Cannabis use frequency may differentially impact cognitive performance in young adults with and without HIV

Introduction

Use and acceptance of medical and recreational cannabis have expanded in the United States, alongside increased legalization at the state level (1, 2). Cannabis use is most common among young adults (age 18–25), with 24.1% reporting use in the past month and 14.4% meeting diagnostic criteria for cannabis use disorder in 2021 (3), raising concerns about possible harm, including impacts on cognitive functioning (4). At the same time, public perception of harm related to cannabis use has decreased (5), and there is evidence that cannabis can be beneficial for managing pain, sleep issues, and stress (6–8). Furthermore, several different compounds in cannabis, including tetrahydrocannabinol (THC) and cannabidiol (CBD), are recognized as having anti-inflammatory properties which may be beneficial (9, 10). In this complex context, study outcomes are influenced by individual differences in motivation for use, variation in product properties (e.g., content of THC, CBD), and route of administration, as well as use characteristics such as frequency. It is critical to better understand the potential consequences of use, particularly within specific populations that may be more susceptible to negative or positive cognitive impacts. Given these complexities, we examined the impact of cannabis use frequency on cognitive performance in young adults, some living with HIV infection, to improve understanding of the potential benefits of cannabis among young individuals living with a chronic neuroinflammatory disease. These data may instruct future studies addressing tailored guidelines for cannabis use among populations experiencing neuroinflammation.

Research on the cognitive effects of chronic cannabis use in adolescents and young adults, i.e., differences observed outside the period of acute intoxication, has reported negative impacts across cognitive domains including learning and memory, attention, processing speed and executive functions (4, 11–14). Dellazizzo and colleagues, in a review of studies of adolescents and adults (4), noted that verbal, but not visual, learning and memory and executive functions were the cognitive domains most consistently demonstrating negative effects of cannabis use, while attention and processing speed outcomes were more variable. Recent studies and reviews document poorer performance associated with frequent or heavy use in learning, attention, executive functions, memory and processing speed (4, 11, 13, 15–17). Frolli and colleagues (11) specifically showed a dose-dependent relationship with heavy chronic users showing significantly worse performance in working memory, processing speed, IQ and visuospatial reasoning relative to occasional or non-users. Overall, heavy use appears to have negative consequences on learning and memory, executive functions, and processing speed, although there are suggestions that mild or moderate use may have fewer consequences.

Given the influence cannabis use has on neuroinflammatory pathways (18), there is potential for some positive impacts on brain health for people living with HIV infection (PWH). Even in the current era of effective systemic viral suppression with antiretroviral medications, HIV affects central nervous system (CNS) functioning in up to 40% of individuals and has the potential to disrupt CNS development (19). Cognitive effects of HIV are most commonly observed in domains considered to represent more “fluid cognition”, including episodic and working memory, executive functions, and processing and psychomotor speed (20–22). The ongoing impact of HIV on the CNS is hypothesized to be mediated by gut dysbiosis and immune activation, established early in infection, and resulting chronic neuroinflammatory processes (18, 23–25), including the release of inflammatory cytokines (23, 24), that contribute to cognitive deficits. Plasma markers of immune activation and cytokines (e.g., sCD163, interleukin-6) are elevated in HIV and associated with neurocognitive dysfunction in adults (18) and youth (25).

Cannabis has anti-inflammatory effects in general through the endocannabinoid system, with cannabinoid receptors densely represented in the immune system and, crucially, in the brain, presenting possible benefits of use against chronic neuroinflammation [see (18) for a review of mechanisms]. Cannabis, then, has the potential to alleviate the cognitive impacts of HIV infection through altering inflammatory pathways and reducing associated neuroinflammation (26). The effects of cannabis on neuroinflammation may vary by cannabis compound (e.g., THC, CBD, terpenes) (18). For example, prior reports include findings that THC can affect macrophage susceptibility to HIV during human monocyte differentiation (25), while CBD has been shown to affect inflammatory gene expression in PWH (27).

Cannabis use is common among PWH (28), who use it to manage symptoms such as pain, anxiety, and sleep issues, as well as using recreationally similar to those without HIV. Findings related to the effects of cannabis on cognitive functioning among PWH have been mixed, however. Studies of adults have found both negative (29–32) and positive (33, 34) associations of cannabis use with cognitive performance in PWH. Notably, Ayoub and colleagues’ (35) recent review identified better performance in executive functions, risk-based decision making, attention, processing speed, and verbal fluency among cannabis-using PWH compared to non-users. In a large cohort of 18–70 year old adults with and without HIV and current cannabis use, people with HIV and cannabis use were less likely to be classified as cognitively impaired and had better learning and verbal fluency than non-using people with HIV (36). In contrast, a recent study (37) found that cannabis use (grams/past 30 days) was linearly associated with lower global cognitive functioning among PWH but not those without HIV (group mean ages 33–40 years). In their review, Ayoub and colleagues noted several factors that may contribute to discrepancies among study outcomes, including differences in demographics (e.g., age) and clinical characteristics of participants; changing cannabis products and usage over time; and varying or unclear cannabis use definitions. Crucially, few studies have focused on young adults with HIV (i.e., ages 18–24; YWH) to explore this potential interaction between HIV infection and cannabis use, despite their high rates of cannabis use and their continued CNS neurodevelopment. Furthermore, most studies have focused only on high levels of use, and few have included mild/moderate users with the goal of establishing harmful vs. protective use levels among people with HIV, including young adults.

This study begins to fill this gap by addressing the effects of varied frequency of cannabis use on cognitive performance in young people with (YWH) and without (YWoH) HIV infection in the young adult age range (ages 18–24). We hypothesize that, while all young adults may be vulnerable to negative impacts of frequent cannabis use due to ongoing neurodevelopment, for those living with HIV, some cannabis use may at least partially offset the negative impacts of HIV through a reduction in inflammation. Particularly within the context of HIV infection, a chronic inflammatory disease, effects of cannabis use may be dose dependent and vary by cognitive domain, with differential outcomes relative to those living without HIV infection. We examined the association of no, mild/moderate, and heavy cannabis use frequency with cognitive performance among young adults, and whether the presence of HIV infection alters this association. These analyses focus on fluid cognition domains that have been implicated as vulnerable to both HIV and cannabis: verbal learning and memory, processing speed, working memory, and executive functions.

