General practitioners in the sentinel surveillance network saw a total of 182,984 patients during the study period in 2007, of which 1,226 (0.7%) had a reported ILI. The ILI rate peaked in weeks 33 and 34 between 2.0% and 2.2% from a nadir of 0.04% in week 18. In 2008 there were 159,030 patients seen and a total of 876 (0.6%) reported to have an ILI. The weekly rate generally climbed steadily from 0.2% at the start of the 2008 study period in week 18 to a peak of 1.3% in week 35.
General practitioners collected nose and throat swabs for testing from 480 (39%) and 407 (46%) patients with ILI in 2007 and 2008 respectively. Of these, 223 (46%) in 2007 and 117 (29%) in 2008 were positive for influenza. The 2007 season was characterised by a high proportion (58%) of type A/H3N2 influenza cases for which limited strain typing data indicated a generally even split between A/Brisbane/10/2007-like and A/Wisconsin/67/2005-like viruses with a further 17% due to type B and 22% due to A/H1N1 for which all of those typed were the A/Solomon Islands/3/2006-like strain (table ). A majority (56%) of influenza cases in 2008 were type B with a further 36% due to type A/H3N2 although like 2007, a high proportion of specimens were unable to be recovered or typed (table ).
Influenza positive swabs by subtype, year and strain, 2007-2008
Following exclusion of cases for whom vaccination status was unknown, symptom onset or specimen collection dates were unknown or more than four days had elapsed between symptom onset and specimen collection, there were 386 (80%) and 330 (81%) study participants in 2007 and 2008 respectively (table ). In 2008, a higher proportion of influenza negative patients (17%) compared to influenza positive patients (6%) were excluded because more than four days had elapsed between symptom onset and specimen collection (p = 0.004) whereas in 2007 there was no significant difference (14% and 8%; p = 0.10). There was no statistically significant difference in whether or not study participants had a specimen collected within four days of symptoms onset by age group in either 2007 (p = 0.90) or 2008 (p = 0.09).
Study inclusion and exclusion criteria by year, 2007-2008
An epidemiological curve of influenza negative and influenza positive patients eligible for inclusion in the study (designated as controls and cases respectively) shows an earlier detection of influenza in 2007 compared to 2008, although there was only two weeks' difference in the time from which influenza positive patients were reported for more than three consecutive weeks indicating the start of each season (Figure ). In addition to a higher number of study participants, the 2007 influenza season was longer (as defined by the number of consecutive weeks in which influenza positive cases were reported) and consisted of a higher proportion of influenza positive study participants (n = 194; 50%) compared to 2008 (n = 106; 32%). The dominant circulating influenza type and subtype varied over the two seasons: 23% of cases in 2007 were A/H1N1, 60% were A/H3N2 and the remainder were type B; the respective proportions in 2008 were 4%, 36% and 57% (table ).
Case and control recruitment from influenza-like illness (ILI) presentations at sentinel general practices by week and year, 2007-2008.
Cases and controls by age group, sex, month of swab collection, year and type/subtype, 2007-2008
Age group, sex and month of swab collection distributions for controls and cases (including type and subtype strata) are shown in table . There was no statistically significant difference in the sex distribution between controls and cases in either 2007 or 2008. In both years, the numbers and proportions of controls and cases were highest in the 20-49 years age group. Influenza type B cases were significantly younger than controls in 2008 (p < 0.001); there were no other statistically significant differences in age distribution between controls and cases. With the exception of subtype A/H1N1 in 2008 for which there were only four cases, stratification by month of swab collection revealed statistically significant differences between cases and controls (range: p < 0.001 to p = 0.02) because of the higher proportion of type A and type B cases identified in August and October 2007 respectively, and subtype A/H3N2 and type B in August/September 2008.
A similar percentage of total study participants were vaccinated in 2007 (19%) and 2008 (17%), although the difference between vaccinated controls and vaccinated cases was generally higher in 2007 (table ). In 2008 a higher proportion of cases of subtypes A/H1N1 and A/H3N2 were vaccinated compared to controls. In both years the proportion of cases and controls that were vaccinated generally increased with age group. Among the study participants reported as vaccinated, only one control in each year (0.5% in 2007 and 0.4% in 2008) and no cases were vaccinated within 14 days of symptoms onset, for which there was no statistically significant difference.
Cases and controls by year, age group, vaccination status and type/subtype, 2007-2008
Following adjustment for month of swab collection and age, there was a statistically significant protective effect of influenza vaccine against all influenza in 2007 (VE = 59%; 95% CI, 25 to 78%) (table ). The absence of vaccinated cases and controls (table ) meant VE was unable to be estimated for several of the five age groups so age was collapsed into three variables: children (0-19 years); working age adults (20-64 years); and the elderly (65 years or older). When stratified by age group, the statistically significant association in 2007 was restricted to the 20-64 years age group. Furthermore, when examined by influenza type and subtype, and after adjusting for age group and month of swab collection, the vaccine was found to only be protective at a significant level against the influenza A/H3N2 subtype (VE = 68%; 95% CI, 32 to 85%), for which a statistically significant protective effect was maintained among the working age adults age group only. In 2008, only the unadjusted measure of VE against type B influenza was statistically significant. Receiving vaccine was positively associated with influenza illness for both A/H1N1 and A/H3N2 subtypes after adjustment for age and month of swab collection in 2008 but neither of these associations was statistically significant.
Crude and adjusted vaccine effectiveness of seasonal vaccine against influenza by year, age group and type/subtype, 2007-2008
Sensitivity analyses were conducted to determine the possible effect of assumptions about timing of swab collection and vaccination status on the VE estimates. The effect of not excluding study participants if more than four days had elapsed between symptom onset and specimen collection was a reduction of the adjusted VE point estimates between 7% and 15% in 2007 and between 5% and 35% in 2008. Study participants who were known to be vaccinated within 14 days of symptom onset (one control each in 2007 and 2008) were classified as not vaccinated in the primary analysis. The effect of excluding these cases or classifying them as vaccinated resulted in variations of 0% to 7% around the VE point estimates, but no changes in their relative statistical significance. However, collection of the 'date of vaccination' field only commenced in 2008, in which it was completed for 86 (91%) of the 94 vaccinated study participants. In 2007, only 16 (22%) of the 73 study participants reported as vaccinated had a recorded date of vaccination.