Objective: Large language models (LLMs) are increasingly used as guideline-decision support in oncology, but most validation has occurred in common cancers, where clinicians can usually recognize errors. The safety-critical setting is the opposite — rare cancers, where local expertise is limited, and errors may go unrecognized. Whether common-cancer concordance reflects the reliability needed there is unclear. Methods: We benchmarked GPT-5.2 against the EAU-ASCO 2025 Penile Cancer Guidelines using 100 guideline-derived questions across five domains and three zero-shot prompting strategies, yielding 300 responses. Two experienced urologists, blinded to prompt identity, scored concordance on a 3-point scale and classified errors as hallucination, omission, temporal, or reasoning. The primary analysis was a mixed-effects ordinal logistic regression with question as a random intercept; a cluster-robust logistic regression served as sensitivity. Reporting followed TRIPOD-LLM. Results: GPT-5.2 was fully concordant in 236 of 300 responses (78.7%; 95% confidence interval, 73.7–82.9; weighted κ=0.914). Prompt strategy did not improve performance (role-play +3 percentage points; 95% credible interval, −2 to +9). The domain effect was not robust once between-question clustering was modeled. Among 64 non-concordant responses, omission predominated (47.7%) and reasoning errors followed (37.5%); in treatment, omission accounted for 68.0% of errors, including loss of guideline-specified detail. Four questions were uniformly discordant across all three prompts. Conclusion: Headline concordance overstated clinical reliability. The clinically informative finding was not the overall accuracy but the structure of failure — omission of decision-critical treatment detail and stable errors. Prompt engineering did not rescue performance. Current LLM output may assist orientation in rare-cancer decision pathways but is not ready as stand-alone guideline-decision support.

Background/Objectives: Adherent perinephric fat (APN) is common in laparoscopic partial nephrectomy (LPN). It is absent from the lists of recognised positive surgical margin (PSM) predictors. We tested whether APN is associated with PSM risk and whether the effect varies by resection technique. Methods: We analysed 401 International Society of Urological Pathology (ISUP)-gradable renal cell carcinomas (31 PSM events; chromophobe RCC was excluded from the primary cohort because the WHO/ISUP grade is not assignable) carried out at three centres by eight surgeons from 2013 to 2024. Nested logistic models added APN, then an APN × enucleation interaction, to a guideline-based covariate set (tumour size, ISUP grade ≥ 3, cT1b stage, centre fixed effects, surgeon-documented enucleation). Firth penalisation and a 2000-replicate bootstrap were used to limit small-sample bias; they cannot add information beyond the 31 observed events. Results: APN was an independent PSM predictor (adjusted odds ratio [OR] 2.68, 95% confidence interval [CI] 1.21 to 5.95, p = 0.02). The effect was technique-conditional: APN × enucleation interaction OR 7.83 (95% CI 1.08 to 56.92, p = 0.04), consistent in direction across estimators but imprecise in magnitude. PSM occurred in 29.4% of APN-positive enucleation cases (odds ratio 10.21, 95% CI 2.07 to 50.30; Fisher exact p = 0.003), against 5.0% in enucleoresection and 10.8% in resection. Conclusions: In this cohort the APN-to-PSM association was technique-conditional. The penalty concentrates in enucleation; the rim-retaining strata gave estimates compatible with both no association and clinically important harm (OR 0.99 and 1.96, both intervals wide). Because the interaction rests on few margin events and the technique was not randomised, this finding is hypothesis-generating: when APN is encountered intraoperatively, retaining a parenchymal rim (enucleoresection or resection) rather than enucleation may be considered, pending prospective validation.

Background/Objectives: When urethral catheterisation fails in acute urinary retention, clinicians may pass a guidewire or go directly to suprapubic cystostomy. No published series compares the two after a failed attempt, and comparative trials enrol men in whom either route remains available. We compared the two. Methods: We studied 352 consecutive men at one centre from 2019 to 2025: 246 went directly to cystostomy, 106 to an ultrasound-confirmed bedside guidewire attempt. Time zero was the documented failure of catheterisation, and patients were analysed by the strategy first attempted. The primary outcome, overall strategy-associated complications, was any new Clavien–Dindo grade II or higher complication within 30 days, whichever route produced it; conditions present at time zero did not count. Overlap weighting on 19 covariates gave risk differences with operator-level bootstrap intervals. Follow-up used the hospital and national health records for every patient. Results: The guidewire attempt placed a catheter in 84.9% of men (95% confidence interval [CI] 76.9 to 90.5). After weighting, overall strategy-associated complications affected 17.6% of the cystostomy arm and 5.3% of the guidewire-first arm (−12.2 points, 95% CI −19.0 to −5.5). Grade III or higher events occurred in 7.7% (19/246) of the cystostomy arm and none of the guidewire-first arm. Shared complications, those either route could produce, affected 9.5% and 5.3%, respectively, a difference of −4.2 (−11.2 to 3.8) after weighting and −0.3 (−6.0 to 5.4) when unweighted. Conclusions: An ultrasound-confirmed bedside guidewire attempt was associated with fewer overall strategy-associated complications at 30 days. For shared complications the difference was smaller and no longer excluded zero. Ascertainment was record-based and asymmetric: cystostomy-specific harms are readily recorded, whereas urethral harms such as false passage or later stricture were not captured, and time to definitive drainage was not measured. These findings describe initial rescue strategies and do not establish that either procedure is intrinsically safer.