Materials and methods

Participants

Young adults (ages 18–24) were recruited in San Diego and Los Angeles counties between 2020 and 2024. Participants with HIV infection were referred from community and clinical HIV treatment sites, with uninfected participants being recruited through service providers to at-risk youth, HIV testing venues, community outreach, and peer referral. Exclusion criteria included more than weekly non-cannabis substance use (methamphetamine, cocaine, and heroin), residing in a drug treatment facility, pregnancy, and neurological disorders including concussion with more than five minutes of unconsciousness. HIV must have been acquired behaviorally, with perinatally acquired HIV excluded due to likely differences in cognitive profiles. Participants had to be prescribed treatment at the time of participation, but viral suppression was not required, to reflect typical YWH. Participants were asked about cannabis use frequency at the time of recruitment to ensure that a range of use was represented. Complete data were available for 99 individuals, 38 YWH and 61 YWoH.

All study protocols were approved by the University of California, San Diego and Children's Hospital Los Angeles Institutional Review Boards and followed the Declaration of Helsinki. Informed consent was obtained from all participants.

General procedure

Study measures consisted of substance use questionnaires and interviews, cognitive tests, and self-report questionnaires assessing demographics and mental health to characterize the sample and capture a range of considerations relevant to young adult populations, including those living with HIV. This battery was typically administered during two visits to reduce burden. A urine toxicology was performed for all participants at the start of each visit to confirm that there were no methamphetamines, other amphetamines, cocaine, or opioids in their system. Participants who failed the urine toxicology were informed of their positive results and rescheduled for another day. Information on time of last use of cannabis, alcohol or medications that might affect cognitive performance (e.g., allergy medications) was collected via self-report. Participants were rescheduled as needed to allow for assessments to be performed without acute influence based on this information and any additional examiner observations of intoxication. Because this study took place during and after the COVID-19 pandemic, COVID mitigation procedures (e.g., masking, plexiglass barriers) were enacted and remained in place for the entirety of the study.

Measures

Substance use

Cannabis and other substance use was assessed with the Timeline Follow-Back (TLFB), capturing days of use for the previous 30, 180 and 365 days (38). Participants were asked on which days in the preceding time period they used cannabis and other substances, using a calendar and anchoring events to facilitate recall. For these analyses, cannabis use was categorized as no use, mild/moderate use (1–<5 days used per week), and heavy use (≥5 days used per week) for each time period examined. The “heavy” use category included daily and near-daily use, consistent with prior literature (39). Participants were also asked how many times they used a substance on days that they used. Days, rather than times, of use was utilized as the primary grouping measure due to likely greater accuracy of recall.

The MINI diagnostic interview (40) for Cannabis Use Disorder and Alcohol Use Disorder was administered for descriptive purposes. The MINI is a short, structured diagnostic interview with sections that screen for cannabis and alcohol use disorders. These sections assess the presence and severity of substance use disorder symptoms as defined by the DSM-V (41), including problems related to use, tolerance, withdrawal, and cravings.

Cognitive functioning

There is considerable overlap in the cognitive domains implicated in both the cannabis and HIV related literature. Therefore, standardized cognitive measures appropriate for young adults were selected to focus on the main domains identified as vulnerable to the effects of cannabis and HIV on the brain: Verbal Learning and Memory, Working Memory, Executive Function (inhibition, problem solving and cognitive flexibility), Processing Speed, and overall Fluid Cognition. Within a given test or battery, some scores reflect more than one overlapping domain (e.g., verbal learning during the initial trial also reflects working memory), therefore, some tests are referenced under different domains.

Scores were analyzed individually, and the statistical approach performed false discovery rate (FDR) correction within domains or functions for our primary analyses when more than one variable was included. We did include one widely used composite measure of fluid cognition from the NIH Toolbox, which did not undergo FDR correction. Cognitive measures and the specific scores used are detailed within each cognitive domain as described below and in Table 1. Analyses used scores standardized by age, where standard scores have a mean of 100 and standard deviation (SD) of 15; scaled scores have a mean of 10 and SD of 3; and t-scores have a mean of 50 and SD of 10.

Table 1. Standardized tests and specific outcome variables included within each cognitive domain.

CVLT-3, California Verbal Learning Test-Third Edition; WMS-IV, Wechsler Memory Scale-Fourth Edition; NIHT, NIH Toolbox; D-KEFS, Delis Kaplan Executive Function System.

Cognitive domains presented with subtests that reflect underlying cognitive processes described in Methods and Results.

All NIHT subtests in Table 1 contributed to Fluid Cognition Composite along with an additional NIHT episodic memory subtest.

Verbal learning and memory domain

Verbal learning and memory were assessed with two measures to capture functions related to less structured, theoretically more effortful, word lists as well as structured story content. Three cognitive functions within the verbal learning and memory domain were analyzed separately: learning (immediate recall), retention (delayed recall), and recognition memory (discriminabilitvy, which reflects encoding and storage without retrieval demands).

California Verbal Learning Test, Third Edition [CVLT-3; (42)]. The CVLT-3 measures word list learning (immediate recall), retention (delayed recall), and recognition memory. On the learning trials (Trials 1–5), participants hear a list of words drawn from four semantic categories and are asked to freely recall the same word list, in any order, over five trials. The standard score for recall of Trials 1–5 is included in Verbal Immediate Recall (see Table 1). Next, a new distractor word list is presented, and participants are asked to recall these words. The initial word list is then queried again for free recall followed by category-cued recall (short delay free and cued recall). Following a 20 min delay, participants are again asked to freely recall the first list of words and then asked to recall the words with category cues (long delay free and cued recall). Both short and long delay free and cued recall of these word lists contribute to the Verbal Delayed Recall measure. Recognition memory is also assessed by presenting both target (first list), distractor (second list), and novel words while querying (yes/no) whether each word was from the first list. The Recognition Discriminability score facilitates detection of retrieval vs. encoding deficits.

Wechsler Memory Scale-Fourth Edition [WMS-IV; (43)] Logical Memory subtest. The Logical Memory subtest of the WMS-IV assesses learning and retention with immediate and delayed recall conditions, respectively, for two stories. During the immediate recall section, participants are asked to recall everything they can remember about each story. Following a 20 min delay, participants are again asked to recall everything that they can from the two stories. The scaled score for immediate story recall (Logical Memory I) was included in the Verbal Immediate Recall function and delayed story recall (Logical Memory II) in the Verbal Delayed Recall function.

