Trajectories of Pre-exposure Prophylaxis Adherence Among Commercially Insured Individuals

Abstract

We used group-based trajectory models to identify 4 distinct trajectory patterns of adherence to preexposure prophylaxis (PrEP) among 20 696 users. Only 44.5% were consistently PrEP adherent, with younger age, being female, or having substance use disorder or depression associated with early discontinuation. Public health efforts are needed to improve PrEP adherence.

Preexposure prophylaxis (PrEP) prevents human immunodeficiency virus (HIV); the iPrEx trial showed 44% risk reduction in HIV incidence with PrEP use compared with placebo among transgender women or men who have sex with men [1]. However, a subsequent analysis showed 99% HIV risk reduction when PrEP was taken daily [2]. The Partners PrEP trial reported 75% risk reduction in HIV incidence among heterosexual couples with daily PrEP compared with placebo [3]. A meta-analysis estimated 86% efficacy with ≥80% PrEP adherence but only 45% efficacy with <80% adherence [4].

Despite the recognized efficacy of PrEP, low PrEP adherence and persistence threaten its effectiveness [1, 2, 4]. This was further demonstrated in the VOICE [5] and the FEM-PrEP [6] trials, which failed to demonstrate PrEP effectiveness, owing to nonadherence; only 29% of participants in the VOICE trial and <40% in the FEM-PrEP trial had detectable drug levels [5, 6]. Several studies assessing PrEP adherence in clinical practice have identified factors associated with nonadherence [7–9]. Lower adherence was observed among patients who were female, <35 years of age, or black and who had substance use disorder (SUD) [9]. However, those studies used a single measure of adherence (eg, proportion of days covered [PDC]) and arbitrarily defined cutoffs to classify an individual as adherent or nonadherent.

That approach may not fully consider the heterogeneity of PrEP adherence patterns over time, limiting the potential to identify patients with differential underlying HIV risk profiles. Hence, the objectives of the current study were (1) to use group-based trajectory models (GBTMs) to identify trajectories of PrEP adherence among commercially insured PrEP users and (2) to identify patient characteristics associated with those PrEP adherence trajectories.

METHODS

Study Design and Data

This retrospective cohort study using data from the 2012–2020 MarketScan Database, a US administrative claims database of >273 million deidentified commercially insured enrollees, was approved by the University of Florida Institutional Review Board.

Study Cohort and Measures

We applied a previously developed algorithm to identify individuals aged 12–64 years who used PrEP. We required persons to be prescribed ≥30 days of PrEP (tenofovir-disoproxil-fumarate with emtricitabine, or tenofovir-alafenamide with emtricitabine) and to be continuously enrolled in commercial health insurance 12 months before and 30 days after PrEP initiation. To ensure PrEP use, we excluded individuals with any diagnosis code (using the International Classification of Diseases, Ninth Revision and Tenth Revision) or prescription for HIV, hepatitis B virus, or HIV postexposure prophylaxis during the 12 months before and 30 days after PrEP initiation. The index date was defined as day 31 after PrEP initiation (ie, day 1). PrEP users were also required to not have HIV and to be continuously enrolled in commercial health insurance for 180 days after the index date to ensure a sufficient PrEP adherence assessment period.

Statistical Analysis

We used GBTM to identify patient groups with similar PrEP adherence patterns across 360 days from the PrEP index date. Medication adherence was measured using the PDC with PrEP for each 15-day interval, calculated as the proportion of days that an individual had medication. We calculated the PDC until the earliest occurrence of HIV outcome diagnosis (defined as 1 inpatient or the first of 2 outpatient HIV diagnosis records), loss of enrollment, or end of follow-up. We selected the final model according to a combination of (1) bayesian information criterion values between models, with the largest value suggesting the best fit, (2) estimated trajectory group proportions sufficiently large (≥10%) to maximize clinical interpretation and utility, (3) visual inspection to identify clinically meaningful trajectories, and (4) Nagin's criteria [10] for assessing the final adequacy of a GBTM [11].

We used multivariable logistic regression to identify patient characteristics (sex, age, geographic region) and comorbid conditions (history of depression, sexually transmitted infection, or SUD) associated with the least adherent group membership, using the most adherent group as the reference. Patient characteristics and comorbid conditions were identified within 1 year before initiation.

RESULTS

Among 20 696 new PrEP users identified, 96% were male, 32% were 25–34 years of age, 6% had SUD, and 12% received a sexually transmitted infection diagnosis within 1 year before PrEP initiation (Supplementary Table 1). Figure 1 shows the final GBTM model with 4 trajectories of distinct patterns of PrEP adherence, including early discontinuation (23.3% of the cohort), rapidly declining (12.2%), gradually declining (20.0%), and consistently adherent (44.5%).

Figure 1.

Adherence trajectories of preexposure prophylaxis use over the 360 days after the index date (N = 20 696). The consistently adherent group is the reference. Dashed lines indicate 95% confidence levels for the predicted average (Avg) proportion of days covered (PDC).

