Infiltrative growth pattern of prostate cancer is associated with lower uptake on PSMA PET and reduced diffusion restriction on mpMRI

Prostate cancer (PCa) represents the second global cause of death from cancer, with an incidence of 190,000 new patients and over 33,000 related deaths per year in the USA [1]. In many institutions, patients with a suspected PCa due to an increased prostate-specific antigen (PSA) value or altered digital-rectal exam results will undergo an ultrasound (US)-guided biopsy procedure. However, US-guided biopsy can miss significant cancer in up to 25% of patients [2]; therefore, the recent joint EAU-EANM-ESTRO-ESUR-SIOG guideline suggests performing multiparametric magnetic resonance imaging (mpMRI)-guided biopsy or MRI-guided US-fusion biopsy [3, 4]. The introduction of mpMRI for PCa detection has induced significant improvement in the early detection of PCa. The mpMRI accuracy was further improved, and inter-reader variability was reduced, thanks to the PI-RADS score system. PIRADS 2.1 criteria are based on a 5-point scale using a combination of mpMRI findings (T2, diffusion-weighted imaging—DWI—with dynamic contrast-enhanced—DCE), to predict the presence of a clinically significant prostate cancer (csPCa). The definition of csPCa on biopsy cores is based on histopathology with a Gleason score (GS) ≥ 4 + 3 (grade 7b and higher) or a maximum cancer core length ≥ 6 mm according to the Epstein criteria [5]. T2-weighted images are mostly used to evaluate prostate gland anatomy and to identify suspicious lesions with a concurrent morphological characterization. DWI represents a core sequence used for PCa able to evaluate the cell density exploiting the diffusion of water molecules due to their thermal energy. A positive correlation between GS and the reduction in apparent diffusion coefficient (ADC) has been displayed in several studies enabling improved csPCa detection [6, 7]. Therefore, it has been shown that performing a mpMRI before a biopsy could avoid ≃30% of unnecessary biopsies and improve the csPCa detection by almost 15% [8, 9]. However, despite the adjusted PIRADS 2.1 version aimed to simplify the interpretation, mpMRI inter-reader reproducibility remains sub-optimal with reported kappa values between 0.37 and 0.48 [10]; false-positive results are also described in almost 20% of the patients [11, 12], and some aggressive tumors are negative on ADC maps [13, 14].

Recently, attention has focused on the new prostate-specific membrane antigen (PSMA) targeting PET tracer for PCa assessment [15]. It is well established that the PSMA uptake of the primary tumor is related to a higher GS and a worse prognosis [16]. However, preliminary results also showed significant PCa without PSMA uptake [17, 18]. Further histopathological analysis of PSMA-negative PCa showed that these tumors have a predominantly infiltrative (INF) rather than expansive (EXP) growth pattern [19]. This observation induces the hypothesis that growth patterns might also affect mpMRI diffusion sequences within cancer tissue and could cause false-negative results. Searching the literature, indeed a few publications mentioning a correlation between morphological patterns and ADC values were present [20‐23]. However, there was no incorporation of tumor aggressiveness and growth pattern in correlation to ADC values; there is still no consensus on how sparse or intermixed benign glands are defined or how reliable such a definition is.

Therefore, we aimed to test the interrater agreement for the new growth patterns on histopathology and to correlate them with ADC values with respect to PSMA uptake and ISUP grades.

Sixty-two (62) patients (mean age 64.2 ± 6.2 years; mean PSA 11.7 ± 12.5 ng/ml) underwent [⁶⁸Ga]PSMA-11 PET/MRI, a median of 50.5 days (72.9 ± 78, range 1–481 days) before RPE. Forty-three patients underwent a PSMA PET/MRI for staging and 19 patients for biopsy guidance. The dominant lesion had an EXP growth in 25 patients (40.3%), while 37 lesions had an INF growth (59.7%). The interrater agreement between both pathologists was almost perfect with κ = 0.81. For further analysis, the results of the more experienced reader were selected.

Lesions with INF growth had a lower ISUP grade compared to lesions with EXP growth (Fig. 4A). On RPE, the ISUP categories were significantly lower for INF growth compared to EXP growth (p = 0.001), with a median ISUP of three for INF growth tumors (mean ISUP = 2.9 ± 0.78), versus a median ISUP of four for EXP growth (mean ISUP = 3.7 ± 0.96) (Fig. 4B). Size on histopathology was measurable in 60 of the 62 lesions (in two patients, the dominant tumor was in the fresh frozen section of the apex). Among the measurable 60 lesions, the INF growth tumors were significantly smaller than the EXP growth (p < 0.001) with a mean tumor max diameter of 9.1 ± 4.8 mm and 17.2 ± 4.5 mm, respectively (Fig. 4C). Patients’ main characteristics are described in Table 1.

It is well-known that there is a strong inverse relationship between tumor cellularity and ADC values [29]. This was also confirmed in our small cohort, with a trend between ISUP and ADC_mean. However, we found that the correlation between ADC values and ISUP grades is lost if the two growth patterns are investigated separately. This is probably due to the fact, that ADC was significantly higher in lesions with INF growth with ADC_mean 1.079 mm²/1000 s compared to EXP growth ADC_mean 0.777 mm²/1000 s, and that tumors with EXP growth had overall higher ISUP grades [20, 30].

