Drug Concentrations in Hair and Dried Blood Spots as Preexposure Prophylaxis Adherence Metrics During Pregnancy and Postpartum

Abstract

We evaluated hair tenofovir (TFV) concentrations as an adherence metric for HIV preexposure prophylaxis during pregnancy and postpartum and compared hair levels with TFV-diphosphate levels in dried blood spots (DBSs). Overall 152 hair samples from 102 women and 36 hair-DBS paired samples from 29 women were collected from a subset of women in a cluster-randomized trial. Having a partner with HIV was associated with higher hair TFV levels (P < .001). Hair TFV concentrations were strongly correlated with DBS TFV-diphosphate levels (r = 0.76, P < .001), indicating hair as a promising cumulative adherence metric for perinatal preexposure prophylaxis assessment.

The World Health Organization recommends offering tenofovir disoproxil fumarate–based daily oral preexposure prophylaxis (PrEP) to pregnant and postpartum women with substantial HIV acquisition risk [1]. Achieving optimal benefit from PrEP hinges on adherence [2]. To date, there are few evaluations of PrEP adherence among pregnant/postpartum women, with most studies utilizing dried blood spots (DBSs) for objective measurement. Two observational studies suggest that adherence is suboptimal in this population [3, 4]. Lower tenofovir diphosphate (TFV-DP) is observed in DBSs among women during pregnancy as compared with nonpregnant periods in directly observed treatment [5].

The investigation into utilizing hair drug levels as a matrix for monitoring PrEP adherence during pregnancy and postpartum remains limited. Hair drug levels predict HIV efficacy in perinatal women with HIV [6]. Hair drug–level testing is noninvasive, does not require specialized transportation and storage, and can be adapted to low-cost measurements [6]. Hair and DBS drug levels have not been compared during pregnancy and postpartum. Therefore, in our study, we aimed to explore the utility of hair as a matrix for evaluating perinatal PrEP exposure, utilizing data from the PrEP Implementation for Mothers in Antenatal Care (PrIMA) Study in Kenya. Specifically, we measured the acceptability of hair collection and TFV hair concentrations during pregnancy and postpartum among PrEP users and compared PrEP levels in hair and DBSs in a subset, with the overall goal of informing evaluation of PrEP exposure during and after pregnancy.

METHODS

Study Population

The PrIMA Study (NCT03070600) was a cluster-randomized trial conducted between 2018 and 2021 that evaluated PrEP delivery models among pregnant cisgender women in western Kenya who were followed through 9 months postpartum [7]. Pregnant women aged ≥15 years who were PrEP naive and HIV negative were enrolled. A validated empiric risk-scoring tool was used to assess HIV acquisition risk at enrollment, with high risk defined as scores >6 [8]. At a subset of study sites, hair specimens were collected at follow-up visits from women who reported using PrEP in the last 30 days and gave consent to specimen collection by trained study nurses. Reasons for not collecting hair samples were ascertained (eg, very short hair). Among PrEP initiators from whom hair was collected, hair specimens were tested for TFV drug concentrations from a random subset of women (25% of all PrEP initiators). Women with a hair specimen tested for TFV and a paired DBS that was collected during the same visit and tested for TFV-DP in prior analyses [4] were included in the comparison of PrEP levels in hair and DBS. All participants provided written informed consent.

Laboratory Methods

Small hair samples (approximately 100 strands) were collected, and TFV hair concentrations in the proximal 1 cm of hair (representing weeks to months of exposure) were analyzed by validated LC-MS/MS methods (liquid chromatography–tandem mass spectrometry) [9]. The assay is validated from 0.002 to 0.400 ng of TFV per milligram of hair. DBSs were tested for TFV-DP in red blood cells with validated ultraperformance LC-MS/MS methods [10]. Values below the lower limit of quantification for TFV-DP (25 fmol/sample) were considered not quantifiable. DBSs provide a marker of cumulative adherence over the prior 6 to 8 weeks.

Statistical Analyses

Summary statistics were generated to compare the characteristics of participants who accepted vs declined hair collection. Reasons for not collecting maternal hair specimens were summarized. Hair weight–normalized TFV concentrations were log₁₀ transformed, and mean concentrations were estimated. Univariate generalized estimating equation regression models were fitted to evaluate the associations of each covariate with TFV concentrations. Covariates with a P < .20 in univariate analyses were included in multivariable models.

