Tertiary Gleason Pattern 5 is An Independent Risk Factor for Prostate Cancer with GS 7: A Retrospective Study from China.

Background: Tertiary Gleason pattern 5 (TGP5) was found to be prognostic in prostate cancer (PCa) after radical prostatectomy (RP), but related data from China was rare. Our study was aimed at nding out the effect of TGP5 on PCa with Gleason score (GS) 7 and supplementing data from China in this eld. Methods: A total of 229 cases met with inclusion criteria during Jan. 2014 to Dec. 2018 were reviewed. Cases were divided into GS 7 without TGP5 and GS 7 with TGP5. We compared age at diagnosis, preoperative PSA level, prostate volume, PSA density (PSAD), GS variation, clinical T staging, pathological T staging, T staging variation, extra-prostatic extension (EPE), positive surgical margin (PSM) and seminal vesicle invasion (SVI) between the groups. Effects of TGP5 on prognosis of PCa with GS 7 were evaluated using biochemical recurrence (BCR) as the primary end point. Results: TGP5 was related to higher PSM rate (P=0.001) and BCR rate (P=0.009) but not related to higher preoperative PSA level, larger prostate volume, higher PSAD, GS upgrade, poorer clinical/pathological T staging, T upstaging, EPE and SVI (all P>0.05). The median follow-up time was 24 months (interquartile range 17.5-45.5). TGP5 was an independent risk factor to PCa with GS 7 after RP using Kaplan-Meier log-rank test (P=0.018). Both univariable and multivariable cox-regression analysis pointed out that TGP5 increased the incidence of BCR in PCa with GS 7 (P<0.05). Stratied analyses were also done. Conclusion: TGP5 is an independent risk factor predicting of BCR after RP in PCa with GS 7 from China. TGP5 is related to higher PSM rate and BCR incidence. It is time to renew the contemporary Grading Group system with the consideration of TGP. interquartile range (IQR). Differences in categorical variables and continuous variables were analyzed with the chi-square test and Kruskal-Wallis test, respectively. BCR-free survival (BFS) was calculated from the date of surgery to BCR or to last follow-up. The Kaplan-Meier analysis and log-rank tests were used to estimate and compare BCR-free survival. Univariable and multivariable analyses were also done with Cox-regression models using hazard ratios (HRs) and 95% condence indexes (CIs). Statistical signicance was considered at P (cid:0) 0.05 (2-tailed test). IBM SPSS Statistics version 26.0 (SPSS Inc., Chicago, IL, USA) was used for analysis. Graphpad Prism version 8.4.3 (GraphPad Software, LLC, USA) was used for curve illustration.

more related to patients' outcome. Tertiary Gleason pattern (TGP) was rst reported in 2000 (5), but there haven't been consensus on its de nition until now. In 2014 ISUP, most of the experts supported the de nition that for specimens after radical prostatectomy (RP), TGP, also called minor highgrade pattern, was de ned as the third most prevalent differentiation pattern consisting of less than 5% of the cancer in the specimen but worse than primary and secondary patterns (4). Is TGP related to PCa prognosis? Many studies aboard have been conducted since TGP was proposed. They got similar conclusions that patients with the existence of TGP had worse outcome than those without it (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20). However, data from China, nation with the largest population, is rare and limited. It is important to know the characteristics of TGP in Chinese population. For GS 7 groups, either 3 + 4 or 4 + 3, TGP speci cally refers to TGP5.
Our study was aimed at: 1. Finding out the basic characteristic of TGP5 in PCa with GS 7 in our institution; 2. Validating TGP5's prognostic effects on PCa patients with GS 7 in China; 3. Supplementing data from China in this eld and doing the rst stratifying analysis considering both TGP and GS in Chinese.

Patients selection and data collection
We identi ed 317 patients who met with following conditions from Peking University First Hospital, which is one of the top institutions in urology eld in China: 1) Underwent prostate biopsy in this hospital and pathological diagnosis were PCa; 2) Underwent RP in this hospital and GS in specimens after RP was 7 either 3+4 or 4+3 ; 3) Surgery time was from Jan. 2014 to Dec. 2018. Patients who received neo-adjuvant, adjuvant therapy or salvage therapy, which were 61 cases in total, were excluded. In the remaining 256 patients, 19 patients were lost to follow-up, 8 patients had incomplete or poor reliability of follow-up data. In total, 229 patients were entered analysis nally ( Figure 1). Ethical audit was permitted before we conducted this study.

