Addressing Gaps in HIV Preexposure Prophylaxis Care to Reduce Racial Disparities in HIV Incidence in the United States

Samuel M. Jenness; Kevin M. Maloney; Dawn K. Smith; Karen W. Hoover; Steven M. Goodreau; Eli S. Rosenberg; Kevin M. Weiss; Albert Y. Liu; Darcy W. Rao; Patrick S. Sullivan

Disclosures

Am J Epidemiol. 2019;188(4):743-752. 

In This Article

Results

Table 1 shows the impact of individual PrEP continuum steps for BMSM at observed and counterfactual values. In comparison with the reference scenario in which no one (of either race) received PrEP, the observed BMSM PrEP continuum scenario projected 8.4% (credible interval (CrI): 7.7, 9.1) of BMSM to be on PrEP across follow-up. This yielded a 3-percentage-point decline in HIV prevalence (39.9% versus 43.4%) and a 23% decline in incidence (hazard ratio = 0.77; CrI: 0.57, 0.99) among BMSM at year 10. The cumulative percent of infections averted was 14.1% (CrI: 8.2, 21.0) for BMSM over the intervention horizon. We then modeled changes to individual steps (holding parameters for remaining steps at observed BMSM values). For awareness, while the observed values were equal for both races, increasing that awareness proportion for BMSM had a strong impact on PrEP use, and with that, declines in incidence. For access, setting the BMSM access parameter to the observed WMSM value (95%) resulted in a smaller decline in incidence than changes to awareness. Conditional on access, empirical differences in the probability of prescription were relatively small. Increasing the proportion highly adherent did not affect the overall proportion of BMSM on PrEP (which includes PrEP users across adherence levels); this also resulted in a relatively small prevention effect. Higher levels of retention on PrEP were associated with greater PrEP prevention benefits because fewer MSM indicated for PrEP were cycled off PrEP during their periods of high sexual risk.

In Table 2, we show projections of the impact of scaling the BMSM continuum parameters jointly on HIV incidence outcomes for both BMSM and WMSM. Compared with the scenario in which all BMSM continuum parameters were set to observed levels for BMSM, when all BMSM parameters were set to levels observed for WMSM, we projected that 17.7% (CrI: 16.8, 18.7) of BMSM would actively be on PrEP. This compares with 23.4% (CrI: 22.2, 24.5) of WMSM, with the difference being due to WMSM's higher level of PrEP indications even when all continuum parameters were equal. In this scenario, where BMSM parameters were set to observed WMSM values, incidence among BMSM would be lower (hazard ratio = 0.53 vs. 0.77) than in the scenario in which BMSM parameters were set to observed BMSM values. Scaling up BMSM continuum parameters to even higher levels (150% or 200% of observed BMSM values) would result in even greater numbers of BMSM on PrEP, with stronger incidence reductions for BMSM. Overall, all levels of PrEP modeled (even those poorer than observed) resulted in a reduction in HIV incidence for BMSM compared with no PrEP, with increasing initiation and engagement associated with incidence declines by greater than three-fourths at optimistic implementation levels.

Figure 1 graphically depicts this relative scaling of the joint BMSM parameters. Changes in outcomes are nonlinear over these relative parameter changes, with the greatest marginal gains from scaling up the parameters in the range of 1.0–1.5 of observed. Although we never modified the WMSM PrEP continuum parameters (see Table 2, with the proportion on PrEP stable in all scenarios), WMSM incidence declined from 0.93 (CrI: 0.68, 1.23) per 100 PYAR in the observed BMSM scenario to 0.69 (CrI: 0.48, 0.99) in the 200% scenario. These are all indirect effects from BMSM PrEP use, possible because 11% of sexual partnerships on average were between-race.

Figure 1.

Empirical distribution of model simulations (n = 250 in each scenario) for race-specific human immunodeficiency (HIV) prevalence (A) and HIV incidence per 100 person-years at risk (PYAR) (B) at year 10 for black and white men who have sex with men (BMSM and WMSM) across relative values of the combined BMSM preexposure prophylaxis continuum (awareness, access, prescription, adherence, and retention). Relative value 1.0 represents the observed BMSM continuum values, 0.5 is half of those observed, and 2.0 is twice those observed. The vertical dashed lines indicate HIV prevalence and incidence among BMSM in the reference "no preexposure prophylaxis" scenario.

