This week’s strongest evidence raises a recurring problem in oncology evidence generation:
A favorable result can depend heavily on how future outcomes, treatment exposure, and implementation are represented.
New work in health economics, comparative effectiveness, treatment modification, and resource allocation illustrates why understanding the assumptions that could change a result may be as important as estimating the result itself.
Peer-Reviewed Evidence
1. Su et al. Cost-Effectiveness of Perioperative Enfortumab Vedotin and Pembrolizumab for Bladder Cancer. JAMA Network Open.
PubMed: PMID 42900471
This economic evaluation compared perioperative enfortumab vedotin plus pembrolizumab with radical cystectomy alone among cisplatin-ineligible patients with muscle-invasive bladder cancer.
Over a 10-year horizon, the intervention generated an estimated additional 1.31 QALYs at an incremental cost of approximately $95,000, producing an ICER of roughly $72,000 per QALY.
The conclusion, however, was highly sensitive to the analytical horizon and assumptions about subsequent treatment. The estimated ICER exceeded $229,000 per QALY using a five-year horizon but fell below $60,000 with lifetime extrapolation. At a $100,000-per-QALY threshold, the intervention had only a 53% probability of being cost-effective.
Why it matters
This is an unusually clear demonstration that cost-effectiveness is not simply a property of a treatment.
It is partly a property of the model used to represent that treatment over time.
The favorable base-case result depended substantially on projected survival and reductions in downstream treatment costs. Those assumptions may become more or less credible as real-world treatment patterns evolve.
For payer-facing evidence generation, models should therefore identify prospectively which assumptions are responsible for the reimbursement conclusion and which will require validation as additional evidence becomes available.
A base-case ICER is useful. Knowing what would cause that ICER to change is more useful.
2. Gebrael et al. Androgen Synthesis Inhibitors vs Androgen Receptor Inhibitors and Survival in Metastatic Hormone-Sensitive Prostate Cancer. JAMA Network Open.
PubMed: PMID 42900475
This US real-world study compared outcomes among 3,404 patients with metastatic hormone-sensitive prostate cancer receiving either an androgen synthesis inhibitor or a next-generation androgen receptor inhibitor.
After inverse probability of treatment weighting, androgen receptor inhibitors were associated with longer overall survival than androgen synthesis inhibition.
The investigators also performed sensitivity analyses using treatment initiation rather than diagnosis as the index date.
Why it matters
The study provides a useful example of the distinction between statistical adjustment and causal identification.
Patients could initiate treatment across a window surrounding metastatic diagnosis, while the primary analysis measured survival from diagnosis. That makes alignment among eligibility, treatment assignment, and time zero particularly important.
Weighting can address measured differences between treatment groups. It cannot independently eliminate every bias introduced by how treatment exposure and follow-up are defined.
For oncology RWE, the methodological sequence therefore matters:
Define the target comparison first. Determine whether the available data can represent it. Then select the statistical method.
Sophisticated adjustment should not substitute for coherent study architecture.
3. Kim et al. Initial Enfortumab Vedotin Dose Reduction and Outcomes in Pretreated Advanced Urothelial Carcinoma. JAMA Network Open.
PubMed: PMID 42905705
This multicenter observational study examined 368 patients receiving enfortumab vedotin for advanced urothelial carcinoma.
An initial dose reduction of at least 10% was not significantly associated with progression-free or overall survival compared with near-standard starting doses. Severe peripheral neuropathy occurred less frequently among patients starting at reduced doses.
Exploratory analyses, however, suggested that reductions exceeding 20% might be associated with shorter progression-free survival.
The authors appropriately caution that these findings do not establish therapeutic equivalence or support routine initial dose reduction.
Why it matters
Treatment-intensity studies are particularly susceptible to reverse causation.
Cumulative dose intensity is affected by subsequent toxicity, disease progression, treatment discontinuation, and survival. Analyzing cumulative exposure as though it were entirely determined at baseline can therefore produce misleading conclusions.
This study instead focuses on the initial treatment decision, before treatment-related toxicity occurs.
That is a more clearly defined exposure.
The broader lesson for oncology RWE is that investigators should distinguish baseline treatment strategies from adaptations that occur because of subsequent clinical events. When those adaptations matter, time-dependent confounding needs to be considered explicitly.
4. Blackstone et al. Allocation Policy Implementation During a Severe Chemotherapy Shortage. JAMA Network Open.
PubMed: PMID 42895749
This study evaluated implementation of a formal carboplatin allocation policy during the 2023 chemotherapy shortage.
Among 483 patients, independent retrospective review found that approximately one-quarter of actual allocation decisions were inconsistent with the intended policy.
Incomplete documentation was a major contributor.
Actual allocation was also associated with patient sex and treatment location. Those associations were no longer evident when allocation decisions were reconstructed according to the policy’s intended criteria.
Why it matters
A defensible decision framework does not guarantee defensible decisions.