In multivariable multinomial logistic regression (Supplementary Table 2), PrEP users were more likely to be in the early discontinuation group than in the consistently adherent group if they were younger (aged 12–24 years [adjusted odds ratio, 5.05 (95% confidence interval, 4.43–5.77) or 25–34 years [1.61 (1.44–1.81)] vs 35–44 years), female (6.62 [5.31–8.24]) versus male, had depression (1.24 [1.11–1.39]) or SUD (1.52 [1.28–1.81]), or started PrEP in 2020 (1.63 [1.36–1.96]) versus 2019.

DISCUSSION

To our knowledge, this is the first study to use a nationally representative sample of the commercially insured US population to study longitudinal patterns of PrEP adherence. Only 44.5% of PrEP users were consistently adherent to PrEP in the first year, while 20.0% had gradually declining, 12.2% had rapidly declining, and 23.3% had early discontinuation PrEP adherence trajectories. Younger PrEP users and those with depression or SUD were more likely to have early discontinuation than consistently adherent PrEP adherence patterns.

We observed lower PrEP adherence (65%) than that reported in a prior study (92%), which used data from Kaiser Permanente Northern California [7]. The higher adherence in that study may be attributable to the use of telemedicine adherence support to mitigate nonadherence, whereas such services may not be readily available to all PrEP users. Another study reported 82% PrEP adherence among transgender PrEP users identified in the Oregon Medicaid claims database [8]. Oregon was one of the first states to approve Medicaid coverage for gender-affirming hormones and surgical procedures; thus, continued interaction with the healthcare system may improve PrEP adherence [8].

Only 2 prior studies applied GBTM to assess adherence among PrEP users: a prospective study in Australia [12] and a study assessing Latino, Asian, or black individuals in a federally qualified health center in Chicago [13]. Similar to our study, the Australian study identified 4 adherence trajectory groups, whereas the Chicago study identified 3. The findings of both those studies are not generalizable to PrEP use in the overall US population, given their study populations, unique clinical settings, and relatively small sample sizes.

Compared with individuals aged 35–44 years, those aged 12–24 years, accounting for 20% in the present study, were 5 times more likely to be in the early discontinuation group than in the consistently adherent group. Prior research has also reported an association of younger age with poor PrEP adherence [12]. This finding is concerning, given the fact that individuals aged 13–24 years accounted for the second-highest new HIV infection rate [14]. Hence, targeted intervention is needed to improve medication adherence within this age group. We also found that PrEP users who were female or had a history of depression or SUD were more likely to be PrEP early discontinuers than to be consistently adherent. This is consistent with previous research showing depression [13] or SUD [15] as predictors for antiretroviral therapy and PrEP nonadherence, respectively; however, further research is needed to understand this association. Finally, individuals initiating PrEP in 2020 during the coronavirus disease 2019 pandemic were >60% more likely to be early discontinuers. The pandemic caused disruptions in PrEP prevention services and healthcare access, including loss of insurance and difficulties obtaining prescriptions that likely affected PrEP adherence [16].

A strength of the current study is that we applied GBTM to assess PrEP adherence using data from a national commercial claims database. Given that most PrEP users are commercially insured, our findings are likely generalizable to commercially insured PrEP use across the United States. This study has limitations. First, we relied on an algorithm of prescription and medical claims to identify PrEP use. It is possible that incomplete, missing, or miscoded claims may affect our study findings. However, the algorithm has been used in previous studies. Second, the PDC was based on pharmacy claims as a proxy measure of medication adherence because we were unable to determine whether patients consumed medication as prescribed. In addition, individuals may use PrEP on demand, which may lead to overestimation or underestimation of the PDC. Third, although we included many covariates, we cannot rule out potential unmeasured confounders (eg, race/ethnicity or sexual behavior), which were not captured in these data. Finally, we included commercially insured individuals only; thus, our findings may not be generalizable to other important subgroups of the US population (eg, uninsured people).

In conclusion, this study identified 4 distinct trajectories of PrEP adherence during the first 360 days of use and observed that <40% of PrEP users belonged to an adherent trajectory group. The early discontinuation trajectory group was associated with younger age, being female, having a history of depression or SUD, and initiating PrEP in 2020. These findings highlight suboptimal PrEP adherence and identify individuals who may benefit from targeted interventions to improve PrEP adherence, ensuring that PrEP users are protected from HIV infection.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author contributions. Conceptualization, data curation, formal analysis, investigation, resources, software, and validation: I. U. and H. P. Methodology: I. U., W. H. L., R. L. C., and H. P. Project administration: I. U., A. G., and H. P. Supervision: I. U., A. G., R. L. C., and H. P. Visualization: I. U., , W. H. L., and H. P. Writing—original draft: I. U. Writing—review and editing: all authors.

Data availability. The data used were from the MarketScan database. Data are not publicly available.