Background

The estimated global prevalence of HIV-associated neurocognitive disorder (HAND) is 52% among young people 18–30 years of age [1]. While several critical periods of cognitive development occur during childhood, neurocognitive development and maturation extend to adolescence and early adulthood. These cognitive processes include complex attention, planning, higher-order processing, and problem solving [2, 3]. Despite the effectiveness of antiretroviral therapy (ART) in achieving viral suppression, cognitive impairment among people living with HIV (PLWH) persists despite suppressive ART [4,5,6,7]. Depending on age administered and HIV status, nutrient-based interventions, such as food or nutrient supplementation, may have unique developmental and cognitive benefits.

Food insecurity (FI) negatively impacts cognitive development and disproportionately affects young PLWH. The global prevalence of FI is particularly high among PLWH [8,9,10] and poses a significant threat to treatment outcomes including cognitive health and development [11, 12]. FI is defined by the World Health Organization (WHO) as experiencing hunger or running out of food at least once during the year and affects an estimated 30% of young people aged 15–18 years globally each month [13,14,15]. FI is associated with impaired cognitive development in young people, often observed via lower academic performance [16,17,18]. A specific joint effect of FI and HIV has been linked to greater cognitive difficulties among PLWH compared to people living without HIV (PLWoH) [19].

FI may lead to poor viral suppression and lower CD4 counts, lower ART adherence, macro and micronutrient deficiencies, treatment interruption, and poor mental health leading to HIV-related cognitive difficulty [20, 21]. Specifically, FI exacerbates inflammation, a key component in neurocognitive difficulty for PLWH [22]. This compounding burden of HIV and FI on cognitive function has been demonstrated among adults [23, 24], however, it is unclear whether and how neurodevelopmental cognitive trajectories may be impacted by FI and HIV status in younger people. Moreover, past evidence suggests that cognitive difficulties are more pronounced among young PLWH living in low-income countries [4], which also experience the highest prevalence of FI [13], making better understanding these relationships a global health priority.

Using data from the African Cohort Study (AFRICOS), we assessed differences in longitudinal cognitive performance across ages among young PLWH on ART and PLWoH. We also assessed FI as an effect modifier between HIV and cognitive outcomes and assessed age-specific effects of FI on cognitive performance.

Methods

Study Cohort

AFRICOS is a multinational ongoing longitudinal cohort study enrolling PLWH and PLWoH since January 2013 [25]. Participants were eligible for inclusion if they were ≥ 18 years of age (≥ 15 years of age after 2019), consented to collection of data and biologic specimens, and received ongoing HIV care at one of the recruitment sites (PLWH only). PLWoH were recruited among persons in the community seeking HIV testing at the clinic sites, including a small number of those who had participated in research studies previously. Recruitment occurred across 5 programs and 12 government-run, hospital-based HIV clinics based in Kayunga, Uganda; South Rift Valley, Kenya; Kisumu West, Kenya; Mbeya, Tanzania; and Lagos and Abuja, Nigeria. These programs are supported by the U.S. President’s Emergency Plan for AIDS Relief (PEPFAR) and deliver HIV prevention and treatment services in collaboration with the U.S. Military HIV Research Program (MHRP). Participants were excluded if they were pregnant at enrollment or presented with any other condition that may compromise study procedures or conduct.

At enrollment and every 6 months, participants completed standardized demographic and behavioral questionnaires, medical history report, physical examination, and phlebotomy. A medical chart review was also conducted by trained clinical staff. An HIV-tailored neuropsychological battery was administered annually. All data collection procedures were conducted using harmonized protocols adapted and translated to each recruitment site.

AFRICOS was approved by the Institutional Review Boards of the Walter Reed Army Institute of Research, Makerere University School of Public Health, Kenya Medical Research Institute, Tanzania National Institute of Medical Research, Nigerian Ministry of Defense, and all collaborating institutions. All participants provided written informed consent or assent prior to any study procedures.

