Specimens were obtained from Western Diagnostic Pathology (Myaree, Western Australia, Australia), which provides outpatient pathology services covering most of the NT. Specimens were eligible for inclusion regardless of indication for testing, provided identifying information was complete and address was within the NT. We accepted only serum tubes with a residual volume >0.5 mL and obtained specimens before routine discarding. Baseline specimens were selected during January–May and all postpandemic specimens from September 2009.

Data obtained for each specimen consisted of date of collection, patient’s age in years at collection, gender, suburb/community of address, and a unique study identifier. Identifying data (name and date of birth) were transferred directly from the laboratory to the Information Services Division of the NT Department of Health and Families for computer-matching to indigenous status. This was successful in 94.1% of cases, and the data were transferred to the investigators linked to the study identifier. Of those cases with a successful match, 59.7% of patients were neither indigenous nor Torres Strait Islander, 39.7% were Aboriginal, 0.1% were Torres Strait Islander, and 0.6% were both Aboriginal and Torres Strait Islander. The suburb of patient’s address for each specimen was linked to 2006 Statistical Local Area (SLA), the Australian Bureau of Statistics’ general purpose base spatial unit, with 82 of 96 NT SLAs represented (11).

After testing, a small number of specimens were redistributed by region, following manual review of suburb of address linkage to SLA. The SLA code was also linked to the 11 statistical subdivisions and 7 health districts in the NT. Three study regions were defined, displayed in Figure 1, consisting of Urban Darwin; Rural Top End (Darwin Rural, East Arnhem, and Katherine districts); and Central Australia (Alice Springs Urban, Alice Springs Rural, and Barkly districts). SLA codes were then linked to the Australian Bureau of Statistics’ Socio-Economic Indexes for Area (SEIFA) (12). These measures use information from census data relating to material and social resources and ability to participate in society to obtain a broad level of relative socioeconomic status for each SLA. For calculation of attack rates by quintile, the SEIFA index of relative disadvantage was used, while for regression analysis, the SEIFA index of relative advantage and disadvantage was preferred, as this index does not incorporate indigenous status.

Antibody responses to pandemic (H1N1) 2009 influenza were assessed at the World Health Organization Collaborating Centre for Reference and Research on Influenza in North Melbourne, Victoria, Australia. Reactivity of serum against pandemic (H1N1) 2009 influenza was measured in 2140 serum samples by using hemagglutination inhibition (HI). Egg-grown A/California/7/2009 virus was purified by sucrose gradient, concentrated, and inactivated with β-propiolactone to create an influenza zonal pool preparation (a gift from CSL Ltd., Parkville, Victoria, Australia). Serum samples were pretreated with 1:4 vol/vol receptor-destroying enzyme II (Deka Seiken Co. Ltd., Tokyo, Japan) at 37°C for 16 h, then enzyme was inactivated by the addition of an equal volume of 54.4 mmol/L trisodium citrate (Ajax Chemicals, Taren Point, New South Wales, Australia) and incubated at 56°C for 30 min. A total of 25 μL (4 hemagglutinin units) influenza zonal pool preparation A/California/7/2009 virus or 25 μL phosphate-buffered saline (“no virus” control) was incubated at room temperature with an equal volume of receptor-destroying enzyme–treated serum. Serum specimens were titrated in 2-fold dilutions in phosphate-buffered saline from 1:10 to 1:1280. After a 1-h incubation, 25 μL of 1% vol/vol turkey erythrocytes wase added to each well. HI was read after 30 min. Any samples that bound the erythrocytes in the absence of virus were adsorbed with erythrocytes for 1 hour and reassayed. Six samples bound erythrocytes in the absence of virus and were excluded from analysis. Titers were expressed as the reciprocal of the highest dilution of serum at which hemagglutination was prevented.

A panel of control and serum samples were run in addition to the test serum samples for all assays. The control panel comprised paired ferret serum samples pre- and postinfection with pandemic (H1N1) 2009; seasonal influeza A (H1N1), A (H3N2), or B viruses; and paired human plasma and serum samples from donors, collected before April 2009 or after known infection with pandemic (H1N1) 2009 or vaccination with the Australian monovalent pandemic (H1N1) 2009 vaccine.

We aimed to estimate the proportion of persons with serologic immunity in each of 12 groups in the post-pandemic sample, consisting of 4 age groups (<14, 15–34, 35–54, and >55 y) within each of the 3 study regions described. In the postpandemic group, we calculated a required sample size of 195 specimens per group (for a total of 2,340 specimens) on the basis of an estimate of 15% immunity with a 95% confidence interval (CI) of 10%–20%.

