{"id":"e1f35364-450e-43ee-8993-70e887e9e815","arxiv_id":"2501.18433","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Proton therapy for cardiac radioablation in ventricular tachycardia gives equivalent target coverage to photons while substantially reducing dose to the heart, lungs, esophagus, and spinal cord.","lead":"This retrospective study compared proton and photon radiation plans for 33 or 34 patients with ventricular tachycardia treated with stereotactic arrhythmia radioablation. The proton plans achieved similar target coverage while significantly lowering radiation doses to the heart, lungs, esophagus, and other nearby organs, which could reduce treatment-related toxicity.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proton plans showed significantly worse target coverage (VAS V25Gy 85.6% vs 93.0%, p<0.001; D99 21.4 vs 22.6 Gy, p=0.022), so the 'equivalent coverage' half of the central claim is internally contradicted and OAR sparing may be a coverage trade-off.","rationale":"Read in good faith, the paper does a reasonable thing: benchmark retrospectively generated proton STAR plans against clinically delivered photon plans for 33 patients. The qualitative direction of the OAR results is plausible and consistent with physical expectations and thoracic proton literature. However, the strongest claim as worded requires both OAR benefit and equivalent target coverage. The paper's own Table 2 fails the second condition. D95 is similar (p=0.294), but D99 and V25Gy are significantly worse for protons. Because the proton plans were normalized to mean target dose rather than to matched coverage, the two arms are not on equal footing. The reader's weakest_assumption about planning asymmetry is valid and likely explains part of the target-coverage gap; my concern is that the internal contradiction is even more direct and more load-bearing, since it does not depend on any assumption about photon reoptimization. I am not alleging any intent to mislead; the reported workflow itself creates the confound. The paper would need either to reoptimize both modalities under matched coverage constraints or to soften the conclusion to describe a coverage-sparing trade-off. Because a targeted reanalysis could support a weakened version of the claim and the OAR dose reductions are physically plausible, conditional acceptance is the appropriate outcome rather than rejection.","tokens_in":11001,"tokens_out":6042,"duration_ms":57142,"concrete_test":"Using the per-patient DVH data behind Table 2, compute ΔV25Gy = photon V25Gy − proton V25Gy and ΔheartDmean = photon heart Dmean − proton heart Dmean for each patient. Regress ΔheartDmean on ΔV25Gy, then repeat the paired OAR comparisons on the subset with |ΔV25Gy| ≤ 5 percentage points. If the heart/lung/esophagus advantages shrink below statistical significance or by more than ~50% in that coverage-matched subset, the reported OAR advantage is a target-coverage trade-off rather than an intrinsic proton dosimetric gain.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim has two conjuncts: proton OAR sparing and equivalent target coverage. Table 2 directly contradicts the second conjunct. Proton plans had lower VAS D99 (21.4±2.8 Gy vs 22.6±2.4 Gy, p=0.022) and substantially lower V25Gy (85.6±10.2% vs 93.0±8.3%, p<0.001). These are target-coverage metrics by the paper's own definitions (§2.4). The proton plans were scaled only to the mean target dose (§2.3) and robustly optimized to spare OARs, while the photon VMAT plans were the delivered clinical plans and were not reoptimized. Thus the reported OAR reductions (heart Dmean 3.6 vs 5.5 Gy, esophagus Dmean 0.3 vs 1.6 Gy) are not cleanly attributable to proton physics; they may reflect accepting lower high-dose target coverage and an asymmetric optimization objective. This is an internal inconsistency, not just a consensus disagreement: even before considering external validity, the data as reported fail to support 'maintaining equivalent target coverage.' A secondary inconsistency (abstract lungs Dmean 1.6/2.1 Gy vs Table 2 lungs Dmean 0.7/1.2 Gy) reinforces the need for reanalysis.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a retrospective dosimetric comparison of proton versus photon stereotactic arrhythmia radioablation (STAR) for ventricular tachycardia (VT) in a cohort initially described as 34 patients. Proton plans were generated retrospectively with robust optimization and scaled to match the mean target dose of the clinical photon VMAT plans. Dosimetric metrics for the ventricular arrhythmogenic substrate (VAS) and organs at risk (OARs) were compared with paired tests and Bonferroni correction. The authors conclude that proton therapy provides significant OAR sparing while maintaining equivalent target coverage, and they discuss clinical implications, patient selection, and limitations.","tokens_in":11258,"tokens_out":2607,"duration_ms":23571,"significance":"If the results were reliable, the study would be a useful planning-level contribution to the emerging STAR literature, providing quantitative evidence for modality