To our knowledge, this is the first investigation of early bacterial carriage conducted simultaneously in young Aboriginal and non-Aboriginal children living in the same setting, in this case in an urban and peri-urban rural area of Western Australia. It is one of very few studies investigating associations between early carriage and subsequent OM, despite evidence to suggest that early URT carriage may affect development of long-term humoral and cellular immunity, which in turn can increase susceptibility to disease [20, 21].
We have reaffirmed the high rates of single as well as multiple carriage of Pnc, Mcat and NTHi from a young age in Aboriginal children and high rates of OM in both Aboriginal and non-Aboriginal children. Early nasopharyngeal carriage of one or more of the 3 otopathogens, and specifically NTHi, increased the risk of OM in Aboriginal children, while in non-Aboriginal children Mcat was associated with an increased risk of OM as determined by full clinical examination but not by tympanometry alone. The important contribution of environmental factors to the high rates of OM in Aboriginal children [13, 14] is emphasized by the observation of reductions in ORs for early carriage (NTHi or any otopathogen) and subsequent risk of OM when controlling for environmental factors (Table 3).
We found no association between early carriage of either Pnc or Mcat and risk of OM in Aboriginal children. This may be because it is only their carriage during the first month (rather than the first 3 months of life) that increases risk of OM, but our data were too sparse to restrict analysis to carriage during the first month of life. Pnc and Mcat carriage begins at a younger age than NTHi: by age 2 months, ~40% of Aboriginal children had carried Pnc and/or Mcat at least once compared with 27% for NTHi .
There were similarities and differences between the current study and the only other comparable study which was conducted among Aboriginal children in the NT . Both studies identified the importance of early onset of URT carriage and risk of OM. However, in the NT study Aboriginal children were seen 2-4-weekly from birth and both carriage and ear health status were assessed at each visit. The primary endpoint in the NT study was time to first documented episode of OM (which occurred in all Aboriginal children in the NT before the age of 3 months), while our primary endpoint was any diagnosis of OM up to age 2 years. Carriage rates of single and multiple pathogens were even higher in the NT study than in ours; hence, in the NT study it was more difficult to investigate the contribution of individual otopathogens to subsequent disease than in our present study. Our lower carriage rates are unlikely to be due to differences in laboratory methods as our laboratory scientist (JB) was trained by colleagues in the NT and used the same methods throughout. The most likely explanation for differences in carriage rates between the two study populations is the somewhat better socioeconomic status in our urban/peri-urban WA community than that on a remote island in the NT, also reflected in the higher prevalence of OM and CSOM in the NT community than in our study population [4, 5].
The positive association between early carriage of Mcat, but not Pnc or NTHi, and OM in the non-Aboriginal children in our study is consistent with findings in non-indigenous North American children . However, we did not find an association between early Mcat carriage and presence of middle ear effusion as determined by type B tympanogram, possibly because we were identifying milder forms of OM using tympanometry at routine follow-up in the field than in the clinic setting. Of the three otopathogens, Faden et al. found that early carriage (defined as <3 months) of Mcat was the most strongly associated with early onset of OM, though the authors also reported that early carriage of any pathogen resulted in earlier onset of OM . In our study, there was a small positive association between early carriage of any pathogen and risk of clinical OM in non-Aboriginal children. The lack of association in non-Aboriginal children between early NTHi carriage and subsequent OM is most likely due to the small number of NTHi isolated before age 3 months (total of 8 of which only 4 were isolated in absence of other otopathogens). Similarly, the lack of association between early Mcat carriage and subsequent OM in Aboriginal children in our study is likely to be the small number (8) of Mcat isolated alone in this population (37% of all Mcat were isolated together with Pncs, 13% with NTHi and 25% with both Pnc and NTHi). This is in contrast to non-Aboriginal children in whom 63% of all Mcat (17/27) were the only otopathogen isolated from an NPA sample. To our knowledge there are no data available to explain in biological terms why one pathogen is more likely to predispose to OM than another.
Our findings suggest that carriage of more than one bacterial pathogen is more likely to result in OM than carriage of a single pathogen (Table 1). However, in view of the small number of isolates detected at an early age and the limited number of subsequent clinical examinations, we were unable to conduct formal statistical analyses to investigate this further. Nevertheless, a strength of our study was that we were able to use type B tympanogram to confirm our analyses using clinical diagnosis of OM as an endpoint.
In addition to the limited sample size, there are several other limitations or potential biases in this study. Firstly, we were not able to conduct a clinical examination every time an NPA was collected due to lack of available specialist personnel, though we were able to conduct tympanometry at the same time as NPA collection from age 3 months onwards. Secondly, in view of the long period of follow-up required, it is possible that we did not enroll children at greatest risk of OM, since we did not enroll children whose families were not intending to stay in the area for 2 years. We previously looked at how representative our KOMRP study population (of which our present data form a subset) was of the general population in the area: there were more teenage Aboriginal mothers, non-Aboriginal parents were older and non-Aboriginal mothers were less likely to smoke in pregnancy than the general population .Thirdly, while all children were asked to attend a clinic 3 times over 2 years for routine examination, it is possible that those who attended were sicker children seeking medical attention. Conducting tympanometry during routine field follow-up visits helped address this potential bias. Finally, we cannot exclude the possibility of residual confounding.
The results reported here were part of a broader study investigating causal pathways to OM, allowing us to control for important environmental factors. It would now be important to consider the role that high bacterial load and/or mixed infections (bacterial-bacterial and viral-bacterial) at a young age may play in subsequent risk of OM  in a larger cohort of both Aboriginal and non-Aboriginal children [4, 22–24].