Robert Grant received 3 grants to support meals and lodging for study personnel to attend annual investigator meetings and 1 grant to support a study video project. Monica Gandhi received payment for participating in a one-time advisory board meeting of Gilead (March 2012). This does not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials.
Conceived and designed the experiments: AL, YH, PB, PLA, KG, RG, SB, RG, and MG contributed to the design, analysis, and interpretation of the data, drafting and/or revising of the article. Performed the experiments: QY and YH conducted the hair assays for this paper. Analyzed the data: AL, CJ, PB and MG were primarily responsible for analyzing the data. Contributed reagents/materials/analysis tools: KS helped provide data relevant to this manuscript. Wrote the paper: All authors (AL, QY, YH, CJ, PB, KS, PLA, KG, RG, SB, RG, and MG) contributed to the drafting or revising of the article and approved the final version of the manuscript. Contributions as above.
Pre-exposure prophylaxis (PrEP) trials using tenofovir-based regimens have demonstrated that high levels of adherence are required to evaluate efficacy; the incorporation of objective biomarkers of adherence in trial design has been essential to interpretation, given the inaccuracy of self-report. Antiretroviral measurements in scalp hair have been useful as a marker of long-term exposure in the HIV treatment setting, and hair samples are relatively easy and inexpensive to collect, transport, and store for analysis. To evaluate the relationship between dose and tenofovir concentrations in hair, we examined the dose proportionality of tenofovir in hair in healthy, HIV-uninfected adults.
A phase I, crossover pharmacokinetic study was performed in 24 HIV-negative adults receiving directly-observed oral tenofovir tablets administered 2, 4, and 7 doses/week for 6 weeks, with a ≥3-week break between periods. Small samples of hair were collected after each six-week period and analyzed for tenofovir concentrations. Geometric-mean-ratios compared levels between each pair of dosing conditions. Intensive plasma pharmacokinetic studies were performed during the daily-dosing period to calculate areas-under-the-time-concentration curves (AUCs).
Over 90% of doses were observed per protocol. Median tenofovir concentrations in hair increased monotonically with dose. A log-linear relationship was seen between dose and hair levels, with an estimated 76% (95% CI 60–93%) increase in hair level per 2-fold dose increase. Tenofovir plasma AUCs modestly predicted drug concentrations in hair.
This study found a strong linear relationship between frequency of dosing and tenofovir levels in scalp hair. The analysis of quantitative drug levels in hair has the potential to improve adherence measurement in the PrEP field and may be helpful in determining exposure thresholds for protection and explaining failures in PrEP trials. Hair measures for adherence monitoring may also facilitate adherence measurement in real-world settings and merit further investigation in upcoming PrEP implementation studies and programs.
ClinicalTrials.gov +
Recent studies have provided new hope for effective HIV biomedical prevention strategies
Importantly, several trials have highlighted the critical relationship between adherence and PrEP efficacy. For example, the efficacy of FTC/TDF in iPrEx rose from 44% overall to an estimated 92% among those with detectable blood drug levels
Incorporating measures of drug exposure as biomarkers of PrEP adherence has been critical to interpreting PrEP trials. In iPrEx, although mean adherence by self-report was 95%, drug was detected in only 8% of seroconverters compared with 54% of matched active-arm controls who remained uninfected
Due to TFV's extended intracellular half-life (∼150 hours)
To widely employ hair as a biological marker of drug exposure in HIV prevention settings, the relationship between different dosing strategies and drug concentrations in this matrix must be established. Moreover, individual variation in pharmacokinetic (PK) parameters (e.g. plasma oral clearance measured as areas-under-the-concentration-time-curve or AUCs) can contribute to variability of antiretroviral drug levels in various compartments
The protocol for this trial and supporting CONSORT checklist are available as supporting information; see
STRAND was an open-label, randomized, three-period PK study conducted between October 2009 and March 2011 at the University of California San Francisco (UCSF) Clinical Research Center and the San Francisco Department of Public Health (SFDPH) Bridge HIV research clinic. To facilitate directly observed dosing, participants were recruited from UCSF (students, staff, and faculty) and from neighborhoods adjoining SFDPH. Tablets of TDF (300 milligrams, mg) were provided by Gilead Sciences (Foster City, CA). As several PrEP trials were evaluating TDF alone as a PrEP agent
Eligible subjects were healthy male or female volunteers age 18 or older without serious or active medical conditions. Participants were all HIV-negative based on HIV rapid antibody testing and were hepatitis B surface antigen negative at enrollment. Enrolled participants could not be on nephrotoxic agents, had to have an estimated creatinine clearance ≥60 mL/min, serum creatinine levels below the upper limit of normal, urine dipstick tests that were negative or trace positive for glucose and protein, adequate hepatic and hematologic function, and negative urine pregnancy tests (for women). Female volunteers were required to use effective contraception during the study.
