CRP concentrations in urban Indigenous participants from DRUID were higher than in the general population sample; Indigenous women had particularly elevated CRP levels. In men and women from both studies, CRP was strongly associated with waist circumference and BMI. Among those in the smallest waist circumference category, CRP levels were not elevated in DRUID women. After adjusting for waist circumference and other potential confounders, the risk of elevated CRP in DRUID was not significantly different to that for AusDiab, although women still had more elevated CRP than men. In men and women, having lower HDL cholesterol, impaired glucose tolerance (IGT), and higher diastolic blood pressure were also associated with having a high CRP, while current smoking was associated with high CRP in men but not women. The results for a CRP of 10.0 mg/l or above suggested that such extreme values were more common in DRUID women than AusDiab women, and in women than men, even after accounting for other risk factors.
The high CRP concentrations in this urban Indigenous population are consistent with data for an isolated Aboriginal community, where over 50% of participants had an elevated CRP . Data from a remote Indigenous Island community were also compared with published data from different countries, demonstrating the relatively high CRP concentrations in the Indigenous group, in which 39% of men and 18% of women had CRP > 10 mg/L .
Wang and Hoy  showed that from around age 10, CRP levels rose with age and were always higher in females than males. Our unadjusted data also suggest increasing CRP with age, but in the multivariate analysis this was no longer seen. It is possible that the association between CRP and age was due to the increase in abdominal obesity with age, and so was lost once waist circumference was accounted for.
Our results, and the results from Shemesh et al , suggest that in the absence of obesity, women do not have elevated CRP relative to men. Similar CRP levels were also reported among lean Chinese men and women (median CRP men 0.84; women 0.93 mg/L) . In a population-based study in Japan, men had higher CRP levels than women, but participants were relatively lean and CRP concentrations were extremely low, with a median for men of 0.16 mg/L and for women 0.09 mg/L . Men and women from the Dallas Heart Study in the lowest tertile of fat mass (by DEXA) also did not show differences in CRP concentrations .
Other studies have shown steeper increases in CRP in association with risk factors in women relative to men. In an Israeli study and in the Framingham Offspring Study , men with no metabolic syndrome components had a slightly higher CRP than women, but as the number of components increased, CRP rose more steeply in women than men, and there were significant interactions between number of components and sex [10, 11]. Similarly in the Dallas Heart Study, CRP rose more steeply across BMI groups in women than in men, so that at BMI > 30 kg/m2, CRP levels in women were about twice those in men . Among Aboriginal Australians from an isolated rural community there was a strong linear association between CRP and BMI in women, but no association in men .
Women using oral contraceptives or hormone replacement therapy tend to have higher levels than other women . The possibility exists that endogenous estrogen is also associated with CRP , hence menopausal status may contribute to differences in CRP between the older AusDiab and younger DRUID women, but we do not have data to examine this. However, the data presented by Wang and Hoy  on CRP levels with age do not indicate any change in the association at an age where most women would reach menopause.
The significance of higher CRP in women than men across two very different populations is not clear. CRP is known to be associated with CHD, and typically men are considered to be at higher risk than women . A recent review suggested that global inflammation, as indicated by elevated CRP, may predict cardiovascular outcomes in women better than the traditional risk factors identified in men . Onat et al on the other hand found that CRP predicted CHD similarly in Turkish men and women, but diabetes in women only . In either case, the extremely high CRP concentrations seen in the DRUID women could presage extremely high rates of diabetes and CHD, although no prospective data is available to confirm this.
It is of interest that in our multivariate models current smoking was associated with elevated CRP in men only. In a Turkish population-based study, smoking in women was actually associated with a lower CRP concentration than in non-smokers, while in men smokers had higher CRP than non-smokers .
In a recent study of Native Alaskans, CRP was found to be associated with pathogen burden, measured as IgG, IgA and IgM antibodies to C. pneumoniae and IgG antibodies to other pathogens . High CRP levels seen in Indigenous Australians, particularly those above 10 mg/L, may be in part attributable to pathogen burden. For example, the prevalence of H. pylori in two Indigenous populations in Western Australia was extremely high, 91% in a remote community and 60% in an urban community, which is still twice what would be expected in the non-Indigenous Perth population . High rates of acute otitis media related to infections with H. influenzae and S. pneumoniae are common in Indigenous children . McDonald et al  found that higher CRP concentrations in a group of remote Australian Aborigines were associated with IgG seropositivity to H. pylori and C. pneumoniae and higher IgG titre for cytomegalovirus.
Our study has several limitations. The data are cross-sectional, and we are unable to comment on the representativeness of volunteer cohorts with low to moderate response rates. There is evidence that people who attended the AusDiab follow-up were healthier than those who did not attend , which would suggest they may have lower CRP concentrations than the more representative original sample. This may explain to some extent their lower unadjusted CRP levels relative to DRUID participants. However, the multivariate analysis accounted for many health related variables, both risk markers (lipids, glucose tolerance, blood pressure, obesity) and behavioural factors (smoking), which appeared to largely explain the differences between DRUID and AusDiab.
We are unable to account for the use of statins, which can reduce CRP levels , in DRUID. However, we do have data for all but 23 men and 18 women on lipid lowering medication, which includes statins. Adding an indicator variable for lipid-lowering medication did not markedly change the ORs for DRUID relative to AusDiab from those presented in Table 2; although in women, such drug use was associated with a reduced risk of high CRP.
The assay methods used for hs-CRP in both studies were different and we do not have data to directly evaluate their comparability. Roberts et al  compared 9 hs-CRP methods, including the DPC and Roche methods, with the DadeBehring BN II assay. Although both showed good concordance with the reference method, the Roche assay tended to give slightly higher results and correctly classified around 75% of 388 blood donors into quartiles based on the Dade method, while the DPC assay correctly classified around 95% . Nonetheless, criteria proposed to identify high risk CRP concentrations do not specify the method, and are widely applied. The difference between assays could potentially contribute to higher hs-CRP values in DRUID using the Roche assay, but is unlikely to explain the large differences in crude CRP concentrations observed.
The main strength of our study is that we were able to combine data, collected using similar methods, from population based studies of the general population and an Indigenous Australian population.