The impact of PrEP on HIV disparities is also shown in Table 2 and Figure 2. The absolute disparity in the no-PrEP scenario was 6.08 per 100 PYAR, depicted by the dashed horizontal line. Each dot in the figure represents 1 simulation, across the range of simulated relative BMSM continuum values (0.5–2.0). The set of points at a given x-axis value therefore represents uncertainty in the relationship between the continuum value and disparity measure as function of the inherent stochastic variation in the model. Implementing PrEP under the observed BMSM scenario (dotted vertical line) would reduce the absolute disparity compared with the scenario with no PrEP (4.95 per 100 PYAR), a 19% decline. If BMSM parameters were set to observed WMSM values, incidence would decline by 47% (hazard ratio = 0.53) among BMSM, with an absolute disparity of 3.30 per 100 PYAR, a 46% decline. The prevention index, the difference in hazard ratios, was effectively zero (0.01) in the scenario with BMSM parameters set to WMSM values, and even lower (indicating a greater individual-level prevention effect for BMSM) as the continuum is scaled up. Reductions in the absolute disparity index coincide with reductions in the prevention index; however, parity in the hazard ratios by race (i.e., the same individual-level effect of PrEP) is not necessary to reduce absolute disparities (i.e., population-level difference in incidence).

Figure 2.

Modeling of the absolute disparity index (human immunodeficiency (HIV) incidence in black men who have sex with men (BMSM) − HIV incidence in white men who have sex with men (WMSM)) and prevention index (hazard ratio (HR) from HIV PrEP for BMSM − HR from PrEP for WMSM) across relative values of the combined BMSM preexposure prophylaxis (PrEP) continuum, at year 10. Each dot represents 1 simulation. Dots were slightly horizontally jittered to reduce overplotting; points presented to the right of the value of 2.0 represent points for the value of 2.0. Dashed horizontal line shows the pre-PrEP disparity index; dotted vertical line shows the empirical BMSM continuum values.

The relative disparity index tells a different story. Under no PrEP, the predicted relative disparity was 4.68, whereas the disparity would increase to 6.32 in the observed BMSM scenario. Relative disparities increased despite higher PrEP use among BMSM because this relative measure is sensitive to changes in its denominator (i.e., WMSM incidence, as a function of their PrEP use). Only when the individual-level benefit of PrEP is greater for BMSM compared with WMSM (i.e., the prevention index is less than 0) do the relative disparities fall below levels in the no-PrEP scenario. Overall levels of effective PrEP care for BMSM would need to be greater or equal to those for WMSM to generate a reduction in the disparity on a relative scale.

Figure 3 aggregates the PrEP continuum into 2 factor groups of initiation (awareness, access, and prescription) and engagement (adherence and retention), with counterfactual levels of BMSM PrEP parameters in each group and outcomes of BMSM percent of infections averted and NNT. In Figure 3A, greater gains in the percent of infections averted for BMSM are projected with an increase in the initiation factors (moving left to right) compared with the same proportional increase in the engagement factors (moving bottom to top), shown by the relatively vertical orientation of the bands at the 1.0/1.0 intersection. At worse than observed initiation levels, little is gained by improving engagement. In Figure 3B, the NNT at observed initiation factor levels ranges from approximately 9 to 13 years of BMSM person-time on PrEP to prevent 1 new BMSM infection. The NNT is lower as engagement is scaled up because adherence increases the per-dose prevention efficiency. The NNT is higher as initiation factors are scaled up because high PrEP coverage leads to substantial declines in the HIV incidence rate, requiring more person-time on PrEP to prevent an infection.

Figure 3.

The percent of infections averted (PIA) for black men who have sex with men (BMSM) (A) and number needed to treat on preexposure prophylaxis (PrEP) for 1 year to prevent 1 new infection with human immunodeficiency virus among BMSM (B) across relative values of the combined BMSM PrEP continuum for initiation (factors = awareness, access, and prescription) versus engagement (factors = adherence and retention), estimated using simulated data.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....