M A J O R A R T I C L EH I V / A I D S
Atazanavir Concentration in Hair Is the Strongest
Predictor of Outcomes on Antiretroviral Therapy
Monica Gandhi,1Niloufar Ameli,2Peter Bacchetti,3Kathryn Anastos,5Stephen J. Gange,7Howard Minkoff,6
Mary Young,8Joel Milam,4Mardge H. Cohen,9Gerald B. Sharp,10Yong Huang,2and Ruth M. Greenblatt1,2,3
1Departments of Medicine, and2Clinical Pharmacy, and3Epidemiology/Biostatistics, University of California, San Francisco, San Francisco, California,
and4Department of Medicine, University of Southern California, Los Angeles, California, and5Department of Medicine, Albert Einstein College of
Medicine, Bronx, New York, and6Department of Medicine, SUNY Downstate Medical Center, Brooklyn, New York, and7Bloomberg School of Public
Health, Johns Hopkins University, Baltimore, Maryland, and8Department of Medicine, Georgetown University Medical Center, Washington, District of
Columbia, and9Departments of Medicine, John Stroger (formerly Cook County) Hospital and Rush University, Chicago, Illinois; and10National Institute
of Allergy and Infectious Diseases/National Institutes of Health, Bethesda, Maryland
assess exposure (eg, querying adherence and single plasma drug level measurements) are limited. Hair
concentrations of antiretrovirals can integrate adherence and pharmacokinetics into a single assay.
Methods. Smallhairsampleswere collected from participants in the Women’s Interagency HIV Study (WIHS), a
large cohort of human immunodeficiency virus (HIV)-infected (and at-risk noninfected) women. From 2003
through 2008, we analyzed atazanavir hair concentrations longitudinally for women reporting receipt of atazanavir-
basedtherapy. Multivariate random effects logisticregression modelsforrepeatedmeasures were used toestimate the
association of hair drug levels with the primary outcome of virologic suppression (HIV RNA level, ,80 copies/mL).
Results.424 WIHS participants (51% African-American, 31% Hispanic) contributed 1443 person-visits to the
analysis. After adjusting for age, race, treatment experience, pretreatment viral load, CD4 count and AIDS status,
and self-reported adherence, hair levels were the strongest predictor of suppression. Categorized hair antiretroviral
levels revealed a monotonic relationship to suppression; women with atazanavir levels in the highest quintile had
odds ratios (ORs) of 59.8 (95% confidence ratio, 29.0–123.2) for virologic suppression. Hair atazanavir
concentrations wereevenmore strongly associated withresuppression of viral loads in subgroupsin which there had
been previous lapses in adherence (OR, 210.2 [95% CI, 46.0–961.1]), low hair levels (OR, 132.8 [95% CI, 26.5–
666.0]), or detectable viremia (OR, 400.7 [95% CI, 52.3–3069.7]).
Conclusions. Antiretroviral hair levels surpassed any other predictor of virologic outcomes to HIV treatment
in a large cohort. Low antiretroviral exposure in hair may trigger interventions prior to failure or herald virologic
failure in settings where measurement of viral loads is unavailable. Monitoring hair antiretroviral concentrations
may be useful for prolonging regimen durability.
Adequate exposure to antiretrovirals is important to maintain durable responses, but methods to
The use of combination antiretroviral therapy (cART) is
the prime determinant of longevity among human im-
munodeficiency virus (HIV)-infected individuals, and
the focus of treatment has now shifted to maintaining
durable responses on existing regimens. Toward this end,
are needed to identify factors that may threaten viro-
logic response and herald resistance. By the time a regi-
men has failed virologically, important opportunities for
adherence interventions and preventing resistance have
mutations can accumulate by the time a regimen fails
clinically or immunologically [1, 2]. A low-cost method
that provides an early predictor of virologic failure would
be useful in both industrialized and nonindustrialized
settings to monitor responses to long-term cART .
Adherence to treatment, commonly assessed by self-
report, is a strong contributor to cART outcomes.
predictive models oftreatment response
Received 24 November 2010; accepted 3 February 2011.
Correspondence: Monica Gandhi, MD, MPH, Divs of HIV/AIDS and Infectious
Diseases, University of California, San Francisco, 405 Irving St, 2nd fl, San
Francisco, CA 94122 (email@example.com).
Clinical Infectious Diseases
? The Author 2011. Published by Oxford University Press on behalf of the Infectious
Diseases Society of America. All rights reserved. For Permissions, please e-mail:
d CID 2011:52 (15 May)
However, limitations of the accuracy of self-reporting and other
commonly used adherence measures are well described .
