Background
Worldwide, prostate cancer (PC) is considered the second most frequently diagnosed cancer in men and the fifth leading cause of cancer death [
1]. Recently, radiolabeled prostate-specific membrane antigen (PSMA) ligands such as
68Ga-PSMA-HBED-CC have been introduced as a promising radiotracer for the PET imaging of PC [
2]. PSMA is a transmembrane protein that is significantly overexpressed in most prostate cancer cells [
3]. Different studies demonstrated that
68Ga-PSMA–PET enables imaging with a higher specificity and sensitivity regarding the detection of metastases, compared to current standard imaging (CT, MRI and bone scintigraphy) and other PET tracers such as
18F-Choline [
4‐
7]. It also improves detection of metastatic lesions at low serum PSA levels in biochemically recurrent prostate cancer [
8].
The liver is considered to be the third most common site for systemic metastases in PC (25%), after bone (90%) and lung (46%), according to autopsy studies [
9]. The prevalence of clinical liver metastases in retrospective studies was 4.3 and 8.0% [
10,
11]. Liver metastases typically occur in systemic, late stage, hormone refractory disease [
10]. However, there are reports of patients with liver metastases as the first site of metastatic disease and the liver representing the only metastatic site [
10,
12,
13]. Especially in this patient collective, early and reliable detection of liver metastases is of high clinical importance for accurate staging and therapy planning.
There is evidence that in PC, liver metastases are frequently associated with neuroendocrine characteristics; in a prospective study of 28 patients with liver metastases, Pouessel et al. measured increased levels of the neuroendocrine serum markers chromogranin A and neurone-specific enolase in 84 and 44% of the patients, and out of six patients with a pathological analysis, two had neuroendocrine metastases [
10]. Neuroendocrine transdifferentiation might lead to loss of PSMA-expression and therefore impede the visualization of liver metastases in
68Ga-PSMA-PET [
14]. Furthermore, the relatively high background activity of the liver might also affect the visibility of liver metastases in
68Ga-PSMA-PET [
14]. Imaging of hepatic PC metastases in
68Ga-PSMA-PET has been reported in case reports, but not been systematically researched in a larger cohort of patients [
12,
15‐
18].
Therefore, the aim of this study was to investigate the 68Ga-PSMA-PET imaging properties of liver metastases in PC patients.
Methods
Study population
For this retrospective study, we obtained approval from our institutional ethics review board. We extracted 739 consecutive patients with confirmed prostate cancer from our local database who underwent at least one
68Ga-PSMA-PET/CT between September 2013 and April 2017. Out of these, we identified eighteen patients with liver metastases, according to the criteria described below. Prostate cancer was histologically proven in all patients. Only patients with no other known type of cancer but PC were included. All available additional information from clinical records were obtained. Patients’ characteristics are summarized in Table
1. Gleason score (GS) was available in eleven, therapy information only in thirteen and PSA level only in twelve patients.
Table 1
Characteristics of the study collective of PC patients with liver metastases
Age (years) | 70.1 ± 8.5 | 71.0 (54.5–81.4) | |
PSA (ng/ml) | 556.3 ± 1398.4 | 124.6 (0.01–4962.0) | |
Gleason score | | 9 (6–10) | |
Therapy | | | 13 |
RP | | | 7 (53.8) |
RT | | | 6 (46.2) |
ADT | | | 11 (84.6) |
CTX | | | 7 (53.8) |
177Lu-PSMA | | | 4 (30.8) |
Positron emission tomography tracer
68Ga was eluted from a conventional
68Ge/
68Ga radionuclide generator (Eckert & Ziegler Radiopharma GmbH, Berlin, Germany) and compounded with PSMA-HBED-CC (ABX GmbH, Radeberg, Germany) according to the method described previously [
19,
20].
Imaging protocol
PET/CT imaging was performed 75.8 ± 18.2 min after intravenous injection of 120.5 ± 25.7 MBq of
68Ga-PSMA. PET scans were acquired using a Gemini Astonish TF 16 PET/CT scanner (Phillips Medical Systems) in 3D acquisition mode [
21]. Axial, sagittal and coronal slices were reconstructed (144 voxels with 4mm
3, isotropic). Before PET scan, a low-dose CT was performed for anatomical mapping and attenuation correction (30 mAs, 120 kVp). Each bed position was acquired for 1.5 min with a 50% overlap. In case contrast-enhanced CT (CE-CT) was performed, 80–120 ml of contrast agent (Ultravist® 370, Bayer Schering Pharma, Berlin, Germany) was injected intravenously with a delay of 70 s for the venous phase.
