Characteristics of PSA Bounce after Radiotherapy for Prostate Cancer: A Meta-Analysis

The rate and characteristics of prostate-specific antigen (PSA) bounce post-radiotherapy remain unclear. To address this issue, we performed a meta-analysis. Reports of PSA bounce post-radiotherapy with a cutoff of 0.2 ng/mL were searched by using Medline and Web of Science. The primary endpoint was the occurrence rate, and the secondary endpoints were bounce characteristics such as amplitude, time to occurrence, nadir value, and time to nadir. Radiotherapy modality, age, risk classification, androgen deprivation therapy, and the follow-up period were extracted as clinical variables. Meta-analysis and univariate meta-regression were performed with random-effect modeling. Among 290 search-positive studies, 50 reports including 26,258 patients were identified. The rate of bounce was 31%; amplitude was 1.3 ng/mL; time to occurrence was 18 months; nadir value was 0.5 ng/mL; time to nadir was 33 months. Univariate meta-regression analysis showed that radiotherapy modality (29.7%), age (20.2%), and risk classification (12.2%) were the major causes of heterogeneity in the rate of bounce. This is the first meta-analysis of PSA bounce post-radiotherapy. The results are useful for post-radiotherapy surveillance of prostate cancer patients.


Introduction
Radiotherapy is a definitive treatment for prostate cancer (PCa). Prostate-specific antigen (PSA) is the biomarker used for post-treatment surveillance of PCa patients [1,2]. In curative cases, PSA levels decrease gradually over a period of more than five years after radiotherapy and reach a nadir. In a subset of patients, however, PSA levels fluctuate and show a temporal increase called the PSA bounce [3]. It is difficult to appropriately diagnose PSA increase post-radiotherapy as the bounce; therefore, the PSA increase post-radiotherapy can be the cause of severe anxiety in both PCa patients and clinicians. Misinterpretation may even endanger patients by leading to unnecessary salvage treatment in cases meeting the definition of biochemical failure. PSA bounce can occur in relation to various radiotherapy modalities, including external beam radiotherapy (EBRT), stereotactic body radiotherapy (SBRT), low dose-rate brachytherapy (LDR-BT), and high dose-rate brachytherapy (HDR-BT) [4,5]. As these radiotherapy modalities use different radiation sources, doses, and fractionation, as well as   A meta-analysis showed that the rate of PSA bounce for all studies was 31% (95% confidence interval (CI), 28-33%) ( Figure 2). The bounce rates according to modality were as follows: 34% (95% CI, 30-37%) for LDR-BT, 36% (95% CI, 29-42%) for HDR-BT, 22% (95% CI, 19-25%) for EBRT, 28% (95% CI, 23-32%) for SBRT, 28% (95% CI, 26-31%) for EBRT followed by boost irradiation, and 56% (95% CI, 47-64%) for carbon-ion radiotherapy ( Figure 2). For all studies, bounce amplitude was 1.3 ng/mL (95% CI, 1.1-1.4 ng/mL); time to bounce occurrence was 18 months (95% CI, 17-20 months); nadir value was 0.5 ng/mL (95% CI, 0.4-0.6 ng/mL); and time to nadir was 33 months (95% CI, 22-43 months). The results of the meta-analysis stratified by modality are summarized in Table 2, and the original forest plots are shown in Figures S1-S4. Nadir value was higher in bounce-positive patients than in bounce-negative patients for EBRT, SBRT, and CIRT, whereas time to nadir was greater in bounce-positive than in bounce-negative patients regardless of modality (Table 3).       The rate and characteristics of the bounce showed significant heterogeneity among the studies ( Table 2). To find the cause of the heterogeneity, we performed univariate meta-regression analysis. Age, radiotherapy modality, use of androgen deprivation therapy (ADT), and risk classification were selected as the covariates for meta-regression based on previous studies reporting that these factors affect the bounce kinetics [4,5]. The heterogeneity in the bounce rate was attributed to modality (29.7%), age (20.2%), and risk classification (12.2%) ( Figure 3A,B, Table 3). Regarding bounce amplitude, age was a significant cause of heterogeneity ( Figure 3C, Table 3). For time to bounce occurrence, modality was a significant cause of heterogeneity (Table 4).   The rate and characteristics of the bounce showed significant heterogeneity among the studies ( Table 2). To find the cause of the heterogeneity, we performed univariate meta-regression analysis. Age, radiotherapy modality, use of androgen deprivation therapy (ADT), and risk classification were selected as the covariates for meta-regression based on previous studies reporting that these factors affect the bounce kinetics [4,5]. The heterogeneity in the bounce rate was attributed to modality (29.7%), age (20.2%), and risk classification (12.2%) ( Figures 3A,B, Table 3). Regarding bounce amplitude, age was a significant cause of heterogeneity ( Figure 3C, Table 3). For time to bounce occurrence, modality was a significant cause of heterogeneity (Table 4).

