The validation of cervical cytology. Sensitivity, specificity and predictive values.

Statistics and methods for the validation of the results of cytologic screening for cervical cancer and its precursors were examined. Many of the methods commonly used, including the calculation of sensitivity and specificity on raw data, contain flaws that undermine their conclusions. Using a large computerized database of 748,871 cytologic screenings of 277,842 women over a ten-year period, the value of screening was examined. Only subsequent histologic examinations within one year were accepted to validate positive initial cytologic diagnoses; only two subsequent cytologic screenings within the next three years were accepted to validate negative initial cytologic diagnoses that had not been followed by a histologic examination. Cases not meeting these criteria were excluded from the initial analysis. From these data, the predictive value of a negative cytologic examination was determined to be 99.8%; the predictive value of a positive cytologic examination was 73.4% for an initial diagnosis of mild-to-moderate dysplasia, 90.6% for a diagnosis of severe dysplasia/carcinoma in situ, 94.5% for a diagnosis of carcinoma in situ or microinvasive carcinoma and 95.5% for an initial diagnosis of invasive carcinoma. Cases with an initial "questionable" cytologic diagnosis had a positive predictive value of only 64.0%. Extrapolation from the validated cases to the entire screened population showed an overall sensitivity of 80% and a specificity of 99.4% for cytologic screening for cervical cancer. The sensitivity was slightly lower for mild and moderate dysplasia (78.1%) and slightly higher for carcinoma in situ and severe dysplasia (81.4%) and invasive carcinoma (82.3%).

[To what extent have early cancer detection and preventive examination for cancer of the uterine cervix fulfilled established expectations?]. / Wieweit haben die Krebsfrüherkennungs- und Vorsorgeuntersuchungen für das Karzinom der Cervix uteri die an sie gestellten Erwartungen erfüllt?

The results of cytological screening examinations of women are reported. The study was performed over a ten-year period and comprised 748,871 examinations of 281,705 women. Many long-term observations are among them. 28,171 women (10%) had more than five cytological repeat findings in this period. Basing on the presented dates the following aspects speak in favour of the effectivity of the screening examinations: 1. The decrease of cytological rates of discovery between first examination and repeated examination, 2. the new increase of rates of discovery in case of longer intervals between the examinations, 3. the shifting of the relation between dysplasias, carcinomata in situ and invasive carcinomas in favour of the less severe pathological changes. The rate of newly discovered cervical carcinomas decreased between 1971 and 1980 from 0.9% to 0.2%, the rate of newly discovered carcinomata in situ and severe dysplasias decreased from 3.5% to 1.2%. For future planning of screening programmes several problems have to be taken into consideration--for example, the low attendance rate of women at regular screening examinations, the high rate of false-negative cytological findings and questions of efficiency of programmes in repeatedly examined groups of population.

Comparison of Papanicolaou staining and Feulgen staining for automated prescreening.

Feulgen staining is considered to be a quantitative DNA-specific cytochemical procedure. The applicability of this staining in high-resolution cytometry was tested in comparison with a regressive Papanicolaou staining. Papanicolaou-stained or Feulgen-stained intermediate and carcinoma cells selected by a cytologist were examined with a Zeiss scanning microscope photometer at 546 and 560 nm, respectively. After cell image segmentation and feature extraction, a statistical data evaluation was carried out by computer. Cell distributions with respect to four selected nuclear features demonstrated the influence of the staining procedure on cell feature measurements. The discriminatory power of the classification system as related to both staining procedures was studied using discriminant analysis. Using only nuclear features, a 7.3% improvement of the overall correct classification rate (from 85.0% to 92.3%) was achieved using Feulgen staining. The misclassification rate was simultaneously reduced by 50%. Using cytoplasmic as well as nuclear features, a 98% rate of correct classification was achieved.

[Automated cytodiagnosis from the viewpoint in the clinical cytologist]. / Automatisierte Zytodiagnostik aus der Sicht des klinischen Zytologen.

Automated prescreening means to replace visual prescreening by cytotechnologists by a machine. The aim is to reach a better diagnostic level by eliminating human failure. All machine positives whether true positive or false positives have to be examined visually (by cytotechnologists or cytopathologists). If there are many false positives the system will be highly uneconomic. From a clinical point of view the number of false negatives is crucial. The false negative rate may in no case be higher than that reached by cytotechnologists and medical cytologists, but should, if possible be lower. Yardstick for the validation of prescreening systems can only be the diagnostic level of well kept cytologic laboratories of conventional cytology. Systems which do not meet these requirements don't have the least chance of being applied in routine diagnostic cytology.

[Necessity for and construction of cytological data banks]. / Notwendigkeit und Aufbau von zytologischen Datenbanken.

