Referee 1's Review (Andrew Hayward):

**********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication, the manuscript has been revised in light of these comments and to address other editorial requirements.
**********

This is an important and well written account of the results of a pilot household intervention trial to reduce transmission of influenza. Even though this only represents the results of a pilot study it provides many valuable insights. I recommend this paper for publication and have a few comments which the authors may wish to consider.

The overall study is very well designed and well described. Although 35% of households dropped out at an early stage, follow up of the remaining households was excellent. The use of laboratory techniques was appropriate. Although inclusion of acute and convalescent serological studies would have strengthened the findings, they would likely have led to unnaceptably low participation rates.

The main value of the work is in describing household secondary attack ratios and secondary attack rates and how these can be dramatically affected by the case definition used. It would be interesting to know how a definition of "Any respiratory illness" compared as following a case of confirmed influenza in the household this may be reasonably predictive.

This data will clearly be valuable for those planning or interpreting similar research or developing mathematical models of transmission. The authors note that the secondary attack rates are lower than reported in other studies and site a number of plausible possible explanations. They do not discuss whether there might be environmental or behavioural differences that could lead to differing secondary transmission rates in their study compared to previous work, e.g differences in use of air conditioning or high background use of facemasks.

The marked differences in secondary attack rates given the different case definitions is in marked contrast to the apparently highly similar area under the ROC using the different definitions comparing definitions to laboratory confirmed cases. It would be helpful for the authors to comment on how this occurs.

The finding that 19% of the virus positive cases were asyptomatic is intriguing - I have seen work showing that viral excretion can preceed symptoms but I'm not sure whether asymptomatic viral excretion is widely recognised - it would be useful if the authors could comment on whether this is a finding that is common to other studies. Asymptomatic viral excretion could have important implications for control. It would also be interesting to know whether these 4 asymptomatic cases were culture confirmed or PCR confirmed as PCR may be more likely to give false positives due to cross contamination.

The potential role of asymptomatic infection could be discussed more fully. For example highlighting that many serologically confirmed influenza infections are asymptomatic. Secondary attack rates may therefore be substantially lower than secondary infection rates.

The uptake of the interventions was predictably incomplete - this is a major challenge for this sort of research and limits our ability to study true effectiveness. However, observational studies suggest that even with incomplete uptake such interventions may be effective (see recent Jefferson review in BMJ) so they may still find an effect. In a pandemic situation compliance is likely be driven by fear and may be much higher. The already high background use of face masks in Hong Kong may also limit the ability of the study to demonstrate effectiveness.

Whilst detailed reporting on uptake of the intervention is appropriate I felt that the amount of material presented on the differnece in outcomes in different arms of the intervention was greater than is needed especially given that the pliot is clearly not powered to assess a difference in these outcomes. I am not sure that the supplementary tables and figures add value.

In table 1 the figures in brackets refer to proportions not percentages as stated

Not sure that Tables S1 and S2 are necessary given the limited power