Document resource
In nature, herding reflects adaptive responses to environmental pressure. In predator-dense African savannahs, animals herd defensively for protection. In resource-scarce deserts, herding occurs around limited resources. Herding in oncology drug development reflects these same evolutionary pressures: defensive clustering around promising targets to reduce risk.1 The analysis by Zhang and colleagues2 provides an empirical test of when this adaptive strategy becomes maladaptive.Using TIGIT inhibitors as an exemplar, the authors quantify the human and financial capital consumed by herding. Zhang et al identified 30 agents across 220 clinical trials which enrolled ~49 000 patients globally with none of the agents receiving regulatory approval to date. This moves the discussion beyond anecdote to evidence, appropriately foregrounding the human cost of large-scale redundancy.The challenge of herding lies in duration and scale. Since the success of pembrolizumab, with at least 14 PD-1/PD-L1 inhibitors now approved, herding has become increasingly structural rather than episodic.3 Early signal amplification, combined with capital market incentives and regulatory precedent, has encouraged rapid programme convergence before biological uncertainty is adequately resolved. The TIGIT experience illustrates how such convergence can magnify attrition rather than mitigate it.Why herding exists Herding in oncology drug development does not primarily reflect poor judgement but predictable behaviour under uncertainty. When preclinical models and biomarkers cannot reliably distinguish success from failure, parallel clinical exploration becomes a rational substitute. At the same time, competitive market structures reward speed, ownership and first-to-market positioning, while offering little incentive for data sharing that could collectively reduce duplication. Not all clustering is equivalent: early simultaneous pursuit of a promising target can function as distributed validation, whereas sequential entry after signal clarity runs the risk of producing escalating redundancy with diminishing informational value, unless lessons learnt from the initial agent(s) can lead to improvement in efficacy, toxicity or both. Herding can de-risk investment, accelerate collective learning, expand global trial access and ultimately drive approvals that increase availability and reduce cost over time ( table 1).Table 1Benefits and emerging costs and risks associated with herding in oncology drug development, from the perspective of patients, clinicians, industry and regulators/payersStakeholderBenefits of herdingEmerging costs and risksPatientsEarlier access to novel therapiesIncreased geographic availabilityHope when standard options are exhausted/not availableCumulative toxicity burden across similar programmesOpportunity cost: enrolment into trials and drugs that may not have efficacy/may not be approvedCliniciansFamiliarity with mechanisms and toxicity profiles before the drug is approvedMore equitable and earlier access to novel therapiesFaster trial activation and greater opportunities for trial participation and contributionTrial capacity consumed by redundant programmesReduced ability to prioritise biologically innovative studiesEthical tension when equipoise erodesIndustry sponsorsDe-risked investment via validated targets Faster development timelines Benchmarking against competitorsAmplified financial loss due to parallel attrition Limited differentiation in crowded markets Escalating development costs with diminishing returnsRegulators and payersLarger datasets across populations Earlier signal validation through replicationCompetitive pricing due to competitionDifficulty distinguishing incremental benefit Downstream reimbursement challengesEfforts to mitigate herding must therefore confront an ethical tension; limiting duplication may improve sustainability but risks constraining access, slowing price competition and prematurely narrowing therapeutic diversity. The challenge is not to eliminate convergence but to recognise when it transitions from necessary exploration to avoidable repetition.The hidden costs From a clinical perspective, prolonged herding introduces additional costs that are not fully captured by trial budgets. Large numbers of individuals were exposed to near-identical regimens, often with overlapping immune-related toxicities and limited mechanistic differentiation. While toxicity was not the central focus of the analysis, the cumulative burden of immune-mediated adverse events across multiple similar programmes is non-trivial. Moreover, as the authors note, the globalisation of trials, while increasing accrual, raises ethical concerns when patients in low/middle-income countries participate in studies with limited likelihood of downstream access. 4These observations point to a central tension in contemporary drug development: access today versus sustainability tomorrow. Herding can expand patient access, but it also consumes finite resources. Investigators, trial sites and eligible patients are increasingly absorbed by an ever-expanding pipeline of near-identical programmes, potentially crowding out more differentiated or higher-risk innovation.This tension is not confined to immunotherapy. The antibody drug conjugate (ADC) field provides a forward-looking parallel. With >300 ADC trials activated globally, intense clustering has emerged around a limited number of payload classes, particularly topoisomerase I inhibitors.5 While this has accelerated ADC maturation, it has also led to trial pipeline saturation, overlapping toxicity profiles and growing challenges in patient selection and sequencing. Without earlier diversification in payloads, biomarkers and trial design, the ADC ecosystem risks repeating the TIGIT trajectory, but at an even greater scale.Towards adaptive dispersion From our perspective in oncology drug development, the solution is not to eliminate herding but to recognise its ecological limits. We need mechanisms enabling timely dispersion once biological uncertainty narrows. Four coordinated mechanisms could help ( figure 1):Smarter trial design: Platform trials, eg, I-SPY6 with adaptive cohorts and clear futility boundaries would allow rapid elimination of ineffective agents while preserving infrastructure.Regulatory efficiency: Initiatives like Project Orbis accelerate approval of truly novel agents7 and could reduce reliance on ‘me-too’ programmes as access vehicles. Regulators might also require explicit differentiation justification for programmes entering crowded spaces.Global coordination: Shared situational awareness regarding emerging pipelines, global coordination via trials consortia and pre-registration with mandatory reporting, enforced through funding and regulatory mechanisms, would prevent repeated failures.8Patient-centred execution: patients are not members of the herd; they are frequently the limiting resource and therefore need to be active participants in trial design.9 Patient advocacy-driven initiatives such as dose optimisation studies demonstrate how incorporating patient priorities (tolerability, quality of life and trial burden) can shift design toward meaningful differentiation.Figure 1A systems framework to mitigate maladaptive herding in oncology drug development. Rather than eliminating herding, sustainable oncology drug development requires coordinated mechanisms that enable timely dispersion once biological uncertainty narrows. Smarter trial design, regulatory agility, global coordination and patient partnership act synergistically to preserve clinical trial capacity, protect patients as the limiting resource and shift incentives toward differentiation, durability, reduced toxicity and meaningful clinical impact.The analysis by Zhang and colleagues provides a crucial empirical foundation for this conversation. The task now is to ensure that future waves of innovation learn from this experience. In nature, herds are not judged by how tightly they cluster but by whether they move in ways that sustain the landscape that sustains them. The future of sustainable oncology innovation depends on recognising when collective exploration must give way to deliberate differentiation.