Table 4.

Results of meta-regression of longitudinal studies with moderator to study practice effects in studies reporting mean raw test scores

Test	k	k1	$\sum n$	b0 (SE)	95% CI b0	z-value (P) b0	b1 (SE)	95% CI b1	z-value (P) b1
MMSE	14	9	1658	−0.630 (0.436)	(−1.485;0.225)	−1.444 (.149)	0.603 (0.482)	(−0.341;1.547)	1.252 (.211)
Coding/substitution	5	2	75	0.039 (0.207)	(−0.366;0.444)	0.187 (.852)	0.037 (0.370)	(−0.689;0.763)	0.100 (.920)
TMT A	5	2	135	0.037 (0.135)	(−0.228;0.302)	0.274 (.784)	0.300 (0.174)	(−0.041;0.642)	1.725 (.085)
Digit span forward	4	3	228	−0.878 (0.360)	(−1.585;−0.172)	−2.437 (.015)*	0.826 (0.428)	(−0.014;1.665)	1.928 (.054)
Semantic fluency	5	3	280	−0.868 (0.389)	(−1.630;−0.106)	−2.233 (.026)*	0.615 (0.504)	(−0.372;1.602)	1.221 (.222)
Phonemic fluency	8	5	368	−0.765 (0.683)	(−2.103;0.574)	−1.120 (.263)	0.960 (0.864)	(−0.733;2.653)	1.111 (.266)
TMT B	5	2	125	0.210 (0.196)	(−0.174;0.595)	1.072 (.284)	0.080 (0.297)	(−0.502;0.661)	0.268 (.789)
Digit span backward	6	3	258	−0.428 (0.460)	(−1.330;0.474)	−0.930 (.353)	0.441 (0.653)	(−0.840;1.721)	0.675 (.500)
Verbal memory delayed recall	8	4	263	0.054 (0.118)	(−0.177;0.285)	0.456 (.648)	0.224 (0.150)	(−0.070;0.518)	1.493 (.135)
Verbal memory immediate	8	4	192	0.089 (0.109)	(−0.125;0.302)	0.812 (.417)	0.070 (0.143)	(−0.210;0.351)	0.492 (.623)
Visual memory figures immediate	3	1	109	0.121 (0.168)	(−0.209;0.451)	0.718 (.473)	0.499 (0.209)	(0.090;0.908)	2.390 (.017)*
Picture naming	6	1	119	0.004 (0.108)	(−0.208;0.215)	0.033 (.974)	0.320 (0.220)	(−0.111;0.751)	1.454 (.146)
Reading	3	2	51	0.392 (0.301)	(−0.198;0.982)	1.302 (.193)	−0.254 (0.372)	(−0.983;0.474)	−0.684 (.494)

Note: The effect sizes of time effects in patients who had no interval testing (ie, b0) are interpreted based on Cohen’s rules-of-thumb¹⁵. Differences in time effects in patients who did have additional interval testing (vs. the ones who did not) (ie, b1) are summed with this baseline time effect to interpret the effect sizes of change in the patients who had additional interval testing (again based on Cohen’s rules-of-thumb). CI = confidence interval, k = number of included studies in the analysis, k1= number of studies that had additional test assessments between baseline and follow-up, $\sum n$ = total number of included patients in a meta-analysis, SE = standard error, * indicates a p-value <.05. Tests of moderate or high effect size are indicated in bold for estimates of b0 and underlined for b0+b1.