selection and supporting ongoing proton-STAR trials. The statistical approach is generally appropriate: normality checks, paired tests, and Bonferroni correction are correctly described. The use of robust optimization for proton plans and scaling to the photon mean target dose are sensible methodological choices that partially address fairness. However, the central claim of 'equivalent target coverage' is directly contradicted by the study's own target-coverage metrics, and several data inconsistencies reduce confidence in the reported OAR advantages. The paper would be clinically valuable only after a careful data audit and reanalysis.","major_comments":[{"comment":"The claim of 'equivalent target coverage' is contradicted by the reported VAS metrics: proton V25Gy is 85.6±10.2% vs 93.0±8.3% (p<0.001) and D99 is 21.4±2.8 Gy vs 22.6±2.4 Gy (p=0.022). These are the paper's own coverage metrics, so the data as reported show a significant coverage trade-off. The abstract and conclusion must be revised, or the analysis must be re-run with corrected data.","section":"§3.2.1, Table 2, Abstract, Conclusion"},{"comment":"The comparison is asymmetric: proton plans were generated with robust optimization and scaled to match the mean target dose, while the photon VMAT plans were the delivered clinical plans and were not reoptimized under comparable objectives. This confounds modality with optimization choices, making the observed OAR reductions (e.g., heart Dmean 3.6 vs 5.5 Gy) not cleanly attributable to proton physics. The authors should either reoptimize photon plans on the same footing or explicitly frame the results as a planning-workflow comparison with this limitation prominently stated.","section":"§2.3 and §2.4"},{"comment":"The cohort is inconsistently reported: the Abstract says 34 patients, §3.1 says '26 males and 7 females' (33 total), and Table 1 says 'Number of Patients 33'. Moreover, the median VAS volume is 29.7 cc in §3.1 but 68.25 cc in Table 1. These discrepancies undermine confidence in the data and must be reconciled.","section":"Table 1 vs §3.1, Abstract"},{"comment":"The Abstract reports lungs Dmean as 1.6±1.5 Gy (proton) and 2.1±1.4 Gy (photon), whereas Table 2 reports 0.7±0.6 Gy and 1.2±0.7 Gy for the same metric. At least one of these values is incorrect; the abstract and table must be reconciled before the manuscript can be considered reliable.","section":"Abstract vs Table 2"},{"comment":"The Methods state that target metrics include D98, but the Results and Table 2 report D99. This inconsistency is more than a typo because the choice of quantile affects the reported coverage difference (D99 p=0.022) and the conclusion of equivalence. The authors should specify which quantile was intended and, ideally, report both D98 and D99 consistently.","section":"§2.4 vs §3.2.1 and Table 2"}],"minor_comments":[{"comment":"The spinal cord doses in Figure 3 are reported as '0.3 cGy vs 0.6 Gy' for Dmean; the mixed units are confusing and likely a typo. Please unify to Gy throughout.","section":"Figure 3 caption and §3.3"},{"comment":"The sentence 'with similar CI and dose coverage metrics, including D95 and D98' is internally inconsistent with the significant D99 and V25Gy differences reported in Table 2; please align the wording with the actual results.","section":"§4 Discussion"},{"comment":"The statistical section states p<0.05 as significance level but does not specify the number of comparisons used for the Bonferroni correction or report the adjusted threshold; please provide this detail.","section":"§2.5"},{"comment":"Some reference citations are grouped (e.g., refs 23–26) without specific attribution; consider citing individual references where a specific finding is discussed.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The internal inconsistencies (patient count, VAS volume, lung Dmean) suggest that a data audit is essential before this manuscript can be considered for publication. If the Table 2 values are verified, the central conclusion must be substantially weakened to reflect a target-coverage trade-off rather than equivalence. The planning asymmetry also needs to be addressed, either by reoptimizing photon plans or by reframing the study as a planning-workflow comparison. These are fixable within the manuscript's scope, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, what's new: this is the first cohort-scale dosimetric comparison of proton and photon STAR for VT, with 33-34 patients, a clinically relevant target (electroanatomical substrate), and a broad set of OAR metrics. That fills a real gap, and the qualitative pattern of OAR sparing with protons is physically plausible and consistent with thoracic literature. The case studies are illustrative, and the statistical approach is reasonable with Bonferroni correction.