In the present study, comparing to conventional predictive models, we proposed a new nomogram by incorporating MTD and PSA nadir, which showed improved accuracy of BCR prediction for patients after RP.
After RP, PSA is expected to be undetectable within six weeks[9] and it is utmost important parameter that should be monitored postoperatively. Elevated PSA level after RP indicates high risk of local recurrence or metastasis[10]. If the postoperative PSA reaches 0.2 ng/mL, patient is assigned the status of BCR[18], which was a signal of cancer activity at visual undetectable level. The relationship between PSA nadir and BCR after RP has been extensively studied. A retrospective study reported that compared with men with PSA < 0.01 ng/mL after RP, the probability of BCRFS at 5 years dropped from 92.4–56.8% in patients with PSA ≥ 0.01 ng/mL[19]. In a study of 582 patients carried out by Matsumoto et al., PSA persistence after RP was associated with increased BCR and overall mortality[20]. These results are in line with the observations in the present study. In current study, 71.8% patients got an undetectable PSA nadir and 28.2% patients had a detectable nadir during follow-up. PSA nadir after RP was found to be an independent prognostic factor (P < 0.001) in predicting BCR in univariable and multivariable analyses. Patients with PSA nadir < 0.01 ng/mL had significantly longer BCRFS in our study cohort (Fig. 1B, log-rank p < 0.0001).
According to our clinical experience, tumor burden should be associated with oncological outcomes. Tumor volume and MTD as the common indicators of tumor burden have been studied by the researchers and have proved to be independent prognostic factors of BCR[13, 21]. However, prostate cancer has been recognized as a multifocal disease[22] and the calculation of tumor volume and MTD is complicated. In 2013, Billis et al. found that the tumor extent in a surgical specimen should be estimated with the dominant tumor and not the total tumor extent. They also reported the association of the dominant tumor with BCR prediction[23]. Even so, the calculation of tumor volume is time consuming and difficult. For the above reasons, we chose MTD as the research target and it was defined as the maximum diameter of the dominant tumor. Unlike previous studies, we measured MTD based on MRI instead of pathological specimen. MRI has better repeatability and less deformation, while on pathological specimen, deformation can vary greatly due to shrinking of tissues after soaking in formalin. Lee et al. measured the diameter of the suspicious tumor lesion on diffusion weighted images of MRI and demonstrated that the diameter of tumor could increase the prediction of insignificant prostate cancer in candidates for active surveillance[14]. In the studies of Kozal et al.[24] and Müller et al.[25], MTD was an independent prognostic factor for BCR, even though they measured MTD on pathological specimens. Based on their findings, we hypothesized that the MTD on MRI could be an independent prognostic factor for prostate cancer, however the relationship between MTD measured on MRI and BCR after RP has rarely been explored in their study as well as other previous studies. As expected, the results of the present study showed that MTD on MRI was an independently significant predictor of BCR (p = 0.034) and the Kaplan-Meier curve depicted that men with MTD > 2.9 cm had shorter BCRFS (Fig. 1C, log-rank p = 0.0003). Interestingly, the median MTD in the present study was larger than that in the previous studies[26]. We attributed this phenomenon to shrinking of tissues after soaking in formalin which might decrease the MTD [27]. Additionally, in our study, pathological tumor stage ≥ T2c was reported in the majority of patients (n = 290, 86%, Table 1) and it might be another reason why we have larger MTD. In the study of Eichelberger et al., MTD was found to be associated with the pathologic stages[28]. With the rapid development of radiographic technologies and artificial intelligence, the identification and measurements of prostate cancer on MRI are more accurate with high repeatability for prognostic evaluation.
The CAPRA-S score is a postoperative score created by Cooperberg et al.[6], based on preoperative PSA, Gleason score, SM, ECE, SVI, and LNI. The prognostic value of these variables was verified in our study cohort as well. All of them were significantly associated with BCR in the univariable analysis and Gleason score, SM, and SVI were independent predictors of BCR in multivariable analysis. The c-index of our newly developed nomogram was slightly higher than that of the CAPRA-S score in our study cohort. Moreover, our nomogram predictions closely approached the actual outcome both at 3 and 5 years after RP, demonstrating good calibration, as depicted in the calibration plot. Comparing these two models, we found that our new nomogram consisted of two parts. One part was composed of the commonly used variables, namely Gleason score, SM, and SVI and the other part was composed of PSA nadir and MTD measured on MRI. In the current study, we observed that both PSA nadir and MTD were significantly associated with BCR in univariable analysis and they were also independent prognostic factors after adjusting preoperative PSA, Gleason score, SM, ECE, and SVI. Kaplan-Meier curve showed that the patients with these two risk factors simultaneously had the shortest BCRFS and patients with none of these two risk factors had the longest BCRFS (Fig. 1D, log-rank p < 0.0001). However, only few of the previous prediction tools used MTD and PSA nadir at the same time. To verify the incremental predictive power of the combination of PSA nadir and MTD, we developed a basic model including Gleason score, SM, and SVI for comparison. The c-index was decreased from 0.76 to 0.66 (p < 0.001) when PSA nadir and MTD were removed from our new nomogram. The time-dependent ROC curves illustrated the advantage of our new nomogram at both 3 and 5 years after RP. The decision curve analyses also showed the advantage of our new nomogram, across the various threshold probabilities, and the new nomogram had greater net benefit than both the basic model and the CAPRA-S score in our study cohort. Our new nomogram is a promising tool to predict BCRFS and guide follow-up and decision-making of adjuvant treatment. In addition, PSA nadir and MTD improved the accuracy of our new nomogram in predicting BCR.
Our study has several limitations. First, it was a retrospective study and the population was relatively smaller compared to the previous studies. Second, the present study has not yet been validated externally and the analysis of overall survival was lacked due to the short-term follow-up duration.