Analytical and quantitative cytology is based upon photometric and morphometric measurements of single cells in stained cytological specimens. It analyzes the features of single cells and cell populations with respect to their cytodiagnostic relevance. One way to do this is the establishment of learning sets of visually selected cells, and the intercomparison of unknown cells with these data sets. Difficulties arise with the visual classification of single cells from PAP-stained gynecological smears and the pooling into classes due to the considerable photometric variation from specimen to specimen. It is demonstrated, that neither for a cancer prescreening apparatus nor an interactive diagnostic machine, highly differentiated learning sets can be dispensed with. The build-up of the TUDAB learning sets and results of single cell classification, visual reclassification, as well as population analysis and specimen classification are shown.

[Early detection of endometrial carcinoma]. / Möglichkeiten der Früherkennung des Endometriumkarzinoms.

Carcinoma of the endometrium has not only seen a relative increase compared to carcinoma of the cervix in the last few decades but also an absolute increase. The only method to improve the cyto-morphological detection of endometrial carcinoma is to obtain the cytological material from the site of the disease which is the uterine cavity. For this purpose a large number of devices were developed within the past 10 years. These permit an approximately 90% detection rate by endometrial cytology. It may be difficult to introduce instruments into the uterine cavity of older women with a narrow cervical canal. Even well experienced cytologists have to relearn the evaluation of cyto-smears from the uterine cavity. The recognition of highly differentiated carcinomas and atypical hyperplasias of the endometrium is especially difficult. Screening examinations of women over age 45 by endometrial cytology from the uterine cavity area apparently capable of detecting a fairly large number of asymptomatic endometrial carcinomas. How far the intra-uterine cytology is capable of detecting pre-cancerous conditions of the endometrium such as atypical glandular hyperplasia is still not known.

Results of cytologic mass screening in the Federal Republic of Germany.

Since July 1971, the national screening program for cancer has provided all women 30 years of age and older in the Federal Republic of Germany with the opportunity of undergoing a medical checkup for uterine and mammary cancer once a year. Within this program, more than 27 million cytologic screenings of cervical smears were carried out by the end of 1976. At the Cytologic Institute of the Bavarian Cancer Society, Munich, 503,870 cytologic screenings in 226,428 women were carried out from 1971 to 1976, and the rate of positive and suspicious cytologic findings declined from 2.6% to 0.9% in that period. Whereas the detection rate of cytologic findings suggestive of invasive carcinoma, carcinoma in situ and severe dysplasia was 0.5% on the first screening, it was 0.16% on the second screening and 0.08% on the third screening. A total of 272 invasive carcinomas of the cervix and 948 carcinomas in situ and severe dysplasias was detected by the examinations and verified histologically. According to the number of women screened per annum, an incidence peak for carcinoma in situ (including severe dysplasias) was observed between 25 and 29 years of age, with a second, lower peak at about 65 years. The curve of invasive carcinoma of the cervix, however, rose continuously to old age. From 1963-66 to 1976, the rate of histologically verified invasive carcinoma of the cervix declined continuously, from 0.36% to 0.03%. At first, the rate of carcinoma in situ and severe dysplasia increased slightly, from 0.44% to 0.5%, but since 1971 it also decreased markedly, reaching 0.16% in 1976. Although similar curves are found in foreign studies, morbidity and mortality from cervical carcinoma has decreased relatively slowly within the total population. The possible reasons for this finding are discussed.

Numerical analysis of cell populations in smear and monolayer preparations from the uterine cervix. I. The proportions of isolated, abnormal epithelial cells in slides from one applicator.

Automated screening systems with a receiver operating characteristic (R.O.C.) demand a minimum number of isolated and diagnostically relevant cells to classify cytologic specimens. Since routine Papanicolaou smears have not proven suitable for this, a new preparation procedure based on sedimentation velocity separation has been developed. A population analysis was made to assess the diagnostic relevance of the new procedure. For each of 21 histologically confirmed cases, three types of slides were made from material obtained by a single cotton-tipped applicator: (1) a routine smear and slides prepared from the (2) supernatant fraction and (3) sedimented fraction of the remaining material after suspension and separation. Results showed that only one-third of the squamous epithelial cells obtained in each case were found on the routine smear. Of those, only 39% were isolated cells, as compared with 81% of the cells in the residual (suspended) material. Further, in cases of slight and moderate dysplasia as well as carcinoma in situ, an enrichment of the number of pathologic cells was noted when comparing the sedimented fractions to the original suspensions. This would seem to prove the utility of the new preparation procedure as the basis for automated analysis based on samples instead of on single cells.

Computer recognition of ectocervical cells: image features.

Machine recognition of ectocervical cells has been achieved with good classification success by a number of research groups. Even though a substantial number of cell image features have been introduced, only a moderate number are required for classification. For them, definitions and measures of their discriminatory potential are presented.

[The effectiveness of cervical cytology screening for cervical cancer (author's transl)]. / Wie effektiv sind zytologische Krebsvorsorgeuntersuchungen bei der Bekämpfung des Zervixkarzinoms?

