In re-analysing our data (Nature 492, 34–36; 2012), George Santangelo with David Lipman, and Steven Salzberg each exclude two-thirds of the top-cited US publications assigned by Scopus to life or health sciences. We consider that their analyses discard most health research that matters.

Clinical trials are primary research that catalyses preventive or therapeutic innovation; reports and guidelines also decisively inform and radically transform health. Extremely highly cited reviews formulate pioneering concepts or synthesize influential work. Also, scholars reaching the top 0.01% with papers of any type predict some further excellence.

The authors' analyses depend on grant acknowledgments (75–83% in both series) but these are a problematic metric because they represent grants that were awarded 10–15 years ago when the NIH budget was expanding and acceptance rates were highest. Also, any of several co-authors can acknowledge a funding source if they are under pressure to demonstrate grant-related productivity, even if that source is irrelevant. And let's imagine a hypothetical funding system that forces all geniuses to quit science: 100% of papers could still acknowledge funding.

Groundbreaking projects account for less than 1% of awarded grants. Students whose papers reach 1,000 or more citations should certainly become principal investigators: stars will abandon systems that stifle independence.