Table 1.

Cost-effectiveness of 10TT under different MNAR assumptions for missing quality-of-life data

Scenario number	MNAR rescaling parametersa		Incremental costb (£) [95% CI]	Incremental QALYs [95% CI]	INMBc (£) [95% CI]	Probability cost-effectivec (%)
	c control	c 10TT
1 (MAR)	1	1	− 35 [− 504 to 434]	− 0.004 [− 0.074 to 0.066]	− 49 [− 1632 to 1534]	48
2	1	0.95	− 35 [− 504 to 434]	− 0.037 [− 0.107 to 0.032]	− 713 [− 2280 to 853]	19
3	0.95	1	− 35 [− 504 to 434]	0.026 [− 0.044 to 0.095]	550 [− 1022 to 2121]	75
4	0.95	0.95	− 35 [− 504 to 434]	−0.008 [− 0.076 to 0.061]	− 115 [− 1670 to 1440]	44
5	0.95	0.90	− 35 [− 504 to 434]	− 0.041 [− 0.109 to 0.027]	− 780 [− 2321 to 762]	16
6	0.90	0.95	− 35 [− 504 to 434]	0.022 [− 0.046 to 0.091]	484 [− 1063 to 2030]	73
7	0.90	0.90	− 35 [− 504 to 434]	− 0.011 [− 0.078 to 0.057]	− 181 [− 1714 to 1352]	41

All results are based on imputed data and comparing the 10TT arm to the control arm (n = 537). For participants with complete cost and effectiveness data (n = 166; 31%), the observed incremental cost was − £65 [95% CI − 924 to 794], incremental QALYs was − 0.040 [− 0.169 to 0.088], INMB was − £741 [− 3645 to 2163], and probability cost-effective was 31%

CI confidence interval, INMB incremental net monetary benefit, MAR missing at random, MNAR missing not at random, QALY quality-adjusted life year, 10TT Ten Top Tips

^aHow missing quality-of-life data are assumed to differ from the MAR-imputed values. c _control = 0.9 means that all imputed quality-of-life values in the control arm have been reduced by 10%

^bMissing costs assumed to be MAR in all scenarios

^cAt a cost-effectiveness threshold of £20,000/QALY