Table 6.

Moderators of light treatment effect on phase shift by using subgroup analyses with mixed-effects models.

(a)
Subgroup	k	Hedges’s g	95% CI		p (Z)a	I2
			Lower	Upper
Intensity of light treatment (lux)
Group 1
< 1000	2	0.944	0.423	1.465	< 0.001	0
1000–5000	8	0.983	0.611	1.356	< 0.001	0
> 5000	2	2.676	1.705	3.648	< 0.001	0
Group 2
< 2000	3	0.677	0.199	1.154	0.005	0
2000–5000	7	1.134	0.829	1.440	< 0.001	0
> 5000	2	2.676	1.705	3.648	< 0.001	0
Group 3
< 3000	4	0.710	0.362	1.058	< 0.001	0
3000–5000	6	1.204	0.853	1.555	< 0.001	0
> 5000	2	2.676	1.705	3.648	< 0.001	0
Treatment duration (daily hr)
≤ 1	3	0.952	0.206	1.698	0.012	0.743
> 1	9	1.166	0.688	1.644	< 0.001	17.873
Treatment duration (total hr)
≤ 10	7	1.182	0.865	1.499	< 0.001	0
> 10	5	1.166	0.282	2.050	0.010	30.014
Number of daily treatment sessions (n)
1	9	1.137	0.607	1.667	< 0.001	18.492
> 1	3	1.130	0.627	1.633	< 0.001	0

(b)
Subgroup	k	Hedges’s g	95% CI		p (Z)a	I2
			Lower	Upper
Timing of treatment
Night time	6	1.528	0.950	2.107	< 0.001	7.083
Day time or mixed	7	0.798	0.390	1.206	< 0.001	0
Treatment day
Same day	4	0.967	0.334	1.599	0.003	0
Different day	9	1.137	0.688	1.586	< 0.001	22.871
Type of treatment light
Single type of light	10	1.141	0.741	1.541	< 0.001	21.004
Mixed type of light	3	0.889	0.062	1.717	0.035	0
Control treatment
Active treatment	11	1.216	0.848	1.584	< 0.001	8.3649
No treatment	2	0.245	− 0.389	0.878	0.449	0
Shift work
Night shift	10	1.164	0.720	1.607	< 0.001	11.668
Rotating shift	3	0.892	0.193	1.592	0.012	5.7388
Simulated shift
Yes	9	1.379	1.041	1.718	< 0.001	7.0895
No	4	0.436	0.071	0.802	0.019	0
Timing of measure
Night time	8	1.194	0.635	1.753	< 0.001	16.256
Day time or mixed	5	1.016	0.592	1.441	< 0.001	0.110
Research design
Randomized controlled trial	7	1.447	0.869	2.025	< 0.001	14.171
Randomized crossover trial	6	0.748	0.423	1.072	< 0.001	0
Data handling
Intention-to-treat	6	1.128	0.398	1.859	0.002	26.446
Per protocol or unknown	7	1.127	0.827	1.429	< 0.001	0

(c)
Subgroup	k	Hedges’s g	95% CI		p (Z)a	I2
			Lower	Upper
Risk of bias
Incomplete outcome data
Low risk	10	1.150	0.718	1.583	< 0.001	15.024
High risk	3	0.857	0.223	1.491	0.008	9.249

^aP value for the Z test of the moderator effect.