CYTYC'S THINPREP CERVICAL CYTOLOGY PREPARATION SYSTEM RECOMMENDED

CYTYC'S THINPREP CERVICAL CYTOLOGY PREPARATION SYSTEM RECOMMENDED for approval by FDA's hematology and pathology devices panel in a June 7 meeting held in Bethesda, Maryland. The five-member panel unanimously endorsed Cytyc's claim that "the instrument prepares slides with equivalent or improved representation of cervical neoplasia or its precursor lesions...compared to the conventional Pap smear." Four panel members voted in favor of approval without conditions, and one member voted for approval with conditions. In the ThinPrep processing system, cervical specimens are collected via a conventional scrape, but then are rinsed into a vial containing a preservative solution. The vial is sent to a laboratory and placed into the ThinPrep processor, where the cells are spun, filtering out the accompanying mucus, blood and other non-diagnostic material. The processor then extracts the diagnostic cells from the filter, dispersing them evenly on a slide in a 20 mm circle. Cytyc submitted a premarket approval application for ob/gyn use of the device in March 1992; FDA filed the PMA in January. The firm says it has since mad several amendments to the application. A version of the ThinPrep has been marketed for non-gynecologic applications since October 1990. Cytyc claims that the automated ThinPrep method significantly reduces false-negative readings. An unblinded study of 3,219 patients at five sites demonstrated a four-to-one reduction in false negatives "compared to conventionally prepared slides made from the same patient sample," the firm said in a June 8 release. Using a single sample, separate slides were made using the conventional an ThinPrep methods. Cytotechnologists evaluated individual slides on a blinded basis, placing each in one of five categories: negative, atypical, low-grade squamous intraepithelial lesions (SIL), high-grade SIL or carcinoma. Following the initial evaluation, senior cytotechnologists and pathologists analyzed the slides on an unblinded, "consensus review" basis, clarifying screening discrepancies between the matched pairs of slides. At two of the five sites, biopsies were performed on all patients designated as low-grade SIL or higher to confirm results. Final diagnosis concurred exactly for 88.3% of the samples. In 101 cases, the ThinPrep indicated a low-grade or more severe lesion while the conventional smear was negative or atypical. In 26 cases, the conventional smear showed a low-grade or more severe lesion, and the ThinPrep was negative or atypical. Investigator Thomas Bonfiglio, MD, University of Rochester, told the panel that "using standard criteria in an unblinded screening,...the ThinPrep processor slide found 13% fewer atypicals, 15% more low-grades and 8% more highgrades, leading to our conclusion that the ThinPrep reveals more disease than the conventional slide." At FDA's request, Cytyc conducted a second, blinded study. All slides from the unblinded study that had been designated as other than negative upon first screening were given, unpaired, to a cytopathologist who made a classification assignment for each. Results from the blinded study were less dramatic but still indicated higher Thin-Prep accuracy. At three of the five sites, ThinPrep slides showed a statistically significant higher detection of positives (defined as precancerous lesions or carcinoma) than the conventional method, at ratios of 59:42, 28:8, and 37:32. The ratio at a fourth site was 86:85. At one site, however, the conventional method detected more disease, at a ratio of 23:7. Commenting on the results, investigator David Wilber, MD, University of Rochester, suggested that "the blinded data less dramatically favors the ThinPrep...[as] a straightforward consequence of introducing inter- and intra-observer variability or error." FDA lead reviewer Linda Magruder asked the panel to address a number of agency concerns, particularly the company's handling of slides designated as "atypical." In its data analysis, Cytyc grouped atypical with negative results. Magruder noted the company's decision to reduce the five-part Bethesda classification system to two categories, "colposcopy assured" or "colposcopy not assured," for calculating data results. "As the atypical category is a complex issue, the data must be provided with the atypias grouped separately, so analyses can be done showing the results with the atypias combined with the negatives, with the abnormals and separately," she said. Panel member Timothy O'Leary, MD, PhD, Armed Forces Institute of Pathology Washington, D.C., echoed FDA's concern, stating "a lot of how you interpret the data seems to depend on whether you toss the ASCUS [atypical squamous cells of unidentified significance] group into the normal category or into the lesioned category." David Zanhiser, PhD, Cytyc's vice president of scientific affairs, defended the company's decision to combine the negative and atypical results into a single "colposcopy not assured" category, explaining that the methodology is "in accord with new National Cancer Institute practice guidelines being developed." According to Zanhiser, NCI notes that "followup of Pap smears without colposcopy is adequate for a diagnosis of ASCUS in which the diagnosis is not qualified further." Zanhiser continued that atypia, in his opinion, "is a category that really means when a pathologist is uncertain as to how this [smear] should be classified." FDA also expressed concern with the company's heavy reliance on the data from its unblinded review. In response, Zanhiser asserted that the two studies demonstrate different endpoints. The "original protocol was designed to look at the possibility of removing the effects of inter-intraobserver variability" in order to reveal "how well the disease is represented on the two slides," he stated. "Complementing this data is the blinded study which shows...what is the detection, rather than the representation of the disease." Panel member O'Leary concurred, stating that he "would agree that one should consider the blind data more strongly than the consensus data, but...you shouldn't ignore the consensus data." Magruder also highlighted the device's "less than optimal" adequacy score for representation of endocervical cells from the "tranformation zone" between the endocervical canal and the vagina. Panel Chairperson David Linder, MD, University of Nebraska Medical Center, Omaha, echoed this concern, asking the company: "Do you foresee a potential problem if substantial increases in numbers of Pap smears do not have the endocervical component?" Zanhiser pointed out that the lost endocervical component was "an artifact of splitting the sample" in the original clinical trials. He argued that the sample on the conventional slide contained the entire endocervical component, with "none remaining to go into the vial." He asserted, however, that in later trials in which the scrape went "direct-to-vial," the endocervical component was maintained. A final concern voiced by FDA was that only seven adenocarcinoma specimens had been included in the clinical trials. The panel, however, appeared satisfied with the company's results because 17 additional cases not contained in the original PMA have since been provided. Panel member Jerome Nosanchuk, MD, Tompkins Community Hospital, Ithaca, New York, stated: "It's a small number, but it seems to me that the detection of cancer cells is the easiest part of this, dysplasias are harder." Although the panel was unified on the approval recommendation, panel member Alice Smith, MD, University of Texas, Dallas, Health Center, voted for approval with conditions. She suggested that the company conduct trials targeting socio-economically disadvantaged minority populations in the southern United States. Cytyc also currently is developing a computer controlled Pap smear analyzer, the CDS-1000. The device scans and analyzes slides produced by ThinPrep and "then presents a ranked summary of specimen information on a computer screen" ("The Gray Sheet" Sept. 16, 1991, p. 16).