Are the Symptoms of Late Urogenital Toxicity After Radiotherapy for Localized Prostate Cancer Transient or Do They Worsen Over Time?

Our aim is to estimate incidence and prevalence rates as well as overall changes in the degree of late urogenital toxicity (UGT) over time in patients treated with 3D conformal radiotherapy (3DCRT) for localized prostate cancer (PC) and to explored possible predictive factors for late UGT. Ninety-four patients with localized PC, treated with 3DCRT with a total dose of 72Gy/2Gy/daily, were examined in this study. The median follow-up was 66 months. Many individual and clinical characteristics were explored as possible predictive factors for late UGT. Incidences of late UGT grade ≥ 1, ≥ 2 and grade 3 were 64.89% (61/94), 22.34% (21/94) and 7.4% (7/94), respectively. The prevalence rates of late UGT grade ≥ 1, grade ≥ 2 and grade 3 were 26.6%-36.3%, 3.2-13.3% and 1.06%-7.89%, respectively. Statistically signicant positive trend was found favoring an increase in the degree of late UGT over time (b=0.028; p<0.001). In the MVA, previous abdominal and pelvic surgery (PAPS) and acute UGT were signicantly associated with late UGT grade ≥ 1 (p=0.007) and (p=0.013), respectively. We can conclude that the degree of late UGT increases over time. PAPS and the occurrence of any grade of acute UGT have a signicant inuence on the development of any grade of late UGT.


Introduction
Technological advances from the end of the last century made it possible to introduce 3D conformal radiotherapy (3DCRT) into everyday clinical practice with the aim of increasing tumor control by dose escalation and reduction of urogenital toxicity (UGT) and gastrointestinal toxicity (GIT). However, randomized trials have shown no signi cant difference in the occurrence of acute or late UGT between conventional radiotherapy and 3DCRT [1,2]. More recently, intensity-modulated radiotherapy (IMRT), volumetric modulated arc therapy (VMAT) and precise image guidance radiotherapy (IGRT) were implemented to improve the conformality of the dose distribution, and thereby facilitate dose escalation for localized prostate cancer (PC) patients. Despite the advanced technique, the results of phase 3 doseescalation trials suggest that there were no signi cant differences between 3DCRT and IMRT for acute or late G2 + or G3 + UGT [3]. Moderate hypofractionation is safe and effective, but longer-term data are still lacking.
In addition to the dose-volume effect reported in the literature, other factors that may affect the incidence of acute and late UGT such as age, smoking, hormonal therapy, diabetes mellitus (DM), certain drugs, genetic and other biological markers have been mentioned in the literature [4][5][6][7][8][9][10][11][12].
The link between the acute and late effects of radiotherapy (RT) in PC has been observed by several investigators [13], and its existence con rmed by the results obtained in the Medical Research Council RT01 TRIAL [14].
Late side-effects of RT for the PC can be irreversible and may decrease the quality of life (QoL) but Schmid et al. demonstrates that the majority of late GIT and UGT symptoms after primary 3DCRT for PC are temporary [15].
In a previously published study [16] we found a statistically signi cant link of any grade of acute UGT with any grade of late UGT by MVA analysis. After considering the results obtained in our previous study, we wanted to examine whether, after a longer follow-up, this link would exist and whether clinical and patient-related factors could in uence the occurrence of late UGT and if so, which ones.
Furthermore, since we observed that symptoms of UGT may be temporary and may occur during followup, we wanted to investigate whether these symptoms worsen over time, remain of the same intensity, or decline in their degree. Therefore, we decided to evaluate changes in the degree of late UGT over time.
For this reason, we conducted a longitudinal study among localized PC patients.

Results
The median follow-up for all 94 patients was 66 (7-101) months. Tumor characteristics such as T-stage, initial PSA, grading and Gleason Score are shown in Table 1.  Table 2 and Table 3. Depending on the type of variables and the normality of the distribution, descriptiveness is presented as N (%), as ± SD or median (min-max).
* Ordinal multilevel regression models with a degree of late UGT as the dependent variable adjacent during observation.  All changes in the degree of late UGT at each time point for each patient during the follow-up are shown in Fig. 2.
As this is a partially retrospective study, we did not have any pre-treatment symptoms recorded in the medical history, so we took the symptoms recorded at the rst check-up during the rst week of radiation treatment.
The changes in the degree of late UGT at particular time points for each patent were examined with a linear mixed effects modeling approach and it was found that there is a statistically signi cant positive trend for an increase in the overall degree of late UGT over time (b = 0.028; p < 001). (Fig. 2) Individual characteristics of patients, clinical and treatment characteristics and dosimetric parameters which are investigated as possible predictors of late UGT by univariate analysis are shown in Table 2 and Table 3. Statistically signi cant predictors for late UGT according to the UVA were DM type II (p = 0.043), PAPS (p = 0.013) and acute UGT (p = 0.014).
Toxicity predictors from univariate analysis which were statistically signi cant were included in the multivariate multilevel ordinal regression model. In a multivariate multilevel ordinal regression model, PAPS (p = 0.007) and acute UGT during radiotherapy (p = 0.013) were statistically signi cantly associated with a higher degree of late UGT (Table 5).  Table 5.