Measures

Annual neuropsychological testing comprised a 30-minute battery administered by staff who underwent re-certification every 6 months for administration consistency across study sites. Staff were also trained to indicate whether neuropsychological testing results were valid and complete. Tests included the WHO Auditory Verbal Learning Test (AVLT), Trail Making Test Part A (TMT-A), Color Trails 1 & 2 (CT1 & CT2), and Verbal Fluency (VF) [26]. Total recall on the AVLT measures verbal learning and memory and was calculated as the sum of the correctly recalled words out of 15 across 5 immediate recall trials and a 30-minute delayed recall trial. Time to completion in seconds on the TMT-A and CT1 measures processing speed. Time to completion in seconds on the CT2 measures set-shifting and executive functions, while total number of named action words (verbs) in 60 seconds on the VF test measures language generativity. The AVLT, TMT-A, and VF were included from study start (2013), while CT1 and CT2 were added in October 2017. All neurocognitive data reflected raw test scores.

Participants responded “yes” or “no” to the question: “Have you had enough food to eat over the past 12 months?” and self-reported the number of meals per day on average. Replicating prior categorization of food insecurity in AFRICOS [27], FI was defined as having < 3 meals per day, not having enough food to eat over the past 12 months, or both.

At each visit, HIV serostatus was confirmed via phlebotomy for PLWoH and viral load was collected among PLWH [28]. For PLWH, self-reported vertical transmission of HIV was also recorded. Demographic data included: age, sex, self-reported ability to read and write (literacy), level of educational attainment, number of persons living in the participant’s household, and weekly household income. Participants also self-reported having ever been diagnosed with tuberculosis, which has high prevalence among PLWH and associations with cognition [29, 30].

Statistical Analyses

We included visits with valid neuropsychological testing results among participants aged 15–27 to specifically model cognitive development during late adolescence and early adulthood. For PLWH, data from visits after at least six months of ART were included to remove influence of ART on cognitive changes after initiating ART. Because the CT1 and CT2 tests were added to the study in 2017, the analytic cohort for CT1 and CT2 was reduced to only include participants who had completed these tests since 2017.

Education was categorized as “none or some primary,” “completed primary or some secondary,” “completed secondary or higher,” and “some vocational or completed vocational.” Household size was included as a binary variable using a threshold of ≥ 5 persons living in the household based on the median response across the entire cohort. Poverty-level household income was binarized using the threshold of 25th percentile of the reported weekly household income by currency type, and a missing value indicator was included.

Descriptive statistics were performed using chi-squared test of independence and analysis of variance (ANOVA) for categorical and continuous variables, respectively. Comparisons were made by HIV status using enrollment visits for PLWoH and first available neuropsychological testing visits for PLWH after being on ART ≥ 6 months. Inferential effects of HIV and FI on each of the cognitive outcomes were estimated using linear mixed models (LMMs) using the Nelder-Mead optimizer [31]. For model building, we tested within-subject random effects and quadratic terms for age in nested models using chi-square likelihood ratio tests, and if supported, these effects were included in final models.

Model Set 1 included HIV status and time-varying FI as independent primary predictors with sex, program site, literacy, education, poverty-level income, household size, and self-reported history of tuberculosis as covariates. Model Set 2 tested the interaction term between FI and HIV status with the same covariates. Model Set 3 tested interactions between FI and age at visit and HIV status and age at visit while controlling for the same covariates. All analyses were performed in R (version 4.2.2) using the lme4 package [32].

Results

Demographic Characteristics

As of October 2023, 4,137 (PLWH n = 3380, (81.7%); PLWoH n = 757, (18.3%)) participants were enrolled in AFRICOS (Fig. 1). The final analytic population included 1,012 participants: 741 PLWH (73.2%) and 271 PLWoH (26.8%). Overall, there were more females than males (n = 575, 56.8%). The largest age category was 21–24 years of age (n = 336, 33.2%), and half completed primary or some secondary education (n = 497, 49.1%). Almost all participants reported being able to read and write (n = 981, 96.9%). Most participants had completed two or more annual visits with neuropsychological testing (n = 702, 69.4%). Among PLWH, the majority were virally suppressed at first available neuropsychological testing (n = 654, 88.3%) with over half having a CD4 count ≥ 500 (n = 443, 60.7%). Approximately half reported vertical transmission of HIV (n = 389, 52.5%). Approximately half of all participants reported having experienced FI across follow-up (n = 476, 47.0%).

