IntroductionNigeria has the second largest global population of people living with HIV
(3.4 million) (1). Successful efforts to
combat the pandemic have led to reductions in HIV related morbidity and mortality
(2). High rates of loss to follow-up
(LTFU) and unplanned interruptions from HIV care have challenged these efforts, and
may be of particular concern among adolescents and young adults (AYA) as they
transition to adulthood (3–5). Indeed HIV/AIDS is now the leading cause of
death among AYA in sub-Saharan Africa, and nearly 1 in 10 HIV-infected AYA worldwide
reside in Nigeria (6, 7).
The WHO defines adolescents as individuals aged 10–19 years (8). This period overlaps with the transition to
independence sometimes defined as youth or young adulthood (15–24 years)
(9). This is a unique time of development
characterized by new psychosocial stresses, desire for autonomy, risk-taking,
concrete thinking, variable levels of social support, and unique perceptions of risk
(10, 11). These factors, coupled with high prevalence of affective disorders
among adolescents with chronic illness, often directly oppose the circumstances
necessary for adherence to complex, chronic medical therapies (12). HIV-infected AYA in particular also contend with important
issues around disclosure and transmission while negotiating the framework of a
chronic, stigmatizing disease (13). Poor
adherence to care among AYA may have serious negative consequences in
resource-limited settings (RLS) with limited access to second and third-line
treatment options (10). Importantly, while
HIV deaths overall have decreased by 30% in Africa over the past 8 years,
deaths have increased by 50% among adolescents (8, 14).
Despite such concerning trends, HIV-infected youth, and particularly
adolescents, remain understudied (15). While
many reports are not disaggregated to highlight HIV outcomes in these groups, AYA
with a range of chronic disease appear to have poorer adherence to care and worse
clinical outcomes than children and older adults (16, 17). The aim of our study was
to determine whether adolescence and young adulthood is an independent risk factor
for inconsistent care after ART initiation, and to compare rates of viremia among
AYA and older adults who remain in care in the first year on antiretroviral therapy
(ART).
MethodsSettingThis study was conducted at the HIV clinic of the Ahmadu Bello University
Teaching Hospital (ABUTH). ABUTH is located in a semi-urban community in Kaduna,
Nigeria where the state’s HIV prevalence is 5.1% (2, 18). With PEPFAR support, ABUTH began providing comprehensive HIV care
in 2006 that was free of charge to all eligible patients. Children are cared for
in the pediatric clinic from birth to 14 years, and in the adult clinic from 15
years of age on. During the study period, the clinic was managed by the AIDS
Prevention Initiative in Nigeria (APIN), a PEPFAR-supported NGO. APIN is also
one of the largest HIV treatment programs in Nigeria.
Study DesignWe conducted a retrospective cohort study of ART eligible patients who
enrolled in the ABUTH “adult” clinic between January 1, 2009 and
December 31, 2011. Data were censored on December 31, 2012. Visit patterns were
assessed during the first year on ART, and HIV viral load after 1 year on ART.
Inclusion criteria included age >14 years at the time of enrollment and
documentation of initiation of ART. Women who were pregnant at enrollment or
became pregnant during the follow-up period were seen in the prevention of
mother to child transmission clinic, and not included in this analysis. All
data, including baseline demographic information, transmission risk factor,
clinical visits and evaluations, laboratory visits with viral load results, and
pharmacy drug pick-up visits were recorded on structured data collection forms
and entered into APIN’s electronic clinical database. These data were
abstracted retrospectively for this analysis.
Visits were most frequent in the first 2 months after ART initiation
when patients are scheduled to be seen at 2, 4, 8, and 12 weeks for adherence
counseling, clinical examination, and Tuberculosis (TB) symptom screening
(clinical visits). Subsequently, patients are seen for ART pick-up every 4 weeks
(pharmacy visits); clinical examination and adherence counseling every 12 weeks
(clinical visits), and laboratory testing (including CD4 count and HIV viral
load) every 24 weeks (laboratory visits) (19).
