AI-Derived PSMA PET-CT Tumor Volume Linked to Mortality
Key Takeaways
- Among patients with prostate cancer undergoing PSMA PET-CT for primary staging or biochemical recurrence at a Swedish center, higher automated whole-body tumor volume was significantly associated with mortality.
- The overall whole-body volume association was significant in primary staging but not in biochemical recurrence.
- Bone TLV showed significant overall associations with mortality in the combined and primary-staging groups.
- Prostate and lymph-node TLV did not show significant overall associations with mortality in the full cohort.
Primary staging of newly diagnosed intermediate- to high-risk disease accounted for 159 patients; 108 were imaged for biochemical recurrence (BCR). A previously developed artificial intelligence (AI) model segmented suspected prostate or prostate-bed, lymph-node, and bone lesions without manual corrections; no study scans were used to train it. Investigators aggregated total lesion volume (TLV) by site and whole body. Site-specific categories separated no detectable TLV from low and high nonzero TLV, divided at the median; whole-body TLV used tertiles. Mortality over five years was assessed with age-adjusted Cox models, overall likelihood-ratio tests across categories, and separate hazard ratios (HRs) for category contrasts.
During follow-up, 48 patients died (18%). Whole-body TLV showed an age-adjusted overall association with mortality across categories (p = 0.001).
In the full cohort, high versus low whole-body TLV carried an age-adjusted HR of 3.20 (95% confidence interval [CI] 1.50–6.80); this category contrast differs from the overall test. The primary-staging group's overall association was significant (p = 0.014).
In BCR, the overall whole-body TLV association did not reach significance (p = 0.061). Bone TLV showed significant overall associations with mortality in the combined cohort and primary-staging subgroup. Overall prostate or prostate-bed and lymph-node TLV associations were not significant in the full cohort.
Retrospective single-center data, age-only adjustment, and limited subgroup deaths constrain interpretation. The authors did not formally compare imaging indications, so the subgroup results do not establish a between-group difference. Unreviewed AI segmentations can include nonspecific findings, especially [18F]PSMA-1007 bone uptake, and shift low-burden patients between TLV categories. AI-derived bone TLV is not an exact measure of bone-metastasis volume, and these associations do not establish causal effects or validated treatment thresholds.
Investigators considered fully automated whole-body and bone TLV possible prognostic imaging markers. They called for larger prospective cohorts, longer follow-up, and assessment of alternative volume thresholds before potential clinical integration.
Clinician Questions
Why can the high-versus-low PSMA PET-CT tumor-volume contrast in biochemical recurrence be significant when the overall association is not?
In BCR, the high-versus-low whole-body AI-derived TLV contrast indicated greater mortality at higher volume, with a CI excluding 1, although the overall likelihood-ratio test across categories was not significant. The tests address different questions; neither establishes a difference between recurrence and primary-staging groups.
Why might men with prostate cancer have no detectable prostate or lymph-node TLV yet fare worse than men with a small measurable TLV?
Zero AI-derived prostate or prostate-bed or lymph-node TLV sometimes coincided with higher mortality than a small measurable volume. The authors proposed disease predominating elsewhere or variation in PSMA expression as possible explanations, not established mechanisms.
Why did investigators use tertiles for whole-body AI-derived PSMA PET-CT TLV instead of a no-tumor group?
Very few patients had zero whole-body automated TLV, so investigators used volume tertiles rather than a separate zero-volume group.
Recommended Reading
- For more on PSMA PET in prostate cancer: Improved Outcomes in mCRPC with PSMA-Directed Diagnostics and Therapies
This content is intended for healthcare professionals for educational purposes. The article and its image were developed with AI assistance and reviewed by members of the ReachMD team prior to publication. Please refer to the original source linked for the complete and most up-to-date information.