How to judge any add-on, using one trial as the lesson
Per transfer or per patient. Almost every add-on looks better on the first.
Before the individual add-ons, one idea that sorts most of them. It is best taught with the trial that demonstrates it.
661 women aged 25 to 40 with at least two usable blastocysts were randomized either to have their embryos screened for chromosome number before transfer, or selected by appearance alone. The results:
| Measure | Screened | Appearance only | Difference |
|---|---|---|---|
| Ongoing pregnancy per embryo transfer | 50.0% (137 of 274) | 45.7% (143 of 313) | +4.3 points, favouring the test, not statistically significant |
| Ongoing pregnancy per patient randomized | 41.8% (138 of 330) | 43.5% (144 of 331) | −1.7 points, favouring no test |
| Reached an embryo transfer at all | 83.0% | 94.6% | −11.6 points |
The numerators are almost identical, 138 against 144. The difference is entirely in the denominator. Do the subtraction: 330 minus 274 is fifty-six women in the screened group who never reached a transfer, against 331 minus 313, or eighteen, in the control group.
Clinic success-rate advertising is almost always per transfer. So is most add-on marketing. When you are shown a per-transfer improvement, the question is: what happened to the patients who dropped out between starting and transferring?
Two related rules, worth carrying through this whole stage:
Intermediate outcomes are not the outcome. Growth hormone raises the number of eggs collected. Screening raises the implantation rate per embryo transferred. You want a baby, not an egg, an embryo grade, or an implantation.
"No evidence of effect" is not the same as "evidence of no effect." One means nobody has looked properly. The other means people looked properly and found nothing. They call for different responses, and they are constantly conflated.