Imaging analysis
Two experienced observers analyzed the PET/CT scans using Visage 7.1 (Visage Imaging GmbH, Berlin, Germany). For the diagnosis of metastases, all available imaging studies including all imaging modalities (CT, MRI,
68Ga-PET) of the patients were taken into consideration. At least two of the following four criteria had to be fulfilled for the diagnosis of liver metastasis: (I) CT imaging with low-to-isoattenuating masses [
22]; (II) MRI with typical presentation of liver metastases according to guidelines [
23]; (III) high focal uptake of
68Ga-PSMA in PET distinctively above normal heterogeneity; (IV) new appearance or significant change of size of lesions according to the RECIST 1.1 criteria compared to previous studies within the same modality with a minimum follow-up interval of six months [
24]. Patients with signs of a malignancy other than PC were excluded. Out of 23 patients with suspected liver metastases, five patients dropped out because they did not fulfill these criteria. Overall 18 patients with hepatic metastases were identified out of 739 patients. Among these, criteria I was fulfilled by all patients, criteria II by four patients, criteria III by 16 patients and criteria IV by 12 patients. Maximum ten metastases per patient were analyzed. In case a patient was imaged more than once, only the most recent
68Ga-PSMA-PET scan was included in this study. As a result, 103 liver metastases were analyzed as part of this study. The sizes of metastases were measured based on the CT scan. Regarding the evaluation of the radiodensity, two groups were formed. One group in which only unenhanced CTs were available (five patients) and another group in which contrast-enhanced CTs were available (13 patients).
To normalize standardized uptake values (SUV) for body weight, they were calculated by the software using with the equation SUV = Ctis/Qinj/BW, where Ctis is the lesion activity concentration in MBq per milliliter, Qinj is the activity injected in MBq, and BW is the bodyweight in kilograms. For PET data quantification, a two-dimensional region of interest (2D ROI), as well as a three-dimensional region of interest (3D ROI), were defined. 68Ga-PSMA-HBED-CC uptake was quantified using maximum standardized uptake values (SUVmax). All values were recorded in the transaxial, attenuation-corrected PET-slice representing the greatest extent of the respective lesion. Regions of interest were defined manually in freehand mode avoiding the periphery of lesions to minimize partial volume effects. SUVmax of the healthy liver was measured in a region with minimal irregularities. An SUVmax-lesion-to-background ratio (LBR) was calculated for all metastases in 3D ROI, using the formula \( LBR=\frac{SUV_{max}\ of\ metastasis}{SUV_{max}\ of\ liver} \). Any tracer uptake 20% or more above liver uptake was considered PSMA-positive, any tracer uptake below that was considered PSMA-negative. The readers were blinded to the results of other diagnostic procedures and the clinical history of the patients.
Statistical analysis
The descriptive statistics are reported as mean, median and/or range when applicable. Nonparametric statistical tests were used as the data contained several outliers. The Mann-Whitney U test was used for the comparison of SUVmax values and mean radiodensity values (HUmean) between the healthy liver and liver metastases. SUVmax values in 2D and 3D ROI were compared using the Wilcoxon signed-rank test. To determine the relationship between SUVmax and size of lesions, patients’ age and PSA serum level, a Spearman’s rank correlation was used. A binomial test was run to evaluate the distribution of liver metastases among the hepatic lobes. The significance level was set to α < 0.05. Statistical analyses were conducted with SPSS 23 for Mac (IBM Corp, Armonk, NY).
Discussion
This study evaluated the imaging characteristics of liver metastases in 68Ga-PSMA-PET. It was demonstrated that the majority of liver metastases highly overexpress PSMA and is therefore directly detectable by 68Ga-PSMA-PET. For the analysis of PET images, it has to be taken into account that also a significant portion of metastases can only be detected indirectly, as these metastases are PSMA-negative.