Cervical cytology screening was performed, between July 1971 and the end of 1976, on 503,870 specimens from 226,428 women. During this period the rate of positive or suspicious findings fell from 2.6% to 0.9%. The number of specimens judged suspicious for invasive carcinoma, carcinoma in situ or with severe dysplasia (among those women who had undergone serial examinations) was 5 per thousand for the first specimen, 2 per thousand for a subsequent second specimen, and 1.6 per thousand for a subsequent third specimen. The screening programme detected (later confirmed histologically) 272 cases of invasive carcinoma and 948 of carcinoma in situ or severe dysplasia. Related to the number of women screended at various ages, the frequency of carcinoma in situ and severe dysplasia had a peak between the 25th and 29th year of life, with a second, lower peak at around the 65th year. For invasive cervical carcinoma the frequency remained high into old age. The proportion of histologically proven cases of invasive carcinoma fell continuously from 1963/66, at 3.6 per thousand, until 1976, at 1.6 per thousand.

Components and results of a new preparation technique for automated analysis of cervical samples.

Components and evaluations of a new preparative procedure for automated high-resolution analysis of cervical samples are presented. This procedure is based on sedimentation velocity separation of samples with subsequent fractionation of the separation column and centrifugal deposition of suspended cells on coated glass slides. A system for specimen collection and mailing of suspended samples is described. A new type of glass slide designed for automated analysis is presented, and centrifugal buckets for cell deposition on a 6-sq-cm area are described. Experimental results with different kinds of coating substances for glass slides as well as different isopyknic media are discussed, and data for differential cell counts are graphically demonstrated. Looking at the diagnostic accuracy and economic feasibility of this system, the authors realize that preparations have to be evaluated quantitatively and that constraints of sample size and processing time have to be taken into consideration for further developments.

Comparison of two Papanicolaou staining procedures for automated prescreening.

A "regressive" (procedure I) and a "progressive" (procedure II) modification of the Papanicolaou stain were tested for their suitability for automated screening procedures. Intermediate squamous cells and carcinoma cells were scanned with a high-resolution system, and digitized data were statistically analyzed after feature extraction. The regressive modification accentuates the textural features of intermediate squamous cell nuclei as well as the contrast ratio of nucleus and cytoplasm. The progressive modification accentuates some criteria of malignancy, such as polychromasia, hyperchromasia and shifting of nuclear to cytoplasmic ratio. Grouping of data according to patients indicates that the differences found are due to the staining procedures and not to individual variability of parameters in different patients.

Progress report of the TUDAB project for automated cancer cell detection.

Two methods for high resolution cell image data acquisition are applied routinely. Cells are either scanned by a computer controlled fast scanning microscope photometer (SMP) or a TV-camera. The software system for digital image analysis was completely revised and implemented on the PR 330 minicomputer. The system contains codes for primary cell data acquisition, segmentation of cells, cell feature extraction and statistical cell analysis. With this system, SMP and TV scanned cell data bases of PAP stained cells in vaginal smears, grouped into several classes, have been built up. Each data base contains 34 primary features and 20 feature combinations for each cell. A linear discriminant analysis is applied routinely for cell classification. The present state of the system and its operation are described, cell features and classification results are shown, and future steps for a prescreening strategy are discussed.

Sedimentation velocity separation: a preparation method for cervical samples.

A preparation procedure, aiming at monolayer deposition of cervical exfoliative material on glass slides for high resolution prescreening has been developed. The main features of this procedure are centrifugal deposition after suspension and sedimentation of samples over isopycnic medium of 1.026 density. Fractioning of the separation column after centrifugation at 50 X g yields two preparations with leukocytes, bacteria and cellular debris predominantly located on the first slide and epithelial cells on the second one. The degree of spatial cellular isolation as well as the amount of diagnostically relevant cells per slide seem to fit the requirements of automated high resolution analysis.

Cytomorphologic results of preparation experiments for monolayer deposition of cervical material.

For automated prescreening methods by high resolution analysis serving as a detecting method in gynecologic mass screening programs a new monolayer deposition method of cervical material has been used. This method will be outlines briefly and the mode of evaluation as well as current cytomorphological findings will be presented. With regard to measurability of the slides prepared according to the new method a number of cytologic criteria were thought to be of particular importance. These criteria are delineated and compiled in a table. A form in which these criteria were listed was filled in by cytopathologists for each slide evaluated. When performing isolation and centrifugation procedures several new morphologic questions arose to the cytopathologist which can only partly be answered by now. If taking into account all criteria of evaluation it may be followed from the present experiences that slides of cervical material are much more suited for automated prescreening methods by high resolution analysis if prepared after isolation and centrifugation in macromolecular liquids than are conventional Papanicolaou smears or slides from suspensions with isolated cells that were not subjected to centrifugation procedures.