DNA-cytometric grading of prostate cancer systematic review with descriptive data analysis

Gleason-score <=6, assessed on core needle biopsies, is an essential prognostic parameter to offer the strategy of Active Surveillance (AS) to patients with locally confined cancers of the prostate. Yet, its interobserver reproducibility is low (48-70%) and its prognostic validity unsatisfactory. An option to complementary assess the malignant potential of these cancers are objective DNA-ploidy-measurements on existing biopsies. For that purpose chromosomal heterogeneity is indirectly quantified by DNA-cytometry resulting in DNA-grades of malignancy 1-4. This review systematically trawls and evaluates all scientific publications on the potential diagnostic and prognostic validity and the heterogeneity of DNA-ploidy measurements in cancers of the prostate between 1966 and 2013. Publications have been classified into Oxford levels of evidence and levels of significance were given for the correlation of DNA-ploidy with different clinical outcomes. 114 scientific articles had to be excluded because of different methodological reasons. All but one of the 67 methodologically acceptable articles report on a significant diagnostic resp. prognostic significance of DNA ploidy measurements in cancers of the prostate. 8 level 1b studies report that DNA-ploidy, assessed on punch biopsies independently predicts organ confinement as assessed after radical prostatectomy. 18 level 2b studies prove that DNA-ploidy measurements add statistically significant information to the Gleason-score. 16 level 2b investigations report a significant correlation of DNA-ploidy with recurrence-free survival. 15 level 2b studies document a significant correlation with overall survival after different types of therapy. 5 level 2b investigations prove a significant correlation with local recurrence or progress after radical prostatectomy. 3 level 2b publications show a significant correlation of DNAploidy with the occurrence of lymph nodeor bone metastases after radical prostatectomy. 1 level 2b study documents the additional prognostic value of DNA-ploidy measurements over conventional subjective grading in prostate cancer patients under AS. All existing 15 narrative reviews on selected articles dealing with prognostic DNA-cytometry in cancers of the prostate are in favor of this method. Prospective level 1b studies, especially those proving the validity of DNA ploidy measurements to predict non-progression in patients with clinically insignificant low-grade low stage cancers of the prostate eligible for Active Surveillance additionally to the Gleason-score are still missing.


Introduction Epidemiology
Mean age of patients facing the diagnosis of prostate cancer in Germany currently is 70 years. 26.1% of all newly diagnosed malignancies among men are cancers of the prostate. Its incidence has risen from 80 in 1993 to 111.4/100,000 men or 65,830 new cases in 2010. 70,100 new cases are prognosticated for 2014. Nevertheless mortality is constantly decreasing, from 30 in 1993 to 20.0/100,000 men in 2010 [1]. Even lethality is low: 11.7% in the USA in 2006 as compared to other cancers [2]. The favorable five-years survival rate of 93% is mainly due to more frequent early diagnoses as a consequence of PSA-testing [1]. doi: 10.7243/2052-7896- [2][3][4][5][6][7] As about 30% of patients who, according to inquest died from prostate cancer, in fact did not according to autopsy [3], the true mortality rates may be significantly lower.

Therapy
Adequate therapy of prostate cancers essentially depends on their individual histological type, stage and grade of malignancy. High grades are associated with early and rapid tumor progression and subsequent metastasis. Low grade and low stage cancers may either locally be treated with curative intention (e.g., by radical prostatectomy, external or internal radiation) or subjected to an Active Surveillance (AS) strategy. Different to "Watchful Waiting", this includes the option for curative therapy if the cancer progresses. About 53% of all newly diagnosed patients with cancers of the prostate in Germany are currently treated by radical prostatectomy, 8% hormonally, 6% by a combination of both, 12% by radiation, 14% by AS and 5% by Watchful Waiting [4].
As the probability of patients with "clinically insignificant microcarcinomas" of the prostate [5] to die from their cancer is very low: 89% overall survival after 8 years [6], 81% overall survival after 10 years [7], the strategy of AS has been designed. About 45% of all screening-detected cancers can be managed with AS [7]. In Germany this strategy is recommended to patients with low-grade Gleason-score(GS)<=6 and low stage (T1c and T2a) cancers, found in <=2 core biopsies with <50% of their volume and a PSA <10 ng/ml. It comprises regular urological examinations, repeated biopsies and PSA-controls but still allows curative therapy if clinical signs of progression can be detected [8].

