Nonadherence to treatment protocol in published randomised controlled trials: a review

Table 5 Statistical methods addressing nonadherence to treatment protocol

	Number (%^b)	Definition of analysis set	Number
Reported a statistical method addressing nonadherence to treatment protocol ^a	51 (52)
Variant of PP
	Primary PP analysis described as ‘PP’		18
		Included only those participants who received full randomised intervention	8
			2
		Minimum degree of adherence required	5^c
		Included only those taking treatment at particular time during trial	1
		Excluded participants if they started disallowed medication	1
			1
	Primary PP analysis described as ‘ITT’ or ‘modified ITT’ analysis		23
		Included only those participants who received full randomised intervention	1
		Included only those participants who received at least one dose of randomised intervention	16
		Included only those participants who received the single treatment	3
		Excluded participants if they deviated from treatment protocol	3
	Sensitivity analysis		12
		Included only those participants who received full randomised intervention	4
		Excluded participants if they received disallowed treatments	2
		Minimum degree of adherence required	2^d
		Censored participants at the point of deviation from treatment protocol	4^e
	Inverse probability of censoring weighted method		1^f
	Subgroup analysis		2^g
	Unlabelled analysis		1^h
As treated analysis			3
Discontinuation of treatment analysed as treatment failure			3
Time to treatment discontinuation included as trial outcome			4
Did not report a statistical method to address nonadherence to treatment protocol	47 (48)

^aNine trials carried out two methods of analysis, two trials carried out three methods and one trial carried out four methods. ^bNumber of trials reporting some form of nonadherence (98) is used as denominator for %. ^cAdherence thresholds used were 60%, 2/3, 75%, 80% and 90%. ^dAdherence thresholds used were 2/3 and 5/6. ^eCensoring times: time of starting disallowed intervention, when participants reported taking less than 2/3 of their medication in the past year, when received treatment out of trial, or censored following six-month lag after receiving less than 80% of drug. ^fCensored when received <80% of drug, weighted by the inverse probability of each participant’s estimated adherence probability. ^gAnalysis split into two groups (according to whether participants had taken more or less than 50% of the prescribed medication) in one trial and into three groups (according to the proportion (>90%, 75-90% or <75%) of their time at risk that they were supplied with drug) in other trial. ^hIncluded if received at least one dose of treatment. ITT: intention to treat; PP: per protocol.

ISSN: 1745-6215