A SPECT-Guided, PET-Sparing Algorithm for PSMA Radiopharmaceutical Therapy Evaluation (SPARE)

[177Lu]Lu-PSMA radiopharmaceutical therapy is delivered as repeated cycles, yet the decision to continue or adapt treatment after the second cycle is not standardized. Posttherapy SPECT/CT is routinely performed but rarely used formally for response assessment. We developed a multimodality, response-adapted algorithm applied after cycle 2 and tested its concordance with real-world decisions. Methods: We studied 142 men with metastatic androgen pathway modulator–resistant prostate cancer (formerly metastatic castration-resistant prostate cancer) who completed at least 2 cycles of [177Lu]Lu-PSMA and underwent posttherapy SPECT/CT, baseline and 12-wk prostate-specific antigen (PSA) testing, and baseline and interim prostate-specific membrane antigen (PSMA) PET and [18F]FDG PET. Response was assessed in accordance with Prostate Cancer Working Group 4 recommendations for PSA and with RECIP 1.0 for SPECT and PSMA PET. The algorithm combined SPECT/CT and PSA, used PSMA PET only when both were stable, and yielded 3 recommendations: continue [177Lu]Lu-PSMA, discontinue [177Lu]Lu-PSMA, or consider [18F]FDG PET. The primary endpoints were the number of interim PSMA PET studies avoided and the concordance between algorithm recommendations (continue vs. discontinue) and real-world decisions. The secondary endpoint was intermodality response evaluation concordance. Results: The algorithm recommended continuation in 74 patients (52.1%), discontinuation in 49 patients (34.5%), and [18F]FDG PET in 19 patients (13.4%). Interim PSMA PET/CT could be avoided in 120 patients (84.5%) (95% CI, 77.7%–89.5%). Among patients with a definitive recommendation, conditional agreement of the algorithm with real-world management decisions was 91.1% (95% CI, 84.7%–94.9%), with a κ of 0.81 (95% CI, 0.69–0.91). All 74 patients (100%) with an algorithm recommendation to continue treatment had matching real-world decisions for treatment continuation, whereas 38 (78%) of 49 patients with an algorithm recommendation to discontinue treatment had matching real-world decisions for treatment discontinuation. Response classification on SPECT agreed only moderately with PSA level (κ = 0.45) and PSMA PET (κ = 0.47), but the combination of SPECT and PSA raised agreement with PSMA PET to “substantial” (κ = 0.67), approaching that of PSMA PET with PSA level (κ = 0.70). Every discordant decision (n = 11) occurred when the algorithm recommended treatment discontinuation in a patient who nonetheless continued treatment, most commonly because [177Lu]Lu-PSMA was being used as a last line of therapy (7/11, 64%). Conclusion: A SPECT-guided algorithm could have avoided most interim PSMA PET studies and reproduced real-world treatment decisions after 2 cycles of [177Lu]Lu-PSMA with high agreement. This algorithm may serve as the basis for a standardized treatment management framework, pending independent, prospective outcome-based validation.