Oncology safety is getting complex

Oncology safety is getting complex Cancer drug safety used to be framed as a simpler question: did the treatment cause the side effect, or did the disease? That question gets harder when the same immunotherapy works differently in different populations, when the same drug moves from metastatic disease into adjuvant use after surgery, and when two active agents are given together from the start. Recent reviews and trial updates suggest oncology pharmacovigilance now has to track not just the event, but also the setting, sequence, and population in which it appears. (cancerbiomed.org) One reason is that treatment effect itself is not uniform across regions. A February 2026 review in *Cancer Biology & Medicine* concluded that first-line immunotherapy outcomes in advanced human epidermal growth factor receptor 2-negative gastric cancer differ between Asian and non-Asian patients, and it linked those differences to variation in tumor biology, molecular subtype, host immunity, *Helicobacter pylori* exposure, diet, and gut microbiome composition. The paper argues that these are not minor background variables; they may be predictors of response. (cancerbiomed.org) That matters for safety because efficacy and tolerability are judged together, not separately. If one population derives more benefit from an immune checkpoint inhibitor, clinicians and patients may accept a toxicity profile that would feel less favorable in a population with smaller gains. Pooled summaries can therefore flatten clinically important regional heterogeneity, especially in multinational development programs where the same aggregate adverse-event rate may sit on top of very different benefit-risk balances. This is an inference drawn from the regional efficacy differences described in the gastric cancer review and from the broader literature on individualized early-stage immunotherapy use. (cancerbiomed.org) The second source of complexity is line of therapy. In metastatic disease, oncologists often tolerate more toxicity because the immediate goal is disease control in a life-threatening setting. In adjuvant therapy, by contrast, treatment is given after surgery to patients who may have no visible disease, so the tolerance for long-lasting immune-related harm is much lower. A 2026 review in *Nature Reviews Clinical Oncology* says adjuvant immune checkpoint inhibition is now standard in several high-risk resected solid tumors, but overall survival benefit has not yet been clearly demonstrated across all tumor types. (nature.com) A separate 2026 meta-analysis summarized in *Bioengineer* pooled 13 randomized trials with nearly 10,000 patients and found that adjuvant programmed cell death protein 1 and programmed death ligand 1 inhibitors improved disease-free survival and distant metastasis-free survival, while not showing a statistically significant overall survival improvement in the pooled population. The same analysis reported higher treatment-related adverse events, including fatigue, nausea, pruritus, and hypothyroidism. That combination of earlier use, uncertain overall survival in some settings, and real toxicity creates a different safety equation than the one used in late-stage disease. (bioengineer.org) This is where pharmacovigilance teams run into a practical problem. An event such as hypothyroidism may be acceptable and manageable in a metastatic setting where a patient is gaining meaningful tumor control, but the same event can look much less acceptable in a curative-intent setting where a patient may live for years with a permanent endocrine consequence. Safety governance therefore has to be context-aware: the same MedDRA term does not carry the same clinical weight in every treatment setting. This is an inference supported by the adjuvant meta-analysis and the Nature review’s discussion of de-escalation and personalization. (bioengineer.org) The third source of complexity is combination therapy. Antibody-drug conjugates are targeted antibodies linked to a chemotherapy payload, and they already bring their own toxicity patterns before immunotherapy is added. When these agents are paired with checkpoint inhibitors, attribution becomes harder because rash, pneumonitis, cytopenias, liver test abnormalities, neuropathy, and fatigue may reflect the payload, the immune activation, overlapping supportive medicines, or the underlying cancer itself. The more active each component is on its own, the less useful simple pooled adverse-event tables become. This attribution challenge is consistent with the combination-treatment landscape described in current breast cancer coverage and broader combination-immunotherapy reviews. (onclive.com) In metastatic triple-negative breast cancer, that shift is already visible. An OncLive interview published in May 2025 described multiple phase 3 programs moving antibody-drug conjugates into the first-line setting, including ASCENT-04, which tested sacituzumab govitecan plus pembrolizumab in previously untreated programmed death ligand 1-positive advanced triple-negative breast cancer, and TROPION-Breast05, which is evaluating datopotamab deruxtecan plus or minus durvalumab against chemotherapy plus pembrolizumab. As these regimens move earlier in care, safety review has to separate what belongs to the conjugate, what belongs to the checkpoint inhibitor, and what emerges only from the interaction between them. (onclive.com) Radioligand combinations add another layer. Radioligand therapy delivers radiation directly to tumor cells through a targeting molecule, which means its adverse effects can overlap only partly with those of immunotherapy. In the phase 1b/2 PRINCE trial reported on April 6, 2026, multicycle lutetium-177 vipivotide tetraxetan plus pembrolizumab in metastatic castration-resistant prostate cancer produced a prostate-specific antigen response of at least 50% in 76% of 37 treated patients, with most treatment-related adverse effects reported as grade 1 or 2 and no grade 4 treatment-related adverse effects or treatment-related deaths attributed to lutetium-177 vipivotide tetraxetan. (cancernetwork.com) Even in that encouraging result, attribution remains unsettled. The PRINCE investigators wrote that radioligand therapy “remains the principal driver of response,” with the immune checkpoint inhibitor potentially adding benefit in a subset of biologically susceptible tumors. That means future safety analysis cannot stop at asking whether the combination is manageable overall; it also has to ask which component is driving benefit, which component is driving harm, and whether biomarkers can identify the smaller group for whom the extra immunotherapy risk is justified. (cancernetwork.com) For pharmacovigilance operations, the implication is a heavier causal burden. Safety teams increasingly have to evaluate adverse events across changing lines of therapy, mixed mechanisms of action, and geographically heterogeneous populations rather than assuming one global label narrative will fit all use cases. Signal detection frameworks built for single agents in refractory disease are under strain when the same molecule is later used after surgery, in biomarker-selected patients, or inside multidrug combinations. (cancerbiomed.org) That likely pushes the field toward more stratified safety governance. Regional analyses may need to sit alongside pooled analyses when efficacy differs by population. Adjuvant and perioperative programs may need lower tolerance thresholds for persistent immune toxicity than metastatic programs. Combination regimens may need component-aware case review, not just regimen-level incidence reporting. Biomarkers such as circulating tumor DNA, which the Nature review highlights as promising for personalizing adjuvant immunotherapy, could eventually help narrow exposure and