Furthermore, interindividual variations in pharmacokinetics
can lead to outcome disparities even when adherence is high,
indicating the utilityofanobjective measure thatcouldintegrate
both combined effects of adherence and individual pharmaco-
kinetic parameters. We have developed methods for monitoring
antiretroviral exposure by means of determination of anti-
retroviral concentrations in small-volume hair samples [5–9].
We previously reported that hair levels of protease inhibitors
(PIs) were the strongest independent factor associated with
short-term virologic response in individuals initiating new PI-
based regimens . Drug levels in hair were more closely as-
sociated with HIV RNA suppression six months after starting
therapy than were self-reported adherence and other commonly
applied factors, including pretreatment viral load and CD4 cell
count, extent of antiretroviral or PI experience, age, or race. A
better test of the ultimate value of drug levels in hair is whether
these measures predict impending virologic failure. We report
here on a longitudinal study that assessed how well atazanavir
levels measured in hair predict treatment outcomes in a multi-
site observational cohort of HIV-infected women.
The Women's Interagency HIV Study Cohort and Study Sample
The Women’s Interagency HIV Study (WIHS) is the largest
cohort of HIV-infected women and at-risk HIV-noninfected
women studied in the United States . This ongoing pro-
spective multicenter study with sites in the San Francisco and
Los Angeles, California; Chicago, Illinois; Bronx and Brooklyn,
New York; and Washington, DC metropolitan areas observes
participants via visits that occur at 6-month intervals.
Interviewer-administered survey instruments, physical exami-
nation, and specimen collection are performed at each visit.
A small sample of hair (?10–20 strands, or 1–3 mg) is cut from
the occipital region of the scalp at each study visit from every
consenting HIV-seropositive woman reporting cART. All par-
ticipants who reported taking an atazanavir-based reporting at
any study visit during the period from April 2003 through
April 2008 were included in this particular analysis. WIHS study
protocols and consent materials were reviewed and approved by
institutional review boards at all participating institutions.
Hair Collection, Processing, and Analysis
A hair specimen is collected in the WIHS if the participant
reports taking antiretrovirals for at >1 month. Field staff at all
WIHS sites have been uniformly trained on the method of
collecting hair. Briefly, a small thatch of hair is cut as close as
possible to the scalp from the occiput and the distal portion
labeled to denote directionality (Figure 1). Methods for ex-
traction and analyses of most of the PIs, nonnucleoside reverse
transcriptase inhibitors, and tenofovir have been developed and
optimized in our laboratory and reported elsewhere [6–11].
Atazanavir and ritonavir levels were measured in hair samples
collected at visits during which the participant reported current
use. Using 2 mg of human hair, atazanavir is detected at levels as
low as 0.05 ng/mg hair and ritonavir is detected at levels of as
low as 0.01 ng/mg hair. The method has been validated in the
range of 0.05–20 ng/mg hair for atazanavir and 0.01–4.0 ng/mg
hair for ritonavir with good linearity and reproducibility. Of
note, we have tested antiretrovirals in hair in this diverse WIHS
cohort and found that median levels and interquartile ranges
among participants with undetectable viral loads do not vary
significantly by race or ethnicity.
The primary outcome was virologic success (or ‘‘suppression’’) at
each study visit, defined as an HIV viral load ef ,80 copies/mL.
Multivariate random effects logistic regression models for re-
levels with the dichotomous outcome of virologic suppression.
We used the hair level at each visit to predict virologic success,
the head (left panel) and the cutting of the hair thatch right at the scalp (right panel).
Hair collection procedure for antiretroviral levels: Picture demonstrates the isolation of the 20-strand hair thatch from the occipital region of
d CID 2011:52 (15 May)
and levels were analyzed both as continuous measures and as
in models were variables that could affect response, including age,
race, viral load at the time of regimen initiation (continuous or
dichotomized into ,100,000 vs >100,000 copies/mL), prior
antiretroviral treatment experience (dichotomized into yes vs no),
and degree of PI experience (categorized into naı ¨ve to PIs, past
experience with 1 PI, or treatment with >2 PIs), nadir and pre-
treatment CD4 cell counts, study year, nucleoside reverse tran-
scriptase inhibitor (NRTI) backbone components, history of
clinical AIDS, and self-reported adherence. Adherence to ataza-
navir was reported as the percentage of prescribed doses con-
sumed during periods of 6 months, 30 days, and 3 days; visual
analog scales aided in estimating percentages . Level of ad-
herence was analyzed either as a continuous measure or catego-
rized into <74%, 75%–94%, or >95% over the time interval
assessed. Because atazanavir is often coadministered with ritona-
vir, hair levels of each of these agents are substantially collinear, so
separate models were run for atazanavir and ritonavir in the
women receiving ritonavir-boosted atazanavir. All analyses were
performed using SAS software, version 9.2 (SAS Institute).
