Background
Self-reported adherence to PrEP may be inaccurate due to social desirability and recall biases. Although biomarkers can provide more reliable information on pill-taking, the knowledge of biomarker collection can itself increase the accuracy of self-report.
Methods
We explored correlations between self-reported PrEP adherence and biomarker data in the PUMA trial in Kenya. HIV-uninfected women ≥18 years using oral tenofovir (TFV)-based PrEP for three months were enrolled (N=100) from 2021–2022 and randomized 1:1 to a urine point-of-care assay with drug-level feedback or standard-of-care. Adherence was measured at enrollment, 3-, 6-, 9- and 12-months via urine and hair assays and self-report. Short-term PrEP adherence was defined as a positive urine assay result (TFV>1500 ng/mL). Long-term PrEP adherence was defined as detectable TFV in hair (>0.002 ng/mg). Self-reported adherence was calculated as the mean score (range: 0-100) on a 2-item modified Wilson scale assessing percentage of doses taken in the prior month and adherence rating. We used t-tests to explore differences in mean score by biomarkers at Month 12. Generalized estimating equations estimated longitudinal associations between the self-reported adherence score and the biomarkers – a positive urine assay or detectable TFV in hair – adjusting for study arm.
Results
At enrollment, 68% had a positive urine assay, 72% had detectable TFV in hair, and the mean self-reported adherence score was 90 (standard deviation [SD]=18). At Month 12, 86 were retained, of whom 59% had a positive urine assay, 69% had detectable TFV in hair, with a mean adherence score of 79 (SD=28); mean adherence score was higher in the groups with a positive urine assay (mean score of 91 vs. 62 for those with a negative assay; p<0.01) and detectable TFV in hair (mean score of 87 vs. 62 for those with undetectable TFV; p<0.01). In longitudinal models, higher adherence score was associated with a greater likelihood of a positive urine assay (adjusted relative risk [aRR]: 1.12; 95% confidence interval [CI]: 1.08-1.17; p<0.01) and detectable TFV in hair (aRR: 1.02; 95% CI: 1.00-1.04; p=0.02).
Conclusions
Associations between self-reported and biomarker-verified PrEP adherence indicate that women in this cohort were able to accurately report pill-taking. Self-reported adherence using validated items may be able to more accurately identify PrEP use patterns in contexts where women know their adherence is being monitored periodically with biomarkers.
Table or Figure
