Upper tract urothelial carcinoma (UTUC) is an uncommon and heterogeneous disease with poor oncological outcomes [1-3]. For patients with localized UTUC, radical nephroureterectomy (RNU) with bladder cuff excision is the definitive treatment of choice [4]. In cases of advanced UTUC, neoadjuvant or adjuvant chemotherapy combined with RNU is considered [5]. However, the role of systemic chemotherapy remains unclear, requiring more concrete evidence from prospective randomized clinical trials [6].

Considering the high risk of recurrence and progression of UTUC, accurate prediction of patient prognosis is of great value for patients as well as clinicians to improve risk stratification and clinical decision-making for further therapeutic plans [7,8]. However, conventional survival prediction is calculated using data that includes patients who have already died, and not exclusively patients who survived for an additional few years after initial assessment [9]. As patient prognosis changes over time with survivorship achieved beyond primary treatment, standard survival estimation may be less relevant for patients with particularly unfavorable prognoses such as UTUC [10].

Conditional survival (CS) probability is the likelihood of surviving for an additional duration after diagnosis or treatment of diseases, assuming an individual patient has already survived for a specified time [11]. Valuable studies of CS estimation for several malignancies, including lung, colorectal, ovarian, prostate, kidney, and bladder cancer have been reported, providing highly accurate predictions of risk dynamics over time [11-14]. However, there is a dearth of evidence exploring the concept of CS in UTUC after RNU. Here, we assessed the CS probability and relevant clinicopathological features of UTUC patients treated with RNU, and thus described the predicting factors of a specific CS estimation in these patients.

The baseline characteristics of the patients are shown in Table 1. The median follow-up period was 58.70 months (interquartile range [IQR], 35.12 to 104.45). The median age of patients was 63.2 (IQR, 56.5 to 70.4), and approximately 80% were male. Fifty-eight patients (17.6%) had a history of previous or concomitant bladder cancer. Half of the patients had UTUCs in the renal pelvis. In pathologic findings, the proportion of advanced stage (pT3 and T4), high grade, and node positivity was 44.2%, 65.5%, and 4.2%, respectively. After surgical treatment, 85 individuals (25.8%) received adjuvant chemotherapy. Of the entire study population, 126 patients (38.3%) experienced intravesical recurrence, and 122 patients (37.0%) died due to various causes while 89 patients (27.0%) died from UTUCs during follow-up. Kaplan-Meier curves were generated to illustrate CSS, OS, and IVRFS in this population (Fig. 1).

We examined whether the CS probability changed due to additional survival time after RNU. As shown in Supplementary Table 1, patients who had already survived 5 years after RNU had the most favorable conditional CSS probability of all given years, compared to those who survived less than 5 years (Fig. 2A). Considering survivorship from baseline to 5 years after surgery, the 5-year conditional CSS increased gradually from 73.7% to 74.7%, 77.4%, 80.9%, 84.5%, and 90.4% each successive year by their survivorship, whereas the actual survival rates decreased over time (Fig. 2B). Conditional OS rates also improved as the patients’ survivorship increased after RNU (Supplementary Table 2). Of note, patients with 5-year survivorship showed the best conditional OS rates in all estimated years, compared to those with poor survivorship (Fig. 3A). Similar to the findings for CSS, the 5-year conditional OS increased from 67.9% to 69.9%, 72.6%, 77.5%, 79.4%, and 83.0% each successive year, in all given survivorships, while the actual survival rates diminished with the progression of time (Fig. 3B). We also observed that conditional IVRFS probabilities improved as the patients’ survival years increased after surgery (Supplementary Table 3). Patients who had survived 5 years after surgery had the highest conditional IVRFS rates in most estimated survival periods compared to those with shorter survival times (Fig. 4A). In addition, the 5-year conditional IVRFS showed a marked increase from 56.5% to 77.6%, 91.0%, 95.8%, 94.6%, and 97.0% each successive year of survival, while the actual IVRFS rates declined over time (Fig. 4B).

