Combination of 99mTc-labeled PSMA-SPECT/CT and Diffusion-Weighted MRI in the prediction of early response after carbon ion therapy in prostate cancer: a prospective study


 Background The purpose of this study was to assess the potential of 99mTc-labeled PSMA-SPECT/CT and Diffusion-Weighted Image (DWI) for predicting treatment response after carbon ion radiotherapy (CIRT) in prostate cancer. Methods We prospectively registered 26 patients with localized prostate cancer treated with CIRT. All patients underwent 99m Tc-labeled PSMA-SPECT/CT and multiparametric MRI before and after CIRT. The tumor/background ratio (TBR) and mean apparent diffusion coefficient (ADC mean ) were measured on the tumor and the percentage changes between 2 time points (ΔTBR and ΔADC mean ) were calculated. Patients were divided into two groups: good response and poor response according to clinical follow-up. Results The median follow up time was 38.3months. The TBR was significantly decreased ( p =0.001), while the ADC mean was significantly increased compared with the pretreatment value ( p <0.001). The ΔTBR and ΔADC mean were negatively correlated with each other ( p = 0.002). On ROC curve analysis for predicting treatment response, the area under the ROC curve (AUC) of ΔTBR (0.867) for predicting good response was higher than that of ΔADC mean (0.819). The AUC of combined with ΔTBR and ΔADC mean (0.895) was higher than that of either ΔADC mean or ΔTBR alone. The combined use of ΔTBR and ΔADC mean showed 91.4% sensitivity and 95.2% specificity. Conclusions Our preliminary data indicate that the changes of TBR and ADC mean maybe an early bio-marker for predicting prognosis after CIRT in localized prostate cancer patients. In addition, the ΔTBR was a more powerful prognostic factor than ΔADC mean in prostate cancer treated with CIRT.


Background
External beam radiotherapy (EBRT) is commonly used as a curative strategy for man diagnosed with localized prostate cancer. Because there are some critical organs at risk (OARs) surrounding the prostate, it is very di cult to deliver a high dose to prostate while minimizing the radiation dose to the adjacent OARs, such as, rectum and bladder. Carbon ion radiotherapy (CIRT) is considered to be the most advanced and promising radiotherapy technique. The physical and biological advantage of carbon ion that allow for the application of a high dose to the prostate while maintaining a steep gradient to the surrounding normal tissue [1]. Shanghai Proton and Heavy Ion Center (SPHIC) started CIRT for prostate cancer in 2014. Until November 2019, there are more than 200 prostate cancer patients have been treated at our center. However, CIRT is a novel and to date not thoroughly investigated technique. Until now, there are only about 3000 patients with prostate cancer received CIRT around the world [2,3]. So, the experience for CIRT is very limited for prostate cancer. In addition, Prostate cancer often has a long natural history, it often takes a decade or more to judge the therapeutic e cacy of prostate cancer. An early prediction of treatment response may allow for therapeutic optimization; such as radiation dose modi cation. Thus, we assessed whether molecular imaging can act as an early predictive tool for these patients' outcome after CIRT.
Prostate-speci c membrane antigen (PSMA), a unique membrane-bound type II glycoprotein, is known to be over-expressed in almost all prostate cancer cells, with only 5-10% primary prostate cancer not having PSMA expression [4]. PSMA -targeted molecular imaging have been approved to be a better diagnostic tool in patients with prostate cancer than conventional imaging [5]. However, the clinical data focusing on the predictive value of PSMA imaging for primary localized prostate cancer patients treated with radiotherapy (especially CIRT) was limited. In addition, our primary study shown that apparent diffusion coe cient (ADC) vales may be an useful imaging bio-marker for early assessment of therapeutic response of prostate cancer to CIRT [6].
To our knowledge, there were limited studies addressed the relationship between PSMA-targeted imaging and diffusion-weighted image (DWI) of prostate cancer [7]. And there was no data comparing the predictive value of these two functional imaging for prostate cancer patients underwent CIRT. Therefore, we designed a prospective clinical trial to evaluate and compare the potential value of 99m Tc-labeled PSMA-single photon emission computed tomography/computed tomography (SPECT/CT) and DWI for predicting outcome after CIRT in prostate cancer.

Patients:
This study was a phase I study evaluating the CIRT for localized prostate cancer in dose escalation at SPHIC. Prior to screening procedures and treatment, signed informed consent was obtained from all patients. This trial is registered with ClinicalTrials.gov, number NCT02739659.
Eligible men were required to be aged 20-85 years and have Karnofsky Performance Score ≥ 70, pathologically con rmed adenocarcinoma of the prostate. And localized patients (T1-4 N0 M0, AJCC 7th ) without pelvic lymph nodes or distant metastasis planned for CIRT were eligible for this study. PSMA-SPECT/CT and magnetic resonance imaging (MRI) examinations were conducted at two time points: before and immediately (1 week after the last irradiation) after CIRT. And the interval between SPECT/CT and MRI examination was less than one week. Men who had received prior chemotherapy or radioisotopes for prostate cancer were excluded. The study protocol was approved by all institutional ethics boards.