For exploratory purposes, weight-normalized TFV hair concentrations were classified into categories according to dosing benchmarks (< 2 and 2–7 doses/wk) from directly observed pharmacokinetic (PK) studies among nonpregnant women, as hair benchmarks for pregnant women are unavailable [9]. The correlations between hair TFV levels and paired DBS TFV-DP levels were calculated per the Spearman coefficient. Calculation of median hair TFV levels was based on benchmarks for PrEP dosing in DBSs among pregnant/postpartum women in the IMPAACT 2009 study and hair benchmarks in nonpregnant women [5, 9]. Our cut point of 2 doses per week was chosen due to categorizations of existing benchmarks and does not indicate protection-effective adherence levels. Analyses were conducted with R version 4.2.3.

RESULTS

Study Population Characteristics and Hair Collection Acceptability

In total, 145 women in the PrIMA Study who reported PrEP use in the last 30 days were approached for hair collection. Among this subset, the median age was 27.0 years (IQR, 23.0–30.7), 91.7% were married, and 26.2% had a partner with HIV (Supplementary Table 1). Overall, 91.0% (132/145) of women accepted hair collection. Among women who declined hair collection (n = 13), the most frequent reasons were concern about disrupted appearance (57.9%) and personal reasons (35.1%). Women who declined hair collection less frequently had a prior pregnancy (P = .04); other characteristics were similar between women who accepted and declined.

TFV Hair Concentrations

Among 132 women who found hair collection acceptable, 102 had hair samples collected from 152 follow-up visits that were tested for TFV levels and included in the primary analysis. Of the 102 women, 34 had hair samples collected at multiple study visits, with 20 contributing 2 samples and 13 contributing 3 samples. The most frequent reasons for not having samples collected and/or tested were logistical concerns (Supplementary Table 1). One-third (32.9%) of samples were from pregnancy visits at a median gestation of 31.5 weeks (IQR, 24.0–36.0) and 67.1% from postpartum visits at a median 3.5 months since birth (IQR, 1.5–9.0). The median time undergoing PrEP was 4.4 months (IQR, 1.7–7.4) between the study visit when PrEP was first prescribed and the sample collection date. Mean hair TFV concentrations did not differ by pregnancy vs postpartum status (P = .75; Table 1). In multivariable models, having a partner with HIV was associated with 425% higher adjusted mean TFV concentrations as compared with having a partner thought to be HIV negative (95% CI, 171.2%–917.7%; P < .001). Having an HIV risk score >6 was associated with 106.2% higher hair TFV levels as opposed to having lower risk scores (95% CI, 112.74%–277.1%; P = .019). Having a positive syphilis result during pregnancy (7/152, 4.6%) or recent sexually transmitted infection diagnosis (14/152, 9.2%) was associated with lower hair TFV levels (P < .05), although precision was limited due to the low frequency of sexually transmitted infections. No other characteristics were associated with TFV levels.

Table 1.

Correlates of Mean Log₁₀ Maternal TFV Hair Concentrations During Pregnancy and Postpartum Among a Subset of Women Reporting PrEP Use in the Past 30 Days Enrolled in the PrIMA Study

