Table 1

Coding method and reliability for benefit paragraphs

		Examples	Reliability
		Examples	Measures	Score
Direct benefits	+1	“This drug may result in tumour shrinkage”	Pearson coefficient based on mean for each form	Benefits possibility: 0.984
		“You may experience relief of your symptoms”
		“Results in animals are promising”
		“This may induce an immune response to your cancer”

	0	“You may or may not benefit”
		“Direct benefit can not be guaranteed”
		“You may not benefit from the procedures”		Direct benefits proportion: 0.982
		“You are the first person to receive this drug”
	−1	“You are unlikely to benefit from the drug”
		“This treatment is not a cure”
		“Despite treatment, your condition will likely worsen”
		“Any improvement will stop when study ends”

Other	A	“Your participation may benefit others with the disease”	Cohen’s kappa	0.981
		“You will be helping scientists to learn whether…”

	P	“The goal of this study is to test the drug’s safety”		0.987
		“The study is designed to determine if…”

	H	“The treatment you receive may even be harmful”		1.00
		“The approach may prove to be toxic”

Agent	+1	“drug”, “gene therapy”, “medicine”, “vaccine”, “therapy”, “treatment”, “immunisation”	Pearson coefficient based on mean for each form	0.987
	0	“approach”, “gene transfer”, “injection”, “method”, “procedure”, “product”, “regimen”, “substance”, “technique”; technical terms (eg “vector”, “virus”, “modified cells”, “Ad-p53”); tradenames (eg “TRICOM”, “TNFerade”), “treated cells”

	−1	Any of the above when preceded by adjectives “experimental”, “investigational”, “research”, “study”, “test” (eg “study drug”, “experimental gene therapy”, etc)