Pathological specimens processing
All prostate specimens after RP were processed according to standard pathological procedures. Reports were given by two quali ed pathologists separately. Pathological T staging was given according to 2009 American Joint Committee on Cancer TNM staging system. GS was de ned as the sum of primary and secondary patterns of prostate specimens. TGP5 was de ned as the third prevalent pattern ( 5%) but worse than primary and secondary patterns in RP specimen with GS 7 according to the de nition most experts supported on 2014 ISUP. GS upgrade was considered when GS in specimens after RP was poorer than GS in prostate biopsy specimens according to Grading Group system from 2005 ISUP, while TGP5 did not change the GS. Surgical margin was considered positive when tumor cells were in contact with the ink on the RP specimen surface.

Follow-up
The primary endpoint of our study was biochemical recurrence (BCR). It was de ned as prostate-speci c antigen (PSA) >0.2 ng/ml by two consecutive (more than 1 week apart) measurement after RP. Generally, patients would be evaluated with PSA level 3 months after RP. If this PSA values exceeded 0.1 ng/ml, we considered that the surgery did not reach the standard of "radical". In this case, salvage therapy would be given instead and these people would also be excluded even PSA >0.2 ng/ml. Patients were followed up quarterly within the rst year, semiannually in the second year, and annually in the following years mainly with PSA evaluation. Among the primary 317 patients, after telephone follow-up and looking up original medical history, we identi ed 61 cases who underwent neo-adjuvant/adjuvant/salvage therapy, which were excluded, and 8 patients who had incomplete or poor reliability of follow-up data, which were not included in nal analysis. There were 19 patients lost to follow up.

Statistical analysis
Categorical variables were presented as numbers and percentages whereas continuous variables as median and interquartile range (IQR). Differences in categorical variables and continuous variables were analyzed with the chi-square test and Kruskal-Wallis test, respectively. BCR-free survival (BFS) was calculated from the date of surgery to BCR or to last follow-up. The Kaplan-Meier analysis and log-rank tests were used to estimate and compare BCR-free survival. Univariable and multivariable analyses were also done with Cox-regression models using hazard ratios (HRs) and 95% con dence indexes (CIs). Statistical signi cance was considered at P 0.05 (2-tailed test GS 7 without TGP5 and GS 7 with TGP5, respectively. There were no signi cant differences between groups in age at diagnosis (P = 0.461, P = 0.708 using chi-square test after stratifying ages into < 65 years old and ≥ 65 years old) and preoperative PSA level (P = 0.054, P = 0.176 using chi-square test after stratifying preoperative PSA level into < 10 ng/ml and ≥ 10 ng/ml). Other characteristic distributions were shown in Table 1. No signi cant difference between these two groups was found when it came to prostate volume (P = 0.264, P = 0.247 using chisquare test after dividing prostate volume into < 50 ml and ≥ 50 ml), PSAD (P = 0.355, P = 0.166 using chi-square test after dividing operation ways (P = 0.728). There were statistically difference in positive surgical margin rate (P = 0.001) between GS 7 without TGP5 and GS 7 with TGP5. Table 1 also showed the biochemical recurrence incidence and cases in our study in different groups.
There was signi cant difference in BCR incidence between GS 7 without TGP5 and GS 7 with TGP5 (P = 0.009), which indicated TGP5 might be a risk factor in prognosis of PCa.
