Point-of-care semi-quantitative test for adherence to tenofovir alafenamide or tenofovir disoproxil fumarate

Abstract

Objectives

Objective measurement of antiretrovirals may aid clinical interventions for improving adherence to HIV prevention or treatment regimens. A point-of-care urine test could provide real-time information about recent adherence to regimens containing tenofovir disoproxil fumarate or tenofovir alafenamide. We developed a lateral flow immunoassay (LFA) and ELISA for urinary tenofovir.

Methods

The intensity of the LFA test line was quantified using an optical reader and visually scored 0–5 by two independent people, using a reference card. The sensitivity and specificity of both the ELISA and LFA were determined for two different tenofovir concentration cut-offs for tenofovir disoproxil fumarate and tenofovir alafenamide adherence—1500 and 150 ng/mL, respectively. To validate the assays, we measured 586 urine samples from 28 individuals collected as part of a study of tenofovir pharmacokinetics in adults, which were also measured by MS for reference.

Results

Both the LFA signal and ELISA signal were each strongly correlated with drug concentrations (0.91 and 0.92, respectively). The LFA signal and ELISA were highly sensitive and specific at both thresholds (LFA sensitivity/specificity: tenofovir disoproxil fumarate, 89%/96%; and tenofovir alafenamide, 90%/96%) (ELISA sensitivity/specificity: tenofovir disoproxil fumarate, 94%/94%; and tenofovir alafenamide, 92%/84%). Visual scoring of the LFA was also highly sensitive and specific at both the tenofovir disoproxil fumarate threshold and the tenofovir alafenamide threshold (sensitivity/specificity: tenofovir disoproxil fumarate, 91%/94%; and tenofovir alafenamide, 87%/90%).

Conclusions

Our rapid semi-quantitative test can measure tenofovir concentrations relevant to both tenofovir alafenamide and tenofovir disoproxil fumarate adherence, which may support adherence-promoting interventions across a range of HIV care settings.

Introduction

Antiretroviral therapy (ART) is highly effective for people living with HIV to suppress viral replication and prevent onward transmission, but monitoring and improving drug adherence have been challenging. The efficacy of ART in suppressing HIV RNA viral load is highly dependent on maintaining therapeutic concentrations of the drug.¹ Also, measurements of drug concentrations may also help to calculate the odds of future viraemia.² As compared with subjective measures of adherence, such as self-reported adherence, clinician assessments, pill counts, pharmacy refill records and clinic visit attendance, direct measurement of drug concentrations is the most accurate and objective method for assessing uptake of oral ART and pre-exposure prophylaxis (PrEP).^3,4 Therefore, accurate real-time information may empower effective targeted interventions to improve adherence.⁵

The drug tenofovir is frequently included in first-line oral ART and PrEP regimens, formulated as one of two prodrugs, tenofovir disoproxil fumarate or tenofovir alafenamide.⁶ Tenofovir in the urine is highly correlated with days since the last dose of tenofovir disoproxil fumarate, with a washout half-life of about 17 h.⁷ Rapid lateral flow tests for tenofovir in urine have recently been reported.^8–10 These tests accurately provide a binary ‘yes/no’ visual readout of tenofovir concentrations above or below the tenofovir disoproxil fumarate threshold concentration of 1500 ng/mL. However, a test that can discriminate lower tenofovir concentrations in the urine would be able to differentiate between moderate and very low adherence to tenofovir disoproxil fumarate and furthermore may be valuable for determining adherence to tenofovir alafenamide.¹¹

We sought to develop a lateral flow assay for tenofovir in urine and evaluated two different readout modalities to enable a semi-quantitative test relevant to both tenofovir disoproxil fumarate and tenofovir alafenamide adherence. It is reported elsewhere that about 98% of individuals with a tenofovir concentration in the urine of less than 1500 ng/mL have not taken a dose in over 24 h and about half of these individuals have not taken a dose in over 3 days.¹² In the case of tenofovir alafenamide, the half-life of tenofovir in the urine is about 30 h and likewise highly correlated with adherence since the biophysical mechanisms of tenofovir clearance are similar to those of tenofovir disoproxil fumarate.¹³ Urine concentrations of tenofovir are roughly 10-fold lower with recommended dosages of tenofovir alafenamide (10–25 mg, once daily) compared with standard doses of tenofovir disoproxil fumarate (300 mg, once daily).^14,15 On the other hand, directly observed therapy studies of tenofovir alafenamide indicate urine concentrations about 74% lower than those for tenofovir disoproxil fumarate.¹¹ The validity of any particular cut-off for tenofovir alafenamide adherence has not yet been demonstrated, so we elected to select a more conservative (i.e. lower) threshold concentration of 150 ng/mL for the purposes of quantifying assay performance.