Outcome MeasuresOutcomes were assessed at the end of the first year on ART. We
categorized patients into two mutually exclusive groups based on their visit
patterns. A visit was defined as any clinic visit for clinical, laboratory, or
pharmacy services. Patients were defined as being in care if
the time between any two consecutive visits was ≤3 months, and the time
between the last visit and censor date was ≤6 months. All other patients
were defined as having inconsistent care. The latter group
comprised both patients who had unplanned care interruption
(UCI) (if the time between any two consecutive visits was ever
>3 months, but they returned to clinic before the censor date) and
patients who were inactive from the clinic (if the time between
any two consecutive visits was ≤3 months, but the time between the last
visit and the censor date was >6 months) (Figure 1). Patients known to have transferred care or died during
the follow-up period were categorized based on their visit patterns prior to
transfer or death. Under routine circumstances, an absence from the clinic of at
least 3 months implied that a patient missed three ART pick-up visits, and at
least one clinical visit. In select circumstances, clinic protocol permitted
dispensing of 2-month ART prescriptions (usually reserved for patients
virologically suppressed on ART for >1 year). We chose a 3-month window
to define UCI to ensure no overlap with this select group of stably suppressed
patients.
Statistical AnalysisBaseline Demographic and Clinical ParametersWe compared baseline demographic and clinical parameters including
sex, level of education, employment status, marital status, co-infection
with tuberculosis, and median CD4 count, between AYA and adults (at the time
of ART initiation). Binary and ordinal variables were compared using Chi
Squared tests, proportions of continuous variables were compared using test
of proportions, and medians of continuous variables were compared using
Kruskal Wallis tests.
Patterns of Care UtilizationWe determined the proportion of AYA and adults in care and with
inconsistent care at the end of the first year on ART to standardize
follow-up time. We used Chi square tests to determine if the proportion of
patients who remained in care differed between AYA and OA.
Risk of Inconsistent Care in the First Year on ARTUsing the entire cohort of patients, (both those who were in care
and those who had inconsistent care), we built bivariate and multivariate
negative binomial regression models to assess the association between
baseline age category (AYA vs. older adults) and the risk of inconsistent
care during the first year on ART. We adjusted for potential confounders
including sex (male vs. female), education level (no primary education vs.
any primary, secondary, or advanced education), marital status (married vs.
single), employment status (employed vs. unemployed vs. student), baseline
TB diagnosis (no vs. yes), and CD4 count at the time of ART initiation
(<100 cells/µL vs. 100–200 cells/µL vs.
201–350 cells/µL vs. >350 cells/µL vs.
missing). Covariates demonstrating marginally significant bivariate
associations (p≤0.10) were advanced to the multivariate model.
Risk of Viremia Among Patients In Care in the First Year on ARTHIV RNA testing is recommended as a “desirable test”
at baseline before ART initiation, and every 6 months on ART, according to
the 2010 Nigerian national guidelines (19). In the subset of patients who remained in care during the
first year on ART, we abstracted HIV RNA values obtained 12 months after ART
initiation (+/− 6 months), dichotomized these values (>1000
copies/mL vs. ≤1000 copies/mL), and compared the risk of viremia
between AYA and adults using a negative binomial regression model. Model
building was approached as described above. A threshold of 1000 copies/mL
was chosen to be consistent with virologic failure (as suggested by the
WHO), and to ensure that we did not include low-level blips in this
definition (20). However, given
limitations on the capacity of viral load testing in this setting, we could
not adhere to WHO’s guidance suggesting repeat HIV RNA testing
within 3 months to confirm this value (20).
Sensitivity AnalysisThere is substantial variation in how consistent and inconsistent HIV
care are defined in the literature (21).
Consequently, we varied the definition of consistent care from having
≤60 days to ≤90 days between laboratory, pharmacy, and clinic
encounters in sensitivity analysis. Statistical analysis was conducted with
Stata Statistical Software (StataCorp. 2013. Stata Statistical Software:
Release 13. College Station, TX, USA).
IRB ApprovalWe obtained IRB approval from Partners HealthCare (Protocol number:
2013P000219) and Harvard T. H. Chan School of Public Health in Boston, MA, USA,
and the Nigerian Institute for Medical Research in Lagos, Nigeria.
