The prognostic value of coagulation factors
Fig. 1 showed the design of current study. The K-M survival analysis was used to judge the recurrent-free survival of patients with different coagulation function status. The median value of these coagulation factors used to separate the patients to high-level and low-level subgroups. We found that the lower value of FDP (HR: 0.55, 95% CI: 0.314-0.951, P =0.033), and TT (HR: 0.54, 95% CI: 0.314-0.920, P =0.024) are the risk factor of the biochemical recurrence of prostate cancer patients (Fig. 2). While the other factors, PT (HR: 1.04, 95% CI: 0.611-1.768, P = 0.89), PTA (HR: 1.01, 95% CI: 0.585-1.751, P = 0.965), FIB (HR: 0.7, 95% CI: 0.404-1.205, P = 0.197), INR (HR: 1.08, 95% CI: 0.629-1.855, P = 0.78), APTT (HR: 1.43, 95% CI: 0.839-2.441, P = 0.189) and DD (HR: 0.84, 95% CI: 0.493-1.446, P = 0.537) are not the predicative factors for the biochemical recurrence for prostate cancer patients. Thus, we considered containing TT and FDP into our nomogram model.
Separation of the training and validation cohorts
With the ratio of 7:3, prostate cancer patients were separated into the training (n = 120) and validation sets (n = 48) randomly, and the patients' characteristics in both cohorts were summarized in Table 1. The distribution of patients with different clinical features consistent among the training and validation cohorts (all P > 0.05). The recurrent-free survival time in the training cohort is 19.66 ± 16.98 months, while the recurrent-free survival time in the validation cohort is 19.66 ± 15.37 months.
Establish and validate the prognostic nomogram for prostate cancer patients in the training cohort
Other than the TT and FDP parameters, we also enrolled the clinical features of age, TPSA, and Gleason Group, which are traditional used for the predicting of the severe status of prostate cancer patients in clinical, to establish the prognostic nomogram (Fig.3). We also tried to enrolled the pathological T stage to establish the nomogram model, but the overall prediction value was less good than without it (data not shown). Patients can obtain the total points along with the value of age, TPSA, Gleason Group, TT, and FDP refer to the nomogram model, and the specific points for each value listed in Table 2. We observe that younger age is a risk factor for the recurrence of prostate cancer patients, while no doubt results that higher TPSA and advanced Gleason Group is the risk factor. Referring to the median risk score, we subclassified the prostate cancer patients into high- and low-risk of recurrence in the training cohort (high-risk group vs. low-risk group: recurrent-free survival time 25.49 ± 19.45 months vs. 14.11 ± 12.23 months, log-rank P-value < 0.0001, Fig. 4A). The stability and predictive value were also proved by calibration plot (Fig. 4B) and C-index analyses (C-index = 0.74 for 1-year and 0.69 for 3-year of recurrent-free survival prediction, Fig. 4C).
ROC and K-M analyses for prostate cancer patients in the validation cohort
We calculated the total points of patients in the validation cohort according to the well-established Nomogram model based on the training cohort. K-M plot and log-rank analysis proved that the patients in the high-risk subgroup had unfavorable prognoses than the low-risk subgroup (high-risk group vs. low-risk group: recurrent-free survival time 25.31 ± 16.50 months vs. 14.67 ± 12.93 months, log-rank P-value = 0.004, Fig. 5A). The predictive value of the Nomogram was confirmed by ROC analyses again in the validation cohort, with the AUC value of 0.651 (Fig. 5B).
To verify whether this nomogram model could be used in different clinicopathological subgroups, we performed subgroup analyses. We found that this model was able to predict the recurrence-free survival of prostate cancer patients in different age subgroups (<= 60, and > 60 years old) and PSA level subgroups (<= 10 ng/ml, and > 10 ng/ml) (Fig. 6). Attributing to the limited samples in the “Gleason score < 7” subgroup, we failed to obtain positive results (Fig. 6). Future studies are needed to verify this finding.
The results of Public databases and GSVA analysis of coagulation function
As mentioned above, the activation of the coagulation system might play important roles during the progression of prostate cancer. We thought to explore the relationship between coagulation function status and recurrent-free survival of prostate cancer patients at genetic levels. We acquired the mRNA expression profile of prostate cancer samples from four public datasets [TCGA-PRAD cohort (495 samples), MSKCC cohort (140 samples), GSE116918 cohort (248 samples), and GSE70769 cohort (93 samples)]. Then, the coagulation pathway activated score for each patient from the online website GSVA was calculated. Our results suggested that these patients with higher coagulation scores mostly had unfavorable prognoses than those with lower coagulation scores in the MSKCC cohort (P = 0.023) and GSE116918 (P = 0.012) (Fig. 7). We failed to obtain positive results for the other two cohorts, even the curves showed good tendency (TCGA-PRAD cohort, P = 0.082, and GSE70769 cohort, P = 0.261, Fig. 7).