Yes, of course, these treatments should not be taken too lightly.
At the same time, some patients do benefit. In Bartlett's earlier work, it seems that at least 15 out of 100 women benefitted from the anthracycline if they had abnormal CEP17:
http://www.gbcc.kr/pdf/GBCC2009_Panel_5-1.pdf
There was no benefit to the normal CEP17 group. Anything that reduces patients getting treatments that will only give them side effects is a good thing, especially when the side effects are so serious.
And yes, there is a great deal of discussion about whether the predictor is her2 or topoIIa or CEP 17. Bartlett's point (see slide 25 above) is that cancers with abnormal CEP17 are more likely to have deleted topoIIa (over four times) or amplified topoIIa (1.5 times) and are twice as likely to be her2 positive. So, some of the connection between topoIIa and her2 and response may actually come from CEP17.
It's like when they found a link between coffee drinking and heart attacks, which seems to suggest that coffee is the problem. Then they discovered that heavy coffee drinkers are more likely to smoke, which turned out to be the cause of the heart attacks.
It would be a bit more convincing if they knew why CEP17 made such a big difference, since right now they only have a correlation without anything in the way of causation. I notice that in the scientific abstract he only says that CEP 17 may be important.