Table 5.

Moderators of between-articles variance of true effects

Moderator	Type	Number of effects	Number of articles	P value	Marginal R2	Between-articles variance
						τ2	I2	Explained
Without moderators		644	101			2.8%	42.9%	n.a.
U-87 MG source	cat.	644	101	0.075	n.s.	2.6%
U-87 MG authentication	cat.	644	101	0.476	n.s.	2.8%
U-87 MG age (cell passages)	cont.	138	11	0.238	n.s.	1.5%
Cell concentration	cont.	113	20	0.323	n.s.	3.8%
Confluence level at cell passaging	cont.	57	11	0.319	n.s.	0.8%
Glucose level of culture medium	cat.	644	101	0.016	7.0%	2.5%*	40.1%	10.9%
Mycoplasma exclusion	cat.	644	101	0.491	n.s.	2.8%
Supplemented antibiotics	cat.	644	101	0.094	n.s.	2.6%
FBS source	cat.	644	101	0.067	n.s.	2.6%
Type of untreated control	cat.	644	101	0.370	n.s.	2.8%
Articles reporting quality	int.	644	101	0.031	3.3%	2.6%†	41.7%	5.0%
TMZ conc.	cont.	644	101	<0.001	38.6%	3.4%‡	64.3%	0.0%
Treatment duration	cont.	644	101	<0.001	6.0%	2.9%§	45.7%	0.0%

Random-effects three-level meta-regressions with the raw data the effects were calculated with as first level, the reported effects as second level and the articles the effects were reported in as third level. Marginal R² indicates the regression model fit⁶⁹; τ²: estimator of the variance of true effects; τ: square root of τ²; I²: proportion of between-articles variance of the total observed variance, including sampling error. τ² estimator: restricted-maximum likelihood. The column ‘Explained’ indicates the reduction of τ² after including the particular moderator compared with τ² without moderators (only applicable if the number of included effects and articles is identical). For some continuous moderators, the number of effects and articles included in the regression is reduced due to non-reporting, which leads to a limited comparability of τ² between parameters with different numbers of belonging articles and effects. ‘Not reported’ was included as a category for categorical moderators. P value is given for the test of the moderator. R², I² and the explained heterogeneity were only calculated for moderators that prove significance in the test of the moderator (α=0.05).

*95% CI of τ²: 1.7% to 3.6%.

†95% CI of τ²: 1.8% to 3.8%.

‡95% CI of τ²: 2.5% to 4.8%.

§95% CI of τ²: 2.1% to 4.2%.

cat., categorical; cont., continuous; FBS, fetal bovine serum; int., interval; n.a., not applicable; n.s., not significant; TMZ conc., temozolomide concentration; U-87 MG, Uppsala-87 Malignant Glioma.