All participants had to be willing to undergo directly observed dosing (DOT), provide personal cell phone numbers for unobserved modified DOT (mDOT) visits, have dark (brown or black) hair (to minimize interindividual variability in this proof-of-concept study), a minimum occipital scalp hair length of 3 centimeters (cm), and not have used hair dyes or hair permanent products in the past 3 months. Participants at high risk of HIV infection, including those reporting injection drug use (IDU); vaginal or anal intercourse with an HIV-positive partner, IDU partner, or with more than 4 partners of the opposite sex; and/or transactional sex in the prior 6 months were excluded.
Subjects were randomly assigned to one of 6 dosing sequences and completed three 6-week dosing periods of 2, 4 and 7 doses per week, with a break of at least 3 weeks between periods (
At enrollment and end-point visits, approximately 150–200 strands of scalp hair were collected; additionally, a similar number pubic hair strands were collected on an opt-in basis. After 4 weeks of dosing during the 7 doses/week period (at steady state for plasma TFV), participants were admitted to the UCSF Clinical Research Center for intensive 24-hour plasma PK evaluations to calculate individual PK parameters (plasma AUC for oral clearance and Cmax). The participant's usual diet was ascertained prior to the PK visit, and simulation of the usual diet was undertaken during PK sampling. An initial blood level was drawn (called the “0” timepoint) prior to an observed dose of TFV, followed by blood collection at 0.25, 0.5, 1, 1.5, 2, 3, 4, 6, 8, 10, 12, 16, and 24 hours post-dose for measurement of TFV plasma concentration. Concentrations could not be measured for one participant at the 0.5 hour time and another participant at the 1.5 hour time; these were interpolated by fitting quadratic functions to the log-concentrations using the two previous and two subsequent times.
Several small thatches of hair (∼150–200 strands) were cut as close as possible to the scalp in the occipital region and/or pubic region, and the distal portion labeled
Plasma was collected in EDTA tubes for drug concentration measurement for the intensive pharmacokinetic studies. Plasma was analyzed for TFV via standard techniques of LC/MS/MS
All analyses were conducted using SAS (Version 9.2, SAS Institute, Cary, NC) and Stata (Version 12.1, College Station, TX). The primary outcome variable was concentration of TFV in hair at each end-point visit. Median drug levels with 95% confidence intervals (CIs) were calculated, and geometric mean ratios with 95% CIs were used to compare levels between each pair of dosing conditions. We used geometric means to avoid overweighting large ratios and to match the logarithmic transformation used for the multivariate modeling. Spaghetti plots were used to depict drug levels for each participant, with each line representing an individual. The chosen sample size (n = 24) was not based on formal power considerations, but was feasible and expected to provide useful new information about typical relationships between TDF dosing and drug concentrations, as well as a preliminary assessment of how well hair levels could distinguish between daily versus intermittent dosing.
Linear mixed models with random intercepts were used for multivariable analyses of predictors of logarithmically transformed hair levels. Plasma AUC0–24, a measure of the subject's oral clearance (dose*F/CL), was calculated using the linear-log trapezoidal method, and Cmax was defined as the largest observed TFV concentration in plasma over the 24 hours. The primary predictor variables were dosing condition (2, 4, and 7 dose/week), log-transformed plasma AUC, and log-transformed Cmax. We examined linearity assumptions for continuous predictors by adding quadratic terms to the models.
Twenty-four participants met eligibility criteria and were enrolled (
Variable | N | % |
Female | 12 | 52% |
Male | 11 | 48% |
Non-Hispanic White | 12 | 52% |
Asian/Pacific Islander | 3 | 13% |
Latino/Hispanic | 7 | 30% |
Other | 1 | 4% |
Some college | 9 | 39% |
Bachelors | 9 | 39% |
Graduate | 5 | 22% |
Variable | N | Mean | SD | IQR |
23 | 34.2 | 9.0 | 26.6–41.1 | |
23 | 129.4 | 31.1 | 107.2–154.6 | |
23 | 73.9 | 13.2 | 64.1–86.3 |
Estimated by the Cockcroft-Gault equation
N = number of participants; SD = standard deviation; kg = kilograms. IQR = interquartile range.