68Ga-PSMA-PET/CT has demonstrated potential to improve the initial staging, lymph node staging, and detection of recurrence of PC, even at low PSA levels. Several studies have indicated that
68Ga-PSMA-PET is more accurate compared to other tracers as such as
18F-choline [
25]. So far, the imaging properties of liver metastases in
68Ga-PSMA-PET have not been systematically researched.
In our cohort, liver metastases were present in 2.4% of patients who underwent
68Ga-PSMA-PET. This was lower compared to the prevalence reported by other studies, likely as a result of the different study designs and the limited sensitivity of PET for the detection of small (< 1 cm) metastases [
10,
11]. In our study population, the majority of patients demonstrated PSMA-positive hepatic metastases, while only a small number of patients demonstrated PSMA-negative or mixed metastases. An explanation for the difference of
68Ga-PSMA-HBED-CC uptake in liver metastases could be the diversity of phenotypes in metastases, predominantly the neuroendocrine trans-differentiation. In PC, liver metastases are frequently associated with neuroendocrine characteristics as well as with advanced state in systemic disease [
10]. It is thought that the degree of neuroendocrine trans-differentiation increases with disease progression and in response to ADT [
26]. A pronounced elevation of neuroendocrine serum markers such as neuron-specific enolase and chromogranin A has been demonstrated in patients with long duration of ADT [
27]. Autopsy studies have confirmed the phenotypic heterogeneity of end-stage metastatic prostate cancer [
28,
29]. A large part of neuroendocrine prostate cancer cells does not express generic PC biomarkers including P501S, PSMA, and PSA [
30]. This is consistent with the histopathologic finding in one of our study patients with PSMA-negative liver metastases, in whom liver and prostate biopsy were performed. Histopathology of the metastasis revealed an infiltration of the liver with neuroendocrine carcinoma cells, which were positive for the neuroendocrine biomarker CD56, but negative for PSA, PSMA and androgen receptor. In the same patient, histopathology of the prostate tissue exposed an acinar adenocarcinoma with 5% of the cells presenting neuroendocrine markers, which can be interpreted as a partial trans-differentiation. The findings of this study are also consistent with a case report by Usmani et al. of a PC patient with an unsuspicious
68Ga-PSMA-PET, whereas a
68Ga-DOTANOC-PET performed ten days later revealed multiple somatostatin-avid hepatic and lymph node metastases, and lymph node cytology confirmed neuroendocrine differentiation [
31]. Overall, neuroendocrine trans-differentiation could explain the loss of PSMA-expression of liver metastases in progressive disease. Vice versa, the detection of PSMA-underexpression in liver metastases could represent trans-differentiation; clinicians need to be familiar with this concept as it may result in treatment adaptation.
Interestingly, the radiodensity of PSMA-negative liver metastases was significantly lower compared to the PSMA-positive metastases, in both unenhanced and contrast-enhanced CTs. This finding could further support the differentiation of liver metastases in PC but needs to be verified in a larger cohort.
Additionally, a significant positive correlation between the serum PSA level at the time of examination and SUV
max of PSMA-positive liver metastases was observed. This could be explained by the fact that both parameters tend to increase within the progression of the disease. The finding is consistent with the studies of Koerber et al. and Sachkepides et al., who reported that patients with higher PSA values demonstrated a significant higher tracer uptake in intraprostatic tumor lesions on PSMA-PET/CTs [
32,
33]. Between the size and SUV
max of PSMA-positive liver metastases, a weak but significant association was found. This might be the result of a proliferative advantage of highly PSMA-expressing cells, as it has been demonstrated in-vitro [
34]. We further observed a weak but significant, negative association between age and SUV
max of PSMA-positive liver metastases. A hypothesis explaining this finding could be that patients who develop liver metastases at a younger age have a more aggressive subtype of PC with higher PSMA-expression. This, however, needs to be investigated in a larger cohort.
A limitation of this retrospective study is that diagnoses of liver metastases were not confirmed histopathologically since no biopsies of most of the metastases were performed. A possible limitation to the lesion-based analysis regarding the calculation of mean SUVmax values could be due to an overestimation of the patients subgroup with multiple metastases compared to the subgroup with few metastases.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.