Proposed Sheffield quantitative criteria in cervical cytology to assist the diagnosis and grading of squamous intra‐epithelial lesions, as some Bethesda system definitions require amendment

Abstract

This study assesses the accuracy of published quantitative and qualitative criteria in the Bethesda System (TBS) for squamous intra-epithelial lesions. Quantitative image analysis was undertaken on illustrations from TBS publications and also from slides in Cytology Training Centre teaching sets. Comparisons were also made with the British Society for Clinical Cytology (BSCC) terminology in cervical cytology, using the illustrations in their terminology publication and amalgamating the results into their proposed new two-tier model. TBS quantitatively defines low-grade squamous intra-epithelial lesions (LSIL) in both conventional and liquid-based cytology (LBC) preparations as showing nuclear enlargement more than x3 the area of a normal intermediate squamous cell nucleus. This study found that the increase in mean nuclear area was limited to only x2 in conventional preparations. In LBC (SurePath preparations, there was only a statistically non-significant x1.2 increase. This study identified a progressive and statistically significant reduction in mean cytoplasmic area from normal intermediate cells to LSIL and then to high-grade squamous intra-epithelial lesions (HSIL) in both conventional and LBC preparations. Furthermore, the most consistent quantitative finding in both conventional and LBC preparations was a statistically significant increase in the mean area and diameter ratios from normal intermediate cells to LSIL and then to HSIL. In all instances this varied from x2 to just below x3. This is in agreement with TBS, which states that the cytoplasmic area in HSIL is decreased leading to a marked increase in nuclear to cytoplasmic (NC) ratio. With the exception of an increase in mean nuclear area in conventional preparations from normal intermediate cells to LSIL, the predominant cause for this increase in NC ratios was a reduction in mean cytoplasmic area. The numerical increase in NC ratio for LSIL identified in this study was greater than implied by the 'slightly increased' statement in TBS. TBS comments that some HSIL cells can have the same degree of nuclear enlargement as in LSIL and that other HSIL cells may have much smaller nuclei than in LSIL. Both of these qualitative comments were supported in this study. The mean diameter NC ratios of 33% and 50% could provide useful diagnostic assistance in the distinction of normal intermediate cells and LSIL and between LSIL and HSIL, respectively. Because of overlapping individual ranges, however, additional diagnostic features such as nuclear morphology must be used in the distinction of normal intermediate cells, LSIL and HSIL. No statistical difference was identified in the mean diameter NC ratios between ASC-US and LSIL in TBS publications. In addition, the proposed new BSCC low and high grades of squamous abnormality were not statistically different from ASC-US/LSIL and HSIL, respectively. This provides support that the proposed BSCC two-tier system of squamous abnormalities is comparable to TBS. This study shows that LBC has variable but major and significant effects on nuclear and cytoplasmic morphology and that quantitative definitions in conventional preparations cannot be automatically extrapolated to LBC methodology. The study shows that some TBS quantitative and qualitative criteria require amendment and that an alternative quantitative approach, such as diameter NC ratio has a more valid scientific evidence base. Furthermore, use of NC ratios avoids the problems associated with the variable changes in nuclear and cytoplasmic areas, occurring between conventional and different commercial LBC preparations. By contrast, classifications based on area comparisons must be tailored to the specific conventional or commercial LBC preparation.