Patient Demographics
From January of 2012-October of 2016, we enrolled n=92 PCa patients with newly diagnosed disease, biochemically recurrent, or progressing disease. Of (n=92) total patients, 15% (n=14/92) were stage I, 30% (n=28/92) stage II, 9% (n=8/92) stage III, and 46% (n=42/92) Stage IV (Table 1). Prostate adenocarcinoma comprised 92% (n=46/50) of the non-metastatic cohort and the histology was undetermined in 8% (n=4/50) of non-metastatic patients. In the non-metastatic group, 72% (n=36/50) were newly diagnosed untreated and 16% (n=8/50) were defined as biochemically recurrent prior to initiation of second line therapy, 14% (n=7/50) had received prior chemotherapy, and 28% (n=14/50) had undergone androgen-deprivation therapy (ADT). In metastatic patients, 7% (n=3/40) were newly diagnosed untreated with 2% (n=1/40) defined as progressive PCa by biochemical recurrence, 71% (n=30/42) had received prior chemotherapy and 93% (n=39/42) had received ADT. Prostate adenocarcinoma was present in 93% (n=39/42) of metastatic patients, 2% (n=1/42) had neuroendocrine PCa, and 5% (n=2/42) were of unknown histology.
CAML Cell Presence Versus Conventional PCa Bioassays
Prior to the induction of treatment for new disease, or progressive disease starting new line therapy, CAMLs were identified in 79% (n=71/90) of available BL blood samples (average: ~5 CAMLs/7.5mL). Two samples failed due to blood clotting during microfiltration. CAML presence had a sensitivity of 78% (n=39/50) in the non-metastatic cohort, and 80% (n=32/40) in the metastatic population, with no statistical difference between groups (p=0.820) (Fig. 1). Additionally, non-metastatic PCa averaged 3 CAMLs/7.5mL, whereas mPCa averaged 6 CAMLs/7.5mL (p=0.108). Though statistically non-significant, the average number of CAMLs in circulation appears to be twice as high in mPCa over patients with non-metastatic.
Stratifying individual pathological stage, CAML presence was 57% (n=8/14) stage I, 82% (n=23/28) stage II, 100% (n=8/8) stage III, and 80% (n=32/40) of mPCa. CAML sensitivity between pathological stages found that CAMLs are less common in stage I disease than stages II and IV (p=0.086 and p=0.073, respectively), and are significantly less frequent compared to stage III (p=0.030) (Fig. 1). Average CAML number in stages I, II, III and IV patients contained 3, 3, 6 and 6 CAMLs/7.5mL blood, respectively, with no statistical difference between groups. Based on these preliminary data, CAMLs appear more common in PCa with progressing or advanced disease, and that the number of CAMLs in circulation does not differentiate individual pathological stages but may differentiate local and advanced PCa.
CAML’s were found to be more sensitive over CTCs in circulation (79% CAMLs vs. 21% CTCs, p<0.001) across all stages of disease. Comparing CAML and CTC presence among individual stages, CAMLs are more common than CTCs in stage I disease (57% vs. 7%, p=0.003), stage II (82% vs. 14%, p<0.001), stage III (100% vs. 38%, p=0.004), and stage IV (80% vs. 28%, p<0.001). When comparing non-metastatic and metastatic cohorts, CAMLs were statistically more sensitive than CTCs in both non-metastatic (78% vs. 16%, p<0.001) and metastatic (80% vs. 28%, p<0.001) cohorts. Overall, CAMLs were found to be the more frequent circulating cell type across all stages of PCa than CTCs (Fig. 1).
