Question 13: How can we do this in CMA2?
We want to do a meta-regression on year of publication.
First we must enter the year as a variable. Go back to the data entry screen. We need to insert another variable. Click on column heading I. Click "Insert" and call the variable "Year". Enter the year of publication in the column.
| Study name | Endarterectomy Events | Endarterectomy Total N | Stenting Events | Stenting Total N | Odds ratio | Log odds ratio | Std Err | Year | J |
|---|---|---|---|---|---|---|---|---|---|
| Naylor 1998 | 0 | 12 | 5 | 11 | 0.047 | –3.052 | 1.554 | 1998 | |
| Wallstent 2001 | 5 | 112 | 13 | 107 | 0.338 | –1.085 | 0.545 | 2001 | |
| CAVATAS 2001 | 25 | 253 | 25 | 251 | 0.991 | –0.009 | 0.298 | 2001 | |
| Brooks 2001 | 1 | 51 | 0 | 53 | 3.178 | 1.156 | 1.645 | 2001 | |
| Brooks 2004 | 0 | 42 | 0 | 43 | 2004 | ||||
| SAPPHIRE 2004 | 8 | 167 | 7 | 167 | 1.150 | 0.140 | 0.530 | 2004 | |
| EVA-3S 2006 | 10 | 262 | 25 | 265 | 0.381 | –0.965 | 0.385 | 2006 | |
| SPACE 2006 | 38 | 584 | 46 | 599 | 0.837 | –0.178 | 0.227 | 2006 | |
| BACLASS 2007 | 1 | 10 | 0 | 10 | 3.316 | 1.199 | 1.693 | 2007 | |
| ICSS 2009 | 43 | 857 | 72 | 853 | 0.573 | –0.557 | 0.199 | 2009 |
Then click "Run analyses". Click the button "Analysis". This includes an option "Meta-regression". Click this. The box marked "No prdictor" has a little down arrow. If you click this you will see a list of moderator variables, just one in this case (if you are following me closely), Year. Click this. You should see a scatter plot with a regression line:
I would call this the regession of log odds ratio on year, not the regession of year on log odds ratio, as CMA2 does.
You can change the size of the circles with the solid square and two arrows buttons, but I cannot see any way to change the scales, which are pretty horrible.
You can also have circles proportional to the study weights with the "Proportional" button. I then made the symbols smaller:
Clicking "Table" gets the details of the regression:
| Fixed effect regression | ||||||
|---|---|---|---|---|---|---|
| Point estimate | Standard error | Lower limit | Upper limit | Z-value | p-Value | |
| Slope | –0.03064 | 0.03947 | –0.10801 | 0.04673 | –0.77624 | 0.43761 |
| Intercept | 61.07079 | 79.18041 | –94.11995 | 216.26154 | 0.77129 | 0.44054 |
| Tau-squared | 0.13360 | |||||
| Q | df | p-value | ||||
| Model | 0.60254 | 1.00000 | 0.43761 | |||
| Residual | 12.16014 | 7.00000 | 0.09541 | |||
| Total | 12.76269 | 8.00000 | 0.12028 | |||
Although there is a slight downward slope, it is nowhere near significant, so there is little to support the theory that surgical techniques have improved.
Back to question.
Back to Systematic Reviews index.
This page maintained by Martin Bland.
Last updated: 19 February, 2010.