There were 167 adverse events (AEs) reported during the study, with each participant reporting at least 1 AE. Most AEs were either mild, grade 1 (128 events, 76%) or moderate, grade 2 (37 events, 22%). The most commonly reported AEs were headache (54%), upper respiratory infection (38%), nausea (29%), abdominal gas (17%), soft/loose stools (17%), and fatigue (17%). Only one participant experienced a grade 3 AE (traumatic jaw fracture) and one experienced a grade 4 AE (metastatic breast cancer), both adjudicated as unrelated to study drug. Four participants withdrew before completing all study procedures: one after the first TDF dose due to a grade 2 hypersensitivity reaction (urticarial rash with tongue tingling) occurring prior to PK data collection, one before completing the 4 and 7 doses/week periods due to headaches, one before completing the 2 doses/week period due to use of post-exposure prophylaxis and elevated pancreatic enzymes, and one before completing the 7 doses/week period due to the new cancer diagnosis described above.
Over 90% of doses were observed according to protocol: 249/267 (93%) doses in the 2 doses/week arm, 473/517 (91%) doses in the 4 doses/week arm, and 773/856 (90%) in the 7 doses/week arm. Of these 1495 doses observed on protocol, 82% were DOT and 18% were mDOT visits. An additional 130 (7.9%) doses were taken from participants' back-up medication supplies (confirmed via phone/text message). Two participants were unable to attend several consecutive DOT visits during their 7 doses/week period due to travel; all 31 of these doses were confirmed on time via phone or text. Combining per protocol doses and confirmed doses from back-up supplies, 1624/1640 (99%) of expected doses were taken according to prescribed schedules Overall, 55/67 (82%) of dosing periods had 100% on-time completion rates, and only 2 periods had <90% completion: one participant missed 3/12 (25%) doses during the 2 doses/week period, and another participant missed 4/36 (11%) doses in the 4 doses/week period.
Median TFV levels in scalp hair increased monotonically from 2 to 4 to 7 doses/week (
Hair (scalp) | ||
Dosing period | N | Median (range) TFV levels (ng/mg) |
2 doses/week | 22 | 0.012 (0.008 to 0.021) |
4 doses/week | 22 | 0.023 (0.011 to 0.042) |
7 doses/week | 21 | 0.038 (0.021 to 0.053) |
Period comparison | Geometric mean ratios (95% CI) | |
4 vs. 2 doses/week | 21 | 1.76 (1.45 to 2.13) |
7 vs. 4 doses/week | 21 | 1.60 (1.34 to 1.90) |
7 vs. 2 doses/week | 20 | 2.85 (2.46 to 3.32) |
N = number; TFV = tenofovir; ng/mg = nanogram/milligram; CI = confidence interval.
Hair TFV concentrations for each subject are shown for each dosing period. Each line represents drug concentration data from one participant (at 2, 4, and 7 doses/week). The red line distinguishes between hair scalp concentrations for 2 vs. 7 doses/week. ng/mg = nanogram/milligram.
Patient ID | Hair TFV (ng/mg) at 2 doses/week | Hair TFV (ng/mg) at 4 doses/week | Hair TFV (ng/mg)at 7 doses/week |
039 | 0.017 | 0.011 | 0.044 |
051 | 0.013 | 0.028 | 0.026 |
063 | 0.012 | 0.042 | 0.028 |
ID = identification number; TFV = tenofovir; ng/mg = nanogram per milligram.
In a model with log(doses per week) and log(AUC) as predictors, a log-linear relationship was seen between doses per week and TFV scalp hair level, with an estimated 76% (95% CI 60–93%, p<0.0001) increase in hair level per 2-fold dose increase. Adding a random dose effect to this model estimated minimal inter-individual variability in dose effect (estimated random effect variance of zero), suggesting very similar dose effects across most subjects. TFV plasma AUCs, ranging from 1210 to 5610 ng×h/mL, were a moderate predictor of TFV hair level, with a 23% increase in hair level per 2-fold increase in AUC (p = 0.035). Allowing for different residual variance at different dose levels (heteroscedasticity) resulted in similar dose and AUC estimates. Also, similar results were observed when restricting analyses to the 7 doses/week period where intensive PK parameters were measured (27% increase in hair level per 2-fold increase in AUC, p = 0.042). Cmax appeared to have little positive effect on hair level when controlled for dose and AUC (−20% per 2-fold increase, 95% CI −43% to +10%, p = 0.16). There was no statistically significant effect of total body weight (p = 0.35), estimated creatinine clearance (p = 0.52), or sex (p = 0.89) on TFV hair concentrations.
Pubic hair TFV levels varied erratically and did not monotonically increase in 5/11 subjects who had at least two on-drug pubic hair samples. In addition, levels were exceedingly high in 6 samples (∼30–300 times higher than TFV levels in scalp hair).
