Table 2. Subgroup analyses of odds ratio (OR) of gastric cancer according to refrigerator use*.
No. of reports | OR | (95%CI) | I2 | P for heterogeneity | P value for interaction | |
---|---|---|---|---|---|---|
Location | ||||||
European countries | 7 | 0.70 | 0.45 to 1.08 | 94.00% | <0.001 | 0.10 |
Italy | 2 | 1.00 | 0.83 to 1.21 | 47.70% | 0.167 | |
Sweden | 1 | 1.93 | 1.39 to 2.68 | NA | NA | |
Netherlands | 1 | 0.95 | 0.73 to 1.24 | NA | NA | |
Germany | 1 | 0.75 | 0.53 to 1.07 | NA | NA | |
Turkey | 1 | 0.22 | 0.14 to 0.34 | NA | NA | |
Portugal | 1 | 0.22 | 0.14 to 0.36 | NA | NA | |
American countries | 1 | 0.70 | 0.48 to 1.02 | NA | NA | |
Venezuela | 1 | 0.70 | 0.48 to 1.02 | NA | NA | |
Asian countries | 5 | 0.68 | 0.50 to 0.93 | 76.60% | 0.002 | |
China | 3 | 0.75 | 0.55 to 0.96 | 49.34% | 0.034 | |
India | 1 | 0.70 | 0.50 to 1.10 | NA | NA | |
Iran | 1 | 0.79 | 0.23 to 3.13 | NA | NA | |
Study quality | ||||||
Score>7 | 6 | 0.60 | 0.29 to 1.23 | 94.40% | <0.001 | 0.53 |
Score≤7 | 7 | 0.80 | 0.68 to 0.94 | 75.10% | <0.001 | |
Publication year | ||||||
Before 2000 | 4 | 1.09 | 0.80 to 1.49 | 84.1% | <0.001 | 0.01 |
2000–2011 | 9 | 0.55 | 0.39 to 0.77 | 90.6% | <0.001 | |
Controlling age in models | ||||||
Yes | 11 | 0.75 | 0.57 to 0.99 | 87.60% | <0.001 | 0.33 |
No | 2 | 0.45 | 0.11 to 1.77 | 97.30% | <0.001 | |
Controlling smoking in models | ||||||
Yes | 7 | 0.59 | 0.41 to 0.84 | 82.10% | <0.001 | 0.35 |
No | 6 | 0.84 | 0.62 to 1.13 | 92.40% | <0.001 | |
Controlling BMI in models | ||||||
Yes | 1 | 1.93 | 1.39 to 2.68 | NA | NA | 0.07 |
No | 12 | 0.65 | 0.52 to 0.81 | 88.40% | <0.001 | |
Controlling SES in models | ||||||
Yes | 8 | 0.83 | 0.63 to 1.09 | 88.90% | <0.001 | 0.28 |
No | 5 | 0.55 | 0.35 to 0.87 | 89.90% | <0.001 | |
Controlling family history of gastric cancer in models | ||||||
Yes | 3 | 0.59 | 0.43 to 0.81 | 89.80% | <0.001 | 0.64 |
No | 10 | 0.74 | 0.57 to 0.96 | 91.10% | <0.001 |
Abbreviations: BMI, body mass index; OR, odds ratio; SES, socio-economic status.
*The ORs were summarized by using random-effects meta-analysis.