 Solution

A valid comparison needs comparable disease stage, risk, systemic treatments and dose prescriptions; otherwise selection bias and confounding can masquerade as treatment benefit. Referral to specialist centres can select younger, wealthier or more complex patients. Randomized controlled trials reduce such biases but can face limited recruitment, patient preference, crossover and lack of clinical equipoise.

Both arms need expert contemporary planning and comparable quality assurance. Centre experience, changing delivery technology and an obsolete comparison arm can distort results. The chosen endpoint must match the proposed benefit: equal tumour control with lower late toxicity may be valuable even without a survival difference. Statistical power, competing mortality, multiple comparisons and long follow-up are crucial for rare late effects, particularly second cancers. Blinding of treatment delivery is difficult, so independent endpoint adjudication and predefined analyses help. Economic analyses must state their perspective and include access and follow-up costs.