We found that there is no significant difference between HIV-infected Aboriginal persons and non-Aboriginal persons regarding the time to HIV plasma viral load suppression of < 500 copies/ml and time to CD4 cell response of 100 cells over baseline after initiation of HAART. There was however, a significant higher mortality risk for Aboriginal persons after the initiation of HAART. After adjustment for confounder variables, Aboriginal persons had mortality rates 3.12 times higher than non-Aboriginal persons. It is interesting to note that the only clinical characteristic associated with mortality risk in this population was adherence during first year of follow-up. The other variables associated with mortality were socio-demographic characteristics of the participants (age and income). We observed that clinical factors were only a significant predictor when we looked at virologic and immunologic responses. Note that among the clinical factors, poor adherence was the strongest predictor of adverse outcomes in all analyses. To explain this paradox, we considered potential confounding effects, given that race/ethnicity are closely intertwined with socioeconomic status and behavioural factors [1, 4]. Our method of selecting confounding factors ensured that the final models were parsimonious while at the same time the estimates were not substantially affected by any potential confounders available in the data.
Other studies have examined the effect of ethnicity on response to HAART but to our knowledge, this was the first study to specifically examine this issue for Aboriginal persons living in large urban areas. Prior studies looked at potential differences among ethnic groups in response to HAART by measuring short-term virologic and immunologic response or differences in survival. In a Danish cohort, race (white versus not white) did not predict differences in virologic suppression and CD4 cell response one year after initiating HAART . Race (white versus not white) also did not independently predict virologic response or CD4 response in a group of American men who have sex with men after 33 months of initiating HAART . Another study also found that race (Hispanic, Black or White) was not a factor in CD4 cell count response among American patients who experienced plasma HIV viral suppression within 6 months of initiation of HAART . In a comparison between an African and a European cohort both on HAART there were also no differences in CD4 response or short term virologic response . Racial differences were found in virologic response after 9 months however, poorer responses in the African cohort were thought to be attributable to lower adherence in this group.
In accordance with these prior studies, our study also showed no racial differences in virologic or immunologic responses to HAART. In other words, both Aboriginal persons and non-Aboriginal persons in our group on average achieved a typical response to HAART. This is characterized by a rapid decrease in HIV viremia to undetectable levels and a gradual increase in CD4 cell count to levels approximating those in uninfected individuals .
Survival has also been used to examine potential racial/ethnic differences after initiation of HIV treatment. Prior studies have found an increased mortality risk for HIV-infected racial/ethnic minority groups who had non-significant differences in clinical management or HIV-treatment [43, 44]. In an examination of trends in survival amongst men who have sex with men and who were diagnosed with AIDS during the HAART era, declines in deaths were smaller among racial minorities (black, Hispanic, Asian/Pacific Islanders, American Indian/Alaskan Native) compared with whites . These studies indicate that race/ethnicity has an unfavorable effect on mortality risk, however, none of these studies directly measured the effects of HAART.
This study has a number of limitations. Our measure of a history of IDU was self- or physician-reported. Because injection drug use is a stigmatized behaviour, a history of IDU may be underreported. A history of IDU was also a baseline measure that would not account for injection drug users who might have become abstinent during follow-up . Secondly, although using refill compliance as a measure of adherence has been previously validated, [37, 46–49] it may not account for individuals who received their medication but did not actually take it. In the two situations described previously, the misclassification present in the variables history of IDU and adherence could potentially bias the associations between Aboriginal status and the three outcomes of interest. The analyses for mortality and CD4 cell recovery showed a similar association between Aboriginal status in both unadjusted and adjusted regressions. The difference between the coefficients for aboriginal in the univariate and multivariate analyses ranged from 0.10 (cell recovery analysis) to 0.25 (mortality analysis), which indicates that misclassification bias or residual confounding did not influence these results. Note that for the analysis of viral suppression the coefficient for Aboriginal status, though not statistically significant, changed the direction of association. We conducted a more detailed analysis for this outcome to assess the effect of confounding and misclassification bias on the relationship of Aboriginal status and viral suppression. We observed that adherence was the strongest confounder in this analysis, since the hazards for aboriginal status across the levels of adherence were substantially different, ranging from 0.58 (95%CI: 0.37–0.92) for <95% adherence to 1.34 (95%CI: 1.01–1.78) for ≥95% adherence. Therefore, the coefficients for aboriginal just controlling for adherence and the coefficient shown in Table 3 changed by 0.05, which shows that our results were not influenced by misclassification bias or residual confounding.
The all-cause mortality rate in this cohort was associated mainly with HIV disease resulting in infectious and parasitic diseases (48%) and injuries due to intentional self-harm (19%). In addition to Aboriginal status, there were several potential clinical and socio-demographic confounders taken into account in our study. However, we did not control for the effect of co-morbidity (e.g., psychiatric illnesses), co-infections with hepatitis C, anemia, and other lifestyle (e.g., cigarette smoking) characteristics known to be related to HIV disease progression [13, 14, 16–18, 21, 48]. These factors could have influenced our results by biasing our results through residual confounding. We believe that controlling for other socio-demographic and lifestyle variables other than age, sex, income, education and history of IDU would not change our results. In our cohort we do not have data on co-infections collected longitudinally, other than hepatitis C. If we decided to include information on hepatitis C in our study, all our analyses and objectives would have to be changed dramatically mainly because of two reasons: (1) being infected with hepatitis C would not be considered a confounder variable, but it would be a factor influencing the definition of our study population; (2) we would need more study subjects to study the association of being aboriginal and clinical outcomes separately according to three distinct disease groups: (i) HIV positive, HCV negative; (ii) HIV negative, HCV positive; and (iii) HIV positive, HCV positive.
Finally, the sample size in our study was small. There were 569 participants with unknown Aboriginal status that were not included in our analyses. Like in all observational studies collecting information on socio-demographic characteristics, there is always a chance for missing information. To date several studies dealt with missing information via assumptions about the missing data or via imputation techniques for missing data. In this study, we decided to not include participants with missing information on the exposure of main interest. As reported in our results, there were significant demographic and clinical differences between participants and non-participants. Therefore, we recommend caution when interpreting the results from our analyses and extrapolating to other minority populations.
Our study highlights the need for continued research on medical intervention for HIV-infected Aboriginal persons, in particular to determine if providing services for Aboriginal drug users to address their addictions can improve survival after the initiation of HAART. Understanding the mechanism by which such health care disparities exist by determining what other aspects of being Aboriginal increase their risk of mortality after initiating HAART can provide potential targets for intervention in this vulnerable population. Results could be further extrapolated to the understanding of health care inequalities amongst other marginalized populations.