\n\nThe problem is the central claim. The abstract and conclusion say proton and photon plans achieved 'comparable' or 'equivalent' target coverage, but Table 2 shows the opposite: proton VAS D99 was 21.4 vs 22.6 Gy (p=0.022) and V25Gy was 85.6% vs 93.0% (p<0.001). Those are the paper's own coverage metrics. So the second conjunct of the headline result is contradicted by the data. The discussion even asserts comparable D95 and D98, but D98 is not reported in Table 2, and D95 is comparable but V25Gy and D99 are not. This is not a minor nuance; it changes the interpretation of the OAR savings. The OAR reductions may in part reflect a trade-off against high-dose target coverage, especially since the proton plans were robustly optimized to spare OARs and scaled only to mean target dose, while the photon plans were the delivered clinical VMAT plans not reoptimized.\n\nThere are also internal inconsistencies that erode confidence: abstract says 34 patients, Table 1 says 33 (and 26+7=33); text says median VAS 29.7 cc, Table 1 says 68.25 cc; abstract lung Dmean values (1.6/2.1 Gy) don't match Table 2 (0.7/1.2 Gy). Individually they could be typos, but together they suggest the data pipeline needs checking. No raw data are provided, so none of this is independently verifiable.\n\nWhat the paper does well is scope a clinically important question and report a reasonable set of metrics. The authors also honestly list limitations (RBE, range uncertainty, neutrons, retrospective design). But as written, the conclusion overstates the case.\n\nRecommendation: send it to peer review, but require major revision. The authors should fix the inconsistencies, report D98 and all coverage metrics, and either match coverage between modalities (e.g., re-optimize photon plans to the same coverage objective) or explicitly present the OAR results as a trade-off rather than an advantage. If the coverage difference is real, the paper is still useful, but the framing must change.","headline":"First cohort-level proton-vs-photon STAR comparison, but the paper's own target coverage data contradict its 'equivalent coverage' conclusion.","tokens_in":11822,"tokens_out":2801,"would_cite":false,"duration_ms":23877,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Proton therapy delivers the same arrhythmia-targeted dose as photon therapy while significantly reducing radiation to the heart, lungs, and esophagus in stereotactic arrhythmia radioablation for ventricular tachycardia.","keywords":["Ventricular Tachycardia","Proton Therapy","Stereotactic Arrhythmia Radioablation","Dosimetric Comparison","Organ at Risk Sparing","Bragg Peak","Robust Optimization","Treatment Planning"],"falsifier":"Re-plan the same cohort with both modalities using identical, pre-registered optimization objectives (same target dose normalization, same organ-at-risk constraints, same number of beams or arcs, no manual fine-tuning), then compare DVH metrics; if the significant organ-at-risk differences persist under this symmetric planning protocol, the claim is supported, and if they shrink to insignificance, the reported advantage is a planning artifact.","tokens_in":10799,"feed_emoji":"⚡","tokens_out":6635,"duration_ms":54832,"temperature":0.7,"pith_summary":"This paper tries to establish that, for stereotactic arrhythmia radioablation (STAR) of ventricular tachycardia, proton therapy can deliver the same prescribed dose to the arrhythmia substrate as photon therapy while exposing the heart, lungs, esophagus, spinal cord, and stomach to substantially lower radiation. The authors generated retrospective proton plans for a cohort of patients who had already received photon STAR, using robust optimization and scaling the proton plans to match the mean target dose of the photon plans. On every meaningful organ-at-risk metric, protons came out lower—mean heart dose fell from 5.5 to 3.6 Gy, mean esophageal dose from 1.6 to 0.3 Gy—without a statistically significant loss of target coverage. The interest for a general reader is that lower organ-at-risk doses should translate into fewer treatment-related complications such as esophagitis, pneumonitis, and long-term cardiac damage, making a promising non-invasive treatment safer and potentially available to sicker patients.","feed_headline":"Proton therapy lowers heart, lung, esophagus doses in STAR","feed_subtitle":"A dosimetric comparison finds protons cut mean heart dose by a third while keeping arrhythmia-target coverage equal.","key_machinery":"The central physical mechanism is the proton Bragg peak: protons deposit most of their energy at a controllable depth and have essentially no exit dose, so beams can be aimed so that the high-dose region stops within or just past the target. On top of that, the study's planning methodology is the other load-bearing piece: proton plans were generated with robust optimization to account for range and setup uncertainties (±3.5% range margin), and they were normalized by scaling to the mean target volume dose of the corresponding photon plans, a step intended to put the two modalities on equal footing for a fair dose comparison.","core_discovery":"The paper reports that when proton STAR plans are compared with clinically delivered photon VMAT plans for the same patient cohort, mean doses to the heart, normal heart (excluding the arrhythmia substrate), lungs, esophagus, spinal