Working memory domain

Working memory refers to holding a limited amount of information in active storage to enable retention and manipulation for more complex processes such as decision-making. Two measures were administered to capture facets of verbal and visual working memory:

CVLT-3. The standard score for word list learning after exposure to the first trial of a targeted word list (CVLT-3 Trial 1) was used as a measure of immediate word list recall.

NIH Toolbox [NIHT; (44)]. The NIHT is an iPad-administered, standardized battery of cognition, motor functioning, sensation and emotion. Age-corrected standard scores for NIHT subtests were used. The NIHT List Sorting subtest captures short-term cognitive storage and manipulation of nonverbal information. Participants are presented with a series of pictures of objects varying in size and category and required to recall them in first size and then size by category order (e.g., animals from smallest to largest, then fruits from smallest to largest).

Processing speed domain

NIH Toolbox. The NIHT Pattern Comparison subtest measures processing speed by requiring the participant to determine as quickly as possible whether two pictures are identical.

Delis-Kaplan Executive Function System [D-KEFS; (45)]. The D-KEFS is a battery of tests designed to measure executive functioning in individuals age 8–89 years. Age-standardized scaled scores from two subtests were used in this study. The Number Sequencing and Letter Sequencing trials of the Trail Making subtest assess visuomotor processing speed and require a participant to connect circles in either numerical or alphabetical order as quickly as possible. Scaled scores for times to complete the two trials were combined into a Non-Switching Trials composite score used as a measure of Processing Speed.

Executive functions domain

Subtests from two gold standard measures were used to capture a range of executive functions, including inhibition, cognitive flexibility, problem solving and set shifting.

NIH Toolbox. Two NIHT subtests assess specific facets of executive functioning. First, the Flanker subtest measures inhibitory functioning by requiring participants to quickly indicate the direction (right vs. left) in which an arrow points. The arrow is located in the center of a row of arrows that point either in the same (congruent) or opposite (incongruent) direction as the center arrow. The unpredictable inhibitory requirement of the incongruent trials results in slower responding for individuals with poorer executive functioning; speed of responding serves as the primary measure from this subtest. Second, the Dimensional Change Card Sort subtest measures cognitive flexibility and problem solving by requiring participants to match a series of picture pairs to a target picture by a varying criterion (shape or color).

D-KEFS. The Trail Making subtest contains several components, allowing psychomotor speed to be parsed from the added cognitive demands of switching mental sets. The Number Sequencing and Letter Sequencing trials are described under Processing Speed above. The Number-Letter Switching trial of this subtest measures cognitive flexibility by asking participants to connect circles in alternating alphabetical and numerical sequences (e.g., 1—A—2—B) as quickly as possible.

The Category Switching trial of the D-KEFS Verbal Fluency subtest measures verbal cognitive flexibility and set shifting by requiring participants to switch between two semantic categories while generating words as quickly as possible.

Fluid cognition domain

NIH Toolbox. The NIHT provides a Fluid Cognition composite score that reflects overall capacity for new learning and information processing in novel situations and is calculated using subtests measuring executive function and cognitive flexibility (Flanker and Dimensional Change Card Sorting), attention (Flanker), episodic memory (Picture Sequence), working memory (List Sorting), and processing speed (Pattern Comparison). This composite score was included as an integrative measure of overall fluid cognition.

Demographics

Demographic, medical, mental health and educational history was gathered via self-report. Participants’ engagement in education past high school was examined as a covariate in analyses. For those with HIV, HIV history was collected via chart review.

Financial Resources Scale [FRS; (46)] includes 30 self-report items that assess participants’ access to resources for basic needs such as food, housing and utilities, benefits including employment and healthcare, social needs/self-care such as the ability to spend time together as a family, and extra resources (47). The FRS uses a 5-point scale (1 = not at all adequate to 5 = almost always adequate). The total score was used as a covariate to account for economic resources and access.

Mental health

Achenbach System of Empirically Based Assessment [ASEBA; (48)] Adult Self Report. The ASEBA Adult Self-Report evaluates behavioral and emotional symptomatology for a range of internalizing and externalizing disorders and DSM-V diagnoses. The instructions were modified by allowing participants to use their current gender identity for any gender-specific questions. The Anxious/Depressed scale t-score was used as a covariate in adjusted analyses to account for the potential influence of anxiety and depression on cognitive performance.

Statistical methods

For the descriptive characteristics, we provide mean (standard deviation) for continuous variables and count (percent) for categorical variables (Table 2). Frequency of Alcohol and Cannabis Use Disorders were compared between those with and without HIV using Fisher's Exact Test. We used semiparametric generalized linear models (GLM) for modeling associations of HIV status and cannabis use frequency with cognitive outcomes. Unlike parametric GLM, semiparametric GLM imposes no mathematical model such as normal distributions on the distribution of the response (dependent variable) of GLM with inference based on a set of generalized estimating equations (GEE) and thereby provides valid inference for virtually all data distributions that arise in common practice (49).

Table 2. Participant characteristics.

Group comparison significance is derived from GLM models with cannabis category and reflect FDR correction within cognitive domain; significant p values are reported in bold; if p > .05, differences are noted as not significant (ns).

“Does not identify” was used to accommodate participants who preferred to identify their race as Hispanic/Latino, a race category absent from standard NIH race categories. Native Am/Pac Isl = Asian/Native American/Pacific Islander.

Anxious/depressed T score is an outcome of the ASEBA.

We first examined cannabis use as an ordinal variable with three levels of categories described above; given a broad but sparse range of use across the mild to moderate use frequency range, we created a single category for mild-to-moderate use to increase power. Given these distributional qualities of our data, we considered categorical analysis more robust and report it as our primary analyses. As secondary analyses, we explored cannabis use as a continuous variable using the same models as above, despite the potential limitations of the broad middle range of use. To test for interactive or synergistic effects, all models included HIV status, cannabis use frequency, and a cannabis use frequency by HIV status interaction; all main effects of HIV and cannabis use frequency (both categorical and continuous) reported herein are those tested within the full models including the interaction. The primary analyses of 30-day cannabis use as a categorical variable used false discovery rate correction within the cognitive domains defined in the Methods section. We also report secondary analyses of 180- and 365-day use.

