Since its introduction in the mid-1990s, highly active antiretroviral therapy (HAART) for the treatment of HIV infection has proved efficacious in reducing morbidity and mortality and suppressing viral replication [1, 2]. However, virologic or immunological failure, drug toxicities, drug interactions, and problems with adherence, often lead to discontinuing, interrupting, or switching HAART regimens. These changes, particularly early modification of initial HAART medications within the first year of therapy, have been associated with poor clinical outcomes [3]. Also, exposure to the finite number of treatment medications available can diminish the effectiveness of related drugs in future therapeutic regimens [4]. Both the International AIDS Society-USA Panel [5] and the Department of Health and Human Services [6] have cautioned against prematurely switching antiretroviral regimens because of limitations placed on future treatment alternatives due to potential or actual development of viral resistance.

The reasons for switching HAART medication regimens are multifactorial and include failure of current therapy, drug toxicity and drug interactions, as well as non-compliance [7]. Studies have shown that patients with lower mean CD4⁺ count and higher HIV RNA levels are more likely to switch HAART [4, 8]. Marital status as well as time from HIV diagnosis to HAART initiation has also been associated with HAART treatment modification [9, 10].

Many individuals infected with HIV and treated with HAART are from U.S. minority race/ethnicity groups [1] and although reports have suggested that race/ethnicity is a risk factor for changing/switching the HAART regimen, few studies have investigated regimen switching in minority populations; however, African-American race has been associated with switching, interrupting [2], and discontinuing HAART [7, 11-14]. The effect size of African American race on discontinuation (HR 1.26) was similar to that of higher pill burden in one report [7] and the adjusted odds of interruption for African Americans (OR 1.77) was similar to the effect of depression in another [12]. Hispanics have been less well studied, but one report that grouped Hispanics with African Americans and Others showed a 2-fold hazard for discontinuation of HAART [13]. Additionally, studies have shown possible differences between races/ethnicities in adherence and ability to manage drug toxicities [15-17]. In these studies, potential confounding due to the association of race/ethnicity with social factors, such as and access to healthcare and medications, has been challenging to address.

The U.S. Military HIV Natural History Study cohort minimizes many of the confounding social factors and so, we sought to address this question. If race/ethnicity is not itself a risk factor for early HAART switching, which is associated with poor clinical outcomes [3], then clinical and public health efforts to improve HAART outcomes can be focused toward access to healthcare and medications, as well as other related factors. The U.S. Military HIV Natural History Study (NHS) provides the ability to evaluate associations between race/ethnicity and clinical outcomes without many of the socioeconomic confounders in other cohort studies. Within the military healthcare system there is free cost and equal access to healthcare and medications, and military members have steady income and a minimum of a high school education or equivalent. Previous studies in the NHS have shown that despite this, African Americans were less likely to achieve viral suppression after initiating HAART [18-19]. However, clinical outcomes after HAART initiation were similar among races/ethnicities in other studies in our cohort [20, 21], although risk factors for adverse outcomes may differ by race/ethnicity.

We sought to test the hypothesis that in a U.S. military cohort, non-Caucasian race/ethnicity HIV-positive patients are no more likely to switch initial HAART regimens within a one-year period of time than Caucasian subjects while also exploring other factors related to HAART switching.

2457 subjects met the inclusion criteria. The population studied was predominantly male (91%) with a median age of 34 years (Table 1). 42% of the cohort were Caucasian, 43% African-American, 9% Hispanic and 6% were Other race. 65% of the subjects initiated HAART during 1996-2001, 19% during 2002-2007 and 15% during 2008-2012. Overall, the most common regimen initiated was unboosted PI (51%) followed by NNRTI (34%) and boosted PI (7%). At the time of HAART initiation 84% were active duty.

There were significant differences in HIV viral load, but not CD4 count at HAART initiation between race/ethnicity groups (Table 1); however, the largest proportion of subjects in each group had viral loads greater than 10,000 log₁₀ copies/ml and CD4 counts above 350 cells/ml. Those in the Hispanic and other groups had a significantly shorter time between HIV diagnosis and HAART initiation and significantly more individuals in these groups started HAART in the 2008-2012 era than the 1996-2001 era compared to Caucasians and African Americans (Table 1).

Caucasians were significantly more likely to have malignancies other than basal cell carcinoma, and coronary artery disease at HAART initiation. Major depression was diagnosed in 24.2% and anxiety disorder in 8.8% of the total population, but each were significantly lower in African-Americans compared with Caucasians, Hispanic or Other racial groups. There were several baseline lab values that were statistically different between ethnicities (AST, total bilirubin, HDL, LDL, triglycerides, creatinine and hemoglobin), but these were not felt to be clinically significant. (Table 1 and Supplementary Table S1).

This study showed lower odds of switching HAART medication regimens in African-Americans and no greater odds in Hispanics compared to Caucasians at one year after initiating HAART, however other races were more likely to switch, but constituted a small proportion of studied subjects. It is important to characterize the frequencies of switching specific HAART regimens as this provides an estimation of the rate at which the regimens may be failing or intolerable. We used the U.S. Military HIV Natural History Study to evaluate the frequency and type of HAART regimen switching over a 15-year period focusing on differences among race/ethnicity groups in a setting with minimal socioeconomic confounding. The proportion of subjects who switched regimens in this study is lower than that generally reported in the literature possibly related to fewer socioeconomic and healthcare availability factors in the military setting. However, elements known to be associated with HAART medication (HIV viral load and CD4 count) demonstrated statistically significant relationships in this study to odds of switching at 12 months [4, 22].

