Letter to the Editor RE: Wheatley et al. “Cost-Effectiveness of Interventions to Improve HIV Pre-Exposure Prophylaxis Initiation, Adherence, and Persistence among Men Who Have Sex with Men”

We were pleased to see another locally-oriented cost-effectiveness article to inform the US’ strategy in reaching the ‘Ending the HIV Epidemic’ (EHE) initiative. As we expressed in our prior work, we believe the EHE strategy is appropriately locally-oriented given the wide differences in the underlying epidemiology, legal and structural differences across urban centers in America.¹

Equally important however, is acknowledging the disparities in HIV incidence across racial and ethnic groups, which are propagated by disparities in access to care^2,3 and reinforced by stigma and racially-segregated social networks.^4,5 Thus far, PrEP uptake has largely been concentrated in populations of white MSM.⁶ Critically, (i) sexual mixing between white and Black/African American (hereafter referred to as Black) or Hispanic/Latinx MSM is low, limiting the benefits of PrEP to non-white MSM; and (ii) the underlying level of unsuppressed virus in the community of white MSM is far lower than that of Black or Hispanic communities. These two factors – structural and social manifestations of racism– combine to drive large differences in new diagnoses across these populations, which we believe represents the key barrier to reaching the EHE goals.

To this end, explicitly accounting for the underlying racial/ethnic disparities in health services access is of paramount importance. While this limitation was noted explicitly by Wheatley et al, the omission of these vital data was influential in their results. Wheatley et al. modeled the initiation of PrEP only among indicated MSM testing negative for HIV. Given this design, the difference in PrEP initiation between racial/ethnic groups was solely determined by the difference in the calibrated HIV testing rate (which resulted in a testing rate ratio of 1.8 for white MSM compared to Black or Hispanic/Latino MSM). While PrEP uptake figures by race/ethnicity are not available at the local level, we applied national-level estimates from 2016⁷ and re-weighted with the state-level MSM-specific race/ethnicity distribution, resulting in an estimate of 77% of PrEP users being white/others, 16% Black and 7% Hispanic/Latino. More recently, the US CDC estimated PrEP users in 2019 to be 67% white (71% white/others), 15% Hispanic/Latinx, and 14% Black⁸ amounting to a four- to five-fold difference in the percentage of white PrEP users. This national estimate may well underestimate the true racial/ethnic PrEP disparities among MSM in southern regions without Medicaid expansion including Atlanta.⁹

Another important distinction lies in the baseline level of PrEP coverage, the number of MSM indicated for PrEP and intervention scaleup. Wheatley et al. assumed 52% of the MSM population were indicated for PrEP, rather than the 25% at highest risk in our application (in concordance with CDC recommendations).^10,11 More specifically, the authors assumed 53,560 MSM were indicated for PrEP in Atlanta (52% of 103,000 MSM) whereas the CDC estimated the number of persons indicated for PrEP in the four EHE counties in Atlanta (Cobb, DeKalb, Fulton and Gwinnett) was 25,720 and 39,030 for the state of Georgia as a whole in 2019.⁸ Furthermore, Wheatley et al. assumed a baseline level of PrEP coverage of 15.1%, compared to the 8.6% we derived for the same MSM population size. These factors combined for a 4-fold difference in the MSM population accessing PrEP at baseline. Despite the similar number of new MSM users as a result of the PrEP interventions in the two models, this scale-up was concentrated among the most sexually active individuals in our application. Assuming that the interventions were to be distributed equally across race/ethnic groups would further blunt the estimated impact of PrEP and the implementation strategies used to promote uptake, adherence and persistence in Wheatley et al. With one of the highest rates of new diagnoses in the nation,¹² we believe Atlanta stands to benefit greatly from expanded testing and PrEP access, particularly if these interventions can be implemented in a way that reduces racial inequities in access.¹³

There are several other distinctions in these applications that contributed to the projected differences in incidence. First, one in every fourteen infections averted was among non-MSM in our model, which captured the full adult population rather than MSM only. Second, Atlanta is one of the fastest-growing urban centers in the nation, with a projected growth of up to 5.13 million people (aged 15–64) to 2040, coinciding with a substantial demographic shift: the Hispanic/Latinx community is projected to nearly double in proportion during this time - from 11.2% to 20.8%.¹⁴ We therefore accounted for these changes, anticipating their impact on longer-term incidence projections. Third, we accounted for poorer ART persistence in the South, and particularly among Black individuals in the South, which increases the size of the unsuppressed population and thus enhances the effects of PrEP in this community.^3,15,16 Fourth, the available literature on the effectiveness and costs of delivering PrEP was limited at the time of our work’s publication. The ongoing HealthMindr,¹⁷ Life-Steps for PrEP,¹⁸ ePREP¹⁹ and PrEP@Home²⁰ trials are sure to be valuable additions to the literature, though their results are required before an evidence-based recommendation can be made. Finally, the costs of PrEP remain a salient issue, particularly with the availability of generic formulations that could greatly expand PrEP uptake among MSM facing cost-related barriers.²¹ Our price estimates accounted for financial support provided by the Gilead Advancing Access program,²² which has the potential to help increase access to populations that may be most in need since far more PrEP users tend to be privately insured.²¹

Models are designed to produce guidance based on health outcomes and expenditures registered far into the future. Uncertainty is inherent, and variance in long-term outcomes between models with different structural assumptions is to be expected. By contrasting and comparing these projections we can draw insights about the most uncertain elements, refine our estimates and advocate for more and better data where it is needed most. We welcome opportunities for such collaboration in support of the EHE initiative.