Participant Demographic Characteristics
Table 1 summarizes demographic and other characteristics for the
424 WIHS participants receiving atazanavir-based cART at any
time from April 2003 through April 2008. The racial and ethnic
distribution of the study sample was 215 African-Americans
(51%), 131 Hispanics (31%), 66 whites (15%), and 12 others (3%
[Native Americans or Asian-Americans]). Each woman con-
tributed 1–9 WIHS biannual study visits to the analysis (median,
3), for a total of 1443 person-visits. Of these 1443 person-visits,
participants reported ritonavir coadministration with atazanavir
in 1136 (79%). One hundred one women in the study cohort
(24%) had never been treated with PIs before starting the ataza-
navir-based regimen, and the median pretreatment CD4 count
was 281 cells/mm3(range, 5–2046 cells/mm3). Adherence was
reported as >95% at 1116 person-visits (77%), 75%–94% at 244
person-visits (17%), and <74% at 83 person-visits (6%). Viral
loads were below the assaythreshold (,80 copies/mL) during 918
(64%) of the 1443 person-visits. On the basis of log-likelihood
statistics, assessment of atazanavir levels as quintiles fit the mul-
tivariate models better than tertile and quartile categorization
schemes and better than continuous measures. Table 1 shows the
hair concentration ranges represented in each quintile (first
quintile, 0.05 to <0.658 ng/mg; second quintile, .0.658 to <1.78
ng/mg; third quintile, .1.78 to <3.13 ng/mg; fourth quintile,
.3.13 to <5.19 ng/mg; fifth quintile, .5.19 ng/mg). A strong
relationship between atazanavir levels in hair and self-reported
adherence (dichotomized into ,95% vs >95%) was observed (F
statistic P , .001).
Association of Atazanavir Hair Levels with Virologic
Figure 2 shows the percentage of person-visits during which
viral loads were ,80 copies/mL according to hair quintile. HIV
RNA levels were undetectable in only 25% of visits in which hair
concentrations of atazanavir were in the lowest quintile. In
April 2003 Through April 2008 (N 5 424)
Characteristics of Women Participating in the Study,
contributing to analysisa
Age, median years (range)43 (21–71)
White (non-Hispanic)66 (15)
African-American (non-Hispanic)215 (51)
Other 12 (3)
Past PI treatment
Pretreatment viral load, copies/mL
HIV viral load, median copies/mL (range)
Pretreatment CD4 cell count, cells/mm3
CD4 cell count, median
Person-visits at which
atazanavir is boosted
Person-visits at which
viral load is undetectableb
Adherence during past 6 months
(self-reported) at person-visit
Person-visits in each atazanavir
hair level quintile (ng/mg), no. (%)
Quintile 1 (0.05 to <0.658 ng/mg)
Quintile 2 (.0.658 to <1.78 ng/mg)
Quintile 3 (.1.78 to <3.13 ng/mg)
Quintile 4 (.3.13 to <5.19 ng/mg)
Quintile 5 (.5.19 ng/mg)
PI, protease inhibitor.
Data are no. (%) of study participants unless otherwise indicated.
aMedian number of visits per patient was 3 (range, 1–9); 105 women had 1
visit; 70 had 2 visits; 53 had 3 visits; 61 had 4 visits; 66 had 5 visits; 33 had 6
visits; 23 had 7 visits; 11 had 8 visits; 2 had 9 visits.
bThreshold of detection, 80 copies/mL.
d CID 2011:52 (15 May)
contrast, for person-visits in which atazanavir concentrations in
hair were in the highest quintile (.5.19 ng/mg), the likelihood
of maximal virologic suppression was 87%. Figure 2 also depicts
the univariate relationshipbetween hairatazanavirlevelsandthe
likelihood of virologic suppression in repeated measures anal-
yses. The odds of viral load undetectability increased with each
quintile; the odds ratio (OR) for viral suppression if the
atazanavir level in hair is in the second quintile (.0.658 to
<1.78 ng/mg) is 4.3 (95% confidence interval [CI], 2.5–7.4; P ,
.001), for example, and the OR for virologic response if ataza-
navir levels are in the top quintile is 63.3 (95% CI, 30.8–130.0;
P , .001).
