Cognitive Effects of a Ketogenic Diet on Neurocognitive Impairment in Adults Aging with HIV: A Pilot Study

Abstract

We assessed a ketogenic diet (KD) intervention protocol and the cognitive effects of KD in older adults with HIV-associated neurocognitive impairment. Adults older than 50 years and living with HIV and mild to moderate neurocognitive impairment were randomized to either a KD or a patient choice diet (PCD) for 12 weeks followed by a 6-week washout period. A neurocognitive battery was administered at baseline, week 12, and week 18. Paired t-tests compared groups at baseline and MANCOVAs were conducted to assess between-group differences on primary outcome variables at weeks 12 and 18. We enrolled 17 participants; 14 completed the study. No between-group baseline differences were noted. The KD group demonstrated improved executive function and speed of processing at week 12, which were negated after participants resumed their usual diets. Our study supports the potential efficacy of a KD for the treatment of HIV-associated neurocognitive impairment.

Widespread availably of combination antiretroviral therapies (cART) has transitioned HIV from a once-terminal condition to a chronic disease manageable across the lifespan (May et al., 2014). However, adults aging with HIV experience many health problems, including cognitive impairment. HIV-associated neurocognitive disorder (HAND) has a range of cognitive impairments: (a) asymptomatic, (b) mild to moderate, and (c) HIV-associated dementia. While the severity of HAND has significantly declined since the pre-cART era, the proportion of those affected by HAND is not significantly different (Heaton et al., 2011). HAND affects roughly half of adult persons living with HIV (PLWH) who are at least 50 years of age, and cognitive impairments may worsen over time (Grant et al., 2014). In addition, cognitive impairments among adult PLWH are commonly observed 10 to 15 years before age-related cognitive impairments in similarly aging populations (i.e., ages 50 years vs 65 years, respectively). Thus, it is plausible that HIV works independently, and perhaps synergistically, with normative aging processes to negatively impact brain health in adult PLWH (Egbert et al., 2019).

The most common cognitive impairments observed in adults with HAND include deficits in attention, executive function, speed of processing, and learning. The decline in these cognitive domains reduces quality of life, limits functional independence (i.e., financial management, transportation, shopping) and may interfere with effective disease management (i.e., medication adherence, attending health appointments; Cody & Vance, 2016; Jacks et al., 2015). In addition, older adult PLWH have more age-related co-morbidities (i.e., cardiovascular disease, diabetes) and are prescribed higher quantities of medications than similar aging populations without HIV, a consistent observation independent of the number of HIV therapies (Greene, Steinman, McNicholl, & Valcour, 2014; Vance, Mugavero, Willig, Raper, & Saag, 2011). As a result of HIV, advancing age, and greater co-morbid disease burden, polypharmacy is a pervasive problem in this population (Ware et al., 2018). Polypharmacy, defined as five or more prescribed medications, is associated with greater risk of falls, higher frailty, and more hospital admissions, and these polypharmacy-associated risks increase with age (Gnjidic et al., 2012). In addition to remembering to take more prescribed medications, adult PLWH are often required to maintain more challenging medication administration regimens to minimize drug-to-drug interactions and abate more serious adverse effects than persons without HIV (Greene et al., 2014). Thus, non-pharmacological treatments may be uniquely beneficial for aging adult PLWH to consider when alternate interventions are available.

HAND and Energy Metabolism

Effective cART management has significantly decreased incidence of the most severe form of HAND, HIV-associated dementia. Individuals who receive early diagnosis and treatment for HIV are more likely to develop asymptomatic or mild neurocognitive disorder in comparison to individuals whose diagnosis and/or treatment are delayed. However, the similar prevalence of HAND in comparison to the pre-cART era suggests that cognitive impairment is likely associated with changes in neuronal function and that neuronal insults begin early in the disease (Lentz et al., 2011). HAND pathogenesis is complex; however, it is widely accepted that aberrant immune system responses to HIV (e.g., activated brain resident cells and perivascular macrophages) as well as the release of neurotoxic HIV proteins (e.g., Tat, Gag, Pol) play a critical role (Cody & Vance, 2016). These brain insults impair normative neuronal processes and alter mitochondrial function, creating a pro-inflammatory environment (i.e., increased reactive oxygen species and oxidative stress; Tiede, Cook, Morsey, & Fox, 2011). Over time, these mechanisms interfere with the brain’s energy efficacy, leading to a hypometabolic state.

Brain hypometabolism is implicated in numerous neurodegenerative conditions (e.g., Alzheimer’s disease, age-related neurocognitive disorder, HAND, Parkinson’s disease; Homenko, Susin, Kataeva, Irishina, & Zavolokov, 2017). The co-occurrence of such a wide range of neurocognitive disorders and metabolic function underscores the fact that metabolism likely plays a central role. Additionally, brain imaging studies confirm that brain hypometabolism occurs, is extensive, and worsens over time independent of viral load (Borjabad et al., 2011), and it is reasonable to conclude that maladaptive brain metabolism after HIV infection plays an intrinsic role in HAND development.

The ketogenic diet (KD) is a low carbohydrate (≤ 50 grams/day)/high fat diet that has demonstrated improved brain metabolism and more optimal cognitive function in cognitive aging disorders (e.g., Alzheimer’s, Parkinson’s, mild neurocognitive impairment of aging). While the precise mechanisms are unclear, KDs have been associated with improved brain metabolism, neural antioxidant effects, reduced expression of crucial genes involved in inflammation, greater adenosine triphosphate production, and increased mitochondrial biogenesis (Henderson et al., 2009; Stafstrom & Rho, 2012). Moreover, KD intervention studies in similar aging populations (e.g., Alzheimer’s, Parkinson’s) suggest the use of ketones for brain fuel improves brain metabolism and increases cerebral perfusion (Hartman, 2012). Thus, in the context of HIV-associated neurocognitive disorder, targeting brain metabolism defects as a plausible treatment for cognitive impairment is reasonable to consider (Croteau et al., 2018; Vázquez-Santiago, Noel, Porter, & Rivera-Amill, 2014).