			Univariate Regression			Multivariable Regressiona
	No. (%)b	Mean Log10 TFV	Absolute Difference (95% CI)	Difference, % (95% CI)	P Value	Absolute Difference (95% CI)	Difference, % (95% CI)	P Value
Maternal age, y
< 22	30 (19.7)	−2.25	1 [Reference]
≥22	122 (80.3)	−2.17	0.08 (−.29, .46)	21.34 (−49.05, 189.01)	.662
Marital status
Currently married	140 (92.1)	−2.20	1 [Reference]
Unmarried	12 (7.9)	−1.94	0.26 (−.25, .78)	83.45 (−43.56, 496.31)	.313
Completed education ≥12 y
No	97 (63.8)	−2.18	1 [Reference]
Yes	53 (34.9)	−2.16	0.02 (−.28, .31)	3.6 (−47.06, 102.74)	.918
Regular employment
No	120 (78.9)	−2.19	1 [Reference]
Yes	30 (19.7)	−2.10	0.09 (−.26, .44)	23.51 (−44.83, 176.51)	.608
Any living children
No	17 (11.2)	−2.35	1 [Reference]
Yes	135 (88.8)	−2.16	0.18 (−.17, .54)	52.93 (−32.87, 248.36)	.312
Pregnancy status
Postpartum	102 (67.1)	−2.2	1 [Reference]
Pregnant	50 (32.9)	−2.16	0.04 (−.19, .26)	8.71 (−35.02, 81.87)	.75
Partner HIV status
Negative	38 (25)	−2.49	1 [Reference]			1 [Reference]
Positive	51 (33.6)	−1.67	0.82 (0.56, 1.08)	561.38 (259.73, 1115.98)	<.001	0.72 (.43, 1.01)	425.33 (171.18, 917.66)	<.001
Unknown	63 (41.4)	−2.42	0.06 (−.21, .34)	15.86 (−38.24, 117.35)	.646	0.05 (−.21, .32)	13.16 (−38.87, 109.46)	.694
Lifetime No. of sexual partners >2
No	46 (30.3)	−2.34	1 [Reference]
Yes	106 (69.7)	−2.11	0.23 (−.05, .51)	68.84 (−11.6, 222.48)	.113
Syphilis status in current pregnancy
Negative	142 (93.4)	−2.14	1 [Reference]			1 [Reference]
Positive	7 (4.6)	−2.79	−0.65 (−1.10, −.20)	−77.58 (−92.02, −37.05)	.005	−0.48 (−.72, −.25)	−67.12 (−80.77, −43.78)	<.001
Recently diagnosed with any sexually transmitted infections
No	136 (89.5)	−2.12	1 [Reference]			1 [Reference]
Yes	14 (9.2)	−2.72	−0.61 (−.99, −.22)	−75.18 (−89.74, −39.95)	.002	−0.38 (−.73, −.03)	−58.44 (−81.49, −6.65)	.033
High risk for HIV acquisitionc
No	34 (22.4)	−2.47	1 [Reference]			1 [Reference]
Yes	118 (77.6)	−2.1	0.36 (0.10, .63)	131.48 (26.71, 322.86)	.006	0.31 (.05, .58)	106.19 (12.74, 277.1)	.019
High self-perceived HIV riskd
No	90 (59.2)	−2.34	1 [Reference]			1 [Reference]
Yes	59 (38.8)	−1.93	0.4 (0.11, .70)	153.5 (29.11, 397.74)	.007	0.21 (−.04, .46)	60.94 (−9.53, 186.3)	.105
Depressione
No	113 (74.3)	−2.21	1 [Reference]
Yes	32 (21.1)	−2.06	0.14 (−.23, .52)	38.52 (−41.48, 227.88)	.459
Intimate partner violencef
No	133 (87.5)	−2.14	1 [Reference]
Yes	17 (11.2)	−2.42	−0.28 (−.61, .06)	−47 (−75.63, 15.25)	.109
Low social supportg
No	104 (68.4)	−2.17	1 [Reference]
Yes	46 (30.3)	−2.17	0.01 (−.34, .35)	1.22 (−54.04, 122.93)	.976

Abbreviations: HITS, Hurt, Insult, Threaten, and Scream; MOS-SSS, Medical Outcomes Study–Social Support Survey; PHQ-2, Patient Health Questionnaire–2; PrEP, preexposure prophylaxis; PrIMA, PrEP Implementation for Mothers in Antenatal Care; TFV, tenofovir.

^aVariables with P < .2 in univariate regression were selected into multivariable regressions. Syphilis and risk perception were adjusted in the multivariable regression of partner status.

^bSummary statistics are presented at the visit level, and missingness is not presented.

^cHIV high risk at enrollment defined as empiric risk score ≥6 [10].

^dSelf-perceived HIV risk was assessed by asking, “What is your gut feeling about how likely you are to get infected with HIV?” “Extremely likely” and “very likely” were considered high self-perceived HIV risk, while “somewhat likely,” “very unlikely,” and “extremely unlikely” were considered low self-perceived HIV risk.

^eDefined as a PHQ-2 score ≥3.

^fDefined as a HITS score ≥10.

^gDefined as a MOS-SSS score < 72.