Table 2 presents the multivariate analysis for virologic re-
sponse in which levels of atazanavir in hair are adjusted for age,
race, extent of PI experience, pretreatment viral load, and self-
copies/mL was associated with a higher odds of achieving
virologic suppression over time, compared with individuals who
started atazanavir-based treatment with viral loads >100,000
copies/mL (OR, 3.2 [95% CI, 1.5–6.9]; P 5 .002). Participants
who were not African-Americanshowed a trend toward a higher
likelihood of virologic suppression while receiving cART than
did African-Americans. Those who had been treated with >2
PIs prior to atazanavir showed a trend toward a lower likelihood
adherence was associated with a higher odds of virologic sup-
pression, with an OR of4.0 (95%CI,1.9–8.6; P , .001)for visits
when participants reported >95% adherence, compared with
those when participants reported ,75% adherence. Adding
pretreatment CD4 cell count, history of clinical AIDS, study
year, or the NRTI backbone agents to the multivariate models
did not significantly alter the results, so these variables were
removed from the final models.
In adjusted analyses, concentrations of atazanavir in hair
were the best independent predictor of virologic suppression.
An atazanavir level in the second quintile, compared with the
first, yielded an OR for virologic suppression that is com-
parable to a self-reported adherence of >95%. Hair atazanavir
levelsin the higherquintiles
progressively increasing odds of virologic suppression;
women whose atazanavir levels were in the highest quintile
had an OR of 59.8 (95% CI, 29.0–123.2; P ,.001) for virologic
When models were repeated focusing on the 1136 person-
visits during which ritonavir was coadministered with atazana-
vir, we found that ritonavir levels in hair were similarly the
strongest predictor of subsequent virologic response. Finally,
when we looked at hair level at one visit as a predictor of
virologic suppression at the subsequent WIHS visit, we also saw
a strong and monotonic relationship between hair level and
Virologic Suppression Rates in Different At-Risk Scenarios
We then investigated the key question of whether higher
atazanavir exposure as indicated by hair levels could reestablish
virologic suppression after a preceding lapse in adherence,
exposure, or virologic suppression. We examined three separate
‘‘at risk’’ subgroups of the 1443 person-visits in our study: (1)
those preceded by a reported adherence level of <95% at any
previous visit (405 person-visits from 152 participants); (2)
those preceded by a hair atazanavir level in the lowest quintile
(0.05–0.658 ng/mg) at any previous visit (377 person-visits from
125 participants); and (3) those with an HIV viral load .1000
copies/mL at any previous visit (356 person-visits from 139
participants). Each subgroup showed lower overall rates of
virologic suppression than did the entire group over time;
quintile (univariate relationship): ATV, atazanavir.
Percent of person-visits in each hair quintile where viral load suppressed (entire study sample) and odds ratio of virologic response per hair
d CID 2011:52 (15 May)
subgroups 2 and 3 showed differences in suppression rates that
were statistically significant (Table 3).
''Redemption'' Through Improved Atazanavir Exposure
Although each at-risk subgroup had lower overall rates of
virologic suppression than did the entire cohort (Table 3), each
subgroup subsequently revealed a pattern of increasing rates
of virologic suppression by increasing quintile of hair atazanavir
level (Figures 3A, 3B, 3C). Indeed, increasing atazanavir levels in
hair became even more determinative of virologic suppression
(or resuppression for subgroup 3) in multivariate models for
eachsubgroupthan forthe overallcohort(Table4).TheORsfor
reaching a viral load of ,80 copies for person-visits in which
hair levelswere in thehighest quintile were 210.2(95% CI, 46.0–
961.1), 132.8 (95% CI, 26.5–666.0), and 400.7 (95% CI, 52.3–
3069.7), respectively, for subgroups 1, 2, and 3.
In this multivariate analysis of a cohort of HIV-infected women
overtime, we revealthatantiretroviral concentrations inhair are
the strongest independent predictor of virologic suppression.
Levels of antiretroviral drugs in hair showed a monotonic re-
lationship to the likelihood of viral suppression (OR for success,
4.3, 12.7, 22.9, and 59.8 for each increasing quintile of hair
Because low hair antiretroviral concentrations can predict
virologicfailureprior toits development,this measurementmay
be useful in designing interventions aimed at prolonging the
durability of cART.
Concentrations of antiretroviral drugs in hair samples may
provide an integrated measure of behavior and biology. Levels
of medications in hair reflect drug uptake from the systemic
circulation over periods of weeks to months  and capture
average, as well as individual, pharmacokinetic information.