cord, stomach, and cardiac devices are all significantly lower for protons, while target coverage metrics D95 and conformity index are statistically equivalent. For example, whole-heart mean dose was 3.6 ± 1.5 Gy for protons versus 5.5 ± 2.0 Gy for photons (p<0.001), and normal-heart V5Gy was 15.7% versus 32.5% (p<0.001). The authors attribute this to the proton Bragg peak, which stops the beam distal to the target and eliminates much of the low- and intermediate-dose bath that photon arcs inevitably create. They conclude that proton therapy provides a dosimetric advantage in STAR that could reduce treatment-related toxicity.","pith_inferences":["The paper does not model normal tissue complication probability; whether the observed dose reductions translate into clinically meaningful toxicity differences depends on the steepness of the dose-response curves, which such modeling could quantify.","The proton plans were normalized to mean target dose; normalizing instead to the same V25Gy or D95 would change the comparison and could reduce the apparent organ-at-risk advantage, so the magnitude of the benefit is partly normalization-dependent.","Proton variable relative biological effectiveness at the distal edge of the Bragg peak could raise the effective dose to organs just beyond the target (e.g., esophagus, heart), partially offsetting the physical dose advantage.","Cardiac substructures (conduction system, coronary arteries) were not contoured; substructure-sparing analysis could reveal whether the whole-heart dose reduction translates into protection of arrhythmia-relevant sites."],"forward_implications":["Proton STAR could reduce the risk of esophagitis and pneumonitis in VT patients, who often have compromised heart and lung function.","Lower mean heart dose may lower the risk of long-term radiation-induced cardiac dysfunction, relevant for patients with prior radiation exposure or structural heart disease.","The improved organ-at-risk sparing provides room for dose escalation to the arrhythmia substrate, potentially improving arrhythmia control.","Pediatric VT patients, who are most vulnerable to late radiation effects, could specifically benefit from proton STAR's reduced low-dose bath.","Patient selection could become anatomy-driven: patients with the arrhythmia substrate close to the esophagus, stomach, or spinal cord may be preferentially referred for protons."],"supporting_citations":[{"why":"Introduces STAR as noninvasive cardiac radioablation for VT, defining the ventricular arrhythmia substrate target concept.","marker":"(10)"},{"why":"First-in-man proton STAR case report, establishing clinical feasibility that this dosimetric comparison extends.","marker":"(17)"},{"why":"Supplies the clinical photon STAR planning methodology, including the target margin expansion used for both photon and proton plans.","marker":"(18)"},{"why":"Provides the single-fraction SBRT organ-at-risk dose constraints used to evaluate both photon and proton plans.","marker":"(19)"},{"why":"Supplies the single-fraction normal-tissue dose limits against which both plans were checked for constraint violations.","marker":"(21)"}],"fun_headline_variants":["Proton STAR cuts heart dose by a third in VT","Protons spare heart, lungs in arrhythmia ablation","Proton therapy beats photons on OAR sparing in STAR","Proton beam lowers cardiac dose in VT radioablation","Proton vs photon STAR: equal target, less heart dose"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The comparison is fair only if the retrospectively planned proton beams were optimized with the same effort and objectives as the clinically delivered photon plans; if protons got more favorable normalization or optimization choices, the organ-at-risk reductions might be an artifact of planning rather than an inherent physical advantage.","fun_headline_variants_meta":{"raw":{"variants":["Proton STAR cuts heart dose by a third in VT","Protons spare heart, lungs in arrhythmia ablation","Proton therapy beats photons on OAR sparing in STAR","Proton beam lowers cardiac dose in VT radioablation","Proton vs photon STAR: equal target, less heart dose"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000725,"raw_usage":{"total_tokens":3352,"prompt_tokens":1146,"completion_tokens":2206,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":762,"completion_tokens_details":{"reasoning_tokens":2125}},"tokens_in":762,"tokens_out":2206,"duration_ms":13904,"temperature":1.0,"reasoning_tokens":2125,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T23:30:20.472082+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-plan the same cohort with both modalities using identical, pre-registered optimization objectives (same target dose normalization, same organ-at-risk constraints, same number of beams or arcs, no manual fine-tuning), then compare DVH metrics; if the significant organ-at-risk differences persist under this symmetric planning protocol, the claim is supported, and if they shrink to insignificance, the reported advantage is a planning artifact.","supporting_citations":[],"review_version":1}