Background: Prostate cancer (PCa) with ISUP Grade Group (GG) 1 or 2 characteristics have favorable outcomes after radical prostatectomy treatment, yet some of them still progress and present with biochemical recurrence (BCR), and some of those (PSA with doubling time of 䁤 1 year) will be considered a high-risk BCR. This might be due to disease heterogeneity and improper cancer staging and grade classification in pre-operative settings. To date, the GG is the most reliable prognostic tool, and has the largest impact on treatment decisions, however the subjective nature of GG assessment and significant inter-observer variability, in particular amongst general pathologists, have therefore highlighted the need for a more robust and objective assessment to guide treatment recommendations. DNA ploidy is an objective prognostic tool, however it requires a large amount of PCa cells that are often not available from limited biopsy tissue. Technology advances allow us to extract numerous tissue features from a small diagnostic area of PCa tissue sections. We hypothesize that combination of tissue architecture features and nuclear morphometry features could improve the prediction of cancer progression in the form of BCR in comparison to GG assessment or DNA ploidy.
Methods: 115 patients from the General Hospital of Celje (GH Celje) with a low– and favorable intermediate risk PCa characteristics that undergone radical retropubic prostatectomy (RRP) between years 2003 and 2009 were included in the study. Pathological evaluation from general pathologists (GGG) was supplemented by an experienced uropathologist from Institute of Pathology in Ljubljana (EGG), who identified diagnostic area with the worst Gleason pattern to be used in the study. DNA ploidy was assessed on prostatectomy samples, which underwent tissue disaggregation, cell isolation and staining with a DNA stoichiometric stain. Using image cytometry, ploidy features were extracted and a Ploidy Score (PS) generated. Tissue features were extracted from thin biopsy tissue sections that were stained with thionin and scanned. Voronoi diagram-based algorithms were applied to diagnostic areas. A linear combination of the most discriminant architecture features generated a MSTA score. A nuclear morphometry features were extracted from perfectly segmented nuclei and LDO score was generated as a linear combination of the most discriminant nuclear features. Using a linear combination of MSTA and LDO score, a global tissue score – QTP score was generated.
Results: In a univariate regression model, the variables of QTP score, LDO score, MSTA score, EGG, PS score and cT were significant predictors of BCR (respective p values: 1,2 x 10–7, 0,004, 0,028, 0,0016, 0,0002, 0,016). In a multivariate regression model, MSTA score was significant predictor of BCR in model with GGG, however not with EGG. LDO score reached independent level in model with EGG (respective p values: 0,01 and 0,003). QTP score was best predictive variable of recurrence in any model, including EGG and cT (respective p values: 3,7 x 10–5, 0,02 and 0,06). Thus patients with a high QTP score were more likely to experience BCR or a high-risk BCR than patients with a low QTP score. Combining MSTA or LDO score with GG resulted in a significant stratification of risk of BCR. Survival analysis showed that QTP score is the best variable in predicting biochemical free survival (log-rank test: p = 4 x 10–12). LDO score or MSTA score were significant predictors of biochemical recurrence free survival, however not better in comparison to EGG or PS.
Conclusion: We have shown that tissue architecture and nuclear morphometry features extracted from PCa biopsy tissue could improve the prediction of cancer progression in comparison to GG or DNA ploidy. The new quantitative method is objective and has potential to identify aggressive PCa recurrences and could thereby introduce more personalized approach to guide treatment recommendations in population with favorable risk PCa.
|