While studies have described frequencies [8, 23, 24], probabilities [25, 26] and risk [27] of discontinuing or switching HAART regimens as well as proportions of subjects changing [28], interrupting [29], or stopping/discontinuing [30] medications, few studies have documented trends in discontinuing or switching HAART medications in minority populations. Data from the Women’s Interagency HIV Study [11] showed that compared to those who did not switch from their initial regimen, those who did switch were significantly more likely to be African American (65% vs 50%). In studies that did not specify HAART-naïve subjects, Li et al. [12] found that Black race was an independent predictor for short interruption of HAART for </= 7 days (OR 2.42, 95% CI 1.23, 4.74); also, Yuan et al. [7]described that Black race was predictive of discontinuation (RR=1.28, 95% CI 1.13, 1.45) while Robison et al. [14] illustrated Black race was associated with discontinuation (OR 1.68, 95% CI 1.80, 2.60) for virologic failure/ non-adherence. In our study, there was a lower odds of switching HAART regimens in African-American compared to Caucasian race in both univariate and multivariate analyses. This may be explained by equal access and lower barriers to medical care seen in the military healthcare system.

There have been few studies that have further stratified other races and thus few data are available on HAART switching among Hispanic populations. In our study they were noted to have no higher rate of discontinuation when compared with Caucasians. Finally, there were increased odds of switching HAART among the Other race group compared to Caucasians even in multivariate analyses that may be related to the relatively small numbers in this group. All of the above differences were noted to be despite similar breakdown by CD4 and viral load and an increased proportion of the group starting HAART in later periods.

In this NHS cohort, while 24.8% of the study subjects switched regimens, the vast majority of those patients switched in the earliest time period of 1996-2001 (76.4%). This finding makes sense as most of the subjects in our cohort initiated HAART in the earliest period, marked by higher pill burden and regimen toxicity, as well as lower potency. Mocroft et al. [28] found the incidence of any discontinuation of HAART was significantly lower after 1999 compared to before [incidence rate ratio 0.43, 95% CI 0.35, 0.53]. Also, Cicconi et al. [26] illustrated that patients who started HAART during a 'recent' period of 2003-2007 were less likely to change their initial regimen because of intolerance/toxicity [adjusted relative hazard 0.67, 95% CI 0.51, 0.89] compared to an ‘early’ period (1997-1999). Similarly, in our study, HAART initiation during the later period 2008-2012 had a much lower rate of regimen switching of 14.2%, suggesting the use of those regimens currently recommended by the DHHS may decrease the risk of switching.

This study demonstrated clear, significant associations between HIV viral load, CD4 count, as well as having depression, and odds of switching HAART regimens. These associations are reassuring as indicators of internal study validity. The association with depression is particularly interesting given associations previously described between depression and HAART uptake and adherence. Depressed patients with HIV infection are an important group receiving HAART prescriptions and this association may suggest this population would benefit from additional provider/clinic-based interventions, including enhanced adherence counseling, support, or closer monitoring after HAART initiation. Furthermore, the significant association between higher viral load at initiation and switching in multivariate analysis is consistent with prior studies demonstrating higher likelihood of discontinuing, interrupting, and modifying HAART regimens compared to lower values [12, 31, 32]. Prior antiretroviral use was associated with a two-fold increased risk of regimen switching (adjusted OR 2.06, 95% CI 1.63, 2.62)) suggesting possible HIV resistance. Van Roon [23] has described non-naiveté for ART as a determinant for discontinuation of initial HAART and Mocroft [31] showed that previously-treatment naïve patients are less likely to modify HAART. This finding also suggests that this population might benefit from closer monitoring after HAART initiation.

There are some study limitations. As the cohort is predominantly male, findings may not be fully reflective of HAART switching among women; additionally, while the cohort is derived from geographically and racially/ethnically diverse individuals from the U.S. population and is likely representative of those with open access to healthcare and medications, unmeasured differences may exist that could limit generalizability, especially to those who are socioeconomically disadvantaged. This was a retrospective analysis covering a broad timeframe and involved multiple sites (sites were not significantly associated with switching in the multivariate analysis) among which practice patterns may have varied. Additionally this was a cross-sectional study at twelve months. We chose this approach to reflect similar analyses in the literature.

This observational, cross-sectional study showed that there was a lower odds of switching HAART at one year between African-American race compared with Caucasian race. There was no significant difference between Hispanic and Caucasian race. Other race/ethnicity had significant effect on switching HAART regimen at 12 months; however, this group only represented a small proportion of the study cohort. Associations previously demonstrated in the literature may be due to socioeconomic or other confounders which are minimized in this cohort [32]. This study reinforces that clinical factors should be noted by providers when prescribing HAART, including low CD4 count, high HIV viral load, and depression, as these were associated with switching the initial regimen; closer monitoring and management of initial HAART may be warranted in these higher risk patients with HIV infection.