Sensitivity and Exploratory Analyses

Overall, 41% of hair samples had TFV levels ≥0.012 ng/mg, indicating ≥2 PrEP doses per week according to hair adherence benchmarks; there was no difference in the frequency of this benchmark in pregnancy vs postpartum status (44% vs 40%, P = .65). In total, 36 hair-DBS paired samples from 29 women were evaluated for correlation and analyzed by hair and DBS benchmarks: 12 (33%) from pregnancy visits at a median 32.0 weeks of gestation (IQR, 27.0–34.0) and 24 (67%) from postpartum visits at a median 3.5 months since birth (IQR, 1.5–6.0). Median time since PrEP initiation was 4.4 months (IQR, 1.6–8.3; n = 36) at sample collection for the subset. Overall, TFV levels in hair were strongly correlated with TFV-DP levels in DBSs (r = 0.76, P < .001), with a strong correlation in postpartum status (r = 0.82, P < .001) and moderate in pregnancy (r = 0.57, P = .05; Supplementary Figure 1). Based on DBS benchmarks for pregnant/postpartum women, 42% of DBS samples had TFV-DP levels indicative of ≥2 doses per week; based on benchmarks from nonpregnant women, 39% of hair samples had TFV levels indicative of ≥2 doses per week. Median hair TFV levels for women who took <2 and ≥2 doses per week were 0.003 ng/mg (IQR, 0–0.006; n = 21) and 0.035 ng/mg (IQR, 0.021–0.039; n = 15), respectively, based on DBS benchmarks; results were similar by pregnancy status (0.029 ng/mg [pregnancy] vs 0.035 ng/mg [postpartum], P = .8, Table 2).

Table 2.

Tenofovir Hair Levels Among Women Enrolled in the PrIMA Study During Pregnancy and Postpartum, by PrEP Dosing Benchmarks Established in DBSs and Hair

	Based on DBS Benchmarka		Based on Hair TFV Benchmarksb
Pregnancy Status: PrEP Doses/wk	No. of Hair Samples	Median (IQR)	No. of Hair Samples	Median (IQR)
Overall
<2	21	0.000 (0–0.006)	22	0.000 (0–0.006)
2–7	15	0.035 (0.021–0.039)	14	0.035 (0.03–0.04)
Pregnant
<2	7	0.000 (0–0.005)	9	0.000 (0–0.006)
2–7	5	0.029 (0.01–0.038)	3	0.038 (0.033–0.039)
Postpartum
<2	14	0.003 (0–0.007)	13	0.000 (0–0.006)
2–7	10	0.035 (0.029–0.041)	11	0.035 (0.03–0.039)

To calculate the hair TFV levels using DBS benchmarks, we first categorized DBS samples into those with TFV-DP levels indicating PrEP use <2 vs 2–7 doses/wk. We then calculated hair TFV levels among hair samples with paired DBS samples indicating <2 vs 2–7 doses/wk.

Abbreviations: DBS, dried blood spot; DP, diphosphate; PrEP, preexposure prophylaxis; PrIMA, PrEP Implementation for Mothers in Antenatal Care; TFV, tenofovir.

^aTFV hair concentrations were classified into categories according to PrEP dosing benchmarks established in DBSs from directly observed pharmacokinetic studies among pregnant and postpartum women in the IMPAACT 2009 study (ie, a TFV-DP concentration in DBS ≥200 fmol/punch and ≥300 fmol/punch was associated with ≥2 doses/wk among pregnant and postpartum women, respectively) [5].

^bTFV hair concentrations were classified into categories according to dosing benchmarks established in hair from directly observed pharmacokinetic studies among nonpregnant women (ie, a hair TFV concentration ≥0.012 ng/mg was considered as taking at least 2 doses/wk) [6, 9], as benchmarks for pregnant women are unavailable.

DISCUSSION

In this cohort of peripartum women using PrEP in Kenya, we found that hair collection was highly acceptable. Having a male partner with HIV was strongly associated with higher maternal TFV hair concentrations. We did not detect a difference in TFV hair concentrations in pregnancy vs postpartum status, although only 41% (n = 15) of participants had drug levels indicating ≥2 PrEP doses per week. Lower hair concentrations in the peripartum period could be a result of inadequate adherence [3, 4] or PK variability in pregnancy [5]. Our results add to the limited data on using objective metrics of PrEP exposure among maternal populations in the settings with high HIV burden, and they have implications for monitoring perinatal PrEP exposure.