We assessed whether the significance of prognostic factors were changed over time by survivorship after RNU. At baseline, well-known prognostic factors, including age, hydronephrosis, pathologic T stage, tumor grade, and lymph node (LN) status, were significantly predictive of 5-year CSS and OS (Supplementary Tables 4 and 5). Tumor location and grade remained as the significant prognostic factors for 5-year conditional CSS, regardless of survivorships, after RNU. For 5-year conditional OS, age was consistently identified as a prognostic factor in all given years of survivorship. Interestingly, patients with poor pathologic features, such as advanced T stage (≥ pT3), high grade, LVI, and LN positivity, showed higher increments of 5-year conditional CSS and OS compared to their counterparts. For 5-year IVRFS, we found that age, BMI, and bladder cancer history were only predictive at the time of surgery, and that only the presence of hydronephrosis remained a significant predictive factor at 1- and 2-year of survival (Supplementary Table 6). However, other clinicopathological variables were not predictive of 5-year IVRFS for any given periods of survival.

Next, using multivariate analysis, we attempted to determine the key predictors for the 5-year conditional CSS, OS, and IVRFS, considering 1-, 2-, 3-, 4-, and 5-year survivorships after surgery. For 5-year conditional CSS, no significant predictors remained in all given survivorship range (Table 2). However, tumor location was found to be a significant factor for 5-year conditional CSS only beyond 2-year survivorship. LVI positivity remained the key predictor in baseline, 1-, 2-, 3-, and 4-year survival times after surgery. These factors tended to increase the hazard ratio (HR) over time. For the 5-year conditional OS, only age was distinguished as a significant predictor in all given survivorships with incremental changes of HR as time progressed (Table 3). In addition, having a high-grade tumor was identified as the substantive predictor in baseline to 4-year survivorships after definitive surgery. However, no significant predictive factors for 5-year conditional IVRFS over time after RNU were identified using multivariate analysis (Table 4).

Accurate prediction of survival probabilities is a cornerstone of counseling regarding prognosis as well as planning therapeutic strategies in cancer patients. Typically, survival estimates have been reported as actuarial survival after definitive therapy at the time of diagnosis or treatment. Due to its dependence on a single time-point, this method of prediction may only provide limited information on survival outcomes [12,14]. However, clinicians should consider the subsequent changes of survival probabilities over time after initial assessment. In this regard, the concept of CS can provide useful and accurate information on risk changes as time elapses after baseline survival prediction [20].

Although CS has been emerging as a clinically reliable predictor for survival probabilities in various malignancies, there is a paucity of evidence in UTUC for the role of CS estimation after surgical resection [11-14]. To date, only one study has been reported, which provided multi-institutional data from 3,544 patients who underwent RNU for UTUC [10]. This study showed that CS of UTUC patients increased with passage of time after surgery. Likewise, our results showed that 5-year conditional CSS and OS increased gradually over time based on continued survival periods after RNU. Accumulated survival time after surgery was critical for improving CS estimates assuming unchanged total survival times. For example, patients who already survived 2 and 3 years after surgery showed CSS probabilities of 82.0% and 87.5% at 3 and 2 years additional survivorships, respectively. Conversely, the typical estimate for 5-year CSS, calculated at the time of RNU, was only 73.7%. Of note, in the study by Ploussard et al. [10] patients with high-stage cancer showed much greater improvement of survival probabilities. Our results also showed that 5-year conditional CSS and OS probabilities increased markedly over time in patients with poor pathologic findings, including high pathologic T stage and high grade. In fact, 5-year CSS increased approximately 6% from baseline (86.1%) to 5-year survivorship (91.2%) in patients with low stage (pT1-2) cancer, while it increased approximately 32% from baseline (57.1%) to 5-year survivorship (88.9%) in those with high stage (pT3-4) tumor.

