Table 2. Summary risk estimates and I2 measures of heterogeneity from random-effects models stratified by selected study characteristics.
| Stratification criterion | Number of included RRs | RR (95% CI) (high vs low PA) from random-effects model | I2 (%) | P-valuea |
|---|---|---|---|---|
|
Methodologic qualityb | ||||
| RRs within upper tertile of quality score | 11 | 0.78 (0.66, 0.92) | 33 | |
| RRs within intermediate tertile of quality score | 12 | 1.00 (0.89, 1.13) | 0 | |
| RRs within lower tertile of quality score |
14 |
0.93 (0.80, 1.07) |
30 |
0.02 |
|
PA assessment | ||||
| RRs based on qualitative PA assessments | 18 | 0.98 (0.85, 1.14) | 35 | |
| RRs based on energy expenditure | 6 | 0.97 (0.84, 1.12) | 0 | |
| RRs based on MVPA duration | 6 | 0.85 (0.69, 1.04) | 43 | |
| RRs based on MVPA frequency |
6 |
0.72 (0.53, 0.97) |
53 |
0.24 |
|
PA domain | ||||
| RRs based on total activity | 4 | 0.95 (0.76, 1.20) | 0 | |
| RRs based on occupational activity | 14 | 0.91 (0.79, 1.04) | 21 | |
| RRs based on recreational activity |
19 |
0.88 (0.77, 1.00) |
40 |
0.84 |
|
Timing in life of PA | ||||
| RRs based on recent PA | 16 | 0.83 (0.74, 0.93) | 28 | |
| RRs based on consistent PA over time | 11 | 0.96 (0.79, 1.15) | 0 | |
| RRs based on past PA |
10 |
1.01 (0.84, 1.20) |
46 |
0.18 |
|
Gender | ||||
| RRs among men | 17 | 0.93 (0.84, 1.02) | 2 | |
| RRs among women | 9 | 0.95 (0.66, 1.36) | 57 | |
| RRs among men and women |
11 |
0.85 (0.73, 0.98) |
42 |
0.41 |
|
Study design | ||||
| RRs from case–control studies | 18 | 0.91 (0.79, 1.04) | 36 | |
| RRs from cohort studies |
19 |
0.89 (0.80, 0.99) |
19 |
0.93 |
|
Study region | ||||
| RRs from studies in North America | 18 | 0.85 (0.77, 0.94) | 0 | |
| RRs from studies in Europe | 16 | 0.95 (0.80, 1.12) | 51 | |
| RRs from studies in Asia |
3 |
1.00 (0.83, 1.20) |
0 |
0.63 |
|
Number of adjustment factorsc | ||||
| RRs within upper tertile of number of adjustment factors | 12 | 0.83 (0.71, 0.97) | 40 | |
| RRs within intermediate tertile of number of adjustment factors | 4 | 0.87 (0.68, 1.10) | 52 | |
| RRs within lower tertile of number of adjustment factors |
21 |
0.96 (0.85, 1.08) |
14 |
0.28 |
|
Adjustment for smoking and obesity | ||||
| RRs adjusted for smoking and obesity | 23 | 0.92 (0.82, 1.03) | 37 | |
| RRs adjusted for smoking but not obesity | 3 | 0.71 (0.54, 0.94) | 0 | |
| RRs adjusted neither for smoking nor obesity |
11 |
0.89 (0.78, 1.01) |
2 |
0.31 |
|
Adjustment for hypertension | ||||
| RRs adjusted for hypertension | 12 | 0.85 (0.73, 0.97) | 30 | |
| RRs not adjusted for hypertension |
25 |
0.93 (0.83, 1.03) |
24 |
0.30 |
|
Adjustment for diabetes | ||||
| RRs adjusted for diabetes | 5 | 0.81 (0.66, 0.99) | 57 | |
| RRs not adjusted for diabetes | 32 | 0.92 (0.84, 1.01) | 14 | 0.18 |
Abbreviations: CI=confidence interval; MVPA=moderate-to-vigorous physical activity; PA=physical activity; RR=relative risk.
a
P-values for effect heterogeneity across strata were obtained from random-effects meta-regression comparing the model including the stratification variable as a single explanatory variable with the null model not including any explanatory variables.
b
The quality scores ranged from 45 to 83 percentage points (out of 100 percentage points), with lower and upper tertile cutoffs of 62 percentage points and 71 percentage points, respectively.
c
The number of adjustment factors (not counting adjustments for age and sex) ranged between 0 and 12, with lower and upper tertile cutoffs of 3 and 5, respectively.