Discussion
Despite the introduction of more advanced techniques such as IMRT, VMAT and IGRT in clinical practice, reductions in acute and late toxicity were observed only in rectal symptoms [17]. On the other hand, results of phase 3 dose-escalation trials suggest that there were no signi cant differences with 3DCRT or IMRT for acute or late G2 + or G3 + UGT [3]. To overcome the limitations in toxicity reporting by using incidence rates, other approaches were proposed in the literature.
Schmid et al. suggests that the combination of incidence and prevalence rates provides a more comprehensive view on the complex issue of late side effects because incidence rates may lead to misinterpretation or overestimation. Their study demonstrates that the majority of late GIT and UGT after primary external beam RT for PC are temporary [15].
In contrast, Olson et al. [19] found that the temporal pattern of symptom occurrence varied, but the prevalence generally changed moderately over time.
In our study the prevalence rate of late UGT grade ≥ 1 increase over time while the prevalence rate of late UGT grade ≥ 2 uctuates.
The increase and decrease of prevalence rate at different time points during follow-up suggests that in some patients the symptoms may be diminished and/or be temporary, while in other patients, symptoms may be developed "de novo" or pre-existing symptoms may have worsened.
Therefore, we wanted to take into account all changes in the degree of late UGT at all-time points for each patient, in order to determine how the overall grade of late UGT changes over time.
Having in mind the uctuation of the prevalence of late UGT symptoms, we expected a decrease in the degree of late UGT symptoms, which would mean that these symptoms are temporary and that they may heal over time.
Surprisingly, we found that there is a statistically signi cant positive trend for an increase in the overall degree of late UGT over time (b = 0.028; p < 001) (Fig. 2). Based on this result, we can conclude that if we follow these patients long enough (longer than 5 years), the degree of late UGT will be higher.
As far as we know, this is the rst time in the literature that late toxicity is assessed taking into account all changes in the degree of late UGT at all -time points for each patient, in order to determine how the overall degree of late UGT changes over time.
The second goal of our study was to estimate the possible in uence of individual and clinical characteristics on the occurrence of late UGT.
In RADAR study, several parameters affecting the development of speci c urinary symptoms after prostate cancer radiotherapy were identi ed. Baseline symptoms and non-insulin-dependent diabetes mellitus (NIDDM) were found to have sustained impacts across all time point, while age and PC1 (PC1 has high correlation to the mean dose) were repeated in more than one time point. Other associations found to be signi cant are not repeated at different time points suggesting poorer or temporary impacts [4].
As this is a partially retrospective study, we did not have any pre-treatment symptoms recorded in the medical history so we could not examine the correlation between baseline symptoms and the occurrence of late UGT.
In the current study NIDDM in UVA was statistically signi cant (p = 0.043) but in MVA was not signi cant (p = 0.09). Similar results we got in our former study: in UVA, NIDDM was statistically signi cant (p = 0.028) but in MVA was border line signi cant (p = 0.056) [16].
In the literature, we found studies which reported a correlation between PAPS and the symptoms of late GIT [20], but we did not nd any which reported correlations between PAPS and the advent of late UGT.
The authors who found a correlation between previous surgery and GIT symptoms assumed that the in ammatory response could increase due to surgery even though the surgical injury was outside the irradiated volume and therefore radiation sensitivity increases. It has also been observed that cytokine expression increases due to surgery, which could lead to a more pronounced in ammatory response of the gastrointestinal mucosa to radiation [21,22].
Christensen et al. [23] observed that the expression of IL1 and IL6 was associated with an increased probability for acute UGT and GIT, respectively.
In contrast, Bedini et al. [24] in their study showed that IL-1b and IL-6 were fairly undetectable in most patients and presented with very low variability among patients. This preliminary study identi ed a correlation between CCL2 levels at the end of radiotherapy and basal CCL2, age, and previous abdominal/pelvic surgery (PAPS), suggesting a different response to RT in older patients and in patients with pretreatment abdominal surgery.
In our prior study PAPS was a statistically signi cant factor for grade ≥ 2 of acute UGT (p = 0.012) but not for late UGT (= 0.108) [16]. Interestingly, in our current study PAPS was a statistically signi cant factor (p = 0.007) for late UGT grade ≥ 1. To our knowledge, we are so far the only one to report a statistically signi cant association between PAPS and acute and late UGT. The relationship between acute and late effects of RT remains controversial.
It was traditionally thought that there was no link between the acute reaction to radiation exposure and the late effects in other organs and tissues, and that acute and late toxicity were the consequences of different independent pathological mechanisms. It has been shown that the acute and late toxicity do not share a similar set of predictors [25,26]. More aggressive RT protocols can result in aggravation, i.e. an increase in severity and duration, of acute radiation effects [27]. The non-healing response can progress directly into a late effect.
In other examples of consequential late effects, there is clinical healing of acute effects and a symptom-free period before the independent development of late effects in the same tissue [26].
In the Medical Research Council RT01 TRIAL [14], acute bladder symptoms correlated with the presence of all late bladder symptoms scores but only with a change in the score for the symptoms of urinary frequency and hematuria. This therefore suggests that patients with symptoms before treatment will probably suffer acute bladder toxicity during treatment and may develop late symptoms.
[28] found a strong dependence of urinary/bowel QoL after radiotherapy on urinary/bowel QoL before radiotherapy. In contrast to absolute scores, QoL score changes (relative to baseline scores) did not correlate with pretreatment scores. They concluded that long-term changes could be well predicted by acute changes.
In our current study applying the multivariate multilevel ordinal regression model we found, apart from PAPS, acute UGT during radiotherapy to be statistically signi cantly associated with a higher degree of late UGT (p = 0.013). The conclusions coincide with the results from our previous study (p = 0.024).
Based on the obtained results, we can conclude that in the examined group the degree of late UGT increased with time. PAPS is a statistically signi cant predictive factor for the onset of late UGT symptoms. There is a statistically signi cant correlation between the frequency and the severity of symptoms of acute and late UGT.