We compared CAML sensitivity to PSA thresholds ≥4ng/mL or ≥10ng/mL to see if they can be used to supplement the PSA “gray zone” for identifying aggressive PCa. Among all patients with BL PSA counts (n=89,), 79% (n=70/89) had PSA levels ≥4ng/mL and 49% (n=44/89) ≥10ng/mL. Single factor ANOVA found that CAML cells were significantly more sensitive in PCa when PSA was ≥10ng/mL (79% vs. 49%, p<0.001), but were similar to PSA values ≥4ng/mL (79% vs 79%, p>0.50). This pattern held in non-metastatic disease, as CAMLs appeared more sensitive than PSA at ≥10ng/mL (78% vs. 35%, p<0.001), and similar to PSA at ≥4ng/mL (78% vs. 76%, p>0.50). In metastatic disease, there was no statistical difference between CAMLs and PSA values. When examining individual stages for CAML sensitivity and PSA values ≥4ng/mL, CAMLs’ sensitivity was higher in stage III disease (100% vs. 75%, p=0.150), but the assays were equally sensitive in other stages. Further, CAMLs were more sensitive than PSA ≥10ng/mL in stage II (82% vs. 37%, p<0.001) and stage III (100% vs. 38%, p=0.004) patients, but not the more sensitive assay in stage I disease (57% vs. 29%, p=0.136) nor stage IV (80% vs 83% p=0.209). Given that CAMLs are cancer-specific and are found ubiquitously in patient blood, with similar sensitivity to low PSA ≥4ng/mL levels, these data suggest that implementation of CAML isolation in tandem with PSA quantification may add diagnostic sensitivity versus PSA alone, though follow up studies will be needed to better elucidate this relationship.
CAML Size Differentiates Local and Advanced Disease
Increasing CAML size (beyond 50μm) has been implicated in solid-tumor pathogenesis and decreased patient survival across multiple of solid cancers32,35-37,39. To examine this pattern in PCa, we compared all BL samples, finding the average max CAML size was statistically larger in patients with metastatic disease over localized PCa (78μm metastatic vs. 35μm localized, p<0.001) (Fig. 2). We then compared CAML size for each stage, finding that average CAML size increased with advancing disease, stage I averaging 23μm, stage II 33μm, stage III 65μm, and stage IV 78μm. Single factor ANOVA comparing sizes based on pathological stage found no statistical difference in max CAML size between stage I and II patients, as well as no difference between stage III and IV patients. However, CAML cells in patients with stage III or IV were statistically larger than in stages I or II by single factor ANOVA (average=75μm vs 30μm, p<0.001). We then divided CAML sizes into three subgroups to better elucidate the stage relationship at BL: 1) 0 CAMLs or CAMLs <50μm, 2) CAMLs ≥50μm-99μm, and 3) CAMLs ≥100μm (Fig 2). We found that stage I and II patients appeared to have a nearly identical distribution of CAML sizes, with most patients having 0 CAMLs present or <50μm CAMLs. In contrast, larger CAMLs (i.e. ≥50μm and ≥100μm) (Fig. 2) were more commonly found in circulation among stage III and IV patients. Interestingly, CTC presence, a phenomenon found in advanced disease, appeared to have a relationship to engorged CAML presence in both the non-metastatic and metastatic settings (Sup. Fig. 1).
Engorged CAMLs Found Prior to Treatment Predict for Early Mortality
At BL sampling, 41% (n=37/90) of patients presented with ≥50μm CAMLs vs those with <50μm CAMLs, which predicted for shorter median progression free survival (mPFS=7.9 vs. >24 months) as well as shorter median overall survival (mOS=17.4 vs >24 months). Cox-Fit Proportional analysis found that engorged ≥50μm CAMLs at BL were able to prognosticate for worsened PFS (HR=7.5, 95%C.I.=3.7-15.3, p<0.001) and worsened OS (HR=13.3, 95%C.I.=5.5-32.5, p<0.001) (Fig. 3).