MRI acquisition:
All MRI examinations were performed using a 3-T MR system (Magnetom Skyra Simens) equipped with a phased-array coil at SPHIC. T1-weighted, T2-weighted, and DWI sequences were acquired, but only DWI sequence was used for analysis in this study. Parametric maps of ADC values were automatically measured by the image software with the use of the two b values.

Images analysis:
SPECT/CT image readout was performed on a work Station and software (Xeleris, General Electric, Waukesha, WI). Two board-certi ed specialists in nuclear medicine, who blinded to patient related medical data, independently read all datasets and resolved any disagreements by consensus. Areas of abnormal tracer uptake within the prostate gland were determined and recorded. For semi-quantitative analysis, the tumor/background ratio (TBR) was calculated for each visually detected lesion or other tissue within each lobe (right / left) of the prostate using the quotient of maximal counts within a circular region-of-interest and mean counts within the obturator muscle [9].
Similarly, all acquired MRI was analyzed and interpreted by two radiologists independently using the manufacturer supplied software (Simens Healthcare). For calculating the mean ADC (ADC mean ) value of tumor, the region of interest (ROI) was manually drawn by two radiologists on single axial image where the tumor shows the maximum dimension. If the two readers disagreed about the exact tumor localization on the MR images, consensus was reached using information from SPECT/CT image or pathological results of biopsies.
To assess the changes of TBR and ADC values after CIRT, percentage changes in TBR and ADC mean were Evaluations of patient outcomes: After the treatments, these patients were followed up every 3 months. Physical examinations and prostate speci c antigen (PSA) were performed at each visit. Naik's report [10] showed that 6 months post-treatment PSA > 0.1 ng/mL in prostate cancer patients treated with EBRT was associated with worse biochemical relapse free survival (bRFS), distant metastasis free survival (DMFS), and prostate cancer speci c mortality (PCSM). Therefore, clinical outcomes were divided into two groups: good response (PSA ≤ 0.1 ng/mL at 6 months after therapy) and poor response (PSA > 0.1 ng/mL at 6 months after therapy).

Statistical Analysis
The data were analyzed using SPSS statistical software (version 22.0; IBM Corp.). All continuous variables were tested for normal distribution using the Kolmogorov-Smirnov test. Clinical data and parametric data from images were compared using the χ 2 test for categorical data, the Student t test for continuous data, and the Mann-Whitney test for nonparametric analysis. We calculated the Spearman rank-order correlation coe cient to characterize correlation strength between imaging features (TBR and ADC mean ) and clinical features (GS, PSA). The correlation between percentage change in TBR (ΔTBR) and ADC mean (ΔADC mean ) were also evaluating using the spearman correlation coe cient. We used receiveroperating -characteristic (ROC) curves and calculated areas under the curves (AUCs) for each parameter. The combinations of parameters that distinguished good responders from poor responders were tested by multi-ROC curve analysis. For all statistical comparisons, a p value of less than 0.05 was considered signi cant.

Results
Clinical characteristics and treatment outcomes A total of 30 consecutive patients with biopsy con rmed prostate cancer being considered for CIRT were prospectively recruited at SPHIC between Apr. 2016 and Mar. 2017. Of them, 4 patients were excluded due to not performed PSMA-SPECT/CT before or after CIRT. Finally, 26 patients with localized prostate cancer, who completely received the CIRT and had adequate 99m Tc-labeled PSMA-SPECT/CT and multiparametric MRI image information at our institution were analyzed in this study.     (Fig. 1). The ΔTBR and ΔADC mean were negatively correlated with each other (Spearman correlation coe cient, -0.586; p = 0.002) (Fig. 2).