Fig. 1
Fig. 1
Full size image

Sample selection flowchart. PLWH, People living with HIV; PLWoH, People living without HIV; NP, Neuropsychological; CT1, Color Trails 1

Table 1 describes differences between PLWH and PLWoH on variables of interest at the first available neuropsychological test, or first test after 6 months on ART for PLWH. PLWH were slightly younger compared to PLWoH (Mean = 20.9, Standard Deviation = 3.6 vs. M = 21.6, SD = 3.6; F (1, 1010) = 8.28, p = 0.004) and a greater proportion were female (59.1% vs. 50.6%; χ2 (1) = 5.92, p = 0.015). A higher proportion of PLWoH completed only one visit compared to PLWH (47.6% vs. 24.4%; χ2 (2) = 50.16, p < 0.001).

Table 1 Participant characteristics at the first neuropsychological testing visit+ stratified by HIV status
Full size table

Reported household income ≤ 25th percentile was more common among PLWH as compared to PLWoH (48.4% vs. 38.4%; χ2(2) = 8.62, p = 0.013). Educational attainment was significantly different across groups with PLWH being more likely to have no or some primary education (21.6% vs. 13.7%; χ2(3) = 18.31, p < 0.001). PLWH were also more likely than PLWoH to report a history of tuberculosis (11.1% vs. 3.3%; χ2(1) = 14.55, p < 0.001). There were no differences by HIV status on ability to read and write, household size, and history of food insecurity.

Differences in Cognitive Performance by HIV Status at First Neuropsychological Test

At first available neuropsychological test, PLWoH had better performance than ART-treated PLWH on all cognitive tests (Table 1). On the test of verbal learning and memory (AVLT), PLWoH recalled more words on average compared to PLWH (M = 17.4, SD = 4.6 vs. M = 16.3, SD = 4.7; F (1, 1010) = 9.74, p = 0.002). PLWoH had faster completion time in seconds on both processing speed tasks than PLWH. These included statistically significant differences on the TMT-A (PLWoH: M = 57.9, SD = 24.9 vs. PLWH: M = 66.3, SD = 32.3; F (1, 1010) = 14.83, p < 0.001) and CT1 (PLWoH: M = 60.3, SD = 20.6 vs. PLWH: M = 68.8, SD = 30.6; F (1, 687) = 10.63, p = 0.001). On a task of verbal fluency (VF), PLWoH named more action items compared to PLWH (M = 11.4, SD = 3.82 vs. M = 10.3, SD = 3.68; F (1, 1010) = 14.44, p < 0.001). PLWoH also completed the task of set-shifting and executive functions (CT2) more quickly compared to PLWH (M = 115.0, SD = 33.7 vs. M = 131.0, SD = 40.1; F (1, 686) = 18.98, p < 0.001).

Effects of Age, HIV Status, and FI on Cognitive Trajectories

Age-Only Models

In age-only LMMs, within-subject random slopes for age were supported in all models except VF. Quadratic terms for age were not significant in any model and therefore were not included in subsequent analyses. In all age-only models, greater age was associated with better performance on all neuropsychological tasks, except for language generativity (VF: β=−0.04 [95%CI: −0.08, 0.01], p = 0.128). These included verbal learning and memory (AVLT total recall: β = 0.10 [95%CI: 0.04, 0.16], p = 0.001), processing speed (TMT time: β=−1.19 [95%CI: −1.56, −0.82], p < 0.001; CT1 time: β = −0.63 [95%CI: −1.06, −0.19], p = 0.005), and executive functions (CT2 time: β=−0.94 [95%CI: −1.56, −0.32], p = 0.003).

Independent and Interactive HIV Status and FI Models

Subsequent LMMs estimated independent effects of FI and HIV status for each neuropsychological outcome, controlling for covariates. In Model Set 1 (Table 2) PLWH had lower performance in language generativity and were slower to complete executive functioning tasks across visits, including VF (β=−0.47 [95%CI: −0.89, −0.04], p = 0.032) and the CT2 (β = 8.10 [95%CI: 2.54, 13.67], p = 0.004) compared to PLWoH. Food insecurity as an independent predictor was not associated with longitudinal neuropsychological testing in these models. In Model Set 2 (Table 3), there were no significant HIV by FI interactions on any of the examined cognitive outcomes. In sensitivity analyses comparing PLWoH to PLWH with viral load < 1000 copies/mL and PLWH with viral load ≥ 1000 copies/mL, results were largely the same. However, PLWH who were not virally suppressed performed significantly worse than PLWoH on TMT-A (β = 7.07 [95%CI: 2.62, 11.52], p = 0.002), and while PLWH generally had lower performance on VF, when assessing separate effects for PLWH who were and were not virally suppressed compared to PLWoH, neither reached significance at the 0.05 level (β=−0.50 [95%CI: −1.10, 0.09], p = 0.098; and β=−0.42 [95%CI: −0.85, 0.01], p = 0.054, respectively).