3. Association of TGP5 with biochemical recurrence after radical prostatectomy.
In univariable cox-regression analysis, pathological T staging, T upstaging, positive surgical margin, Gleason score and TGP5 were found as independent risk factors to BCR after RP. Compared to T2 stage, T3a and T3b had 5.065-fold and 10.116-fold risk to experience BCR with P = 0.005 and P < 0.001, respectively. Patients with T upstaging had 3.226-fold risk to experience BCR than non-upstaging. As a well-acknowledged risk factors in prostate cancer, GS 4 + 3 had 2.896-fold risk to BCR than GS 3 + 4 (P = 0.013). Patients with PSM had 2.362-fold risk to BCR than negative surgical margin (P = 0.042). Patients with TGP5 had about 2.640-fold risk to BCR after RP than patients without TGP5 (P = 0.023). Elder age at diagnosis (≥ 65 years old), higher preoperative PSA level (>10 ng/ml), higher prostate volume (≥ 50 ml), higher PSAD (≥ 0.15 ng/ml/ml), GS upgrade and higher clinical T staging did not increase the risk to BCR compared to age 3.  (4,(21)(22)(23). GS plays an important role in risk classi cation so as in treatment strategy making. To those patients with non-regional and non-distant metastases prostate cancer, if GS ≤ 6 (low-risk), active surveillance or surgical treatment are both available while if GS ≥ 8 (high-risk localized), multimodal treatments are preferred, for example, radical prostatectomy or radical radiotherapy as initial treatment, androgen deprivation therapy (or radiotherapy) as adjuvant therapy. If GS = 7, either 3 + 4 or 4 + 3, radical prostatectomy or radiotherapy is recommended. However, one should receive adjuvant therapy after RP depends on whether he has lymph node metastases or adverse features, which now include positive surgical margin, seminal vesicle invasion and extracapsular extension (see NCCN). Identifying the risk factors for PCa patients with GS 7 is extremely important because it helps identify the certain group of patients who need adjuvant therapy and can bene t from it maximumly in prolonging survival times, meanwhile helps avoid over treatment and unnecessary side effects for those who don't need adjuvant therapy.
GS was rst proposed by Gleason and de ned as the sum of primary and secondary patterns of prostate specimens (2, 3), which indicated that prognosis of prostate cancer was mainly decided by its primary and secondary Gleason patterns. Primary Gleason pattern was found to play an even more important role in predicting malignancy degree of PCa with more investigation being conducted. However, researchers were wondering what role tertiary but higher Gleason pattern played in PCa since GS was proposed until Pan C.C., et al. rst proposed TGP was also a risk factor to PCa (5).
The main purpose of this study was to decide whether TGP5 was an independent risk factor to PCa with GS 7 in Chinese population and whether clinician should take it into account as adverse feature when assigning further treatment for PCa patients with GS 7 after RP.
There have been many studies published concerning TGP's effects on prostate cancer. Pan, C. C., et al. rst brought tertiary Gleason pattern to the public. They analyzed 114 radical prostatectomies with tertiary components ( 5%), which were compared with a prostatectomy database comprised of 2,276 cases without a tertiary component. In their study, GS 7 with TGP5 showed signi cantly worse pathologic stages than GS 7 without TGP5 and were not different statistically from typical GS 8 (4 + 4) tumors.
GS 7 with TGP5 revealed signi cantly higher progression rates than GS 7 without TGP5 (5). Adam, M., et al. compared 2,396 patients with (22.4%) and 8,260 without (77.5%) a tertiary Gleason pattern for adverse histopathological features (extra-prostatic extension, seminal vesicle invasion, positive surgical margins and lymph node invasion) and analyzed the effect of a tertiary Gleason pattern on biochemical recurrence. They found TGP was statistically signi cantly associated with all evaluated histopathological parameters and it was an independent predictor of biochemical recurrence (HR 1.43, p < 0.001). On subanalysis, TGP independently predicted biochemical recurrence in patients with GS 3 + 4 and 4 + 3 after RP, respectively (6). Borhan, W. et al. analyzed a total of 4060 specimens with a GS 7 with and without TGP5. Cases were subdivided into 3 + 4, 3 + 4 with TGP5, 4 + 3, and 4 + 3 with TGP5. They compared prostate-speci c antigen, clinical stage, pathologic stage, and surgical margin status between the groups. The impact of TGP5 on biochemical recurrence was also assessed. They found TGP5 was related to higher PSA level, pathologic stage, positive surgical margin and worse prognostic outcome (8). Although many studies got the same results that TGP5 related to one or more clinicopathological characteristics and PCa outcome (4,(21)(22)(23), some studies failed. It remains uncertain that what variables can predict TGP5 and whether 3 + 4 + TGP5 has same prognostic outcome as 4 + 3 without TGP5.