Finally, we performed exploratory analyses of linear and non-linear associations between continuous measures of cannabis use and cognitive performance. That is, we examined whether or not the pattern of performance became worse, for example, with greater cannabis use frequency in a linear fashion, or whether there might be an inflection point with best performance in YWH with mild/moderate use. To account for non-linear relationships when cannabis use was used as a continuous variable, we also used quadratic or cubit functions, if such a non-linear relationship was suggested by locally estimated scatterplot smoothing (LOESS) curves, and again tested synergistic effects by examining interactions of cannabis and HIV. To help interpret any interaction effect, we also performed analyses by stratifying by HIV group.

To control for covariates, we initially included all potential confounders that influence cognitive performance and applied backward elimination to remove them in the order of language (English as first language), mental health variables of anxiety/depression (T score), FRS 30 score, and education (beyond high school or not) for each outcome, since this variable selection approach provides more reliable variable selection than other methods such as forward and stepwise variable selection (50). Because of the low percentage of YWH who were female at birth (5.3%) compared to YWoH (41%), sex differences could not be addressed. We report analysis results without controlling for any covariates under “Unadjusted models” and analysis results by controlling for the covariates under “Adjusted models” in the Results section. We used two-sided type I errors α = 0.05 for all analyses and controlled for multiple comparison using the false discovery rate (51) within cognitive domains.

Results

Sample characteristics

Participant characteristics are presented in Table 2. On average, both YWH and YWoH were 21 years old and had English as their first language (73% in both groups). Sex at birth differed significantly between groups (41% of YWH, 5.3% of YWoH were female at birth). The majority of participants described their race as White (32.1%) or chose not to identify a race using standard categories, preferring to use Hispanic/Latino; race did not differ significantly between groups. Significantly more YWH (65.8%) reported Hispanic ethnicity compared to YWoH (42.6%). Participants in the 30-day no, mild/moderate, and heavy cannabis use categories did not differ significantly in age, sex, race, or ethnicity.

Mean days of cannabis use (during the prior 30, 180 and 365 days) and distribution of cannabis use across frequency categories did not differ between YWH and YWoH (Table 2). Among participants reporting any lifetime cannabis use, 49% of YWoH and 61% of YWH reported using cannabis within the prior 3 days; this difference was not statistically significant, and this variable was not considered further. Mean age of initiating any cannabis use did not differ between groups (YWoH: 16.98 years, YWH: 16.18 years), nor did mean age of initiation of regular cannabis use (> 1x/week for 52 + weeks) where applicable (YWoH: 18.00 years, YWH: 17.64 years). Cannabis Use Disorder was present in 44% of YWoH and 50% of YWH, and Alcohol Use Disorder in 39% of YWoH and 34% of YWH; groups did not differ significantly, nor did 30-day cannabis use categories differ for Alcohol Use Disorder. For those who used these substances during the past 30 days, the number of days having used nicotine products (YWoH: 5.76, YWH: 10.69) or alcohol (YWoH: 3.78, YWH: 3.53) did not differ between groups. Groups did not differ significantly on co-use of cannabis and tobacco products over the prior 180 days. YWH were more likely than YWoH to have used, at some point in their lives, amphetamines, opiates, or cocaine, but groups did not differ in the number of days using those substances over the past 180 days. YWH were also more likely to have some lifetime use of benzodiazepines and MDMA.

The YWH group had significantly lower available financial resources (FRS; p = .04) and educational attainment beyond high school (p = .02) and higher ASEBA Anxiety/Depression t-scores, reflecting more problems (p = .025; Table 2) than YWoH, leading to the inclusion of these variables as covariates in adjusted analyses.

The majority of YWH (74%) were prescribed Biktarvy as their antiretroviral regimen, with others prescribed Cabenuva injectable, Genvoya, or Triumeq. Current viral load data were available for 37 participants; among these, 26 had undetectable levels, and 11 had levels of 20–100 copies/mL. The mean current CD4 count was 702 cells/mm³. The majority (59%) of the 34 participants with available data on CDC staging were Stage 2, reflecting chronic HIV infection. This generally reflects having had a CD4 count between 200 and 500 at some point since their diagnosis but not having experienced severe levels of immunosuppression (e.g., below 200 cells/mm³). Thirty-two percent were Stage 1 (asymptomatic), and 9% were Stage 3 (history of CD4 < 200 or an AIDS-defining diagnosis).

Unadjusted generalized linear models

Note that primary analyses included 30-day cannabis use as a categorical variable (no use, mild/moderate, and heavy use) with FDR correction within cognitive domain (see domains in Table 1). Analyses of similar models using cannabis use as a continuous variable, examination of 180- and 365-day use, and the continuous use trend analyses examining linear and non-linear cognitive performance trends across level of use were considered exploratory and did not undergo FDR correction. As noted above, all models included HIV status, cannabis use frequency (either categorical or continuous), and a cannabis use frequency by HIV status interaction; all main effects of HIV and cannabis use frequency reported below are those tested within the full models including the interaction.

HIV status effects

Standard scores for all cognitive domains are presented in Table 3 along with significance levels derived from unadjusted GLM models for 30-day cannabis use category by HIV status. Independent of cannabis use frequency, cognitive functioning was within the average to low average range across domains for YWH, with the exception of verbal immediate recall of stories (WMS-IV Logical Memory I), which fell within the mildly impaired range. Compared to YWoH, YWH had significantly lower scores for Fluid Cognition (β = −20.7, p = .01). Within the verbal immediate recall function, YWH had significantly lower performance for learning of word lists (CVLT-3 Trials 1–5; β = −17.4, p = .02) and word list retention (CVLT-3 Delayed Recall; β = −17.7, p = .02); the groups did not differ significantly on recognition memory or either immediate or delayed recall of the WMS-IV Logical Memory stories. YWH also performed lower for both measures of processing speed [NIHT Pattern Comparison (β = 18.7, p = .04); Trail Making Non-Switching Trials (β = −2.6, p = .04)]. They also had worse performance on two measures of executive functions reflecting inhibition (NIHT Flanker, β = −15.5, p = .02) and cognitive flexibility/problem solving (Dimensional Change Card Sort, β = −20.5, p = .02).

Table 3. Unadjusted mean (standard deviation) standardized scores for cognitive measures by HIV status.

CVLT-3, California Verbal Learning Test-Third Edition (standard scores; M = 100, SD = 15); WMS-IV, Wechsler Memory Scale-Fourth Edition (scaled scores; M = 10, SD = 3); NIHT, NIH Toolbox (standard scores); D-KEFS, Delis Kaplan Executive Function System (scaled scores).