Table 2 Model Set 1: Associations between HIV status and FI on neuropsychological outcomes adjusted for covariates
Full size table
Table 3 Model Set 2: Interaction between HIV status by FI on neuropsychological outcomes adjusted for covariates
Full size table

Interactive HIV Status by Age and FI by Age Models

In Model Set 3 (Table 4) interaction terms of HIV status by age at visit and FI by age at visit were tested. There was a significant FI by age interaction for AVLT total recall (βFI*age = 0.17 [95%CI: 0.07, 0.27], pFI*age=0.001), indicating that food insecure participants performed worse on a verbal memory test compared to their food secure counterparts, specifically at younger ages. These differences were attenuated at older ages. Model predicted trajectories for AVLT total recall stratified by HIV status are presented in Fig. 2. Predicted slopes demonstrate that those with FI had a greater increase in AVLT total score across ages, regardless of HIV status. We also found a significant FI by age interaction on CT2, a test of executive functions (βFI*age = 1.63 [95%CI: 0.56, 2.70], pFI*age=0.003). These results indicated that FI was associated with poorer executive functioning over time, regardless of the HIV status. Importantly, these negative effects of FI on executive performance increased in magnitude over time as illustrated in Fig. 3. While PLWH had slower time to completion on CT2 overall, across both PLWH and PLWoH, FI reflected slower completion times as age increased. Results did not differ when PLWoH were compared to PLWH who were virally suppressed and to PLWH who were not virally suppressed in sensitivity analyses.

Table 4 Model Set 3: Interaction between HIV status by age at visit and FI by age at visit on neuropsychological outcomes adjusted for covariates
Full size table
Fig. 2
Fig. 2
Full size image

Predicted slopes for time-varying food insecurity status on Auditory Verbal Learning Task (AVLT) total recall across age stratified by HIV status. Food insecure participants exhibited poorer verbal learning and memory performance than their food secure counterparts at younger ages, and these effects were attenuated with increasing age. PLWH, People Living with HIV; PLWoH, People Living without HIV

Fig. 3
Fig. 3
Full size image

Predicted slopes for time-varying food insecurity status on Color Trails 2 Task time to completion in seconds across age stratified by HIV status. Food insecure participants showed worse executive functioning performance than their food secure counterparts, and the magnitude of this effect increased with age. PLWH, People Living with HIV; PLWoH, People Living without HIV

Discussion

We examined longitudinal cognitive trajectories and their associations with HIV status and FI in a large multinational African cohort of 15–27-year-old PLWH and PLWoH. We observed longitudinal gains in cognitive performance across age, consistent with current theories of continued cognitive development in late adolescence and young adulthood [2]. Nevertheless, PLWH on ART consistently exhibited poorer cognitive performance compared to PLWoH across all ages, with no interaction between HIV status and age. We also did not observe effect modification between HIV status and FI on cognitive trajectories. However, FI was negatively associated with verbal memory at younger ages and exerted an increasingly negative effect on executive functioning across older ages. Overall, while living with HIV was consistently related to poorer cognitive performance across ages, FI may have unique associations with cognitive functioning depending on the developmental window of exposure.

No difference in performance across age by HIV status was observed, suggesting that the impact of HIV on cognition persists even among participants who had received at least 6 months of ART. Similar patterns were demonstrated among virally suppressed women with and without HIV aged 25–77 years in the Women’s Interagency HIV Study. In this study, virally suppressed women living with HIV had consistently poorer cognitive performance over four years compared to women living without HIV with no interactions by age [33]. Another longitudinal study of adolescents 8–18 years of age demonstrated similar patterns between perinatally exposed youth living with HIV and matched controls without HIV [34]. Specifically, youth living with HIV exhibited worse performance across multiple cognitive tests than matched PLWoH at baseline and 4 years later, and as in our study, persisting consistent across age [34]. These continued impacts of HIV on cognitive performance reinforce the critical importance of primary prevention strategies and early ART interventions.

