Table 2. Moderator analyses.

Parameter	k (n)	d	95% CI	I2 (%)	p-value
Depression measurement technique					< 0.0001
Interview-based	18 (19,988)	-0.337	-0.485 to -0.188	34	< 0.0001
Self-report	20 (14,359)	-0.078	-0.217 to 0.008	40	0.076
Telomere measurement technique					< 0.0001
Other method than qPCR	7 (1,420)	-0.717	-1.14 to -0.294	0	0.001
qPCR	31 (32,927)	-0.111	-0.175 to -0.046	35	0.001
Source tissue					<0.0001
Leukocytes	33 (22,465)	-0.210	-0.297 to -0.120	44	< 0.0001
Other	5 (11,882)	-0.181	-0.428 to 0.065	20	0.15
Chronic comorbid medical condition					<0.0001
Condition	8 (2,086)	-0.252	-0.455 to -0.050	27	0.015
Other	30 (32,261)	-0.197	-0.290 to -0.123	45	< 0.0001
Age	37 (22,677)	B = 0.005	-0.0002 to 0.011	80	0.06
Gender	37 (34,323)	B = 0.176	-0.141 to 0.493	82	0.28
Smokers in study (%)	19 (12,838)	B = -0.073	-0.969 to 0.823	69	0.87

The effects of all categorical variables were determined using the continuous random effects model in subgroup analyses. The effects of age and percent male gender, which were imputed as a continuous variable, were analyzed using the method of moments random-effects meta-regression model, resulting in a meta-regression coefficient (B). Highlighted p-values represent significant moderator effects on the relationship between depression and telomere length. MDD = major depressive disorder; k = number of studies within subgroup; n = number of subjects represented from all studies k; d = effect size reported as Cohen's d for subgroup; CI = confidence interval; I² = test of heterogeneity within the subgroup, with 0% no observed heterogeneity, 25% low, 50% moderate and 75% high heterogeneity.