The Swedish cervical screening program is changing from cytology to HPV based screening in 2017. To document the performance of the screening program and to provide a basis for evaluating the effect of changes in the program we performed a nationwide audit of the cervical cancer cases from 2002 to 2011. Our aim with this analysis is to compare different sources of information for the tumor histological classification, and to determine how robust this classification is for subsequent analyses of data.
Data on all 4254 cases of cervical cancer or unspecified uterine cancer diagnosed between 2002 and 2011 was identified from the Swedish National Cancer Registry. Tumor histopathology is provided as a SNOMED classification code by the local pathologist and the clinician at the initial diagnosis. For the Audit an experienced gynecologist reviewed all medical records to identify cervical cancer cases with primary, invasive, epithelial tumors of cervical origin and to extract relevant data on e.g. treatment, mode of detection and histological classification of the tumor. Additionally we collected diagnostic slides and tissue blocks: an external review of the tumor was performed by a senior pathologist and HPV-genotyping was done on the tissue material. The Swedish National Cervical Screening Registry (NKCx) provided data on screening for all cases and matched population controls.
One of 26 different SNOMED codes was reported for all except 6 cases. The systematic external review on 86% of the cases classified the tumors into following groups: Squamous Cell Carcinoma, Adenocarcinoma, Adenosquamous carcinoma and rare histological types (i.e. neuroendocrine, small cell, undifferentiated cancer). The medical record provides information on histology in various levels of detail and in a non-systematic form, so we could only extract this information for 69% of cases, in the same categories as above. There is a high concordance (87%) between all three sources of histology classification. For 9.8% of cases with discrepancies in the histology classification or missing data we decided for a final classification based on: 1) external histopathology review of the sample (in cases of good quality samples), 2) medical record and 3) initial SNOMED diagnosis (this was changed only in 7.5% of cases) for subsequent analyses of data.
Thorough nationwide ascertainment from multiple sources found limited ambiguity regarding tumor histology. For most uses, any one of the sources can be used, as the concordance between different sources of data is high.