Shortcomings of Gleason-grading
Grading the malignant potential of individual cancers of the prostate currently is performed according to the modified Gleason-score according to the International Society for Urologic Pathology (ISUP) [9] on histological sections of biopsies or resected tissue.
Grading the malignant potential of cancers should be reproducible among different pathologists, representative for the tumor as a whole and, most importantly, prognostically valid. Grading the malignancy of cancers of the prostate should predict outcome of patients even after different types of therapy. Neither the original [10] nor the revised GS [9] reveal sufficient inter-observer reproducibility to rely clinical decisions of the significance of radical prostatectomy vs. AS on this subjective prognostic index only. [11] report a reproducibility of 58-69%, [12] of 48%, [13] of 70% and [14] of 47% for the revised score.
The main cause for the revision of the Gleason-system by the International Society for Urologic Pathology (ISUP) was to enhance its representativity on punch biopsies for the tumor as a whole (as observed in radical prostatectomies). Yet, contrary to what was expected, [15] found an agreement of only 72%.
Two groups furthermore demonstrated that the revised Gleason-grading could neither differentiate the survival of score 7a-and 7b- [16] or GS <=6-and GS7-patients after radical prostatectomy [17].
Prognostic DNA-cytometry Cancers  It is based on measurements of the Integrated Optical Density (IOD) of stoechiometrically and specifically DNA-stained nuclei and internal calibration with normal, diploid reference cells. Measurements of nuclei under UV-light, previously stained with DNA-specific fluorescent dyes, like DAPI, in liquids flowing through a capillary are called "DNA Flow Cytometry". Its disadvantage is that the cells are lost after analysis, thus control measurements are not possible. Furthermore cancer cells cannot be differentiated from non-epithelial cells without additional immunocytochemical markers. Measurements on Feulgen-stained nuclei [35] on glass slides, using TV-imageanalysis systems are called "DNA Image Cytometry". It has the advantage that it can repeatedly be performed on prestained and specifically restained slides on individually preclassified cells. Its performance has been highly standardized by a task force of the European Society for Analytical Cellular Pathology, ESACP [30,[32][33][34]. For the purpose of grading the malignant potential of selected solid tumors, four grades of increasing malignancy have been agreed upon: peridiploid (grade 1), peritetraploid (grade 2), x-ploid (grade 3) and multiploid (grade 4) (Table 1 and Figure 1).
• "It is difficult to understand why these well documented data have not yet gained access to treatment protocols" [46].
• "DNA-ploidy is of value in treatment decisions, particularly when surveillance is an option". "DNA-ploidy should uniformly be studied in clinical trials, particularly in • "In retrospective studies … any sample shown to contain representative tumor can provide meaningful information" [48].
• "DNA-diploid tumors have a better prognosis than tu mors of a similar stage and grade that are non-diploid" [49].
• "Flow cytometry has much to tell us about the natural history and biologic behavior of prostate cancer" [50].
• "DNA-cytometry is a powerful tool for grading the malignant potential of prostatic carcinomas, superior to histological and cytological evaluation" [51]. As inclusion of patients with locally confined cancers of the prostate into the strategy of AS requires a valid prognostic assessment of individual tumors and the subjective Gleasonscore suffers from low inter-observer reproducibility and insufficient prognostic validity, more reliable prognostic biomarkers are required. So far no systematic review exists on the prognostic validity of DNA-ploidy measurements, that have to be considered to supplement the Gleason-score on identical specimens. This study provides the first systematic review on that subject.

Review Systematic review of the literature
A query has been performed in PubMed for publications between January 1966 [52] and August 19 th , 2013 with the following key words: "prostate cancer and (DNA-ploidy or DNA-aneuploidy or DNA-cytometry or DNA-image-cytometry)". Studies were classified into different levels of evidence according to their design, applying the criteria of the Oxford Center for Evidence Based Medicine [53]: • Level 1b, diagnosis: Validating cohort studies with good reference standards or clinical decision rule, tested within one clinical center.
• Level 2b, diagnosis: Exploratory cohort studies with good reference standard or clinical decision rule after derivation or validated on split samples or data bases.
• Level 1b, prognosis: Individual inception cohort studies with >80% follow-up or clinical decision rule, validated in a single population.
• Level 2b, prognosis: Retrospective cohort studies or follow-up of untreated control patients in a randomized controlled clinical trial. Derivation of a clinical decision rule or validated on split samples only.
• Level 3b, prognosis: Retrospective cohort studies with insufficiently defined inclusion criteria or less than 80% of follow-up. A. B. has performed the review. No reports were excluded because of their status of publication. A systematic assessment of publication bias had not been performed.
The following features were considered as "good reference standards": For the correlation with diagnosis, the results of histological examination of radical prostatectomies, especially concerning extracapsular spread and infiltration of seminal vesicles. For the correlation with prognosis, the recurrencefree-or overall survival time, the occurrence of lymph node-or bone metastases, clinical proof of local progression or recurrence or a so-called biochemical recurrence.
The diagnostic accuracy of specific indices of nuclear DNA distribution obtained on pretherapeutic biopsies, e.g., to render spread beyond the capsule more likely, should be compared with that of the Gleason-score in studies meeting the criteria of Oxford level of evidence 1b. Similarly the prognostic validity of indices of nuclear DNA-distribution should be investigated in comparison with the Gleason-Score, specific for different therapeutic settings, in Oxford level of evidence 1b studies.

