ca Kawai 2003.

Methods	Prospective cohort study carried out during 2001 to 2002 in 38 practices in Japan (staffed by participating members of the Japanese Physicians Association). Doctors enrolled consenting vaccinated participants on an Internet‐based register from 1 October to 31 December 2001. Unvaccinated participants were selected by the researchers from the same clinic and matched by age and sex. By 31 May 2002, researchers had added data on symptoms of ILI and AE experienced by the participants. Information was elicited on the basis of self reported questionnaires, emails, or phone calls.
Participants	Children aged 0 to 15 years (older children participated, but from 16 years are not separable from 16 to 64 years' age group), adults, and elderly up to the age of 104. A total of 8841 participants took part in the cohort study.
Interventions	Inactivated influenza vaccine containing A/New Caledonia/20/99 + A/Panama/2007/99 + B/Johannesburg/5/99 once or twice. History of previous year's exposure was also elicited. A sliding scale of doses was used for age groups. Results are presented by 1, 2, or no immunisations.
Outcomes	Serological Rapid kit testing was carried out in 75 of the 124 participants with ILI symptoms; 64 of these were positive (A viruses recovered from 3 participants). Paired sera were positive in 5 of the 6 participants in whom they were taken. Effectiveness ILI (sudden onset, temperature over 38 °C, sore throat and fatigue). Influenza was defined as ILI plus rapid test diagnosis, or serum antibody increase or viral isolation. Safety Data for 96 participants are reported for the vaccinated arm but not for participants in the UV arm.
Funding Source	Institutional
Notes	The authors conclude that the vaccines were 67.6% and 84.5% effective, respectively against ILI (1 or 2 immunisations), and 54% and 79.8% effective against influenza (1 or 2 immunisations). No protection against ILI was conferred by immunisation the previous season. Despite an extensive baseline description of the 3 arms, the study has so many problems that the results are difficult to interpret: selection of participants, practices and controls, lack of specification of viral circulation and matching, non‐random serological testing, loss of safety data. Particularly non‐random kit testing makes a nonsense of the conclusions of the study. It is very strange that 64/8841 had influenza and yet had 84% efficacy.
Risk of bias
Bias	Authors' judgement	Support for judgement
PCS/RCS‐Selection Exposed cohort	Unclear risk	Volunteer, non‐information on number of doses
PCS/RCS‐Selection Non Exposed cohort	Unclear risk	Volunteer
PCS/RCS‐Comparability	Unclear risk	Matching by clinic age, sex
PCS/RCS‐Assessment of Oucome	High risk	No information on follow‐up length, self reported
Summary assessments	High risk	Despite an extensive baseline description of the 3 arms, the study has so many problems that the results are difficult to interpret: selection of participants, practices and controls, lack of specification of viral circulation and matching.