Background: The CENTERSTONE study demonstrated that treatment of influenza cases with baloxavir significantly reduced the transmission of influenza viruses to household contacts. Because transmission was defined based on the timing of test positivity and subtype match, it is possible that some events reflected co-incident community infection rather than within-household transmission. Such misclassification could affect the significance and magnitude of the baloxavir treatment effect.

Methods: Specimens from CENTERSTONE households with ≥2 influenza virus positive individuals were selected for whole genome sequencing. Consensus sequences and intrahost single nucleotide variants >2% frequency were determined. Phylogenetic trees of concatenated consensus sequences were constructed by subtype and study site country. Patristic differences between sequences were used to establish thresholds defining household transmission pairs.

Results: We obtained complete, high quality consensus genomes for 376/393 (96%) influenza A specimens and 39/64 (61%) influenza B specimens. Out of the 259 putative transmission pairs analyzed, we had reliable viral sequence data from both individuals in 145. 15/144 (10.4%) pairs had a distance ≥ 0.0005 (one was indeterminate), 17/145 (11.7%) pairs had a distance ≥0.0003, and 21/145 (14.5%) had a distance ≥0.0002. When excluding pairs based on the most stringent 0.0002 cut-off, the adjusted transmission incidence was 9.2% (95.38% CI, 7.2 - 11.8) and 12.9% (95.38% CI, 10.2 - 16.2) [aOR 0.69 (95.38% CI, 0.50-0.95)] in baloxavir vs. placebo groups, respectively.

Conclusions: Application of a stringent, sequence-based cutoff for defining true household pairs supports the outcome of the primary endpoint of CENTERSTONE. Baloxavir treatment of influenza cases can reduce influenza virus transmission in households.