Abstract
Evidence-based surgery has traditionally been framed through drug-development paradigms in which randomized controlled trials (RCTs) occupy a central position. Yet surgical innovation raises specific methodological challenges, including operator dependence, evolving techniques, learning curves, centre effects, and the centrality of patient-centred outcomes. This perspective article examines how scientific evidence in surgery has been constructed historically at the intersection of mechanistic reasoning and empirical observation, and argues that contemporary practice requires a pluralistic approach to methodology. Drawing on the history and philosophy of medicine, we trace the emergence of modern evidence-based medicine and the limitations of transposing its conventional hierarchies uncritically to surgical care. We then discuss complementary frameworks for evidence generation in surgery, including the IDEAL model for staged evaluation of innovation, prospective observational studies and registries, Bayesian methods, artificial intelligence and big-data approaches, qualitative and mixed-methods research, and innovative trial designs adapted to surgical realities. Rather than opposing these approaches to randomized trials, we propose a coherent model in which different methods address distinct dimensions of "surgical truth", from physiopathological plausibility and causal inference to implementation, patient experience, and real-world effectiveness. In this perspective, scientific validity in surgery is best approached through the integration of multiple forms of evidence within adaptive methodological pathways, particularly in oncologic surgery where interventions are complex, innovation is rapid, and the balance between survival, function, and quality of life is central.