Tissue-Based MicroRNAs as Predictors of Biochemical Recurrence after Radical Prostatectomy: Beyond the Abstract

Based on our intensive research on microRNAs (miRNAs) for prognosis and prediction of urological cancers, we conducted a PubMed survey on studies that included biochemical recurrence data from prostate cancer (PCa) patients after radical prostatectomy. We believe that tissue based studies on whole prostatectomy specimen are the most accurate studies for a more accurate risk stratification of PCa patients. Therefore, data from biopsies are not considered. 
After the first miRNA was discovered in 1993, the number of publications has been exponentially increased. Approximately 10 years ago, in 2007 to 2008 the first studies were published on miRNAs in prostate tissue on small sample numbers but also in biological fluids such as serum/urine etc. In 2010, first own data in 24 tissue pairs (Schaefer et al.) indicated that several miRNAs correlated with the Gleason score and especially miR-96 had, combined with Gleason scores, prognostic potential as biomarker. 

Based on a literature search in PubMed until May 2017, we selected from preliminary 148 articles 53 relevant publications and evaluated the relationship between biochemical relapse and their significance as predictive biomarkers. We found, that only few miRNAs were independently described from different researchers. With differences in analytics, partial insufficient statistical power, different statistical evaluation methods and a wide heterogeneity of studies, we saw a limited consistency in the results. Further, the definition of a biochemical recurrence (BCR) with a PSA increase higher that 0.1 or 0.2 µg/L seemed to be clear and easy but in the American Urological Association (AUA) Prostate Guideline Update Panel, more than 50 different definitions have been found in 145 studies. According to the guidelines of the European Association of Urology (EAU), the BCR definition based on a renewed PSA increase to >0.2 µg/L that is confirmed by two consecutive elevated values seems to be more appropriate. Other factors further “dilute” a possible comparability of different studies: 

First the metrological traceability between different PSA assays with biological (PSA heterogeneity) and methodological reasons (e.g. different antibodies), second, the possibility of clinicopathological particularities with different risk groups or Gleason Scores or different interpretation of an R1 status and third, the different duration time of the respective follow-up influence a good comparison of studies. It should be noted that 23% of our studies did not specify the PSA cutoff for BCR at all. The exact specification of the important risk variables "resection margin status (R0/1)" and "lymph node status (pN0/1" was not given in 60 and 70%, respectively. In contrast, the pathological tumor stage and Gleason score were generally indicated. 

A limited validity of several studies is visible in those with sample sizes of less than 50 patients, only 10 to 20 BCRs or a mean follow-up period under five years. In this regard, it is significant that power and sample size calculations were presented in only two studies (no. 2 and 9). Further, only eight studies (15%) performed an internal or external validation of data. However, validation is an important criterion of the development phase.

In all reviewed studies, a total of 58 distinct miRNAs were used as potential BCR markers. Of these 58 miRNAs, only 15 miRNAs were examined in at least two studies, whereas 43 miRNAs were determined in only one study (Table 4, Table S2). The miRNAs miR-221, miR-21, miR-145, miR-1, and miR-96 were the most frequently analyzed miRNAs. However, there is now a great interest in using this wealth of information not only for selecting single markers but also for combining multiple markers into a specific panel or signature together with clinicopathological data. In this review, the studies by Nam et al. (Study 37: 5-miR signature with miR-139-5p, miR-223-3p, miR-301a-3p, miR-454-3p, and miR-652-3p) and Kristensen et al. (Study 46: 3-miRNA prognostic classifier with miR-185-5p, miR-221-3p, and miR-326) support these ideas. Based on the multiple-miRNA approach as a signature combined with rigorous validation processes (three validations in Study 46) or a high sample size and a high number of BCR events (n = 491, 167 BCRs in Study 37), the two studies yielded promising results. Both studies are among the most convincing studies evaluated in this review and can be considered future-oriented examples.

Even if we consider all critical points and several limitations of the selected reviewed studies, we see first promising results that can be considered as proof of the true potential of miRNAs as BCR predictors. In this review we learned from the deficiencies of the conducted studies and can draw corresponding conclusions for future studies. There are two essential conclusions:

1.     No study fulfilled all suggested requirements specified in the final development phase to establish a robust BCR tool for clinical practice using miRNAs. 

2.     Critical study deficiencies were identified. 

Thus, further studies are urgently needed to address these deficiencies in multi-institutional studies to validate the clinical usability of miRNAs in combination with the conventional clinicopathological variables. In the future, miRNA-based BCR tools may contribute to a personalized treatment.

Written by: Zhongwei Zhao, Carsten Stephan, Sabine Weickmann, Monika Jung, Glen Kristiansen, Klaus Jung

Read the Abstract