Single adherence measures or plasma antiretroviral concen-
trations provide ‘‘snapshots’’ of exposure, but a level measured
in hair synthesize adherence and pharmacokinetic variability
over time to provide a robust exposure measure in a single assay
. The value of single plasma antiretroviral levels is further
limited by the so-called ‘‘white coat effect,’’ in which adherence
transiently improves prior to clinic appointments , and by
an inability to define meaningful therapeutic antiretroviral
Table 3.Rates of Virologic Suppression in Overall Cohort Compared With Rates in 3 Scenarios (Subgroups) of Failure
Group No. of visits
P (compared with visits
not in subgroup)a,b
Entire cohort: all visits 144364c
Subgroup 1: self-reported
adherence <95% at a previous visit
Subgroup 2: hair value in lowest
quintile at a previous visit
Subgroup 3: viral load .1000 copies/mL
at a previous visit
aP value was obtained by comparing visits in the subgroup of interest with visits not in the subgroup (accounting for repeated measures).
bOverall P value for the chance of virologic suppression for a person-visit in any of the 3 subgroups (compared with not being in any of the subgroups) is .02.
cMedian HIV RNA level for person-visits where viral load is not suppressed in the entire cohort, 1800 copies/mL (interquartile range, 310–17,000 copies/mL).
and Virologic Success (Entire Study Sample)
Multivariate Model of Hair Atazanavir Concentrations
OR of virologic
response (95% CI)P
Age, per decade1.05 (0.78–1.42) .75
Not African-American1.6 (0.99–2.7).06
RNA level, copies/mL
Past PI treatment
0.56 (0.29–1.08) .09
,.001 4.0 (1.9–8.6)
2 4.3 (2.5–7.4)
3 12.7 (7.1–22.8)
4 22.9 (12.2–43.1)
5 59.8 (29.0–123.2)
study year, and/or nucleoside reverse transcriptase inhibitor backbone
components used with atazanavir did not significantly alter the results of the
analysis, and these variables were therefore removed from the final model. AA,
African-American; CI, confidence interval; OR, odds ratio; PI, protease inhibitor.
Adjusting for pretreatment CD4 cell count, a history of clinical AIDS,
d CID 2011:52 (15 May)
(Figure 3A: subgroup 1; Figure 3B: subgroup 2; Figure 3C: subgroup 3). ATV, atazanavir.
Percent of person-visits in each hair quintile in which viral load was suppressed for 3 subgroups in which ''failure'' was previously revealed
d CID 2011:52 (15 May)
ranges because of substantial interindividual pharmacokinetic
variability [15, 16]. Therefore, despite the importance of en-
suring adequate exposure to the components of HIV regimens,
Our models show that antiretroviral exposure as measured
in hair far surpasses commonly used covariates to predict HIV
treatment outcomes. Failed antiretroviralregimens resultin
substantial long-term adverse effects, including increased drug
and diagnostic testing costs, as well as avoidable clinical and
transmission events. Because patients who experience virologic
failure on a regimen demonstrate attenuated rates of immune
reconstitution on future regimens , substantial efforts
should be made to optimize first-line cART. Preserving
responses to first-line regimens are of particular import in
resource-limited settings, where the average annual cost of
second-line cART regimens canbe up to8 times that of first-line
regimens . The risk of viremia on therapy is highest in the
first year after initiating cART and can be linked to increased
HIV transmission rates . Therefore, when initiating HIV
treatment, the incorporation of an effective antiretroviral
exposure measure, such as hair concentrations, during initial
monitoring may avert early virologic failure, blunted responses
to subsequent regimens, and the need for expensive or
inaccessible second-line regimens.
Previous models of outcomes with atazanavir-based regimens
have failed to define precise parameters of atazanavir exposure
that increase the likelihood of virologic suppression [15, 21].
This failure could be a result of using single plasma atazanavir
concentrations in these models to define exposure instead of a
longer term measure. A recent report revealed that average ad-
herence to dosage with boosted PI regimens was a better pre-
dictor of virologic suppression than was duration or frequency
of missed doses . Hair antiretroviral concentrations average
daily exposure variability in a manner analogous to that of
glycosylated hemoglobin A1C providing information on mean
daily glucose levels in diabetic patients. A previous analysis by
our group demonstrated that hair levels of antiretrovirals are
more closely correlated with areas under the curve from in-
tensive pharmacokinetic studies than are single plasma levels
. Therefore, it is not surprising that hair antiretroviral
measurements predict treatment outcomes with greater accuracy
than do single plasma levels. Another analysis by our group, di-
to predict treatment responses, demonstrated the superiority of
hair levels .