After examining CAMLs’ relationship to clinical outcomes, we then analyzed the non-metastatic PCa and metastatic PCa cohorts separately. In non-metastatic, 23.5% (n=12/50) of patients had ≥50μm CAMLs. mPFS and mOS could not be calculated, as too few patients had clinical events within the 2-year endpoint, PFS (12%, n=6/50) and OS (6%, n=3/50). Although the mPFS and mOS could not be calculated, it was found that ≥50μm CAML presence in non-metastatic disease did significantly predict for worsened PFS (HR=20.9, 95%C.I.=2.7-159.7, p=0.016), but not OS (HR=9.7, 95%C.I.=0.7-135.1, p=0.306) (Fig. 3). The lack of significance in OS was likely a result of too few patients dying within the study time frame, and larger longer-term studies may be required to provide a significant endpoint based on CAML engorgement.
In the metastatic cohort, patients with mPCa were found to have ≥50μm CAMLs in 63% (n=25/40) of samples, which predicted for shorter mPFS (4.7 vs. 12.8 months) as well as shorter mOS (16.2 vs. >24 months). Comparative analyses found that engorged CAMLs significantly predicted for expedited patient progression (HR=2.4, 95%C.I.=1.2-4.9, p=0.031) as well as expedited death (HR=5.4, 95%C.I.=2.2-13.4, p<0.001) in mPCa) (Fig. 3).
To explore previous literature32,35-37,39 on hyper-enlarged CAMLS (≥100μm) for predicting severely worse outcomes in patients, we ran analysis to investigate how these massively-engorged CAMLs relate to outcomes in PCa. First, we examined ≥100μm CAML presence among all patients, and found that 14% (n=13/90) of BL samples had the hyper-enlarged CAML subtype. Survival analysis of ≥100μm CAMLs were found to predict for shorter mPFS (4 vs >24 months) and shorter mOS (8.7 vs >24 months) when compared to patients with CAMLs <100μm in diameter (Fig. 4). Further analysis then confirmed that ≥100μm CAMLs predicted for worsened PFS (HR=12.1, 95%C.I.=4.1-40.0, p<0.001) and worse OS (HR=71.1, 95%C.I.= 18.0-280.3, p<0.001) (Fig. 4). We then examined non-metastatic patients for evidence of massively-engorged CAMLs, which were seen in 2% (n=1/50) of all samples, preventing survival analytics from being run. Interestingly, this individual was diagnosed with stage IIIb biochemically recurrent adenocarcinoma who progressed and died within 16 months after identifying the ≥100μm CAML. In the mPCa population, we found ≥100μm CAMLs in 30% (n=12/40) of BL samples which predicted for shorter mPFS (3.9 vs 9.1 months) and shorter mOS (7.5 vs 19.9 months). In evaluating ≥100μm CAMLs, it was found that ≥100μm CAMLs were statistically significant predictors for worse OS (HR=3.7, 95%C.I.=1.3-10.1, p=0.025), but not for worse PFS in metastatic PCa (Fig. 4).
CAML Size Tracking Throughout Treatment
Following new treatment induction, n=36 patients volunteered a midpoint treatment (T1) sample, and n=11 volunteered samples at treatment completion (T2). Engorged CAMLs were present in 28% (n=10/36) at T1, which predicted for shorter mPFS (5.1 months vs >24 months) and mOS (16.5 months vs >24 months), as well as shorter overall PFS (HR=12.2, 95%C.I.=3.4-43.0, p<0.001) and shorter OS (HR=17.8, 95%C.I.=3.9-80.5, p<0.001). Due to insufficient samples available at T1 for non-metastatic patients (n=17), survival analysis was not possible. However, metastatic patients with engorged CAMLs at T1 (47%, n=9/19) had shorter mPFS (4.6 vs 11.7 months) and shorter mOS (15.6 vs >24 months), as well as significantly predicting for worsened PFS (HR=4.1, 95%C.I.=1.3-12.9, p=0.030) and OS (HR=6.3, 95%C.I.=1.6-24.8, p=0.024) at T1 (Fig. 5).
