oncol. 2005;99(3 suppl 1):S38-S52. Obstet Gynecol. 2006;195:1260-1265.
4. Parker MF, Zahn CM, Vogel KM, et al. Discrepancy in the 18. Siegler EE. Microdiagnosis of carcinoma in situ of the uterine
interpretation of cervical histology by gynaecologic cervix. a comparative study of pathologists diagnoses. Cancer.
pathologists. Obstet Gynecol. 2002;100:277-280. 1956;9:463-469.
5. Stoler MH, Schiffman M, and the Atypical Squamous Cells 19. Cohen J. A coefficient of agreement for nominal scales.
of Undetermined SignificanceLow-grade Squamous Educ Psychol Measurement. 1960;20:37-46.
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.