In this study, we retrospectively analyzed 229 patients who met with inclusion criteria and had complete and reliable data from Jan. 2014 to Dec. 2018 in Peking University First Hospital. Clinical and pathological characteristics distribution between different groups with and without TGP5 was being analyzed. Differing from some researches abroad, our study found that TGP5 wasn't related to elder age at diagnosis (≥ 65 years old) and higher preoperative PSA level (≥ 10 ng/ml). Besides, no statistical difference was found between GS 7 without TGP5 and GS 7 with TGP5 when comparing prostate volume, PSAD, GS variation, clinical T staging, pathological T staging, T staging variation, extra-prostatic extension, seminal vesicle invasion and operation ways.
Recently, a study by Jiakun Li et al. from China was published on OncoTargets and Therapy. In their study, they included 350 patients in total from 2009 to 2017, among which ~ 10% (n = 34) had TGP5. They also found that TGP5 wasn't related to elder age at diagnosis, higher preoperative PSA levels and pathological staging (24). In our study, ~ 25.3% (n = 58) patients had TGP5, which was higher than Jiakun Li's study and also higher than some studies from other countries. Cases were all within 6 years, strictly followed the recommended criteria proposed during ISUP 2014 on TGP5, which meant our study were of more time-effectiveness and reliability. So, even we were surprised by the nding that TGP5 was not related to those variables, we could not fully rule out the possibility that TGP5 was indeed not associated with those variables mentioned above. We believed race and genetics difference might account for and explain it. But given that our sample size was still small, more investigation needs to be done in the future.
To be noted, the PSM rate in the entire cohort was 21.4% and we also found the presence of TGP5 was related to higher PSM with statistical signi cance (P = 0.002). In GS = 7 with TGP5, PSM rate was 36.2% while in GS = 7 without TGP5, PSM rate was only 16.4%. In Sauter G, et al.'s study, the PSM rates were 16.09%, 13.18% and 48.54% in GS = 7, GS = 7 without and with TGP5, respectively (25). In Li, J. et al.'s study, the PSM rates were 20.17%, 19.81% and 23.53% in GS = 7, GS = 7 without and with TGP5, respectively (24). Some other research also had similar ndings that the presence of TGP5 was related to higher PSM rate while there weren't many explanation or discussion about it (8,16). Compared to these studies with larger population, our comparison results were consistent with them. However, PSM rate was higher in our cohort. Especially for group GS = 4 + 3 with TGP5, PSM rate was even higher as 48.1%. There were many factors including pathological T staging, operation ways, experience of surgeons Biochemical recurrence was the primary endpoint in our study, which is now considered as a sign of prostate cancer relapse. To identify the predictors of BCR is crucial since it will help to select patients needing multimodal therapy while spare the others. In our study, for prostate cancer with GS 7, the incidence of BCR in the group with TGP5 was signi cantly higher than that in the group without TGP5, indicating TGP5 was a risk factor to PCa. Kaplan-Maier survival analysis and log-rank test validated it with biochemical recurrence-free survival (BFS) curve showing that the presence of TGP5 was negatively correlated with BFS.
Univariable cox-regression analysis got the same result that TGP5 was a risk factor to prostate cancer with GS = 7, HR 2.640 (95%CI 1.143-6.097). In univariable cox-regression analysis, GS composition and pathological T staging were also risk factors to PCa prognosis. Given that the composition of GS and pathological T staging might interfere with TGP5 in survival analysis, multivariable cox-regression analysis was also conducted, showing TGP5 was an independent risk factor to BCR, with HR 2.360 (95%CI 1.019-5.464).
This study had many limitations. Firstly, like other retrospective studies, recall bias was the biggest source of bias in this study. Secondly, sample size was small compared to previous studies, with only 229 cases nally entered into our analysis and only n = 23(10%) patients experiencing BCR. However, our cohort was of more time-effectiveness and reliability since the pathological reports of specimen accorded with the latest criteria from 2014 ISUP and variables affecting the accuracy of survival analysis were considered as exclusion criteria. Thirdly, there were some variables not recorded such as pelvic lymph nodes dissection (PLND) during operation. In this cohort, there were only 67 cases (29.26%) experiencing PLND during RP, and all were negative ndings. According to the existing studies comparing oncological and non-oncological outcomes between no PLND and PLND, the bene ts and harms of PLND during RP still remain controversial (27). Urologists in this hospital also have different opinions on whether performing PLND during RP or not. We nally decided to discard this variable to ensure the accuracy of the survival analysis. But we believed that lymph nodes dissection during RP was theoretically related to the survival of prostate cancer and might affect some results of this study.
