About this example
Hackshaw, Law and Wald (BMJ 1997) pooled 37 epidemiological studies of lung cancer in lifelong non-smokers living with a smoker. The funnel plot is a textbook example of asymmetry in observational data: small studies with large odds ratios sit at the bottom right, while the matching small studies with odds ratios below 1 are missing.
Egger’s regression test is significant (P ≈ 0.02) and trim and fill imputes seven studies on the left, which lowers the random-effects odds ratio from about 1.24 to 1.19. The contours show that most of the missing studies would have fallen in the non-significant white region, which is what publication bias would predict. Asymmetry can have other causes, though, such as differences in study quality between small and large studies.
The data
| Study | OR | Lower | Upper |
|---|---|---|---|
| Garfinkel 1981 | 1.18 | 0.9 | 1.54 |
| Hirayama 1984 | 1.45 | 1.02 | 2.08 |
| Butler 1988 | 2.02 | 0.48 | 8.56 |
| Cardenas 1997 | 1.2 | 0.8 | 1.6 |
| Chan 1982 | 0.75 | 0.43 | 1.3 |
| Correa 1983 | 2.07 | 0.81 | 5.25 |
| Trichopolous 1983 | 2.13 | 1.19 | 3.83 |
| Buffler 1984 | 0.8 | 0.34 | 1.9 |
| Kabat 1984 | 0.79 | 0.25 | 2.45 |
| Lam 1985 | 2.01 | 1.09 | 3.72 |
| Garfinkel 1985 | 1.23 | 0.81 | 1.87 |
| Wu 1985 | 1.2 | 0.5 | 3.3 |
| Akiba 1986 | 1.52 | 0.87 | 2.63 |
| Lee 1986 | 1.03 | 0.41 | 2.55 |
| Koo 1987 | 1.55 | 0.9 | 2.67 |
…and 22 more rows. Open the example to see all of them.
Open the example, replace the data with your own, and export the figure as SVG, PDF, PNG or TIFF.