Skip to main content
. 2019 Jun 5;6(6):190420. doi: 10.1098/rsos.190420

Table 2.

Results from random-effects meta-analytic models on data from 18 cohorts.

model summary SMD
heterogeneity statistics
forest plot
estimatea [95% CI] p-value Q17 p-value τ2 I2 (%) H2
1 difference in LTL between time-points −0.49 [−0.62, −0.36] <0.0001 333.73 <0.0001 0.07 95.89 24.31 electronic supplementary material, figure S2a
2 assn. between smoking and baseline LTL −0.10 [−0.16, −0.05] 0.0003 26.55 0.0650 0.01 44.70 1.81 figure 2a
3 assn. between smoking and follow-up LTL −0.15 [−0.29, −0.01] 0.0312 95.13 <0.0001 0.06 89.78 9.78 figure 2b
4 difference in assn. between smoking and LTL between time-points −0.05 [−0.16, 0.06] 0.3884 73.45 <0.0001 0.04 90.41 10.43 figure 3a
5 combined assn. between smoking and LTL across time-points −0.13 [−0.22, −0.03] 0.0075 85.62 <0.0001 0.03 83.96 6.24 not shown
6 assn. between smoking and LTL attrition −0.02 [−0.07, 0.04] 0.5311 24.68 0.1021 0.00 36.48 1.57 figure 4a

aNegative parameter estimates for the summary standardized mean difference (SMD) correspond to: model 1—shorter LTL at follow-up; models 2, 3 and 5—shorter LTL in smokers; model 4—stronger assn. between smoking and LTL at follow-up; model 6—faster attrition in smokers. Significant associations are shown in italics.