Group comparison significance is derived from GEE models with cannabis category and reflect FDR correction within Cognitive Domain; significant p values are reported in bold; if p > .05, differences are noted as not significant (ns).

In analyses using cannabis as a continuous variable, the pattern of significant HIV differences was similar except that scores on measures of learning and retention of stories (Logical Memory I and II) also were significantly worse for YWH.

Unadjusted categorical cannabis use models

Categorical cannabis use effects

Cognitive performance by 30-day cannabis use category and HIV group is presented in Table 4. In categorical analyses using the full model, cognitive performance independent of HIV status for those with mild/moderate cannabis use did not differ from those with no use. In contrast, participants with heavy use performed significantly worse than those with no use on Fluid Cognition (β = −15.7, p = .01), word list learning (CVLT-3 Trials 1–5; β = −9.8, p = .04) and retention (CVLT-3 Delayed Recall; β = −13.0, p = .02), Pattern Comparison processing speed (β = −16.3, p = .04), and several aspects of executive functions: inhibitory functioning (Flanker; β = −10.0, p = .046), nonverbal cognitive flexibility/problem solving (Dimensional Change Card Sort; β = −16.0, p = .04), and verbal cognitive flexibility/set shifting (Verbal Fluency Category Switching; β = −2.7, p = .03; Table 4 and Figures 1A–D). Heavy users also performed worse relative to mild/moderate users for some executive functions, including nonverbal cognitive flexibility/problem solving (Dimensional Change Card Sort, β = −15.03, p = .02) and verbal cognitive flexibility/set shifting (Verbal Fluency Category Switching, β = −1.98, p = .03), and, in addition, had worse story learning (Logical Memory I; β = −1.9, p = .04) and retention (Logical Memory II; β = −2.1, p = .02).

Table 4. Unadjusted standardized scores [marginal mean (standard error)] for cognitive domains by 30-day cannabis use category.

SS, standard score; CVLT-3, California Verbal Learning Test-Third Edition (standard scores; M = 100, SD = 15); WMS-IV, Wechsler Memory Scale-Fourth Edition (scaled scores; M = 10, SD = 3); NIHT, NIH Toolbox (standard scores); D-KEFS, Delis Kaplan Executive Function System (scaled scores).

Consistent with findings from 30-day use analyses, results for 180-day use indicated that those with mild/moderate cannabis use did not differ from those with no use for any variables. Young adults with heavy use performed significantly worse than those with no use on Fluid Cognition (β = −19.2, p = .004), word list learning (CVLT-3 Trials 1–5; β = −11.3, p = .04), Pattern Comparison processing speed (β = −21.7, p = .003), inhibition (Flanker; β = −13.1, p = .01), and nonverbal cognitive flexibility/problem solving (Dimensional Change Card Sort; β = −18.6, p = .02). Additionally, considering 180-day use, those with heavy use performed significantly worse than those with no use on working memory reflected by CVLT-3 Trial 1 (β = −2.7, p = .03) but did not differ on word list retention (CVLT-3 Delayed Recall) or verbal cognitive flexibility/set shifting (Verbal Fluency Category Switching). Consistent with findings for 30-day use, heavy users at 180-day use performed significantly worse than those with mild/moderate use on story learning (Logical Memory I; β = −1.9, p = .02) and retention (Logical Memory II; β = −2.1, p = .008), nonverbal cognitive flexibility/problem solving (Dimensional Change Card Sort; β = −17.1, p = .01), and verbal cognitive flexibility/set shifting (Category Switching; β = −2.4, p = .01). In contrast to findings at 30-day use, heavy users at 180-day use also performed significantly worse than those with mild/moderate use on inhibitory functioning (NIHT Flanker; β = −9.8, p = .03), as well as the composite Fluid Cognition score (β = −15.5, p = .01).

For 365-day use, consistent with 30-day findings, those with mild/moderate cannabis use did not differ from those with no use for any variables. Those with heavy use performed significantly worse than those with no use on Fluid Cognition (β = −16.2, p = .03), Pattern Comparison (β = −19.7, p = .01), inhibitory processing (Flanker; β = −11.3, p = .05), and verbal cognitive flexibility/set shifting (Category Switching (β = −2.5, p = .04), and additionally performed significantly worse on word list immediate recall/working memory (CVLT-3 Trial 1; β = −2.5, p = .04) and visuomotor processing speed (Trail Making Non-Switching Trials; β = −2.5, p = .048). In contrast to 30-day use, heavy users did not perform significantly worse than non-users on word list learning (CVLT-3 Trials 1–5) or retention (CVLT-3 Delayed Recall) or on nonverbal flexibility/problem solving (Dimensional Change Card Sort). Similar to 30-day analyses, heavy users performed significantly worse than those with mild/moderate use on story retention (Logical Memory II; β = −1.7, p = .04), nonverbal flexibility/problem solving (Dimensional Change Card Sort; β = −15.3, p = .03), and verbal flexibility/set shifting (Verbal Fluency Category Switching; β = −2.0, p = .03), but, in contrast, did not differ for story learning (Logical Memory I).

HIV x categorical cannabis use interactions

While there were no significant interactions of HIV status with 30-day cannabis use as a categorical variable, for 180-day use, significant interactions were found for Fluid Cognition (β = 23.4, p = .04), inhibitory functioning (Flanker; β = 21.2, p = .03) and nonverbal flexibility/problem solving (Dimensional Change Card Sort; β = 23.4, p = .04) for heavy vs. no use. Performance was worse for YWoH with heavy use vs. no use while performance for YWH was slightly better for heavy use vs. no use.

For 365-day use, there were significant interactions for inhibitory functioning (Flanker; β = 22.0, p = .04), nonverbal flexibility/problem solving (Dimensional Change Card Sort; β = 24.5, p = .03) and verbal flexibility/set shifting (Verbal Fluency Category Switching; β = 3.6, p = .046) for heavy vs. no use; the pattern was similar to that seen for 180-day use, i.e., heavy use compared to no use was worse for YWoH and better for YWH. An additional finding was seen for immediate word list recall/working memory (CVLT-3 Trial 1; β = 3.5, p = .04), where YWoH had lower performance and YWH better performance for mild/moderate vs. no use.