Evidence-based allocation policies, MCDA frameworks, treatment pathways, and payer decision tools all depend on implementation.
Criteria may be explicit and evidence-based while actual decisions diverge because of documentation, workflow, interpretation, or organizational context.
Decision-framework evaluation should therefore extend beyond whether the criteria are theoretically appropriate.
It should also ask:
Did the framework actually govern the decisions that were made?
Implementation fidelity is part of decision validity.
Preprints and Emerging Evidence
5. Sakai et al. SPATNIC Enables H&E-Grounded Cancer-Cell Identification Across Spatial and Single-Cell Transcriptomic Modalities in Colorectal Cancer. bioRxiv. Preprint.
View the bioRxiv preprint
The investigators developed a computational method that uses pathologist-defined annotations from matched H&E sections to identify malignant cells across spatial and single-cell transcriptomic platforms.
The important feature is the use of histopathological ground truth to inform molecular classification rather than relying exclusively on transcriptomic signatures.
Why it matters
Molecular abundance cannot always be interpreted independently of cellular location.
A biomarker signal measured in bulk or spatially complex tissue can arise from malignant epithelial cells, immune populations, stromal cells, or mixtures of those compartments.
Methods that explicitly connect molecular information with histopathological localization may therefore improve biomarker interpretation and reduce the risk of attributing a biological signal to the wrong cellular source.
The broader principle is relevant well beyond transcriptomics:
Where a biomarker is located may sometimes matter as much as how much of it is present.
Evidence status: Preprint; not peer-reviewed. Independent validation, cross-platform reproducibility, and clinical utility remain to be demonstrated.
DCE and DCE-TTO Watch
No newly published oncology DCE or DCE-TTO identified during this review period met our threshold for inclusion.
This week’s evidence nevertheless raises two questions relevant to preference and decision-science research.
The enfortumab vedotin study highlights a potentially important trade-off between treatment intensity, toxicity, and efficacy.
The carboplatin allocation study demonstrates that stated decision criteria and observed choices can diverge substantially during implementation.
Both could inform future preference-elicitation or decision-framework research even though neither is itself a DCE.
What This Means for Evidence Strategy
Three lessons stand out this week.
Economic value is conditional. The bladder-cancer analysis demonstrates how strongly cost-effectiveness can depend on the time horizon, projected downstream treatment, and assumptions about future outcomes.
Treatment exposure must be represented carefully. The comparative-effectiveness and dose-reduction studies show why treatment initiation, subsequent modification, and cumulative exposure cannot automatically be treated as equivalent analytical constructs.
Decision frameworks need post-implementation evaluation. The chemotherapy-shortage study demonstrates that a well-designed policy can produce unintended differences when translated into practice.
Together, these studies reinforce a broader principle:
Decision-grade evidence should identify not only the preferred estimate, but also the conditions under which that estimate would change.
A favorable result is useful.
Understanding its contingencies is more useful.
Worth Reading First
For health economics, Su et al. is the strongest paper this week. Its scenario analyses make the consequences of structural modeling assumptions unusually transparent.
For RWE methodology, Gebrael et al. and Kim et al. are useful companion papers. Both illustrate why defining treatment exposure and time before statistical adjustment is essential to interpretation.
For decision science, Blackstone et al. is particularly valuable because it moves beyond evaluating a decision framework in principle and asks whether the framework actually determined decisions in practice.
References
Su ZT, Garman TS, Rivera Lopez CA, et al. Cost-effectiveness of perioperative enfortumab vedotin and pembrolizumab for bladder cancer. JAMA Network Open. 2026;9(10). doi:10.1001/jamanetworkopen.2026.40370. PMID 42900471
Gebrael G, Jo Y, Nandakumar V, et al. Androgen synthesis inhibitors vs androgen receptor inhibitors and survival in metastatic hormone-sensitive prostate cancer. JAMA Network Open. 2026;9(10). doi:10.1001/jamanetworkopen.2026.39458. PMID 42900475
Kim IH, Kim CG, Jo J, et al. Initial enfortumab vedotin dose reduction and outcomes in pretreated advanced urothelial carcinoma. JAMA Network Open. 2026;9(10). doi:10.1001/jamanetworkopen.2026.39846. PMID 42905705
Blackstone EC, Wang Y, Inyang EA, Shuman AG, Casey M, Abel GA, Hantel A. Allocation policy implementation during a severe chemotherapy shortage. JAMA Network Open. 2026;9(10). doi:10.1001/jamanetworkopen.2026.33164. PMID 42895749
Sakai SA, Nomura R, Ohira M, et al. SPATNIC enables H&E-grounded cancer-cell identification across spatial and single-cell transcriptomic modalities in colorectal cancer. bioRxiv. Posted October 22, 2026. doi:10.64898/2026.10.21.760612. Preprint; not peer-reviewed. Preprint record