I would call that number not “needed to treat” but something like “person-year needed to avert one lung cancer death”. And remember in NELSON the participants only received 4 CT scans in this 10 year period (and not everybody got 4 scans, in the final round only 67% of the participants were scanned). Had all participants received CT ten times (every year), the reduction of lung cancer mortality would likely have been quite a bit larger. For NLST it was 3 screens and less years of follow-up. So one cannot directly compare these numbers, and one cannot translate them easily to the situation in an actual screening where subjects get a CT scans every year for a long period. To do that, one needs to do modeling. For NLST that was done in this study.