In addition to showing that atazanavir concentrations in
hair predict virologic responses more strongly than self-re-
ported adherence or other factors, we demonstrate that lapses
in adherence, antiretroviral exposure, or virologic suppression
are all associated with an increased likelihood of subsequent
failure. Since adherence difficulties or the presence of de-
tectable virus during therapy are well-known contributors to
virologic failure, either state is likely to trigger corrective
measures in the clinical setting. However, the inaccuracy of
self-reported adherence, the lack of routine virologic mon-
itoring in many resource-limited settings, and the fact that
detectable viral loads when available may already indicate
mutations  all increase the appeal of finding another tool
to prospectively predict failure. Our models show that low
antiretroviral hair levels portend a high risk of virologic fail-
ure; hair measures in the clinical setting could therefore
trigger interventions to correct either adherence or low
pharmacokinetic levels (eg, through assessing drug-drug in-
teractions or diet) to extend regimen durability (Figure 4). Of
note, only participants with hair atazanavir measurements in
the higher quintiles during WIHS visits following a failure
scenario had rates of virologic suppression similar to those in
clinical trials. This supports the concept that the reasons for
low drug levels in hair must be investigated and addressed by
adherence intervention, change in regimen, or possibly dose
increase (the latter requires additional study). Our group is
currently planning a clinical trial assessing adherence inter-
ventions, regimen change, or dose modification of anti-
retrovirals based on hair measurements of anchor regimen
components in a clinical setting.
Table 4. Multivariate Models of Hair Atazanavir Concentrations and Virologic Success (3 Subgroups)
Subgroup 1Subgroup 2 Subgroup 3
Person-visitsOR (95% CI) Person-visits OR (95% CI)Person-visits OR (95% CI)
1 83 Reference116Reference 112 Reference
2 91 14.1 (4.7–42.8)83 9.6 (3.4–27.2)78 43.8 (8.9–215.7)
3 99 26.3 (8.4–82.0)7230.0 (8.8–102.5) 6147.2 (9.3–240.2)
4 7360.5 (17.2–211.7)61 24.1 (7.0–82.4)60 82.5 (15.5–439.3)
5 59 210.2 (46.0–961.1)45 132.8 (26.5–666.0)45 400.7 (52.3–3069.7)
adherence. Subgroup 1: self-reported adherence <95% at a previous visit (n 5 405); subgroup 2: hair value in lowest quintile at a previous visit (n 5 377); subgroup
3: viral load .1000 copies/mL at a previous visit (n 5 356). CI, confidence interval; OR, odds ratio.
All P values are ,.001. All models adjusted for age, race, previous treatment with protease inhibitors, pre-atazanavir HIV viral load, and self-reported
d CID 2011:52 (15 May)
As we proposed previously , one possible algorithm for
testing would involve measuring antiretroviral levels in hair
soon after starting a new antiretroviral regimen and performing
HIV viral load testing only if the hair levels fall into the lower
quintiles as defined above (Figure 4). Data showing that the risk
of viremia on therapy is highest during the first year after ini-
tiating cART  supports the use of these measures soon after
regimen initiation. Data from the resource-limited setting
showing that routine viral load monitoring decreases rates of
the use of hair measures as a surrogate for the former . If the
atazanavir level in hair is <1.78 ng/mg (first or second quintile),
the rates of virologic failure approach 50% and intensive ad-
herence interventions (vs a pharmacokinetic evaluation for low
exposure if adherence is deemed adequate) should be triggered.
After a patient is receiving stable HIV therapy, antiretroviral
measurements using hair need not be performed routinely but
only when clinical disease progression is observed (for settings
where routine CD4 cell count or viral load monitoring are not
available) or when an alteration in drug exposure is predicted,
such as a new drug-drug interaction, pregnancy, change in di-
etary patterns, change in liver or renal function, and so on.
Unlike phlebotomy, hair collection is noninvasive and does
not require specific skills, sterile equipment, or specialized
storage conditions. The collection of hair samples for analysis
merely requires a pair of scissors, and storage is at room tem-
perature. Hair can be stored for long periods of time prior to
analysis, shipped without precautions for biohazardous mate-
rials, and analyzed economically in a high-throughput hair
analysis laboratory. These features may make this monitoring
tool particularly advantageous in the resource-poor setting,
especially when routine viral load monitoring is prohibitively
expensive. We recently applied these hair measures in a nested
case-control study in 2 South African public health clinics and
found that low concentrations of lopinavir in hair had a high
predictive value for virologic failure in that setting . We are
currently working on developing a lower cost, point-of-care
method of analyzing antiretroviral levels in hair for resource-
constrained settings to increase the feasibility of this tool. The
results of the analyses presented here argue for the possibility of
hair antiretroviral concentrations serving as a method of HIV
therapeutic drug monitoring that may increase the durability of
current antiretroviral regimens in a variety of settings.
