Table 3

Physicians' assessment of a hypothetical drug (‘neostatin’) by effect formats (number needed to treat [NNT] versus survival gain) and effect level

	Survival gain scenarios	NNT scenarios	Both scenarios combined	Differences
	Proportions recommending ‘neostatin’, % (95% CI)			% (95% CI) χ² statistics

Level 1, NNT = 34, survival gain = 9 months	39.5 (28.7 to 50.3)	25.6 (15.8 to 35.4)	32.7 (25.4 to 40.0)	13.9 (-0.5 to 28.3) χ² = 3.5, P = 0.062

Level 2, NNT = 17, survival gain = 17 months	49.3 (37.9 to 60.8)	31.3 (21.0 to 41.5)	40.0 (32.2 to 47.8)	18.1 (2.9 to 33.3) χ² = 5.3, P = 0.022

Level 3, NNT = 9, survival gain = 32 months	52.4 (41.5 to 63.3)	43.2 (32.3 to 54.1)	47.9(40.1 to 55.6)	9.2 (-6.0 to 24.5) χ²= 1.4, P= 0.238

Trend analysis^a
Odds ratio per level (95% CI)	1.3 (1.0 to 1.8)	1.5 (1.1 to 2.1)
Odds ratio for interaction term format × level (95% CI)	1.2 (0.7 to 1.8)

	Evaluation of ‘neostatin’ on a 0-10 scale,^b score (95% CI)			Mean (95% CI), rank sum statistics

Level 1, NNT = 34, survival gain = 9 months	5.2 (4.6 to 5.8)	4.7 (4.1 to 5.3)	5.0 (4.6 to 5.4)	0.5 (-0.3 to 1.4) \|z\| = 1.3, P= 0.189^c

Level 2, NNT = 17, survival gain = 17 months	6.2 (5.7 to 6.8)	5.0 (4.4 to 5.6)	5.6 (5.2 to 6.0)	1.2 (0.4 to 2.0) \|z\| =3.1, P= 0.002^c

Level 3, NNT = 9, survival gain = 32 months	6.1 (5.6 to 6.6)	5.5 (4.9 to 6.2)	5.8 (5.4 to 6.2)	0.6 (-0.3 to 1.4) \|z\| = 1.1, P= 0.264^c

Trend analysis^d
Regression coefficient per level	0.42 (0.03 to 0.81)	0.41 (-0.03 to 0.85)
Regression coefficient for interaction term format x level (95% CI)	-0.01 (-0.59 to 0.58)

↵a Logistic regression analysis of trend across the three levels of effect size.
↵b 0 = ‘a very poor choice’, 10 = ‘a very good choice’.
↵c Mann-Whitney rank sum test.
↵d Ordinary least squares regression analysis with robust variances.