Cowling 2008.

Methods	Cluster‐randomised controlled trial carried out in Hong Kong SARS between February and September 2007. The study assessed the effects of non‐pharmaceutical interventions on the household transmission of influenza over a 9‐day period. ILI cases whose family contacts had been symptom‐free for at least 2 weeks rapid tested for influenza A and B were used and randomised to 3 interventions carried out. Randomisation was carried out in 2 different schedules (2:1:1 for the first 100 households and subsequently 8:1:1) but it is unclear why and how
Participants	946 index subjects aged 2 years or more in 122 clusters (households). 116 households were included in the analysis, 6 were excluded because subsequent laboratory testing (culture) were negative. There were 350 household contacts in the analysis but there 370 household contacts at randomisation. Attrition is not explained. Index cases were defined as subjects presented with at least 2 influenza‐like symptoms of at least 48 hour duration (such as fever more or equal to 38 degrees, cough, headache, coryza , sore throat, muscle aches and pains) and positive influenza A+B rapid test
Interventions	Households were randomised to either wearing face masks with education (as the control group plus education about face mask use) or handwashing with special medicated soap (with alcohol sanitiser) with education (as the control group plus education about handwashing) or education about general healthy lifestyle and diet (control group). The soap was distributed in special containers which were weighed at the start and the end of the study. Interventions visits to the households were done on average 1 day after randomisation of index case household
Outcomes	Laboratory: QuickVue RTI MDCK culture Samples were harvested using NTS, but the text refers to a second procedure from June 2007 onwards testing for non‐influenza viruses but no data were reported Effectiveness: secondary attack ratios (SAR): SAR is the proportion of household contacts of an index case who subsequently were ill with influenza (symptomatic contact individuals with at least 1 NTS positive for influenza by viral culture or PCR) Three clinical definitions were used for secondary analysis: 1. Fever more or equal to 38 degrees or at least 2 of following symptoms, headache, coryza , sore throat, muscle aches and pains 2. At least 2 of the following S/S: fever more or equal to 37.8 degrees, cough, headache, sore throat and muscle aches and pains 3. Fever of more or equal to 37.8 degrees plus cough or sore throat Safety: no harms were reported in any of the arms
Notes	The authors conclude that “The secondary attack ratios were lower than anticipated, and lower than reported in other countries, perhaps due to differing patterns of susceptibility, lack of significant antigenic drift in circulating influenza virus strains recently, and/or issues related to the symptomatic recruitment design. Lessons learnt from this pilot have informed changes for the main study in 2008” Although billed as a pilot study the text is highly confusing and at times contradictory. The intervention was delivered at a home visit up to 36 hours after the index case was seen in the outpatients. This is a long time and perhaps the reason for the failure of the intervention. Practically, the intervention will have to be organised before even seeking medical care – i.e. people know to do it when the kid gets sick at home
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was computer‐generated by a biostatistician "A pre‐specified table of random numbers will be used to assign one of the three interventions to the household of the index case."
Allocation concealment (selection bias)	Low risk	The households of eligible study index patients were allocated to 3 groups in a 1:1:1 ratio under a block randomisation structure with randomly permuted block sizes of 18, 24 and 30 by using a random‐number generator. Allocation was concealed from treating physicians and clinics and implemented by study nurses at the time of the initial household visit
Blinding (performance bias and detection bias) All outcomes	High risk	Participants and people who administered the interventions were not blinded to the interventions, but participants were not informed of the specific nature of the interventions applied to other participating households
Incomplete outcome data (attrition bias) All outcomes	High risk	Dropout was accounted for. Dropout from randomised population was high: 32% in control group, 37.5% in hand hygiene group and 39.4% in the face masks and hand hygiene group. Reasons for dropout distributed evenly over the 3 groups Authors report follow up as proportion of patients remaining in the study after initial dropout
Selective reporting (reporting bias)	High risk	The choice of season, change in randomisation schedules and unexplained dropouts among contacts, the use of QuickVue which proved unreliable, reporting bias on non‐influenza isolates make this study at high risk of bias