We would like to thank the WIHS participants who contributed data to
We thank the Women’s Interagency HIV Study (WIHS) participants who
contributed data to this study. Data were collected by the WIHS Collaborative
Study Group with centers (Principal Investigators) at New York City/Bronx
Consortium (Kathryn Anastos, MD); Brooklyn, NY (Howard Minkoff, MD);
Washington DC, Metropolitan Consortium (Mary Young, MD); The Connie
Wofsy Study Consortium of Northern California (Ruth Greenblatt, MD); Los
Angeles County/Southern California Consortium (Alexandra Levine, MD);
Chicago Consortium (Mardge Cohen, MD); and Data Coordinating Center
(Stephen Gange, PhD).
The contents of this publication are solely the responsibility of the
authors and do not necessarily represent the official views of the National
Institutes of Health.
Authors’ contributions: M.G. developed the study protocol, provided
study oversight, designed the analysis plan, interpreted the data, and wrote
Possible algorithm for use of atazanavir hair levels in the clinical setting. ATV, atazanavir.
d CID 2011:52 (15 May)
the paper. R.M.G. contributed to the study concept and the analysis plan
and interpretation. N.A. and P.B. provided data management, contributed
to the analysis plan, and performed most of the statistical analyses. R.M.G.,
N.A., P.B., K.A., S.J.G., H.M., M.Y., J.M., M.H.C., and G.B.S. collected
WIHS participant data, helped design protocols, and critically revised the
manuscript. Y.H. developed the laboratory methods for analysis of anti-
retroviral levels in hair.
Financial support. This work was supported by the National Institute
of Allergy and Infectious Diseases (NIAID RO1-AI-65233) with data col-
lected in the Women’s Interagency HIV Study (WIHS). The WIHS is
funded by the National Institute of Allergy and Infectious Diseases (UO1-
AI-35004, UO1-AI-31834, UO1-AI-34994, UO1-AI-34989, UO1-AI-34993,
and UO1-AI-42590) and by the Eunice Kennedy Shriver National Institute
of Child Health and Human Development (UO1-HD-32632). The WIHS is
cofunded by the National Cancer Institute, the National Institute on Drug
Abuse, and the National Institute on Deafness and Other Communication
Disorders. Funding for WIHS is also provided by the National Center for
Research Resources (UCSF - CTSI Grant UL1 RR024131). M.G. was ad-
ditionally supported by a Mentored Patient-Oriented Research Career
Development Award (K23 A1067065) from NIAID.
Potential conflicts of interest. R.M.G. has received payment previously
for manuscript preparation from HRSA and all authors have received
funding from the National Institutes of Health. All other authors: no
1. Bisson GP, Gross R, Strom JB, et al. Diagnostic accuracy of CD4
cell count increase for virologic response after initiating highly active
antiretroviral therapy. AIDS 2006; 20:1613–9.
2. Bennett DE, Bertagnolio S, Sutherland D, Gilks CF. The World
Health Organization’s global strategy for prevention and assessment of
HIV drug resistance. Antivir Ther 2008; 13(suppl 2):1–13.
3. Bertagnolio S, Kelley K, Hassani AS, et al. Surveillance of transmitted
and acquired HIV drug resistance using WHO surveys in resource-
limited settings. 18th Conference on Retroviruses and Opportunistic
Infections, Boston, MA 2011.
4. Berg KM, Arnsten JH. Practical and conceptual challenges in
measuring antiretroviral adherence. J Acquir Immune Defic Syndr
2006; 43(suppl 1):S79–87.
5. Gandhi M, Greenblatt RM. Hair it is: the long and short of monitoring
antiretroviral treatment. Ann Intern Med 2002; 137:696–7.
6. Huang Y, Gandhi M, Greenblatt RM, Gee W, Lin ET, Messenkoff N.
Sensitive analysis of anti-HIV drugs, efavirenz, lopinavir and rito-
navir, in human hair by liquid chromatography coupled with tan-
dem mass spectrometry. Rapid Commun Mass Spectrom 2008;
7. Gandhi M, Ameli N, Bacchetti P, et al. Protease inhibitor levels in
hair strongly predict virologic response to treatment. AIDS 2009;
8. Liu A, Vittinghoff E, Gandhi M, et al. Validating measures of
tenofovir drug exposure in a U.S. Pre-exposure prophylaxis trial. 17th
Conference on Retroviruses and Opportunistic Infections. San Fran-
cisco, CA, 2010.
9. Huang Y, Yang Q, Gandhi M, Greenblatt R, Gee W, Lin E. Sensitive
analysis of ritonavir, lopinavir, atazanavir and efavirenz in human hair
by LC/MS/MS for monitoring antiretroviral drug long-term exposure.