ResultsBaseline Demographic and Clinical ParametersThere were 3,137 patients who enrolled at the ABUTH clinic between 2009
and 2011. Fifty- four patients were excluded because there was no documentation
of ART initiation during the study period; 589 patients were excluded because
they initiated ART after the censor date (Figure
1). Our analysis cohort was comprised of 354 AYA (15–24
years) and 2,140 older adults (>24 years) who initiated ART at the ABUTH
clinic during the study period (Table 1).
Most patients (95%, n=2,366) identified heterosexual sex as a
transmission risk factor. No AYA identified perinatal transmission as their mode
of infection, though 5% (n=18) reported an unknown risk factor, and
2% (n=6) reported transfusion-related transmission. The vast majority of
AYA were female (89%) compared to 65% of older adults
(p<0.0001). Thirty-percent of AYA (n=107) reported having no education
compared to 22% (n=464) of older adults (p<0.0001). Students and
unemployed made up a greater proportion of the AYA population than the older
adult population (23% vs. 6% and 49% vs. 26%
respectively, p<0.0001). Half of AYA were married compared to
62% of older adults (p<0.0001). AYA started ART with a higher
median baseline CD4 count [252/µL; IQR 107, 404/µL] than older
adults [204/µL; IQR 96, 447/µL] p=0.0024].
Patterns of Care UtilizationAt the end of the first year on ART, fewer AYA remained in care than
older adults (46% vs. 53%, p=0.011). By the end of the follow-up
period (median 1.8 years) only 33% of AYA remained in care compared to
43% of older adults, p=0.001. Thirty percent of AYA (n=106) and
25% of older adults (n=526) had periods of UCI lasting more than 3
months, and subsequently returned to care within the first year on ART (p=0.03).
In contrast 25% (n=87) of AYA and 23% (n=485) of older adults
remained inactive in care for more than 6 months, and did not return to care
within the first year on ART (p=0.43). As such, the difference in inconsistent
care for both AYA and older adults was driven primarily by brief interruptions
in care, and not by longer periods of inactivity or “loss to
follow-up” from clinic (Figure
2).
Risk of Inconsistent Care in the First Year on ARTIn bivariate analysis, AYA had increased risk of inconsistent care
compared to older adults [RR 1.15, p=0.008] (Table 2). Male sex [RR 1.13, p=0.003] was associated with increased
risk of inconsistent care; however in stratified analysis of AYA, the data did
not suggest that male youth were at increased risk of inconsistent care compared
to female youth [RR 1.15, p=0.317]. In addition to AYA, being a student [RR
1.17, p=0.029], single [RR 1.17, p<0.001], co-infected with TB [RR 1.20,
p<0.038] and having a baseline CD4 count greater than 350/µL [RR
1.36, p<0.001] or missing value for baseline CD4 count [RR 1.33,
p<0.001] were also associated with increased risk of inconsistent care
in the first year on ART. Having any education [RR 0.91, p=0.039] and having
baseline CD4 count 100–200/µL [RR 0.85, p=0.026] or
201–350/µL [RR 0.73, p<0.001) were associated with
decreased risk of inconsistent care. In multivariate analysis AYA remained at
increased risk for inconsistent care compared to older adults [RR 1.11, p=0.03]
even after adjusting for sex, educational level, marital status, baseline TB
diagnosis, and baseline CD4 count. The final multivariate model was not adjusted
for employment category, as student status was found to be collinear with AYA
age. The relationship between age group and risk of inconsistent care remained
robust when the definition of UCI we varied from 60 to 90 days in sensitivity
analysis.