Patients and methods

Study population

The TARGET study was a randomized, open-label, clinical pharmacokinetic study of tenofovir in healthy adult volunteers without HIV or hepatitis B infection in Thailand.¹⁶ Urine samples were collected from 28 participants taking oral tenofovir disoproxil fumarate (300 mg) and emtricitabine (200 mg) randomly assigned to one of three regimens: once daily, 4 times weekly or 2 times weekly. Urine samples were collected during the 6 week treatment phase, as well as a subsequent 4 week washout phase.

Laboratory procedures

Tenofovir quantification

Tenofovir concentration in each urine sample was measured previously by LC-MS/MS.¹⁷ Tenofovir urine concentrations were below the LC-MS/MS lower limit of quantification (LLOQ) of 50 ng/mL in 79 samples and above the upper limit of quantification (ULOQ) of 23 200 ng/mL in 21 samples. De-identified sample aliquots were stored frozen at −80°C.

Immunoassay development

We previously identified antibodies that bind tenofovir with high specific affinity.¹⁸ We developed and optimized a competitive ELISA, as well as competitive lateral flow immunoassay (LFA), for tenofovir in urine. The competitive LFA uses a single antibody against tenofovir alongside a tenofovir conjugate reporter, similar to an immunoassay described previously.¹⁸ The LC-MS/MS tenofovir concentrations of 74 representative urine samples were unblinded during development to guide optimization of the ELISA and LFA reagents and parameters, and were not included in validation studies.

Assay validation

Urine samples were thawed overnight prior to testing. LFA strips were read after 10 min using a commercial strip reader (Axxin Ax-2x-S). Test line brightness was also assessed by two people. Visual assessments were performed independently using the same visual intensity reference card on a graduated scale from 0 (invisible) to 5 (very dark).

Statistical analysis

For each ELISA microwell plate, the limit of detection (LOD) and LLOQ were determined as 3× and 6× the standard deviation of the blank sample, respectively. The ELISA standard curve for each plate was fitted with a three-parameter logistic regression and used to interpolate tenofovir concentrations in those samples.

The Spearman correlation coefficient was determined between the tenofovir concentration as measured by LC-MS/MS and the concentration measured by ELISA, as well as by LFA test line optical intensity. To determine the semi-quantitative performance of the LFA, the average visual score and optical intensity were both evaluated at 1500 and 150 ng/mL thresholds. For both readout methods at each threshold concentration, the signal cut-off to maximize accuracy (i.e. percentage of all samples that were correctly classified at that threshold) was determined. We calculated the sensitivity (percentage of samples with tenofovir below threshold classified correctly) and specificity (percentage of samples above threshold classified correctly) of each method at the two thresholds. We also calculated the sensitivity and specificity for subsets of data excluding samples with a tenofovir concentration within the LC-MS/MS coefficient of variation of the threshold (i.e. 15%), as measured by LC-MS/MS. These ‘corrected’ metrics should be interpreted as an estimate of the performance upper bound that is unbiased by reference measurement error or sample distribution near the threshold.

Results

For the ELISA, the LOD, LLOQ, IC₅₀ and ULOQ were found to be 135, 1055, 1925 and 23 200 ng/mL, respectively. The Spearman correlation coefficient between ELISA and LC-MS/MS was 0.92 (Figure 1a). For the rapid test, the Spearman correlation coefficient between the LFA test line optical intensity and LC-MS/MS was 0.91 (Figure 1b; the negative correlation is an expected characteristic of a competitive immunoassay). For the optical readout, the optimal threshold intensity for differentiating tenofovir levels above or below 1500 ng/mL was 1320 AU and, similarly, for tenofovir above or below 150 ng/mL it was 3270 AU.

Figure 1.