Discussion
Our study demonstrated that ΔTBR and ΔADC mean after CIRT were negatively correlated. And both of them provide a noninvasive imaging biomarker for the early assessment of treatment response (Figs. 4 and 5). In addition, the ΔTBR was a more powerful prognostic factor than ΔADC mean in prostate cancer treated with CIRT. The combined use of ΔTBR and ΔADC mean served to better distinguish the good responders from poor responders.
PSMA-based molecular imaging has rapidly emerged as a potential new standard of care for imaging prostate cancer, with images demonstrating relevant protein expression levels [11]. It is reported that PSMA expression is a relevant factor for tumor aggressiveness [12]. However, it is still unclear whether a receptor-targeting radiopharmaceutical, instead of a metabolic tracer, would have the same value for treatment response monitoring in prostate cancer [13]. Hillier's study laid the theoretical foundation for response evaluation, which found that 123 I-MIP-1072 may allow monitoring of tumor progression in patients before, during, and after chemotherapy [14]. Seitz's preliminary results concluded that the concordance rate was high between biochemical response and 68 Ga-PSMA PET/CT response in patients with metastatic prostate cancer undergoing chemotherapy [15]. Similar results were con rmed in another study using 99m Tc-MIP-1404 SPECT/CT [16], suggesting a possible role of that imaging tool for monitoring treatment in metastatic prostate cancer. However, these clinical studies were all focused on metastatic prostate cancer. There is no data reported in literature for the prediction of radiotherapy response in primary prostate cancer with PSMA ligands imaging. Although PSMA imaging has been used in prostate cancer radiotherapy, they always research whether this novel nuclear imaging modality can be used to direct a local boost to the lesions or to plan salvage radiotherapy [17,18]. Therefore, we designed a prospective clinical trial to explore the value of PSMA imaging in the early evaluation of CIRT and got very good results. The results showed that TBR signi cantly correlated with baseline PSA (p = 0.002). The change of TBR signi cantly decreased in the patients with good response (p = 0.003), but not in the patients with poor response (p = 0.325). Moreover, ΔTBR showed very high sensitivity (80.0%) and speci city (95.2%) in predicting the response of carbon ion radiotherapy. These indicated that the 99m Tc-labeled PSMA-SPECT/CT could serve as an early biomarker for predicting prognosis after CIRT and in uence planned clinical management in a high proportion of patients with prostate cancer.
Our previous study showed that the mean ADC value of prostate tumor was signi cantly increased after CIRT [6]. In a similar study, Wolf et al [19] also found that particle therapy induced a measurable and continuous increase in the ADC value of prostate cancer during and after therapy. In this study, the change in tumor of the ADC value after CIRT were consistent with these previous studies. The increase of ADC value after CIRT likely indicates the alterations in water diffusivity due to necrosis and apoptoticinduced cell death.
Positron Emission Tomography/Computed Tomography (PET/CT) and DWI have complementary roles in the assessment of prostate cancer [20]. Recently, combined PET/MRI imaging systems have been explored in the clinic, and literature describing the initial experiences with PSMA -PET/MRI imaging in prostate cancer is already available [21]. However, relatively little early date is available regarding PSMA -PET and MRI for assessment of therapeutic response in prostate cancer. In this study, the TBR was inversely correlated with ADC mean before CIRT, however, there was no correlation between TBR and ADC mean after CIRT. Interestingly, the ΔTBR and ΔADC mean were negatively correlated with each other.
Consistent with our data, previous studies in osteosarcoma have shown a signi cant negative correlation between ΔSUV (standard uptake value) and ΔADC mean after chemotherapy [22]. This is partly explainable because the different effectiveness of treatment in uenced by radiosensitivity or chemosensitivity between these patients. Effective treatment may substantially increase tumor necrosis and apoptosis. Accordingly, greater changes of both ADC and TBR/SUV values after treatment potentially suggest that the tumors are more sensitive to treatment.
Prostate cancer often has a long natural history, so it can take many years to determine whether a new treatment strategy for prostate result in improved prostate cancer patients' survival. Combined use of 99m Tc-labeled PSMA-SPECT/CT and DWI imaging modalities can provide various biological information and thus may overcome the limitations of SPECT/CT and DWI. The higher predictive power achieved by a combination of DWI and 99m Tc-labeled PSMA-SPECT/CT parameters enable early predict the treatment response and then optimize the prescription dose, fraction size or hormone therapy during time. Our study also indicated that PSMA high expressed area is potential biological target volume for radiotherapy dose escalation in the future.
In this study, endocrine therapy combined with CIRT should be taken into account. All of those patients in our study were concurrently treated endocrine therapy with CIRT. So the change of these image ndings might contribute to endocrine therapy beyond the CIRT. However, in our clinical practice, Dose-escalated radiotherapy RT with endocrine therapy is a standard de nitive treatment of localized prostate cancer. So our results still have clinical implication.
There were several limitations in our study. First, the number of enrolled patients was relatively small and the follow-up duration was short. The short follow up of 38.3 months is not su cient to identify the true recurrent patients with Phoenix criteria. Primary study [10] showed that 6 months post-treatment PSA > 0.1 ng/mL in prostate cancer patients treated with concurrent radiotherapy is associated with worse bRFS, DMFS, and PCSM. Therefore, our study use PSA response at 6 months after therapy as our clinical outcome endpoint instead of biochemical relapse free survival or overall survival rate. Meanwhile, by the date of follow-up, 2 of the 5 patients in the poor response group had progressed, 1 patient had biochemical recurrence and another one had bone metastasis. These results indicated that this surrogate endpoint was credible. A future study with larger population and longer follow-up is necessary for validating our preliminary results. Second, we investigated the utility of tumor ADC mean and TBR out of  Scatter-plot showing relationship between Δ TBR and Δ ADCmean after CIRT.

Figure 3
ROC curves used to evaluate good response to CIRT with Δ TBR, Δ ADCmean and combined used of Δ TBR and Δ ADCmean, AUC of Δ TBR (0.867) was higher than that of Δ ADCmean (0.819). AUC increased with combined used of Δ TBR and Δ ADCmean (0.895).

Figure 4
A patient with pathology con rmed prostate cancer of Gleason score 4+4 (T3bN0M0, very high risk) who showed good response after CIRT. PSMA uptake (A) was visualized on the prostate with TBR of 34.9.