Hair is a promising PrEP adherence measure. Acceptability of hair collection in our study was similarly high to previous antiretroviral therapy exposure studies among pregnant women in East Africa (>90%) [6], although lower rates of acceptability are seen in studies among children and men who have sex with men [11]. We found that concern about disrupted appearance was a frequent reason for declining hair collection. Qualitative studies could elucidate how hair collection acceptability can be improved, perhaps with less disruptive collection procedures and reassurance of the small amount of hair collected. Although current hair analysis is expensive, low-cost techniques and rapid hair assays that can be performed in resource-limited settings are under development [12]. The acceptability, feasibility, and cost may influence the choice of PrEP adherence metrics in research and practice.

About 60% (n = 21) of PrEP users in our study had low peripartum hair TFV levels indicating <2 PrEP doses per week, suggestive of suboptimal exposure. Studies among pregnant women show reduced TFV-DP levels in DBSs [5]. A directly observed study showed that DBS TFV-DP levels among African women were 31% to 37% lower during pregnancy than postpartum, suggesting that lower exposure is a result of PK changes during pregnancy [5]. More PK studies, including the pregnancy period, are needed to establish dosing thresholds of TFV exposure in hair. In our analysis, we found a strong association between having a partner with HIV and higher hair TFV levels, similar to prior PrEP adherence evaluations [3]. Women with partners with HIV are motivated to use PrEP during pregnancy to protect their infants from HIV [13]. However, women with partners of unknown HIV status may be at higher risk of HIV acquisition [14], and we did not observe an association between unknown partner HIV status and TFV levels in hair. Implementing strategies to enhance awareness of partners’ HIV status is crucial in guiding PrEP decisions among peripartum women.

Our study has limitations. We found generally low levels of hair and/or DBS levels, which may limit our ability to observe patterns between them. We are also unable to assess the contribution of PK variability vs adherence to low TFV concentrations in this study. PrEP use was assessed at monthly visits, but we did not ascertain precise dates of first PrEP ingestion, which may influence the timing of steady-state concentrations. We tested hair samples collected only from women with reported PrEP use in the last 30 days, which are not representative of all PrEP users. To inform potential selection bias, we compared baseline characteristics of PrEP users who initiated PrEP, who had hair samples, and who had hair and DBS samples (Supplementary Table 2). We found a higher proportion of women with hair samples reporting living with a partner with HIV, perceiving a high HIV risk, and having social support when compared with women who ever initiated PrEP. These factors may influence PrEP adherence behaviors overall. Yet, we did not expect the PK data to differ according to these characteristics among women using PrEP, thus minimizing the impact of selection bias on our results. We assumed a linear hair growth rate of 1 cm per month; however, hair may grow at a different rate during pregnancy and postpartum [15], potentially affecting the accurate interpretation of drug exposure levels.

In conclusion, our study shows that hair TFV levels were generally low but comparable among women reporting recent PrEP use during pregnancy and postpartum. Our findings also suggest that hair PrEP measures are well correlated with DBSs, indicating that hair can serve as a promising cumulative adherence metric for assessing PrEP levels over the perinatal period. Future studies should examine the adherence-concentration-efficacy relationship for oral PrEP and other PrEP modalities in cisgender women including the pregnancy period.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. We gratefully thank the study participants and PrIMA staff for their time and contributions. We thank the Kenyan Ministry of Health nationally, the Kisumu County Department of Health, and the facility heads and in-charges for their collaboration with this study.

Disclaimer. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Financial support. This work was supported by the National Institute of Allergy and Infectious Disease (R01AI125498 and MG-3R01AI098472 to G. J.-S.) and Eunice Kennedy Shriver National Institute of Child Health and Human Development (R01HD094630 to G. J.-S.; R01HD100201 to J. P.; R01HD108041 to J. P.) and the National Institute of Nursing Research (R01NR019220 to J. P.) of the National Institutes of Health; and the University of Washington's Center for AIDS Research Behavioral Sciences Core and Biometrics Core (P30AI027757) and the Global Center for the Integrated Health of Women, Adolescents, and Children.