Previously published data from participants of all ages in AFRICOS demonstrated no differences in FI prevalence by HIV status [27]. We replicated these findings in the present analyses that included only late adolescent and young adult participants. Interestingly, we did not observe attenuation or exacerbation of HIV effects on cognitive trajectories by FI, which is somewhat unexpected given prior reports from independent cohorts. At the same time, many studies that identified exacerbating effects of FI on cognition in PLWH included participants ≥ 40 years of age in contrast to our younger cohort [12, 19, 22]. Thus, while we did not find significant interactions between FI and HIV status on longitudinal cognitive performance in this study of late adolescents and young adults, additional studies in younger and older cohorts are needed to better understand relationships among HIV, FI, cognitive development and aging. Additional research could also consider the role of social adversity or school dropout as mediators between FI and cognitive outcomes for young people.

We also found age-specific effects of FI on cognitive trajectories in both PLWH and PLWoH. Regardless of HIV status, food insecure participants showed poorer performance on tests of verbal learning and memory at younger ages and increasingly poorer executive functions at older ages. These findings suggest FI in late adolescence and young adulthood may compromise development and maturation of distinct cognitive functions depending on the age and duration of exposure. Our findings are consistent with and contribute to the growing body of literature on the unique effects of FI across neurodevelopment. For example, a study of Mexican adults > 50 years assessed how early life FI may be associated with late-life cognitive performance [35]. FI before age 10 was associated with poor verbal learning but not visual scanning tasks, while experiencing FI in the past 2 years was associated with worse visual scanning but not verbal learning [35]. Furthermore, among Medicare beneficiaries in the United States ≥ 65 years of age in the National Health and Aging Trends Study (NHATS), FI was associated with negative changes in executive functions but not immediate and delayed memory [36]. In another study of Uruguayan children 6–8 years of age, a nutrient-dense diet (i.e., dark leafy vegetables, eggs, beans, and peas) was associated with better reading but not math scores [37]. While not well established, it is possible that exposure to food insecurity at different stages of development may have differential effects on memory and executive function. These patterns of verbal- and memory-specific effects of experiencing FI at younger ages and executive function-specific effects of experiencing FI at older ages highlight the need for development of domain-specific interventions and cognitive health support programs among people with FI, especially during periods important for neurodevelopment and cognitive maturation. Support programs might include cognitive screening, school feeding programs, or cash transfers to aid those with neurocognitive delay [38].

A strength of this work is sample size and longitudinal data among young PLWH and PLWoH targeting an important neurodevelopmental stage. However, our study has several limitations. Importantly, learning effects may be present in this test-naive cohort. As our analyses focus on the primary effects of HIV, FI, and age across visits, however, the observed differences are likely not completely attributable to learning effects since comparison groups have similar frequency of exposure to the testing procedures. The outcomes assessed in these analyses may also be strongly correlated; future work may consider a multivariate analysis that models the correlations between these outcomes. We were also unable to assess how HIV acquisition at varying developmental windows may impact cognitive development. Future studies might include age- and sex-matched controls to better assess effects of HIV acquisition across developmental windows. Additionally, while we reported comparisons between PLWH and PLWoH, PLWoH enrolled in AFRICOS are not clinical controls and are typically recruited via other testing services provided at PEPFAR clinics. Also, our operationalization of FI was based on self-reported FI and does not consider variations in exposure duration or severity. More rigorous dietary recall or food logging may provide more accurate estimates of nutrient-specific effect modifiers for HIV and cognitive outcomes. Finally, there may be potential for differential loss to follow-up bias by FI. However, given the effects we have seen here, additional follow-up data for those with FI may only strengthen observed associations.

Conclusions

Despite ART experience, young PLWH consistently demonstrated cognitive weaknesses across ages compared to PLWoH. This suggests that additional cognitive support and interventions may be needed for young PLWH. Neuropsychological monitoring may elucidate which interventions are best suited for cognitive aid. Specifically, our findings suggest that timing of nutrient-based interventions may have unique impacts on cognitive domains depending on age. Our findings provide support for primary prevention strategies for HIV coupled with targeted interventions addressing food insecurity in people living with and without HIV.