Rates of Viremia In the First Year on ARTAmong all patients in care at the end of the first year on ART
(n=1,292), 12-month HIV RNA values were available for 86% of AYA and
87% of older adults. Forty percent of AYA had an HIV viral load
>1000 copies/mL at 12-months compared to 26% of older adults
(p=0.033). In univariate analysis, being an AYA was the only independent
predictor of having an increased risk of viremia at 12-months [RR 1.54,
p=0.002], (Table 3). None of the other
covariates (sex, education level, employment status, marital status, TB
co-infection, or baseline CD4 count) were associated with risk of viremia after
1 year on ART for patients who remained in care. Consequently, an adjusted
analysis was not performed. The relationship between age group and risk of
viremia remained robust when we varied the definition of UCI from 60 to 90 days
in sensitivity analysis
DiscussionWe examined consistency of clinical care, and rates of viremia among AYA and
OA in a large clinic in Nigeria. Our findings highlight that compared to older
adults, AYA have more than a 10% increased risk of inconsistent HIV care
after starting ART. Moreover, even AYA who remain in care after starting ART are
50% more likely than older adults to have ongoing viremia at one year. High
rates of viremia among patients in our cohort who consistently attended clinic, lab,
and pharmacy visits suggest medication adherence (but not medication pick-up)
challenges, especially among AYA. These data substantiate the growing evidence that
HIV-infected AYA have poorer use of life-saving HIV care services and clinical
outcomes relative to older adults (10, 11, 22).
In our cohort, consistent engagement in HIV care in the first year on ART
was low for all patients, but much worse for AYA than older adults. Despite our
findings, outcomes of HIV-infected youth may not be directly comparable due to
varying definitions of retention. In our cohort, the proportion of patients who
became inactive from care during the first year on ART, approximately 1 in 4,
remained fairly consistent across age groups. As such, the differences in
inconsistent care between AYA and OA seemed to be driven primarily by differences in
rates of UCI. In one systematic review representing 23 low and middle-income
countries, pediatric ART retention was similar to that of older adults (23). This analysis included studies in which
median age ranged from 3 to 9 years. In contrast, another large analysis, from 7
sub-Saharan African countries, reported higher rates of LTFU among adolescents and
young adults aged 15–24 years (24).
Unlike most other studies of retention in this group, our definition of retention
distinguished between transient interruptions in care and longer periods of
inactivity from clinic. Failure to account for these transient interruptions may
yield mixed findings on retention or LTFU.
Data on the outcomes of HIV-infected youth may also not be directly
comparable due to disparate age classification. Studies from South Africa and Malawi
have found that older adolescents and young adults (15–24 years), but not
younger adolescents (<15 years), were more likely to be lost to follow-up
than older adults (>24 years) (25,
26). Our results are consistent with
these findings. While some have hypothesized that younger adolescents may have more
parental or caregiver support to promote retention than older adolescents, few data
exist to support this (27).
Despite varied reports of retention among HIV-infected AYA, most studies
report poor rates of virologic suppression in this age group relative to older
adults (22, 28). Our analysis is consistent with these findings, but further
emphasizes an important difference in rates of virologic suppression even among
youth who remain in care. The absolute rate of virologic suppression among patients
who consistently attend clinical, pharmacy, and lab visits was 14% lower for
youth compared to older adults. Notably, being an AYA was the only independent
predictor of significant viremia one year after starting ART. Taken together, these
findings suggest discordance between adherence to clinical visits and adherence to
ART, a phenomenon which seemed to be more pronounced in youth compared to older
adults. This poses the question of whether youth who are consistently engaged in HIV
care are less likely to have optimal adherence to antiretroviral medication than
adults who are consistently engaged in HIV care. If so, they may warrant distinct
adherence interventions.
The effect of gender and CD4 count on interruptions from HIV care is
inconsistent across multiple studies (3, 26, 29).
In our cohort, females were over-represented amongst AYA compared to older adults,
and AYA had a higher median CD4 count at baseline than older adults. A prior
analysis by our group showed that high baseline CD4 is an independent risk factor
for UCI from HIV care suggesting that patients who initiate ART with higher CD4
counts may be less likely to comply with care because they do not feel
“sick” (30–32). In our Nigerian cohort, it is possible
that the high baseline CD4 in the younger population may be accounted for by a
predominance of females, who tend to be diagnosed with higher CD4 counts than males
(33, 34). The reasons for this are unclear, but may reflect earlier diagnosis
in the context of routine antenatal care (35). Young females (15–25 years) are an important target population
for HIV prevention, representing more than 40% of new infections among all
women >15 years (1). Furthermore,
recent surveys from sub-Saharan Africa show that girls in the same age group are
five times more likely to be infected with HIV than young boys (6). Despite these data, adolescent and young
girls were not at increased risk for inconsistent care compared to young boys in our
cohort. Importantly, even after adjusting for sex, baseline CD4 count, TB
coinfection, marital status, and education level, youth remained at increased risk
of inconsistent care in the first year on ART.
