Table 3.

Pooled effect sizes of randomized controlled trials of the effect of treatment on lymphedema volume.

	Sample sizea		Heterogeneityb				Mean diff (ml)	Combined effect sizec
	K	N	Q	p	I2	Tau2		Hedge’s gd	95% CI	p	95% PIe
Overall effect (all interventions vs. all comparisons)	40	1970	184.1	<0.001	0.79	0.31	−6.1	0.20	0.00–0.40	0.047	−0.95 to 1.35
Overall effect (w. Sensitivity analysis)f	37	1857	106.4	<0.001	0.66	0.16	−4.5	0.24	0.08–0.40	0.004	−0.59 to 1.07
Overall effect vs. active control	30	1573	165.6	<0.001	0.83	0.38	−10.5	0.20	−0.05 to 0.44	0.12	−1.09 to 1.49
Overall effect vs. non-active control	10	397	18.3	0.032	0.51	0.11	−3.8	0.23	−0.06 to 0.52	0.12	−0.61 to 1.07
Exercise vs. all comparisons	13	684	21.4	0.022	0.44	0.06	−7.2	0.26	0.04–0.47	0.022	−0.33 to 0.85
Kinesio taping vs. all comparisons	4	146	15.9	<0.001	0.81	0.50	+453.4	−0.46	−1.23 to 0.31	0.24	−3.94 to 3.02
Laser vs. all comparisons	4	179	11.8	0.008	0.75	0.28	−5.4	0.50	−0.11 to 1.12	0.11	−2.15 to 3.15
CDP (Complex Decongestive physiotherapy) (vs. all)	13	685	102.3	<0.001	0.88	0.62	−19.6	0.31	−0.15 to 0.77	0.18	−1.49 to 2.12
MLDg + compression vs. compression	6	336	30.1	<0.001	0.83	0.38	+5.7	−0.20	−0.75 to 0.35	0.48	−2.08 to 1.68
MLD + compression + pump vs. MLD + compression	3	159	12.5	<0.001	0.84	0.56	−4.6	0.36	−0.57 to 1.29	0.44	−10.90 to 11.62

P-values of <0.05 are marked with bold.

^aK indicates number of published studies.

^bQ statistic: p values <0.1 taken to suggest heterogeneity; I² statistic: 0% (no heterogeneity), 25% (low heterogeneity), 50% (moderate heterogeneity), 75% (high heterogeneity).

^cEffect size = Hedge’s g. Standardized mean difference, adjusting for small sample bias. A positive value indicates an effect size in the hypothesized direction, e.g., larger reduction in volume or a relative smaller increase in volume in the intervention group. To ensure independency, if a study reported results for more than one measure, the effect sizes were combined (mean), ensuring that only one effect size per study was used in the calculation.

^dIf analyses (Eggers test: p < 0.05) had indicated the possibility of publication bias, missing studies would have been imputed and an adjusted effect size rate calculated with a random effects model. However, for the effects with a sufficient number of studies (K ≥ 10), there were no robust indications of publication bias.

^ePI = 95% prediction interval: The interval in which 95% of future observations will fall, given the observed data, calculated for heterogeneous ESs (I² > 0).

^fSensitivity analysis(Hedges’ g): Omitting three studies (Andersen et al., 2000, Smykla et al. 2013, and Sitzia et al., 2002) with effect sizes 2 standard deviations away from the pooled effect size (0.22 ± 1.57: <−1.35 OR >1.79). Sensitivity analyses (mean diff. ml): Omitting one study (Smykla et al., 2013).

^gMLD = manual lymphatic drainage.