Table 3.

Outcomes after Crossover (N = 82 patients with complete data)

Outcome Measures	Pre-Post	Unmatched Treatment		Group Effect	Interaction	Time Effect (n=82 with
		Groups		(F df)	(F df)	complete data) (F df)
		Opposite	EBC			Intention to Treat* (n=96)
		Mean	Mean			(F df)
		(SD)	(SD)
Back pain/10	Pre	5.09	5.33	0.05 1,80	1.41 1,80	125.11 1,80(p<0.001)
		(2.28)	(2.18)	(p=0.83)	(p=0.52)	14.62 1,94 (p<0.001)
	Post	2.41	2.33
		(1.92)	(1.77)
Leg pain/10**	Pre	3.27	3.11	0.01 1,75	0.70 1,75	63.29 1,75 (p<0.001)
		(2.95)	(3.21)	(p=0.91)	(p=0.41)	55.12 1,93 (p=0.001)
	Post	1.00	1.28
		(1.86)	(1.99)
RMDQ: Total yes/24	Pre	15.09	17.76	3.18 1,78	0.01 1,78	34.61 1,78 (p<0.001)
		(6.75)	(5.90)	(p=0.08)	(p=0.93)	42.37 1,94 (p<0.001)
	Post	10.96	13.50
		(7.74)	(8.20)
Medication: Total	Pre	2.56	2.07	0.001 1,29	1.38 1,29	8.24 1,29 (p=0.008)
pills/day**		(2.76)	(1.83)	(p=0.98)	(p=0.25)	10.46 1,48 (p=0.002)
	Post	0.81	1.33
		(1.33)	(1.76)
During the past week,	Pre	3.13	3.34	1.01 1,79	0.04 1,79	78.12 1,79 (p<0.001)
pain interference		(1.09)	(0.97)	(p=0.32)	(p=0.85)	70.88 1,94 (p=0.001)
with usual work, in	Post	2.17	2.34
and outside home:		(0.85)	(0.99)
rating (0–5)
BDI-11	Pre	6.31	4.71	1.51 1,74	0.23 1,74	34.19 1,74 (p<0.001)
		(7.16)	(4.17)	(p=0.22)	(p=0.63)	33.67 1,94 (p=0.001)
	Post	3.74	2.53
		(5.22)	(3.18)
Taking medication (Yes)	Pre	10	8	McNemar test used for time effect (p=0.040)
		(52.6%)	(66.7%)	Groups collapsed due to small n (p=0.002)
	Post	6	5
		(31.6%)	(41.7%)

Opposite = evidence-based advice and exercises opposite to DP; EBC = evidence-based advice and common back exercises; RMDQ = Roland Morris Disability Questionnaire; BDI = Beck Depression Inventory

*Intention-to-treat analysis using carry-forward of last known data point

**Floor effect: those with no leg pain/no medication at baseline were excluded from analysis.