(a) Correlation between ELISA and LC-MS/MS, for samples with a tenofovir concentration within the ELISA concentration range of 1055 to 23 000 ng/mL (log–log axes). The ideal ‘y = x’ line is shown for reference. (b) Correlation between the LFA test line (optical reader) and LC-MS/MS (log–log axes). (c) Correlation of the intensity of the LFA test line (measured by the optical reader) with the visual score (log10 y-axis). This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

The visual score of the LFA measured independently by two people was concordant within one grade for 96% of samples. As expected, the average visual score was correlated with the measured optical intensity of the test line (Figure 1c).

The corrected and uncorrected sensitivities and specificities of all three modalities for both the 1500 and 150 ng/mL thresholds are given in Table 1. Altogether, the corrected values were 0 to 4 percentage points higher than the uncorrected values and do not substantially affect the interpretation of the data. The LFA was no less accurate for differentiating above or below the 150 ng/mL threshold than at the 1500 ng/mL threshold. At the 1500 ng/mL threshold, the corrected performance of the scorecard method was similar to the reader method. Notably, the visual readout of the LFA was not statistically worse at either threshold than ELISA.

Table 1.

Uncorrected and corrected sensitivity and specificity of the LFA (by average visual score and by optical reader) and the ELISA at classifying samples above or below two tenofovir thresholds

	LFA, visual (95% CI)	LFA, reader (95% CI)	ELISA (95% CI)
1500 ng/mL tenofovir threshold
uncorrected
sensitivity	87% (83%–91%)	85% (81%–89%)	92% (89%–95%)
specificity	92% (88%–95%)	96% (92%–98%)	90% (86%–93%)
correcteda
sensitivity	91% (87%–94%)	89% (85%–93%)	94% (90%–96%)
specificity	94% (90%–97%)	96% (93%–98%)	94% (90%–96%)
150 ng/mL tenofovir threshold
uncorrected
sensitivity	84% (77%–90%)	87% (81%–92%)	90% (89%–94%)
specificity	89% (85%–91%)	95% (92%–97%)	83% (79%–86%)
correctedb
sensitivity	87% (80%–93%)	90% (84%–95%)	92% (86%–96%)
specificity	90% (87%–93%)	96% (93%–97%)	84% (80%–87%)

Corrected values represent estimates of performance upper bounds in the absence of reference measurement error.

^aThirty-two samples within 1275–1725 ng/mL excluded.

^bNineteen samples within 127.5–172.5 ng/mL excluded.

Discussion

We found, by using an optical reader, that the LFA was just as accurate at the 150 ng/mL ‘tenofovir alafenamide’ threshold as at the 1500 ng/mL ‘tenofovir disoproxil fumarate’ threshold. This result, along with the high correlation between intensity and tenofovir concentration, also supports the potential feasibility of a future fully quantitative LFA. As mentioned earlier, the 150 ng/mL tenofovir alafenamide threshold has not yet been formally validated for tenofovir alafenamide adherence and was selected as a rough and conservative estimate based on measurements of tenofovir alafenamide in urine across various settings.

Visual grading using the reference card was also found to be an accurate way to estimate tenofovir concentration and may be useful in settings where an optical reader is not available. In the case of tenofovir disoproxil fumarate, the use of a visual scorecard may be a scalable method to provide additional information about lower rates of adherence. A measurement of 1500 ng/mL can reasonably occur in individuals who have had a most-recent dose of tenofovir disoproxil fumarate any time between 2 and 4 days ago.¹² Likewise, a measurement of 150 ng/mL can reasonably occur for tenofovir disoproxil fumarate dosages 4.5 to 6.5 days ago, based on the half-life of 17 h (i.e. 10-fold reduction after about 2.5 days). Thus, the inclusion of semi-quantitative information about tenofovir levels across this range may provide additional information about adherence out to a maximum of about 7 days.

In conclusion, the LFA is an accurate semi-quantitative test of tenofovir across concentrations relevant to both tenofovir disoproxil fumarate and tenofovir alafenamide adherence. When combined with appropriate adherence counselling and interventions, this may be a valuable tool to help reduce HIV transmission and end the global HIV/AIDS pandemic.

Funding

This work was supported by National Institutes of Health NIAID R01AI13664.

Transparency declarations

None to declare.