Unadjusted continuous cannabis use models

Continuous cannabis use effects

Using cannabis use as a continuous variable, higher cannabis use was significantly associated with worse Fluid Cognition (β = −0.5, p = .007), retention for word lists (CVLT-3 Delayed Recall; β = −0.5, p = .007) and stories (Logical Memory II; β = −0.6, p = .03), word list recognition memory (CVLT-3 Recognition Discriminability; β = −0.07, p = .02), processing speed as measured by Pattern Comparison (β = −0.5, p = .02) and Trail Making Non-Switching Trials (β = −0.1, p = .047), along with three facets of executive functioning: inhibition (Flanker; β = −0.4, p = .03), nonverbal cognitive flexibility/problem solving (Dimensional Change Card Sort; β = −0.6, p = .02), and verbal cognitive flexibility/set shifting (Category Switching; β = −0.1, p = .003). In summary, in comparison with the categorical analyses, additional significant associations of greater cannabis use with worse cognitive performance were seen in story retention, word list recognition memory, and an additional measure of processing speed, while immediate memory for word lists and stories was no longer significant. Results were similar for 180 (with story learning additionally significant) and 365-day use.

HIV x continuous cannabis use interactions

With cannabis as a continuous variable, there was a significant interaction of HIV and 30-day cannabis use for inhibitory functioning (Flanker; β = 0.1, p = .04). For 180-day use, there were significant interactions for story retention (Logical Memory II; β = 0.02, p = .03) and inhibitory functioning (Flanker; β = 0.1, p = .04). Similarly, for 365-day use, there were significant interactions for story retention (Logical Memory II; β = 0.009, p = .03) and inhibitory functioning (Flanker; β = 0.05, p = .04). For all of these interactions, YWoH showed worse and YWH better performance as use increased to the mild/moderate range from no use.

Adjusted generalized linear models for 30-day use

Cannabis use as categorical variable

Adjusted analyses included English as first language, anxious/depressed t-score, FRS-30 score, and education (beyond high school or not) as covariates for 30-day use. After adjustment, HIV status was significantly associated with differences in word list learning (CVLT-3 Trials 1–5; β = −14.6, p = .04) and retention (Delayed Recall; β = −14.2, p = .049) as well as immediate recall of Trial 1 (β = −2.96, p = .03), inhibitory functioning (Flanker; β = −13.9, p = .02), and nonverbal cognitive flexibility/problem solving (Dimensional Change Card Sort; β = −16.7, p = .04), while Fluid Cognition and processing speed variables (Pattern Comparison and Trail Making Non-Switching Trials) were not significant. After adjustment, significant differences for 30-day categorical use remained for heavy vs. no use on Fluid Cognition (β = −13.0, p = .03), Pattern Comparison processing speed (β = −14.2, p = .03), nonverbal (Dimensional Change Card Sort; β = −15.5, p = .03), and verbal cognitive flexibility/problem solving (Verbal Fluency Category Switching; β = −2.6, p = .01); other differences were no longer significant. An interaction for HIV with mild/moderate vs. no cannabis use was significant for word list retention (CVLT-3 Delayed Recall; β = 18.2, p = .03). For 180-day use, differences included heavy vs. no use for Fluid Cognition (β = −16.3, p = .02), immediate recall of Trial 1 (β = −2.3, p = .04), Pattern Comparison processing speed (β = −19.7, p = .008), inhibitory functioning (Flanker; β = −11.9, p = .03), and nonverbal cognitive flexibility/problem solving (Dimensional Change Card Sort; β = −18.2, p = .03). For 365-day use, significant effects remained for heavy vs. no use for Fluid Cognition (β = −14.5, p = .05), Pattern Comparison processing speed (β = −19.4, p = .01), and nonverbal cognitive flexibility/problem solving (Verbal Fluency Category Switching; β = −2.6, p = .02). Additionally, a cannabis by HIV status interaction emerged for immediate recall of Trial 1 (β = 3.5, p = .04) with worse performance among mild/moderate use relative to no use.

Associations of covariates with cognitive variables in the primary analyses for 30-day use categories were examined. Compared to those without education beyond high school, those with education beyond high school had significantly higher scores for numerous variables: Fluid Cognition (β = 17.0, p = .003), word list retention (β = 10.3, p = .02) and recognition (β = 1.8, p = .03), story learning (β = 2.9, p < .001) and retention (β = 2.3, p = .002), immediate recall of word list Trial 1 (β = 9.0, p = .046), cognitive storage and manipulation (List Sorting; β = 8.6, p = .03), Pattern Comparison processing speed (β = 19.7, p = .003), and nonverbal cognitive flexibility/problem solving (β = 11.5, p = .04). English as first language was associated with higher scores for List Sorting (β = 8.3, p = .006) and nonverbal cognitive flexibility/problem solving (β = 2.1, p = .006) compared to those without English as a first language. Current financial resources were significantly associated with higher word list learning (β = 3.9, p = .03) and Trial 1 (β = 0.7, p = .04). Anxiety/depression was not significant in any models.

Cannabis use as continuous variable

Following adjustment, analyses using 30-day cannabis use as a continuous variable showed significant results for cannabis in Fluid Cognition (β = −0.4, p = .04), CVLT-3 Delayed Recall (β = −0.3, p = .048), and Verbal Fluency Switching (β = −0.1, p = .003), with more variables (Logical Memory II, Pattern Comparison, Trail Making Non-Switching Trials, Flanker, and Dimensional Change Card Sort) showing attenuated effects compared to the categorical analyses. The Cannabis x HIV interaction was not significant for any variables.

Among covariates, in these analyses education beyond high school was significantly associated with numerous variables similar to above, and English as first language was significantly associated with two verbal variables [List Sorting (β = 7.8, p = .008), Verbal Fluency Switching (β = 1.9, p = .007)]. Anxiety/depression and current financial resources showed no significant associations.

