Background Current guidelines set out when to start anticancer treatments but not when to stop as disease progresses and death approaches. Older cytotoxic agents are administered intravenously. Most newer drugs, such as Tyrosine-Kinase Inhibitors (TKIs), are oral and widely used in incurable disease with evidence of a few months' survival benefit. New generation agents are considerably more expensive, typically around £5K per month, and perceived as easier to start than to stop.

Aims Systematic literature review examining decisions to stop oral and parenteral anticancer agents in clinical practice, focusing on: a) How are decisions made? b) When are they made? c) Why are they made? d) Who makes them?

Methods Seven electronic databases (Embase, Medline, PsycINFO, CINAHL, ASSIA, Web-of-Knowledge, Cochrane Collaboration), grey literature, references and citations were searched.

Results Forty-nine eligible papers were located. No studies were found concerning criteria for stopping TKIs in practice: most studies examined palliative chemotherapy generally. Key findings: The complexity of chemotherapy regimes means that decisions are not one-off, but rather an on-going process, before all treatments are finally stopped. This may involve switching to alternative therapies and taking breaks. Treatment was stopped when side-effects or burdens became too great. Non-clinical factors were also influential, including physicians' personal experiences, emotions and history. While decision-making was part of the on-going physician-patient interaction, there was variation in who made the final decision: in some studies decisions to continue were patient-led, in others physicians continued treatment if there was evidence of benefit. Some studies presented stopping as ‘biological fact’, treatment being stopped when the patient became too ill.

Conclusions Evidence to inform decision-making concerning stopping palliative anticancer therapy is variable and absent for new generation oral agents such as TKIs. Further research is urgently needed in this area to inform optimal patient care decisions in a resource-constrained NHS.