Table 1.

Case studies illustrating scenarios requiring causal estimation

Causal estimation scenario	Case study	Treatment changes adjusted for	Causal question being addressed
Interest in efficacy despite pragmatic trial design
	MRC hypertension trial	Addition of alternative treatments due to inadequate blood pressure control, or switches to alternative treatments due to side effects	Treatment efficacy in absence of treatment switches or additions
	SANAD trial	Withdrawal from randomised treatment, addition of alternative treatments, or switches to alternative treatments due to inadequate seizure control	Treatment efficacy in absence of treatment changes due to inadequate seizure control
Trial protocol differs from practice
	Vitamin A trial	Non-receipt of trial drug due to failure of the drug distribution system	Treatment efficacy among those who would have complied with active treatment if randomised to receive it
Contamination (control arm receives intervention)
	Honey trial	More extreme forms of treatment, such as antibiotics, surgery or radiotherapy Patient or clinician decision to switch from conventional dressings to honey	Treatment efficacy in absence of any treatment changes
	PACIFICO trial	Switches from standard to new treatment on disease progression	Treatment efficacy in absence of treatment switches from standard to new treatment on progression
Interest in efficacy if taken as prescribed
	Coronary Drug Project	Failure to take treatment according to prescribed schedule	Treatment efficacy among those who complied with treatment protocol

MRC Medical Research Council, PACIFICO Purine-Alkylator Combination In Follicular lymphoma Immuno-Chemotherapy for Older patients, SANAD Standard and New Antiepileptic Drugs