Excluded papers
1.819 titles had been listed and 1 found through other sources. After exclusion of 40 duplicates and reading the respective abstracts, 1.573 records have been excluded and full texts of 207 publications that seemed to deal with the above mentioned subjects were ordered and reviewed (Figure 2). 114 of these have been excluded from further evaluation due to different types of methodological shortcomings .
• 25 dealt with methodological aspects of cytometry only .

Methodologically sufficient papers
Papers that did not reveal the above mentioned shortcomings were considered as "methodologically sufficient". 66 publications reported statistically significant correlations between various DNA-ploidy parameters and one of the above-mentioned patient-relevant endpoints. These comprised 15.693 patients (Tables 3-8): • 8 level 1b studies reported a significant correlation of DNAcytometric features with histologically proven cancer spread beyond the capsule as detected after radical prostatectomy [167][168][169][170][171][172][173][174]. 4 of them document a significant improvement of diagnostic accuracy concerning the prediction of organ confinement by DNA-ploidy features over Gleason-score alone ( Table 3).

Correlation of DNA-ploidy on biopsies with extracapsular spread (ECS) after radical prostatectomy (RPE). Bold p-values refer to Cox multivariate regression analysis.
an univariate analysis [180,182,183] Table 4. [193] found the same after external radiation in a multivariate analysis. 4 level 3b studies [187][188][189][190] proved a significant correlation of DNA-ploidy parameters with recurrence free survival time after radical prostatectomy on multivariate analyses ( Table 4).
• 2 level 3b studies [110,169] proved an independent correlation of DNA-ploidy parameters with overall survival time under AS apart from histological or cytological grading in a multivariate design ( Table 5). 1 level 2b study did the same multivariate for recurrence free survival time [197] ( Table 4).
• 3 level 2b [170,186,229] and 1 level 3b study [173] report on a significant correlation of DNA-ploidy parameters with the occurrence of lymph node-or bone metastases after radical prostatectomy. 2 level 3b studies report the same after hormonal therapy [224,235] Table 8.

Tumor heterogeneity
The following publications dealt with aspects of heterogeneity of DNA-ploidy patterns in cancers of the prostate and representativity of punch biopsy for the tumor as a whole.
• 122 simulated punch biopsies had been investigated from nine prostatectomies containing cancers of unknown stage (mean 12 samples). Five (56%) showed heterogeneity of the DNA pattern (diploid, tetraploid, aneuploid). All four cases having a homogenous DNA content were DNA diploid in all samples. In those cases with a heterogeneous pattern, the areas having abnormal DNA-patterns could not be predicted by histologic pattern or grade [228].
• These authors compared DNA-ploidy patterns (diploid vs. non-diploid) in punch biopsies and subsequent prostatectomy specimens in 12 cases with cancer. Four sections per resected cancer of unknown stage had been investigated. The concordance was to 92% [230].
• 123 DNA-histograms from 48 men with prostatectomy due to cancers of unknown stage (mean 2,6) had been compared with those of six preoperative biopsies (diploid, non-diploid). In 34 men (71%) DNA-ploidy in prostatectomies was correctly predicted as either diploid or nondiploid on biopsies. Under-estimation occurred mainly when only one or two biopsies were analyzed [232].

Conclusions
All twelve reviews on diagnostic or prognostic DNA-measurements in prostate cancer published so far are merely "narrative" and not systematic ones. Yet, they all conclude that DNA-ploidy is of either diagnostic or of prognostic value, without considering the methodologic quality of addressed papers.