American Association of Pharmaceutical Sciences (AAPS) Annual
Meeting. San Diego, CA, 2007.
10. Yang Q, Liu A, Gandhi M, Greenblatt R, Gee W, Huang Y. LC/LC/MS
Assay of Tenofovir in Human Hair for Pre-Exposure Prophylaxis.
Association of Pharmaceutical Sciences (AAPS) Annual Meeting (New
11. Huang Y, Yang Q, Koon K, Lei Y, Gee W, Lin ET, Greenblatt RM,
Gandhi M. Microanalysis of the Antiretroviral Medication Nevirapine
in Human Hair by LC/MS/MS for Monitoring Long-Term Drug Ex-
posure. In press .
12. Barkan SE, Melnick SL, Preston-Martin S, et al. The Women’s
Interagency HIV Study. Epidemiology 1998; 9:117–25.
13. Giordano TP, Guzman D, Clark R, Charlebois ED, Bangsberg DR.
Measuring adherence to antiretroviral therapy in a diverse population
using a visual analogue scale. HIV Clin Trials 2004; 5:74–9.
14. Beumer J, Bosman I, Maes R. Hair as a biological specimen for
therapeutic drug monitoring. Int J Clin Pract 2001; 55:353–7.
15. Liu A, Gandhi M, Bacchetti P, et al. Validating hair as a biological
marker of tenofovir drug exposure in HIV pre-exposure prophylaxis
(PrEP). 18th Conference on Retroviruses and Opportunistic In-
fections. Boston, MA, 2011.
16. Podsadecki TJ, Vrijens BC, Tousset EP, Rode RA, Hanna GJ. ‘‘White
coat compliance’’ limits the reliability of therapeutic drug monitoring
in HIV-1-infected patients. HIV Clin Trials 2008; 9:238–46.
17. Lescure FX, Poirier JM, Meynard JL, et al. Factors predictive of
virological failure on atazanavir in 310 HIV-infected patients. AIDS
18. Ray JE, Marriott D, Bloch MT, McLachlan AJ. Therapeutic drug
monitoring of atazanavir: surveillance of pharmacotherapy in the
clinic. Br J Clin Pharmacol 2005; 60:291–9.
19. Cescon A, Cooper C, Chan K, et al. Factors associated with virological
suppression among HIV-positive individuals on highly active
antiretroviral therapy in a multi-site Canadian cohort. HIV Med 2010.
20. Trotta MP, Cozzi-Lepri A, Ammassari A, et al. Rate of CD41cell count
increase over periods of viral load suppression: relationship with the
number of previous virological failures. Clin Infect Dis 2010;
21. World Health Organization, Joint United Nations Program on HIV/
AIDS, United Nations Children’s Fund. Towards universal access:
scaling up priority HIV/AIDS interventions in the health sector: pro-
gress report, September 2009. World Health Organization website.
Accessed 21 November 2010.
22. EngsigFN, Omland LH, Larsen MV, et al. Risk of high-level viraemia in
HIV-infected patients on successful antiretroviral treatment for more
than 6 months. HIV Med 2010; 11:457–61.
23. Fabbiani M, Di Giambenedetto S, Ragazzoni E, et al. Mid-dosing
interval concentration of atazanavir and virological outcome in
patients treated for HIV-1 infection. HIV Med 2010; 11:326–33.
24. Parienti JJ, Ragland K, Lucht F, et al. Average adherence to
boosted protease inhibitor therapy, rather than the pattern of missed
doses, as a predictor of HIV RNA replication. Clin Infect Dis 2010;
25. Gandhi M, Ameli N, Gange S, et al. Concentrations of efavirenz in
hair correlate strongly with 24-hour intensive PK measurements and
with virologic outcomes. 17th Conference on Retroviruses and
Opportunistic Infections; San Francisco, CA:16–19 February 2010.
26. Hawkins C, Murphy RL. Management of antiretroviral failure
and resistance in developing countries. Curr Opin HIV AIDS 2009;
27. Reynolds S, Sendagire H, Newell K, et al. Routine viral load monitoring
reduces the rate of genotypic resistance to commonly used ART in
Uganda. 18th Conference on Retroviruses and Opportunistic In-
fections. Boston, MA, 2011.
28. van Zyl GU, van Mens TE, McIlleron H, et al. Low lopinavir plasma or
hair concentrations explain second line protease inhibitor failures in a
resource-limited setting. J Acquir Immune Defic Syndr 2011. [Epub
ahead of print].
d CID 2011:52 (15 May)