Trend analyses for continuous cannabis use

Finally, we explored linear and non-linear associations of continuous cannabis use frequency with cognitive performance to examine whether we saw a pattern or trend of performance becoming worse in a linear fashion with more frequent cannabis use, or, instead, of nonlinear relationships being observed. Due to the exploratory nature of these analyses, we include results with p < .10 in this relatively small cohort. The linear trend was significant, reflecting worse performance with more frequent cannabis use, for 10 variables within the YWoH group. In the Verbal Learning and Memory domain, word list learning (CVLT-3 Trials 1–5; β = −0.3, p = .07), word list retention (CVLT-3 Delayed Recall; β = −0.5, p = .01) and recognition memory (β = −0.1, p = .03), and story retention (Logical Memory II; β = −0.1, p = .08) showed linear associations. Also linearly associated were Working Memory (CVLT-3 Trial 1; β = −0.1, p = .099), Processing Speed [Pattern Comparison processing speed (β = −0.5, p = .04) and Trail Making Non-Switching Trials (β = −0.1, p = .03)], and Executive Function [Verbal Fluency Category Switching (β = −0.1, p = .01); Flanker inhibitory functioning (β = −0.4, p = .04), and Dimensional Change Card Sort (β = −0.6, p = .02)]. In addition, Fluid Cognition (β = −0.5, p = .01) had a linear trend. However, the linear trend was not significant for any variables in the YWH group.

With all participants included (regardless of HIV status), models with a cubic (non-linear) term differed from models without (only a linear term) for the following Learning and Memory Domain measures: CVLT-3 Trials 1–5 (β = 0.01, p = .046) and Delayed Recall (β = 0.01, p = .02); Working Memory measure CVLT-3 Trial 1 (β = 0.003, p = .03); Executive Functioning measures: Trail Making Number-Letter Switching Trial (β = 0.002, p = 0.08) and Flanker (β = 0.01, p = .07); and Fluid Cognition (β = 0.01, p = .099). Then, for two of these cognitive measures (inhibitory function measure Flanker and Fluid Cognition), in HIV-stratified analyses, YWH and YWoH tended to differ in direction; however, the coefficients for the cubic terms were not significant for either, perhaps due to less power resulting from smaller sample sizes compared to modeling the two groups together. The cubic term was significant for YWH for CVLT-3 Trials 1–5 (β = 0.008, p = .09), CVLT-3 Delayed Recall (β = 0.008, p = .09), and Trail Making Number-Letter Switching Trial (β = 0.002, p = .045), and for YWoH for CVLT-3 Delayed Recall (β = −0.007, p = .09), however. Thus, the pattern of associations between cannabis use frequency and cognitive performance tends to differ for those with and without HIV, with primarily linear associations for YWoH and non-linear for YWH. This is consistent with our analyses of categorical use where YWH had best performance with mild/moderate use for some variables.

Discussion

Efforts to understand emerging adulthood are critical as many among this young demographic struggle to meet key milestones, such as gainful employment, education, and independent living, in the context of changing legalization and patterns of cannabis use. In this context, our cross-sectional study examined associations of cannabis use frequency with cognitive performance (verbal learning and memory, processing speed, executive functions, fluid cognition) among young adults age 18–24, some of whom were living with HIV infection (YWH), a disease characterized by chronic neuroinflammation. Our findings indicate that overall greater frequency of cannabis use was associated with lower performance in verbal learning and memory, processing speed, and executive functions, as well as a composite measure of fluid cognition which taps their capacity for new learning and information processing in novel situations—a critical developmental skill for transition to adulthood and independence.

Among all young adults, the significant impact of use frequency was apparent primarily for those with heavy use (five or more days per week over the past 30 days) who demonstrated worse performance in multiple cognitive domains relative to those with no use. It is noteworthy that mild to moderate use (between 1- < 5 days of use per week) was not associated with significantly lower cognitive performance compared to no use when those with and without HIV were combined; however, among those reporting mild to moderate use, different patterns of effects were observed for YWH and YWoH. Specifically, heavy use may place young adults at risk for worse cognitive performance more broadly, while mild to moderate use may have some benefits for individuals with HIV, an inflammatory disorder; this demonstrates the need for understanding effects of the full range of cannabis use for our most vulnerable young adults.

When the period of self-reported use was expanded to 180 or 365 days, the pattern of results was similar to the 30-day findings. Cannabis effects were particularly consistent across lengths of measurement for NIHT measures capturing processing speed, inhibitory executive skills, and fluid cognition. Findings across different observation periods were less consistent for verbal learning and memory, working memory, and executive functions of cognitive flexibility/set shifting and problem solving.

Overall, YWH had worse cognitive performance than YWoH across multiple measures reflecting fluid cognition, including processing speed and executive functions of inhibitory control and cognitive flexibility, as well as learning and retention of a word list. However, they did not show significantly lower recognition memory, suggesting effects of HIV on retrieval rather than encoding or storage of material. Thus, in agreement with previous studies of young adults with behaviorally acquired HIV (52, 53), these findings showed a potential ongoing effect of HIV on cognitive functioning among young adults despite a shorter duration of infection than in studies of older adults and ongoing antiretroviral treatment. Although YWH in our study had lower educational attainment than YWoH, adjustment for this discrepancy left many observed differences intact. Similarly, YWH had higher levels of anxiety and depression than YWoH; however, this did not account for any of the HIV effects. While the majority of YWH participants’ scores were within the average range, their Fluid Cognition overall as well as some individual scores fell at least one standard deviation below the standardization mean. Reduced ability to learn and retain information, process information quickly, and inhibit and shift cognition flexibly as needed could affect day-to-day academic, occupational or social functioning as well as self-management of healthcare. This is particularly critical during the young adult phase where knowledge and skill acquisition and mastery of independent living tasks are key developmental goals.

Secondary analyses examined whether associations of cognitive performance with cannabis use frequency as a continuous variable followed linear, quadratic, or cubic trends. We found that cognitive performance in YWoH was worse with higher cannabis use in a linear fashion across several cognitive domains; however, this was not the case for YWH. Consistent with our hypothesis, the effects of cannabis on cognition appear to be influenced by the presence of HIV infection. Among YWH, mild/moderate use appeared to be associated with higher performance in fluid cognition and some aspects of memory and executive functioning, but this pattern was not seen consistently and did not meet criteria for significance for 30-day use, although consideration of cannabis use over longer time periods interacted significantly with HIV for some measures of executive functions and working memory. As noted by others (e.g., 36), previous studies of HIV and cannabis use have focused largely on heavy, such as daily, use. In our analyses, heavy use was associated with worse performance in both YWH and YWoH, and study of mild/moderate use was necessary to discover different patterns for YWH, including an examination of a range of cannabis use frequency which is critical for understanding population-specific impacts on cognitive functioning.