Despite a sparse number of patients at the T2 timepoint, we conducted preliminary analyses (Sup. Fig. 2). Among these patients, 27% (n=3/11) of had CAMLs ≥50μm at T2 which had shorter mPFS (5.1 vs. 7.9 months) and shorter OS (8.6 vs. 20.8 months). Further, these patients trended towards worse OS (HR=8.8, 95%C.I.=1.1-70.9, p=0.133), but did not predict for worse PFS (HR=3.4, 95%C.I.=0.6-24.8, p=0.342) (Sup. Fig. 2). While this patient cohort was too small to for proper clinical outcome analysis, these initial finding appear promising and follow up studies with larger populations with additional time points appears promising.
Multivariate Analysis
A multivariate analysis was used to compare all known significant variables for patient PFS and OS. Parameters for multivariate analysis were defined to be age ≥70 years, pT ≥T2 (locally confined to the prostate), pN ≥N1 (metastasis in a single regional lymph node <2cm), pM ≥M1 (distant metastasis), Gleason score ≥8, PSA ≥50ng/mL, ≥1 CTCs present, ≥3 CAMLs present, and CAMLs ≥50μm (Supplementary Table 1). Among all patients, CAMLs ≥50μm at BL were the most statistically significant independent predictor of PFS (p=0.004), with PSA ≥50ng/mL (p=0.012) and metastasis (p=0.045) also being significant independent indicators. In addition, engorged CAMLs (≥50μm) were statistically significant independent predictors for worse OS (p<0.084) behind nodal spread (p=0.022), and age ≥70 years old (p=0.044). Due to limited patients reaching time to event, as well as limited population sizes multivariate analysis could not be run separately for the non-metastatic and metastatic cohorts though to better elucidate the significance of CAMLs larger validation studies in refined cohorts is warranted.
Analysis of PSA for Predicting PFS and OS
To best identify PSA thresholds for prognosticating PCa, we evaluated PSA levels ≥10ng/mL, ≥20ng/mL, and ≥50ng/mL. Among all patients, 97% (n=89/92) had available PSA counts at BL, with three patients lost during unblinding. Pretreatment PSA levels among all patients found 17% (n=15/89) had 10-20ng/mL, 10% (n=9/89) had 20-50ng/mL, and 22% (n=20/89) had ≥50ng/mL. The combination of both non-metastatic and metastatic diseases found that all three PSA cutoffs were statistically significant predictors for worsened PFS and OS, with increasing PSA predicting for shorter survival (Fig. 5).
Among non-metastatic patients with available PSA counts (n=49), 22% (n=11/49) had PSA between 10-20ng/mL, 4% (n=2/49) between 20-50ng/mL, and 8% (n=4/49) >50ng/mL. We found that all three PSA cutoffs were not significant predictors for worsened survival and all groups had mPFS and mOS >24 months.
Metastatic patients had (n=40) available BL PSA counts, with 10% (n=4/40) presenting PSA levels between 10-20ng/mL, 18% (n=7/40) between 20-50ng/mL, and 40% (n=16/40) having ≥50ng/mL. IT was found that PSA levels ≥10ng/mL had shorter mPFS (6.3 vs. 14.7 months) and mOS (13.7 vs. 23.2 months) but were not statistically significant predictors for PFS or OS (Fig. 5). Similarly, patients with PSA ≥20ng/mL had shorter mPFS (7.3 vs. 11.5 months) and mOS and (13.7 vs. 23.2 months) but did not reach statistical significance. However, patients with PSA ≥50ng/mL had shorter mPFS (4.2 vs 13.6 months) but statistically trended toward worsened PFS (HR= 2.5, 95%C.I.=1.1-5.7, p=0.051). Further, PSA ≥50ng/mL appeared to have shorter mOS (10.1 vs >24 months) and was a statistically significant predictor for worsened OS (HR=3.9, 95%C.I.=1.4-10.5, p=0.015) in the metastatic setting (Fig. 5). Follow up monitoring of increasing PSA at T1 and T2 found no statistical significance in predicting worsened survival.