Table 4. Meta-analysis results.
| Type of study | No. of studies (patients) | RR (95% CI) random effects | RR (95% CI) fixed effects | Pheterogeneity | I2heterogeneity (%) | Psignificance between groups |
|---|---|---|---|---|---|---|
| All studies (adjusted) |
8/5446 |
0.64 (0.53–0.77) |
0.64 (0.53–0.77) |
0.51 |
0 |
|
| All studies (EAC only) |
5/3036 |
0.70 (0.52–0.95) |
0.70 (0.52–0.95) |
0.24 |
28 |
|
| All studies (unadjusted) |
7/3950 |
0.70 (0.56–0.86) |
0.69 (0.57–0.85) |
0.36 |
7 |
|
|
Study design | ||||||
| Cohort studies | 6/4353 | 0.61 (0.49–0.76) | 0.61 (0.49–0.76) | 0.93 | 0 | 0.41 |
| Case–control studies |
2/1093 |
0.86 (0.35–2.07) |
0.74 (0.50–1.09) |
0.04 |
76 |
|
|
Study design in cohort studies | ||||||
| Retrospective | 3/1892 | 0.62 (0.44–0.87) | 0.62 (0.44–0.87) | 0.72 | 0 | 0.92 |
| Prospective |
3/2461 |
0.61 (0.46–0.81) |
0.61 (0.46–0.81) |
0.71 |
0 |
|
|
Medication type | ||||||
| Aspirin | 4/2152 | 0.63 (0.43–0.94) | 0.63 (0.43–0.94) | 0.64 | 0 | 0.44 |
| Non-aspirin COX inhibitors |
3/1416 |
0.50 (0.32–0.78) |
0.50 (0.32–0.78) |
0.49 |
0 |
|
|
Duration response | ||||||
| Longer duration (>1 year) | 4/2002 | 0.54 (0.36–0.79) | 0.54 (0.36–0.79) | 1.00 | 0 | 0.42 |
| Shorter duration (<1 year) | 2/1382 | 0.67 (0.46–0.97) | 0.67 (0.46–0.97) | 0.67 | 0 | |
| Longer duration (>2 years) | 3/1190 | 0.53 (0.33–0.87) | 0.53 (0.33–0.87) | 0.99 | 0 | 0.46 |
| Shorter duration (<2 years) | 2/1382 | 0.67 (0.46–0.97) | 0.67 (0.46–0.97) | 0.67 | 0 | |
| Longer duration (>3 years) | 2/694 | 0.54 (0.30–0.99) | 0.54 (0.30–0.99) | 0.95 | 0 | 0.63 |
| Shorter duration (<3 years) | 3/1726 | 0.64 (0.46–0.90) | 0.64 (0.46–0.90) | 0.75 | 0 | |
Abbreviations: CI=confidence interval; COX=cyclooxygenase; EAC=esophageal adenocarcinoma; RR=relative risk.
P-values and I2 for heterogeneity were calculated using the DerSimonian and Laird Q test. Differences between groups were considered to be statistically significant if Psignificance<0.05.