In the baseline group, we aimed to provide an age-specific, NT-wide estimate of preexisting immunity and calculated a single sample size for each of the same 4 age groups described. We did not stratify by region and assumed increasing prepandemic immunity with age (2% in those <14 y, 5% in those 15–54 y, and 15% in those >55 y).

Samples were chosen at random from each stratum and checked for representativeness of the NT population by gender and region before testing. Data on indigenous status were obtained from Information Services only after final selection of specimens.

For all analyses, immunity was defined as an HI titer ≥40, consistent with published data (13) and the observation that titers of this order develop in 90% of persons <21 days of illness (14). Attack rates were calculated as the difference in proportion immune between the September category and the total baseline group, except for age-specific attack rates where the age-specific baseline proportion was used as the reference. All attack rate calculations were population standardized, with weights calculated separately for pre- and postpandemic samples, based on the demographic characteristics of the 2009 NT population by age-group, indigenous status, and study region. Regression models and proportions immune are displayed unweighted. Statistical analysis was performed with Stata version 11.0 (StataCorp LP, College Station, TX, USA).

We obtained ethical approval from the Menzies School of Health Research Human Research Ethics Committee and the Central Australia Human Research Ethics Committee. We continued to liaise with the Aboriginal and Torres Strait Islander subcommittees of both ethics committees throughout the study.

A total of 445 specimens taken January 10–May 29 were selected from 10,575 available serum tubes (Table 1). Within each age group sampled, the baseline sample was representative of the 2009 NT population (15) by gender, indigenous status, and study region, except that higher proportions of indigenous Australians were seen in the 2 older age brackets. There was no tendency toward an increase in the proportion of specimens with protective immunity over the 5 months from which baseline specimens were taken (p = 0.79, by χ² test for trend).

A total of 34 of 445 baseline specimens (7.6%, 95% CI 5.2%–10.1%) had HI titers >40. Multivariate logistic regression revealed no difference in baseline immunity by gender, indigenous status, study region, or index of socioeconomic disadvantage (p>0.05), with increasing age in years the only significant independent predictor of prepandemic immunity (p = 0.003). Although not statistically significant on the regression model, the proportion of specimens with titers >40 appeared higher in Central Australia (14.0%) than Urban Darwin and Rural Top End (5.6%). Immunity was nevertheless evenly spread geographically within these regions.

A total of 1,689 specimens collected September 3–30, 2009, were selected from 3,228 available. Because of insufficient numbers of specimens, the required sample size was not achieved in 5 of 12 postpandemic groups. The September samples were representative of the 2009 NT population by gender but again included higher proportions of specimens from indigenous Australians in the older age brackets. An HI titer >40 was seen in 329 specimens (19.5%, 95% CI 17.6%–21.4%), with proportions by study group shown in Table 2, geometric mean titers in Figure 2, and reverse cumulative distributions in Figure 3. There was a nonsignificant trend toward a decreasing proportion of specimens with protective immunity over the 5 weeks from which the September specimens were taken (p = 0.20, by χ² test for trend).

Table 3 shows the results of multivariate logistic regression analysis for the 1,592 postpandemic specimens for which the indigenous status of patients was known. No association was detected between immunity and gender, socioeconomic status, or study region. However, younger age and indigenous status were independently associated with immunity. A measure of remoteness was examined as a possible exposure variable, but colinearity with region meant that it was not a useful predictor variable and therefore was not included in regression analysis (16).

The proportion immune in September was geographically heterogeneous across the 3 study regions (p<0.001, by χ² test). The same pattern was seen for Statistical Subdivisions (p<0.001, by χ² test), with proportionate immunity ranging from 7.5% to 42.9%, as illustrated in Figure 4. The picture of heterogeneity was also seen for the indigenous population considered separately. However, the prevalence of postpandemic immunity was more homogeneous for the nonindigenous population considered by either geographic classification and for urban Darwin considered separately. Figure 5 demonstrates that while postpandemic levels of immunity were relatively homogeneous by SLA in less disadvantaged, generally urban areas, comparatively disadvantaged areas had more variable levels of immunity.

As shown in Table 4, attack rates by age group were markedly higher in younger groups, reaching approximately 1 in 3 among children <14 years of age. Indigenous Australians were also disproportionately affected, with attack rates of ≈1 in 4, which were 1.85-fold higher than those seen in nonindigenous Australians. No differences in attack rates were seen by gender, region, or socioeconomic quintile. Given these attack rates, we estimate that 15,600 (95% CI 10,900–20,300) of 67,820 indigenous and 19,500 (95% CI 12,700–26,300) of 157,028 nonindigenous persons in the NT acquired pandemic influenza during May–September 2009.