Other groups already observed that dense tumors had significantly lower ADC values than sparse tumors; however, they considered only peripheral zone lesions and did not assess the correlation with GS: Rosenkrantz et al. found an association between histopathology growth pattern and detectability of tumor on mpMRI, with solid tumor growth (defined as at least 5 mm of a continuous tumor with loose stroma or minimal prevailing benign glands) associated with tumor detectability with an odds’ ratio of 37.6 [22]. Langer et al. also observed that tumors with a “spares” growth pattern (defined as more than 50% of the cross-sectional area with normal glandular tissue) had no reduction in ADC or T2 values compared to normal tissue [21]. Later, the same group used semiautomatic tissue segmentation to correlate histopathological features with MRI findings. They found that ADC and T2 values were negatively related to the percentage area of nuclei or cytoplasm but positively related to the percentage area of luminal space [23]. Recently, Shiradkar et al. used deep learning to explore the relationship between histopathology tissue composition and MRI fingerprinting within regions of PCa, prostatitis, and the normal peripheral zone. They confirmed that T2 and ADC values dropped with increasing ISUP and that the lumen ratio induced significantly higher values for T2 and ADC [31].

Classical solid tumor growth is the fundamental attribute to classify GS 5 disease [32]. A strong correlation between solid growth and ADC, therefore, agrees with the expected tumor behavior. In our cohort, 25 lesions showed an EXP growth, but among them, only eight had solid growth patterns, 12 lesions showed cribriform characteristics, six had an intraductal component, and five showed more than two morphology patterns (e.g., acinar, intraductal, and cribriform). Interestingly, in our cohort, ISUP grade 1–4 lesions with EXP growth and non-classical solid tumor components had low ADC values.

We, therefore, believe that growth pattern characteristics are not yet well understood and need further investigation. We defined an INF growth for entrapped benign glands within the carcinoma complexes, whereas an EXP growth showed pure carcinoma glands within an area of at least 3 circles 5 mm² of each (radius 1.26 mm). With this simple and easy to implement definition, a robust, almost perfect, classification between EXP and INF growth was possible. The strong correlation between growth pattern and imaging parameters confirms that the morphology characteristics of histopathology have an impact on imaging parameters. Again, higher PSMA uptake was associated with EXP growth (SUV_max 19.2) compared to INF growth (SUV_max 9.3). This positive correlation has already been shown in a previous study [19].

The fact that cancer lesions are intermingled by benign glands in an INF growth was a potential explanation for lower SUV values regardless of the grading of the membranous or cytoplasmic immunohistochemistry staining [19]. According to Woythal et al., benign glands have lower PSMA expression and lower PSMA SUV_max than PCa. The same is true for diffusion restriction and ADC values on mpMRI; therefore, mixed tissue probably has higher ADC values [33]. Alternatively, a cumulative reduction of diffusion is typical of the aggressive disease being related to a dense growth pattern (Supplemental Fig. 4A-D). We showed in this study that there are PCa lesions with high ISUP but INF growth and high ADC values. The overall reduced density of those lesions might represent the substrate of the biologic principle able to explain false-negative ADC maps (Supplemental Fig. 4E-H).

Not all lesions with INF growth were difficult to localize on imaging, such as in one patient shown in Fig. 6A–D, with an ISUP 3 tumor and INF growth, the PSMA uptake was intense (SUV_max 16.7) and the diffusion restriction was very prominent with ADC_mean 0.69. This tumor had a maximal diameter of 8 mm, and there was an association with inflammatory changes that might have influenced the imaging characteristics. All EXP growth lesions had an ADC_min < 0.800 and were visible on ADC maps. Only two lesions with EXP growth had a SUV_max < 5 on PSMA PET; one is presented in Fig. 6E–H; both had large PSMA-negative tumor areas on immunohistochemistry.

Overall, we found that the correlation of ADC values was stronger between tumor size on histopathology than ISUP grade (r =  − 0.523, p < 0.001 vs r =  − 0.192, p = 0.135), while PSMA uptake was more dependent on ISUP grade than tumor size (r = 0.508, p < 0.001 vs r = 0.342, p = 0.006). These observations will need validation in larger external cohorts.

Limitations of this study are the retrospective design and the selection bias, given the inclusion of patients who underwent RPE and PSMA PET/MRI only. Furthermore, we lack the whole-mount preparation of histological slides, which would have simplified the correlation between imaging and histopathology. Also, 43/62 patients received a [⁶⁸Ga]PSMA-11 mean dose of 129.8 ± 21.1 MBq (81–160), while 19/62 patients (biopsy guidance) received 85 MBq. Another limitation is that the proposed growth pattern is not an established feature on histopathology analysis for PCa. The exact definition of both patterns might still be subject to further adjustments (e.g., 3 × 5 mm² is indeed the ideal size to define an EXP growth pattern). However, we observed that the proposed growth patterns are strongly associated with imaging parameters both on mpMRI and PSMA PET and could show a high reproducibility between pathologists.

This warrants further investigation, especially regarding the potential translation of growth pattern analysis to core biopsies. The awareness of the growth pattern based on biopsy might alter the interpretation of mpMRI, given the reduced reliability of ADC values.

PCa with INF growth showed significantly lower SUV_max and higher ADC_mean values compared to EXP growth PCa and overall lower ISUP grades. The established negative correlation between ADC_mean and ISUP grade was not confirmed for separate analyses of both growth patterns.