Yes, this is the main, common counterargument that the OP addresses. This does a pretty good job of showing why that argument is overstated (if not entirely in bad faith), so I direct you back to it.
> There was a subgroup analysis on those that actually received screening, which showed a decent improvement in hazard ratio.
I can't emphasize this enough: read the article. Per-protocol analysis (what you're describing) is incredibly biased. OP shows that it's an overestimate of true effect, and I'd go so far to say that it's completely useless (in general; not just for this paper).
People who argue that NORDICC was wrong because per-protocol showed a bigger effect size should be ignored with prejudice. They don't know how to read or interpret scientific studies, and have a poor understanding of statistics.
> There was a subgroup analysis on those that actually received screening, which showed a decent improvement in hazard ratio.
I can't emphasize this enough: read the article. Per-protocol analysis (what you're describing) is incredibly biased. OP shows that it's an overestimate of true effect, and I'd go so far to say that it's completely useless (in general; not just for this paper).
People who argue that NORDICC was wrong because per-protocol showed a bigger effect size should be ignored with prejudice. They don't know how to read or interpret scientific studies, and have a poor understanding of statistics.