Table 2.
Sensitivity analysis of meta-analysis to the coding of years since cessation
| Model: | 3 (original results) | 3 (New dose coding) | 6 (original results) | 6 (New dose coding) |
|---|---|---|---|---|
| Dose (1 year of not drinking) | 1.07 | 1.00 | 0.93 | 0.92 |
| Dose squared | 1.00 | 1.00 | ||
| Controls for: | ||||
| Study characteristics* | No | No | Yes | Yes |
| Radiation | No | No | Yes | Yes |
| Quit in last ten years | Yes | Yes | No | No |
| years since cessation*heterogeneity | No | No | Yes | Yes |
| Q statistic # | 15.9 | 18.38 | 6.6 | 7.95 |
* Study characteristics = 1 if the study was in Japan, a prospective cohort design, had no minimum gap between giving up drinking and joining study, and no controls for hepatitis. # pr>Chi2 that there is no heterogeneity according to the Q statistic. Bold represents statistically significant figures at the 95% confidence level, or Q statistic finds no heterogeneity