Methods
Patients were treated at a single institution with 3DCRT in the period between September 2009 and September 2013 and a total 94/225 patients seen in this period have been included. Study criteria and treatment schedule have already been reported [16,29]. Patients with histologically proven, localized PC with low or intermediate risk, were included. Patients with estimated risk of lymph node involvement > 15% according to Roach formula and with signs of lymph-nodes positive on CT (computerized tomography) / MRI (magnetic resonance imaging) and with previous pelvic irradiation were excluded.
Neo-adjuvant hormonal therapy was not an exclusion criterion. Informed consent was obtained from participants in the study and the institution's ethics committee approved the study.
Patients were divided into two dose volume groups according to the estimated risk of seminal vesicle involvement according to the Roach formula [30]. In the rst dose volume group CTV (clinical target volume) included the whole prostate. In the second dose volume group CTV1 encompassed the whole prostate with the entire seminal vesicle and in second phase CTV2 encompassed the whole prostate only. The dose that was prescribed to the ICRU reference volume covered at least 95% of PTV (planning target volume): 72Gy to PTV, 66Gy to PTV1 and 6Gy to PTV2. All patients received total dose of 72Gy/2Gy daily fraction. The safety margin around the CTV and CTV1 were de ned according our institutional protocol [16,29]. Bladder and rectum dose constraints were taken from Emmami and coworkers rst [31]. Since the end of 2010 we have used Quantec recommendations for dose limitation [32].
After the completion of RT patients were seen every 3 months during the rst year, then every 6 months and after 5 years annually. Follow ups were done until the occurrence of death or the end of the study (30.4.2018). All symptoms of late UGT were recorded during follow up visits and graded according to the EORTC/RTOG scoring scale slightly modi ed by Peters and coworkers [25].
In our previously published study [16] our primary goal was the assessment of acute and late UGT after a median follow-up of 27 months while the second was the estimation of the possible in uence of individual/clinical characteristics on the occurrence of acute and late UGT.
Later, we continued to monitor these patients and decided to re-evaluate late UGT after a median followup of 66 months and determine whether the same or some other individual and clinical factors in uence the onset of late UGT.
In our previous study we estimated incidence of acute and late UGT only. Beside the incidence of late UGT de ned as maximal grade of any symptoms at any times during follow-up, we decided to estimate the median time of the onset of late UGT grade 1 ≤ and a prevalence rate for every grade of late UGT at each time point during the follow-up. In addition, we wanted to estimate whether the grade of late UGT worsened over time or whether the symptoms were temporary, taking into account all changes in the degree of late UGT at each time point for each patient since the last check-up after the end of RT.

Statistical analyses
Depending on the type of variables and the normality of the distribution, results were presented as frequency (percent), median (range) and mean ± standard deviation. Changes in the degree of late UGT during the follow-up were examined with a linear mixed effects modeling approach using R package lme4 version 1.1-23. For the modeling of the relation of the ordinal dependent variable (late UGT) in repeated measurements with potential predictors, a ordinal regression was used. Toxicity predictors from univariate analyzes which were statistically signi cant were included in the multivariate ordinal regression model. All p-values less than 0.05 were considered signi cant. Statistical data analysis was performed using IBM SPSS Statistics 22 (IBM Corporation, Armonk, NY, USA) and R-3.6.3 software (The R Foundation for Statistical Computing, Vienna, Austria). Figure 1 Time of the onset of any symptom of late UGT grade ≥1 After the end of RT, patients were followed up and symptoms of late UGT grade ≥1 were estimated every 6 months until 96 months. The estimated median time of the onset of symptoms of late UGT grade ≥1 is 18 months