Discussion
The strength of this study is that this is the first meta-analysis to investigate the characteristics of PSA bounce post-radiotherapy. We report the rate, amplitude, nadir, and time course of the bounce for different modalities including brachytherapy, EBRT, SBRT, and CIRT. We also report that the bounce occurs more frequently and with greater amplitude in brachytherapy than in EBRT, and a younger age is associated with a higher incidence and greater amplitude of the bounce. These findings have been extensively reported in mono-institutional studies, e.g., the large-scale study by Romesser [46], which were validated here for the first time by meta-analysis. From this standpoint, the results of the present study are useful for post-radiotherapy surveillance of prostate cancer patients to help oncologists and patients interpret temporal PSA increases post-treatment.
The limitations of this study, on the other hand, are the following. First, the studies analyzed were extremely heterogeneous regarding clinical factors such as dose, fractionation, bounce rate according to ADT usage, and risk classification, which was difficult to control in a meta-analysis design. In particular, the ADT strategy (i.e., the presence or absence of adjuvant or neoadjuvant use) should have affected post-radiotherapy PSA kinetics to a large extent, which was difficult to adjust by study design. Second, we were not able to analyze the PSA kinetics post-radiotherapy stratified by bounce positivity except for nadir and time to nadir. This was because extraction of the corresponding data from the original articles was technically impossible; i.e., the original articles did not contain the PSA kinetics data linked to specific clinical variables (e.g., age and risk) in a form that we can compute in the meta-analysis. Third, we were unable to perform multivariate meta-regression analysis because of the small number of studies. Fourth, most of the studies included had a retrospective design, and no randomized studies were identified. Finally, studies on particle therapy were rarely identified (i.e., one study on CIRT and no studies on proton therapy).
The molecular mechanisms underlying PSA bounce remain to be elucidated. Studies have shown that PSA is released from both tumor tissues and the normal prostate glands after irradiation [48]. Radiation-induced antitumor immunity may contribute to the release of PSA from tumor tissues. For example, Yamamoto et al. reported intra-tumoral infiltration of CD3-and CD8-positive lymphocytes in bounce-positive patients [56]. In the present meta-analysis, the bounce was more prevalent after brachytherapy and SBRT than after EBRT. In addition, the bounce rate for CIRT was strikingly high, although only one study was analyzed. These findings may be explained by the highly concentrated dose delivery by brachytherapy, SBRT, and CIRT compared with that of EBRT. Evidence suggests that a high, single-fractionated dose induces antitumor immunity efficiently [57], partially by promoting DNA damage response signaling [58]. In addition, the properties of carbon ions as high linear energy transfer radiation to efficiently induce antitumor immunity (e.g., induction of HMGB1 [59], OX40L, CD40, ICAM-1, and MHC-1, and suppression of PD-L1 [60]) might contribute to the high bounce rate for CIRT. Another possible explanation for the higher bounce rate associated with brachytherapy, SBRT, and CIRT is that the highly concentrated doses delivered by these modalities destroy the normal prostate glands more efficiently. Kirilova et al. showed an increase in metabolism indicative of inflammation in the normal prostate gland of patients experiencing bounce, which supports this notion [61].
In addition to modality, the meta-regression results indicated that younger age is associated with greater bounce occurrence and amplitude. This is consistent with the findings of the systematic literature review, in which 29 of the 50 papers analyzed identify younger age as a predictor of bounce. Yamamoto et al. suggested that this may be related to the higher immunocompetency in younger patients [56]. Further research is warranted to elucidate immunologic responses of PCa and the prostate glands after radiotherapy.

Endpoint Definition
The primary endpoint of this study was the rate of PSA bounce. Secondary endpoints included the characteristics of bounce, i.e., bounce amplitude, time to occurrence, nadir value, and time to nadir. Definitions of these endpoints are listed in Table S2.

Inclusion and Exclusion Criteria
The inclusion criteria were as follows: (i) an original clinical study reporting on radiotherapy for PCa; (ii) available rate of PSA bounce; and (iii) bounce defined as an increase in PSA over a cutoff of 0.2 ng/mL followed by a spontaneous decrease to or below the pre-bounce nadir [19]. The exclusion criteria were as follows: (i) manuscript written in languages other than English; (ii) full manuscript not available; (iii) subgroup analysis of a given reported cohort; (iv) follow-up shorter than 24 months.

Study Selection
A systematic literature search based on preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines [62] was performed on 20 March 2020, using two databases, Medline and Web of Science. The search strategy and population-intervention-comparison-outcome metrics [63] are described in Tables S3 and S4, respectively. The search results were combined using the bibliographic management software Mendeley Desktop version 1.19.4 (Mendeley, London, UK), and duplicates were eliminated. Two investigators (N.D.M.D. and T.Oi.) independently reviewed all records in the following three steps. In step 1, the titles of all records were reviewed to detect potentially relevant records. In step 2, the abstracts of all records that passed step 1 were reviewed to detect potentially relevant records. In step 3, the entire manuscripts of all records that passed step 2 were examined if they contained extractable data for the primary endpoint.