Due to the high progression and recurrence rates of UTUC, proper use of good prognostic parameters is required to determine optimal therapeutic strategies [21]. Various prognostic markers of UTUC have been identified in recent decades. Tumor stage, grade, and LN status are considered the most important predictive factors of prognosis in these patients [7,22]. Other pathologic predictors including carcinoma in situ and LVI, as well as several clinical factors such as age, tumor location, concomitant bladder cancer, and presence of hydronephrosis, have also been identified as potential prognosticators of survival outcomes [7,22]. However, considering CS, evidence that the effect of well-established prognostic factors decreases over time has accumulated, as determined by the diminished HR in multivariate analysis. For example, in bladder cancer patients, pT3-4 stage was significantly predictive for overall mortality from the time of baseline to 3 years after radical cystectomy [14]. Its HR decreased from 2.10 to 1.25 as time elapsed, and it eventually lost its statistical significance [14]. In UTUC, pT3-4 stage was identified as the prognostic factor for cancer-specific and overall CS, yet HR decreased from 4.0 to 1.7 for cancer-specific CS, as well as from 1.4 to 0.5 for overall CS [10]. Our results also indicated that well-known prognostic factors including pathologic T stage and tumor grade lost their impact on 5-year conditional CSS after 2-year survivorship post-surgery. For 5-year conditional OS, significance of pathologic T stage was not sustained after 2-year survivorship. However, we found prognostic factors for 5-year CS probabilities with HR increments over time. In particular, HR of LVI on 5-year conditional CSS significantly increased from 2.71 at the time of surgery to 3.75 at 4 years after surgery. HR of tumor location, whether ureter only or both pelvis and ureter, showed significant increments from 1.49 to 2.19 and from 2.15 to 32.80, respectively, for 5-year conditional CSS. For 5-year conditional OS, only age remained a significant predictor for all given survivorships. HR of age (≥ 65 years) increased gradually from 2.87 at baseline to 5.31 at 5 years after RNU, representing the significance of age on 5-year OS. Therefore, we believe that careful observation as well as proper adjuvant therapy during follow-up periods should be considered for patients with poor prognostic factors for 5-year CS, particularly those showing increased HR over time.

Like a previous report indicating that sex may not affect cancer-control outcomes in patients with UTUC [23], our results noted that patient’s sex was not associated with 5-year CS in all given survivorships. Obesity is also a potential prognostic factor of UTUC and has been associated with poor oncological outcomes such as disease-free survival [24]. However, we did not find a significant relationship between BMI and CS outcomes in this study. More interestingly, the presence of preoperative hydronephrosis has recently been suggested as the worst prognostic factor in UTUC patients [25], and our study also showed poor 5-year conditional CSS at baseline and 1-year survivorship after surgery for patients with hydronephrosis. However, at 2-year survivorship onwards, there was no difference in 5-year conditional CSS attributable to the presence of hydronephrosis. In addition, hydronephrosis was not identified as the predictive factor for 5-year CS in multivariate analysis. Instead, in patients with tumors located at the ureter or both the ureter and pelvis, high grade tumor and LVI tended to increase the risk of cancer-related mortality. Given the limited number of our study population, more convincing evidence is needed in order to properly determine the role of these prognostic factors on the CS probabilities of UTUC patients.

We acknowledge several limitations in the current study. This study was conducted retrospectively at a single tertiary medical center. Thus, our results should be validated by prospective and multi-center randomized clinical trials. Second, due to lack of data, not all potential prognostic factors, such as smoking status, or the type of surgery (open vs. laparoscopic) could be examined. Therefore, the impact of these parameters on CS should be determined in subsequent studies. Finally, we speculated that patients in better general health condition might have received adjuvant chemotherapy selectively; therefore, the role of adjuvant chemotherapy on CS probabilities remains to be determined.

In sum, 5-year conditional CSS, OS, and IVRFS increased in tandem with survivorship in patients with UTUC after RNU, while the actuarial survival rates decreased over time. In addition, patients with poor pathologic characteristics showed the highest increase of 5-year conditional CSS, OS, and IVRFS probabilities compared to their counterparts. Although there were no persistently significant prognostic factors for 5-year conditional CSS after RNU, several variables including high grade, tumor location, and LVI were largely associated with increment of risk over time. For 5-year conditional OS, only advanced age remained a significant predictive factor, with risk increasing in all given survivorships, and higher grade and tumor location tended to elevate risk. Our study offers valuable information on CS and its prognostic factors in UTUC patients after surgical treatment, and expands the current knowledge of risk stratification of these patients.