Adherence to care and to antiretroviral therapy may be influenced by a range
of structural, socio-cultural, and patient-level barriers that uniquely impact young
people, who are themselves in the midst of important physical and psychological
development (11). Structural problems
standing in the way of successful adherence include work and school obligations
conflicting with clinic appointments, housing or food instability, and inability to
afford transportation to the clinic or clinic-based fees (11, 13, 36). Additionally, the lack of trained
healthcare professionals in adolescent healthcare management may further impact
strategies adopted to guide the transition to adult care for HIV-infected
adolescents in RLS (4, 11). Studies of perinatally and behaviorally infected youth in
the US shower lower rates of retention and greater rates of ART discontinuation
among youth cared for in pediatric compared to adult clinics (11, 37). Socio-cultural
factors, especially around perceived and enacted stigma and comfort with disclosure
and parental or family support may also influence adherence behaviors (38, 39).
One study from Botswana of mostly perinatally infected adolescents found that
absence of a parent from clinic visits was associated with a 4-fold increased risk
of virologic failure (40). Patient-level
factors such as altered perceptions of risk, treatment fatigue, and emotional
unpreparedness to cope with a stigmatizing illness may lead to poor adherence (11, 39).
There are few data to guide interventions to improve HIV outcomes among AYA,
especially in resource limited settings. However, successful models of care for
HIV-infected AYA may require multi-level approaches to address structural factors
such as provider expertise, youth-friendly adult clinics, and both food
supplementation and transportation assistance for needy youth; socio-cultural
factors such as support around coping with HIV disclosure and managing stigma; and
finally patient-level factors such as continued parental engagement in care, and
less intensive ART administration (e.g. weekends off, long acting ART (7, 13,
36, 38). Behavioral economics interventions relying on conditional cash
transfers or social protection have been successfully utilized to reduce risk of new
HIV infection among youth in sub-Saharan Africa, and may hold promise for promoting
adherence to ART (41, 42). Integration of social media, and mobile technologies for
“e-health” beyond SMS or text reminders may also be quite important
to engage youth in consistent HIV care (43,
44). While some of these factors are
unique to youth, others are not, but might need to be addressed differently in youth
compared to their adult counterparts.”
Our study had some important limitations. Since our study was conducted in
an urban environment at one university-affiliated clinic, with a small population of
AYA relative to older adults. As such, it may not be fully generalizable to other
settings. Additionally, our definition of inconsistent care combined fewer patients
who were inactive from clinic with more patients who interrupted but returned to
clinic, and may have masked factors that were more prevalent in the inactive group.
Nevertheless, use of APIN’s robust electronic health record allowed us to
adjust for potential confounders between these groups, and quantify inconsistent
care by relying on a range of healthcare visits (clinical, laboratory, and pharmacy)
over the first year on ART. Additionally, our analysis is one of the first in
Nigeria to compare rates of viremia between AYA and older adults.
Globally, the magnitude of differences in HIV clinical outcomes and the
determinants influencing them have not been well quantified or described (6, 22,
28). While there is an important but
small minority of perinatally infected patients who are approaching adolescence and
young adulthood in sub-Saharan Africa, the majority of HIV-infection among young
people is sexually acquired (45). This study
of horizontally infected AYA in Nigeria underscores that age certainly matters for
important HIV health outcomes, and that AYA need additional focused attention to
ensure they maximally benefit from HIV care. We found not only that AYA are more
likely to have inconsistent HIV care after starting ART, but also that even those
engaged in care had a greater risk of virologic failure. These disparate outcomes
suggest a need to better understand the obstacles to engagement in care and
medication adherence in this setting. Our data also suggest that youth-centered
clinics and strategies, as earlier described, may be important to optimally address
the unique health needs and ensure the best clinical outcomes in this age group.