Data Extraction
From the studies identified in Section 4.3, two investigators (N.D.M.D. and T.Oi.) independently extracted the following data: primary and secondary endpoints, radiotherapy modality, age, risk classification [64], the use of ADT, and follow-up period.

Quality Assessment
Two investigators (N.D.M.D. and T.Oi.) independently confirmed that the methodological quality of the included studies was adequate based on the Quality Assessment Tool for Case Series Studies published by the National Heart, Lung, and Blood Institute-National Institute of Health, U.S. [65]. For Section 4.3, Section 4.4, and Section 4.5, decisions were made based on discussion by the two investigators to resolve disagreements on the review results.

Statistical Analysis
Radiotherapy modalities were classified into six groups as follows: iLDR-BT ( 103 Pd, 125 I, or 131 Cs), HDR-BT ( 192 Ir), EBRT (three-dimensional conformal radiotherapy or intensity-modulated radiation therapy), SBRT (using CyberKnife or linac), EBRT+boost (using LDR-BT, HDR-BT, or SBRT), and CIRT. Meta-analysis of bounce (binomial data) was performed using metaprop, a command of Stata (MP 13, StataCorp, College Station, TX, USA) [66]. Meta-analysis of the characteristics of bounce (continuous variables) was performed using metan, a Stata command. For the datasets that lacked the mean and standard deviation to be pooled, these values were estimated from the sample size, median, range, and/or interquartile range, as reported previously [67]. A random-effects model was used considering a high extent of inter-study heterogeneity examined using X 2 and I 2 statistics [68]. Meta-regression was performed to analyze the effect of clinical factors on inter-study heterogeneity in effect size using metareg, a Stata command [69]. To construct the metareg command for bounce rate, logit prevalence and its standard error were used [70,71]; for the remaining PSA kinetics outcomes, mean and standard error were used [72]. Results with a p-value < 0.05 were interpreted as significant.

Conclusions
This is the first study to report the results of meta-analysis and meta-regression of PSA bounce post-radiotherapy. Meta-analysis of 50 studies including 26,258 patients showed that the rate of PSA bounce for all studies was 31% (95% CI, 28-33%); bounce amplitude was 1.3 ng/mL (95% CI, 1.1-1.4 ng/mL); time to bounce occurrence was 18 months (95% CI, 17-20 months); nadir value was 0.5 ng/mL (95% CI, 0.4-0.6 ng/mL); and time to nadir was 33 months (95% CI, 22-43 months). The bounce occurred more frequently and with greater amplitude in brachytherapy than in EBRT. Univariate meta-regression showed that younger age is associated with a higher incidence and greater amplitude of bounce. These data will be useful for post-radiotherapy surveillance of PCa patients to help oncologists and patients interpret temporal PSA increases post-treatment.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/12/8/2180/s1. Table S1: The number of studies and patients according to radiotherapy modality, Table S2: Definition of endpoints, Table S3: Search strategy, Table S4: PICO metrics, Figure S1: Meta-analysis of the amplitude of prostate-specific antigen (PSA) bounce after radiotherapy. ES, effect size; CI, confidence interval; LDR-BT, low dose-rate brachytherapy; HDR-BT, high dose-rate brachytherapy; EBRT, external beam radiotherapy; SBRT, stereotactic body radiotherapy; CIRT, carbon ion radiotherapy. Figure S2: Meta-analysis of the time to occurrence of prostate-specific antigen (PSA) bounce after radiotherapy. ES, effect size; CI, confidence interval; LDR-BT, low dose-rate brachytherapy; HDR-BT, high dose-rate brachytherapy; EBRT, external beam radiotherapy; SBRT, stereotactic body radiotherapy; CIRT, carbon ion radiotherapy. Figure S3: Meta-analysis of prostate-specific antigen (PSA) nadir values after radiotherapy. ES, effect size; CI, confidence interval; LDR-BT, low dose-rate brachytherapy; HDR-BT, high dose-rate brachytherapy; EBRT, external beam radiotherapy; SBRT, stereotactic body radiotherapy; CIRT, carbon ion radiotherapy. Figure S4: Meta-analysis of the time to prostate-specific antigen (PSA) nadir after radiotherapy. ES, effect size; CI, confidence interval; LDR-BT, low dose-rate brachytherapy; HDR-BT, high dose-rate brachytherapy; EBRT, external beam radiotherapy; SBRT, stereotactic body radiotherapy; CIRT, carbon ion radiotherapy. Funding: This work was supported by Gunma University Heavy Ion Medical Center. This work was also supported by Grants-in-Aid from the Ministry of Education, Culture, Sports, Science, and Technology of Japan for programs for Leading Graduate Schools, Cultivating Global Leaders in Heavy Ion Therapeutics and Engineering.

Conflicts of Interest:
The authors declare no conflict of interest.