Accurate adherence measures are critical to the interpretation of PrEP clinical trials and may play an important role in monitoring PrEP use in clinical practice. An effective pharmacologic measure of PrEP adherence in a particular matrix must demonstrate a clear relationship between dose and drug levels. Our analysis provides evidence of a proportional relationship between dose and hair levels of TFV in small samples of scalp hair from healthy, HIV-uninfected individuals. Specifically, there was an estimated 76% increase in TFV hair level with doubling of dose, with similar dose effects across subjects. Hair concentrations exhibited linearity with dose in the range of 2 to 7 doses per week, with minimal inter-individual variation in dose effect. A hair concentration of 0.021 ng/mg distinguished between 2 versus 7 doses/week dosing. This discrimination ability may have useful clinical applications in monitoring PrEP use, as a previous PK modeling study demonstrated that estimated protective efficacy for taking FTC/TDF 2 doses/week was 76% (95% CI 56 to 96%), compared with a projected 99% (95% CI 96 to >99%) for 7 doses/week
Pubic hair samples were not useful in measuring TFV concentrations and varied erratically in this study, with several high outlier values. Similar results have been reported for pubic hair levels of opiates
Antiretroviral drug levels in hair have been shown to be highly correlated with treatment outcomes in HIV-infected patients, demonstrating stronger predictive value than self-reported adherence
As tenofovir's long intracellular half life could afford TDF-based PrEP regimens some PK forgiveness
Our study had several strengths, including use of a randomized cross-over study design, high adherence to mDOT procedures ensuring dosing fidelity, and intensive PK sampling to evaluate key PK parameters that could influence drug concentrations in various compartments. This study also had several limitations. First, participation was limited to individuals with dark hair, as we aimed to reduce inter-individual variability and maximize the ability of hair to discriminate between different adherence patterns in this proof-of-concept study. While this may limit the generalizability of our findings, many PrEP studies are conducted in international settings where most individuals have dark hair. Second, oral clearance at steady state was only measured once during the 7 doses/week period to reduce participant burden, which did not allow for assessment of intra-subject variability at the other dosing regimens. However, the predictive value of plasma AUC on tenofovir hair levels did not substantially improve in analyses predicting hair levels only after the 7 doses/week period. Third, weekend or holiday doses were not directly observed due to feasibility issues; however, these doses were confirmed by text message or phone. This mDOT approach has been demonstrated to be effective in prior studies
In summary, this study reports a consistent relationship between tenofovir dose and hair in HIV-negative men and women. These measures have already demonstrated utility for other ARVs in the HIV treatment setting and the finding of a strong dose relationship for tenofovir in hair paves the way for evaluation of its use in monitoring adherence in tenofovir-based PrEP demonstration projects and real-world PrEP settings. Unlike phlebotomy, hair collection is inexpensive, noninvasive and does not require specific skills, sterile equipment, or specialized storage conditions. While this hair PK study collected 150–200 strands of hair, ∼100 strands are typically needed for TFV level analysis. Hair can be stored for prolonged durations prior to analysis at room temperature and shipped without biohazardous precautions or cold chain requirements. These features may make this monitoring tool particularly promising in resource-poor settings, where high acceptability of hair collection has been demonstrated
Quantitative, time-integrated measures of long-term drug exposure in hair may provide more salient adherence information than qualitative (e.g. detectable/undetectable) results or drug levels in single plasma samples. Furthermore, our results may be useful in interpreting data from intermittent PrEP trials. Such dose reduction strategies are expected to lower costs and toxicity, but would be difficult to evaluate without validated biomarkers of adherence. Given that adherence is the “Achilles' heel” of PrEP
(DOCX)
(DOCX)
We would like to acknowledge Joyce Ycasas; Allison Futeral; Bernadette Perez; Hailey Gilmore; Deawodi Ladzekpo, Kerry Murphy, ANP; Elizabeth Faber, FNP; Sally Holland, PA; Risha Irvin, MD; Tim Matheson, PhD; Eric Vittinghoff, PhD; Teri Liegler, PhD; Joseph Guglielmo, PharmD, Lane Bushman; Brandon Klein, Caitlin Rower, JiaHua Zheng, Alfonso Diaz; Jeff White; Brian Martin; Delia Molloy; Alexa Burrell, Martin Soto; Samar Sharifi; Patricia von Felton; Andrew Forsyth; Jim Rooney, MD; Rich Clark; and the study volunteers who participated in this study.
A portion of the results of this paper was previously presented as a poster at the 18th Conference on Retroviruses and Opportunistic Infections, Boston, MA, 2011. Liu A, Gandhi M, Bacchetti P, et al. Validating Hair as a Biological Marker of Tenofovir Drug Exposure in HIV Pre-exposure Prophylaxis (PrEP). Data from this study will be made freely available upon request.