We conducted a strati ed analysis considering and eliminating the in uence of GS composition on prostate cancer. The cohort was divided into 4 groups as 3 + 4 without TGP5, 3 + 4 with TGP5, 4 + 3 without TGP5 and 4 + 3 + TGP5. Results showed that overall differences among 4 groups were found in terms of GS variation, pathological T staging, prostate extracapsular invasion, positive surgical margins, and biochemical recurrence rate. But these differences mainly existed between group 3 + 4 without TGP5 and group 4 + 3 with TGP5. In 3 + 4/4 + 3 group alone in our study, the existence of TGP5 was not related to poorer clinicopathological features than 3 + 4/4 + 3 without TGP5. K-M survival analysis among 4 groups also showed statistical difference only between group 3 + 4 without TGP5 and group 4 + 3 with TGP5. BFS in group 3 + 4 with TGP5 were not statistically different from group 4 + 3 without TGP5. Whether it indicated that 3 + 4 with TGP5 had same risk as 4 + 3 without TGP5 was worthy of more investigations. Although the stratifying analyses did not nd many positive ndings or consistent with previous studies, due to the small sample size, this study was still the rst stratifying analyses grouped by GS and TGP5 in China, which was the novelty of this study.
While exploring the prognostic effect of TGP on prostate cancer, researchers are also working on how to integrate tertiary Gleason pattern (5) into new grading standard in prostate cancer (with GS 7). Sauter, G., et al. proposed the integrated quantitative Gleason score (IQ-Gleason) to evaluate the risk of prostate cancer. It ranges from 0-117.5 and is calculated as follows: percentage of unfavorable Gleason pattern (Gleason 4 + Gleason 5) + 10 score points if any Gleason 5 pattern was seen + another 7.5 score points in case of Gleason 5 quantities of > 20%. For example, the IQ-Gleason of a Gleason 3 + 4 = 7 cancer with 40% Gleason 4 is 40, the IQ-Gleason of a Gleason 3 + 4 = 7/tertiary grade (TG) 5 cancer with 40% Gleason 4 and 5% Gleason 5 is 40 + 5 + 10 = 55. The IQ-Gleason of a (Gleason 4 + 5 = 9) cancer with 60% Gleason 4 and 40% Gleason 5 is 60 + 40 + 10 + 7.5 = 117.5. This model was published on European Urology and got some initial and promising results (25). Its quanti cation idea originated from Gleason Score but got more. Other models were also created by different researchers but lack of large population-based validation for now. These models remind that urological pathologists should keep an idea of quantitation when processing and reporting specimens after RP. We are looking forward to establish a new grading system for PCa concerning TGP to assign best treatment for patients.

Conclusion
Based on data from the top medical institution in urology eld in China, despite the limitations mentioned above, we still achieved some persuasive conclusions: (1) TGP5 is associated with poorer pathological characteristics like positive surgical margin in prostate cancer specimens but not related to elder age at diagnosis, preoperative PSA level, prostate volume, PSAD, GS variation, clinical T staging, pathological T staging, T upstaging and seminal vesicle invasion in our cohort ; (2) TGP5 is an independent risk predictor of prostate cancer with GS 7, which should be considered when assigning further treatment for these patients after RP in a clinical scenario;(3) Race and genetics difference might exist in TGP5 between China and other countries. More studies with larger population, more possible variables and strati ed analysis are needed to conducted in China. It is time to renew the contemporary Grading Group system, with the consideration of TGP and being quantitative. Abbreviations PCa, prostate cancer; GS, Gleason score; ISUP, International Society of Urological Pathology; TGP, tertiary gleason pattern; RP, radical prostatectomy; BCR, biochemical recurrence; PSA, prostate speci c antigen; IQR, interquartile range; BFS, BCR-free survival; HR, hazard ratios; CIs, con dence indexes; PSAD, prostate speci c antigen density; PSM, positive surgical margin; EPE, extraprostatic extension; SVI, seminal vesicle invasion; LSRP, laparoscopic radical prostatectomy; RARP, robot-assistant laparoscopic radical prostatectomy. 3. Availability of data and materials