Chronic immune activation and inflammation along with release of inflammatory cytokines are thought to underlie continuing CNS impacts of HIV (18, 23–25), and studies of YWH have shown that some markers of macrophage activation are not normalized despite successful viral suppression using antiretroviral therapy (54). Because of this, altering immune and inflammatory processes through non-antiretroviral medications or interventions is seen as a key potential therapeutic mechanism to address the ongoing effect of HIV on the CNS and a rationale for considering cannabis among possible therapies. Recent reviews of the literature on cannabis and HIV have focused on the role of inflammation in mediating their combined effects on CNS functioning (18) and the blood-brain barrier (10). In a study of adults with HIV (55), those using cannabis daily had lower markers of inflammation in cerebrospinal fluid than those with no use, and these markers were associated with learning performance. Our results are consistent with this line of thinking in showing that increasing cannabis use frequency in YWH does not have the linear association with decreasing cognitive performance observed in YWoH. Rather, for some variables, results suggested better performance for YWH with mild to moderate cannabis use compared not only to those with heavy use, but also to those with no use. This pattern would be consistent with possible anti-neuroinflammatory effects of mild to moderate use, with heavy use possibly exerting negative effects as seen in YWOH via other pathways. Future association studies will examine biomarkers of inflammation as mediators of outcomes to explore causality and may shed light on differences in findings for different cannabis use time periods reflecting shorter term vs. more chronic use. Importantly, however, our study was cross-sectional; longitudinal studies assessing inflammation as a possible mediator are needed to determine causality.

Individual characteristics related to educational and language background, availability of financial, social and medical resources, and other social determinants of health, as well as mood disorders, are important influences on performance on standardized tests. Comparable adjustments for sociodemographic variables are not available for the measures used in this study; in addition, it is increasingly recognized that adjusting for these factors, rather than examining their effects explicitly, impedes understanding of the multifaceted influences on cognitive performance and results in lost opportunities for intervention. Thus, adjusted analyses allowed us to examine attenuation of HIV and cannabis effects and also to look at the effects of the covariates themselves. Rather than using race and ethnicity, we chose to use variables for which race and ethnicity are sometimes considered proxies, financial and educational resources and language background. As anticipated, despite attempts to match the groups, YWH had fewer financial resources or post-secondary education and had higher levels of depression and anxiety. Inclusion of these variables, as well as English as a first language, attenuated cannabis effects; however, significant effects remained for fluid cognition as well as processing speed and executive function variables. Similarly, fewer HIV differences were observed but some remained particularly for memory and executive functions. Numerous variables across domains were related to post-secondary education; however, whether this is due to those with higher test scores being more likely to pursue further education, the effect of continued learning engagement on cognitive performance, or a third variable cannot be addressed in this study. Few significant effects were seen for current financial resources or English as a first language, and, surprisingly, none for anxiety and depression. In general, our results support the importance of including a wider range of variables reflecting a whole-person approach when considering standardized cognitive performance among young adults.

Although other recent drug use was screened out, use of alcohol was allowed in order to recruit a representative sample of young adults. As anticipated given that 2/3 of the sample was selected for ongoing cannabis use, the rates of cannabis use disorder were high, and alcohol use disorder was also common. Alcohol use disorder did not differ between cannabis use categories and, in exploratory analyses, was associated only with higher performance on one working memory variable and was thus not seen as explaining the observed cannabis effects. Future analyses will examine the characteristics of our participants’ cannabis use, impacts on daily life, and cannabis use motives and history. In particular, the current literature (4, 56, 57) reflects concern about initiation of cannabis use during adolescence, citing studies showing, for example, that those who began using cannabis prior to age 16 had greater residual effects in executive functions; a review of animal and human studies (58) suggested underlying alterations in the endocannabinoid system. The mean age of initiating regular cannabis use of 17 in our sample will allow us to examine the association of younger initiation with cognitive performance as well as endocannabinoid levels among young adults with and without HIV in future analyses.

Strengths of this study include the representation of a full range of cannabis use frequency, balanced across HIV groups, which may explain differences between our results and studies focused on heavy use or substance use disorders. Standardized, clinically meaningful tests were used that have been shown to be sensitive to the effects of HIV in other studies. Participants were diverse and representative of Southern California HIV prevalence. Limitations include the small sample size, and replication of these findings in a larger sample will be critical. This is particularly important for findings related to the interaction of HIV and cannabis in affecting cognition and for the potential mitigating influence of cannabis for YWH. To reduce impact of acute cannabis use on cognitive outcomes, participants were required to abstain the night before and day of their study visit. However, unlike some previous studies examining the chronic effects of cannabis use, this study did not require a longer period of abstinence, which is a limitation. Despite attempts to recruit a larger sample of individuals female at birth, this proved difficult among YWH since most young adults with HIV at our recruitment sites were male or transgender women; thus, the groups were unbalanced by sex, precluding examination of sex in this study and limiting generalizability of the study. A larger sample size with a balanced sex distribution would provide a more comprehensive analysis of the cognitive impact of cannabis use among YWH in comparison to YWoH. Finally, other potential influences on cognition that may differ between those with and without HIV, such as sleep quality, exercise, and other physical conditions were not included and call for further study.

Summary

This study examined the impact of cannabis use frequency on cognitive performance in young adults and whether this impact differs in the presence of HIV infection, a chronic disorder that affects the CNS through immune and inflammatory processes. Our results suggest an association between greater cannabis use frequency over the prior 30 days and lower performance across several cognitive domains in young adults without HIV; similar effects on cognition were also seen when considering longer use periods up to a year. Importantly, these effects appeared to be altered among YWH; for them, patterns were less consistent and, on some tests, the best performance was seen with mild to moderate frequency of cannabis use. This is consistent with the hypothesis of possible neuroprotective effects of cannabis in the context of a neuroinflammatory disorder such as HIV infection, while cautioning that this protection may be dose-dependent, with heavy use having deleterious effects similar to those seen in YWoH. Future analyses will examine biomarkers of inflammation and immune activation, as well as endocannabinoid levels, as potential mediators of the cognitive associations seen herein and will explore the impact of cannabis use characteristics such as amounts and methods of use. Larger studies will be needed to confirm the observed interactions between cannabis and HIV, address possible sex effects, and establish guidelines regarding dose-dependent neuroprotection for YWH within a whole-person context with consideration of social determinants of health. Furthermore, longitudinal studies will be important to determine the direction of causality between cannabis use and cognition and the role of inflammation.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Mikki Schantell, Columbia University, United States

Reviewed by: Kelly Sagar, McLean Hospital, United States

George Panagis, University of Crete, Greece