Shortcomings of published papers
The most frequent cause for exclusion of papers (n=32) was an inadequate study design (not enough patients:<50), mixture of different therapies, lacking therapeutic, clinical or follow-up information, selection of patients. In 13 publications DNA-measurements were methodologically insufficient (inadequate sampling or calibration, measurements of sections of different thickness, paucity of cells). Correlation with nondiagnostic or prognostic features (n=25) and dealing with methodological aspects only (n=24) cannot be criticized. Many scientists did not obey existing respective international and interdisciplinary methodological consensus reports [30,32-34], especially concerning problematical types of specimens (sections), missing performance standards (<300 nuclei) and individual prognostic interpretation of data.
Algorithms for DNA-grading of prostate cancer [52] have been the first to propose an objective alternative for grading prostate cancer malignancy based on DNA-measurements in cancer cells. Our group has published on "DNA-grading of prostatic carcinoma: Prognostic validity and reproducibility" [233]. Up to 1998 no standardized, internationally agreed algorithms existed, on how to derive prognostically different groups from DNA-histograms of prostate cancers. Each author individually defined at least two, up to five different categories. The only common aspect was that they all comprised a DNA-diploid category as the prognostically most favorable one. In 1998 and 2001 the European Society of Analytical Cellular Pathology (ESACP) "Taskforce on Standardization of Diagnostic DNA-Image Cytometry" has published a detailed proposal how to derive four prognostically relevant groups, resp. grades of malignancy, from DNA-measurements of malignant tumors: peridiploid, peritetraploid, x-ploid and multiploid [33,34] (Figure 1 and Table 1). Unfortunately, not many authors have adopted the respective standardized algorithms since then. Thus, their results concerning the prognostic validity of DNA-grading the malignancy of prostate cancer are hardly comparable. Nevertheless, the main, clinically relevant differentiation refers to DNA-diploidy vs. DNA-non diploidy. During tumor progression, peridiploid cancers primarily increase their rate of proliferation [64,234]. Later on during tumor progression, additional peritetraploid clones evolve [235]. Thus, concerning diploidy vs. non-diploidy, it is not relevant which c-value the peridiploid peak exactly has, but if there is a second peak at 4c or elsewhere. According to [210,234] a prognostically relevant proliferation rate >5% can be stated in peridiploid DNA-histograms, supposed a reasonable number of >1000 of nuclei had been measured to obtain representative results [34].

Diagnostic accuracy
The fact that DNA-ploidy-parameters are able to nearly exclude cancer spread beyond the capsule as detected after radical prostatectomy significantly more precise than the Gleason-Score, has been proven in 8 level 1b studies [167][168][169][170][171][172][173][174], Table 3. Thus DNA-ploidy should additionally be taken into consideration, whenever organ confinement is a prerequisite for certain therapeutic strategies, like AS.

Prognostic validity
For untreated patients with early prostate cancer under Active Surveillance the following results have been published: • Documented for 120 untreated patients in a multivariate level 2b study the significant superior ability of DNA-ploidy over the histological WHO-grade to predict tumor-specific survival time [219].
• Proved in a multivariate level 1b-study with a statistically significant correlation of DNA-ploidy with recurrence-free survival time in 146 untreated patients in comparison with the cytological grade [197,236].
• Proved for 106 untreated patients in a multivariate level 2b study a statistically significant correlation of DNAploidy with overall survival time in comparison with the Gleason-score [169].
• Proved for 287 primary untreated patients in a multivariate level 2b study significant correlation with overall survival time in comparison with the cytological grade [235,236]. Brachytherapy is another standard treatment for organ confined prostate cancer. Patients that most likely reveal cancer spread beyond the capsule have to be excluded from this approach. Using core biopsy material, [223] could correctly predict the majority of failures and non-failures, while Gleason-score failed (Figure 3). DNA-diploid patients had a significantly lower rate of disease recurrence as compared with DNA-aneuploid patients. Thus, DNA-grading of prostate cancer malignancy can be used to further specify the inclusion criteria for brachytherapy.
The fact that DNA-ploidy-parameters could prove in 17 retrospective level 2b studies to add significant prognostic doi: 10.7243/2052-7896-2-7 information to the Gleason-score independent from the type of therapy ( Table 6) should encourage scientists to conduct studies in order to confirm these findings on a higher level of evidence as this had already been proposed by a WHOworking group [48]. Yet, level of evidence 1b studies, proving independent prognostic validity of DNA-ploidy over Gleasonscore to predict non progression of clinically insignificant prostate cancers under Active Surveillance in a prospective setting are still missing. We recommend to perform these.

Heterogeneity
Data on the representativity of DNA-ploidy measurements on biopsies for the cancer as a whole are heterogeneous and depend on the number of samples investigated. While [168,228] and [230] found discrepancies in only 3.8%, 5.0% and 8.0%, [210,232] reported different ploidy-levels in 24.1% and 29.0%. These figures are lower than comparable ones for the Gleason-score (30%:11). Because DNA-ploidy is inhomogenously distributed within prostate cancers, especially of advanced stages, as histopathological grades are, it is advisable to investigate all cancer foci in biopsies, either separately or pooled.

Why is DNA-cytometry not used more widely?
Some critical comments on this method overlook the enormous technological input that computer science, digital image analysis and informatics have meanwhile contributed to develop this method, becoming a biologically well founded, fast and valid prognostic technology. The fact that the procedure up to the recent development of digital nuclear classifiers has been too laborious and too time consuming and pathologists have not been sufficiently reimbursed, further prohibited its clinical acceptance.
While retrospective studies proving the independent prognostic validity of DNA-ploidy measurements have been published for all main types of treatment modalities of prostate cancers, prospective level 1b studies are still missing. As no other treatment decision in cancers of the prostate is so much dependent from an objective, reproducible and valid prognostication of an individual cancers behavior as Active Surveillance, prospective studies should especially focus on patients under this strategy.

Competing interests
The authors declare that they have no competing interests.