The purpose of our randomized, exploratory pilot study was to evaluate if cognitive gains observed while consuming a KD in other cognitive disorders of aging would translate to cognitively impaired, older adult PLWH. We had three specific aims: (a) to examine the feasibility of a KD intervention protocol, (b) to explore the preliminary cognitive effects of a 12-week KD intervention as well as to investigate whether cognitive changes observed at week 12 (if applicable) were sustained after a 6-week washout period (i.e., participants resumed their usual diets), and (c) to examine the effects of a 12-week KD on rudimentary body composition assessments and cardiometabolic biomarkers in cognitively impaired, older adult PLWH.

Methods

Our intervention and feasibility study was a two-arm, randomized clinical study with a target enrollment of 20 participants with HIV-associated neurocognitive impairment. The protocol required participants randomized to the intervention group to maintain a KD for 12 weeks and then return to their usual diets for 6 weeks (i.e., washout period; week 18). Participants randomized to the control group completed all assessments but were instructed to continue their preferred diets (Patient Choice Diet [PCD]). The study was approved by the University of Alabama at Birmingham (UAB) Institutional Review Board.

Participants

Adult PLWH at least 50 years of age with self-identified forgetfulness were recruited through the university’s HIV clinic by brochures, flyers, and word of mouth. Those interested in the study called to schedule a telephone screen to determine eligibility. Mild to moderate cognitive impairment was a study requirement and was determined using the Modified Telephone Interview for Mild Cognitive Impairment (TICS-M). Scores from 17 to 26 on the TICS-M reflect mild-to-moderate cognitive impairment similar to mild-to-moderate cognitive impairment as defined by the widely used Mini-Mental Status Exam (Seo et al., 2011). Individuals with severe cognitive impairment (i.e., TICS-M score <17) were not eligible for participation due to the complex intervention; two individuals who scored less than 17 on the TICS-M were considered too cognitively impaired to independently follow the prescribed diet and were excluded. The TICS-M was used due to the flexibility of a telephone screen. Individuals who scored between the TICS-M a-priori cut points were further screened during the telephone interview for the remaining inclusion criteria: at least 50 years of age, stable HIV (CD4+ T cell count ≥ 350 cells/mm³ for ≥ 2 years and prescribed a cART regimen for ≥ 6 months), and proficient in English. Exclusion criteria included: homelessness, non-HIV related cognitive co-morbidities (e.g., intellectual disability, Alzheimer’s, Parkinson’s), history of traumatic brain injury with loss of consciousness of 30 minutes or more, undergoing chemotherapy or radiation, preexisting metabolic or renal co-morbidities (e.g., diabetes, any stage of renal insufficiency), and pre-existing neuropsychological co-morbidities (e.g., schizophrenia, bipolar disorder, current drug or alcohol abuse). In addition, to reduce confounding effects between studies, individuals were excluded if they were enrolled in another study. A total of 188 individuals completed the telephone screen; 17 met enrollment criteria (Figure 1).

Figure 1.

Flow chart of participants screened and enrolled.

Note. cART = combination antiretroviral therapy; TICS-M = Modified Telephone Interview for Cognitive Status; OGTT = oral glucose tolerance test.

Screening

After the telephone interview, 17 eligible participants were scheduled for a screening oral glucose tolerance test (OGTT) and baseline data collection visit. Medical chart extraction was completed to re-affirm participant eligibility. Because this was a diet intervention, it was not possible for participants or study personnel to be blinded to group assignment. In order to minimize bias, the randomization schedule was sequentially-numbered in opaque sealed envelopes prior to participant enrollment. Envelopes were unsealed at the baseline visit to determine group assignment.

Oral and written consent were obtained on arrival to the screening exam. To reduce confounding effects of pre-existing metabolic disease, a 2-hour OGTT screen was performed. All but two participants met the 2-hour OGTT criterion (≤ 200 mg/dL). Participants with OGTT findings greater than 200 mg/dL were advised of their test results and encouraged to contact a primary health provider; test results were sent to their providers at the university HIV clinic consistent with institutional review board protocol. One participant was unable to complete the OGTT and did not continue in the study. Fourteen participants (62.5% female, 87.5% African American) were invited to continue the study. Participants were compensated $50 for their time for each attended on-site visit (baseline, week 12, week 18).

Intervention

Individuals assigned to the KD arm had all meals and snacks delivered to their homes each week. These participants selected meals from an 8-day KD menu. Menus were developed by a registered dietician well-versed in diet intervention research. Food and snacks were prepared by a metabolic kitchen specializing in food preparation for nutrition research. The Harris-Benedict formula with an activity factor of 1.35 * 10% was used to estimate total energy needs (Batterham, 2005). An additional 10% of total daily energy requirements were provided to maintain eucaloric conditions given the heightened energy needs observed in HIV infection independent of viral load (Sitole, Williams, & Meyer, 2013). A registered dietician provided nutrition education at the baseline visit regarding foods and beverages consistent with a KD as well as foods and beverages to avoid. Persons in the control group were advised to continue their usual diets without restriction. All participants were encouraged to maintain baseline physical activity practices. An example KD menu is provided in Table 1.

Table 1.

Example day of meals for the ketogenic diet

Ketogenic diet		Daily Nutrients
Breakfast		KCAL	1800
Eggs, boiled	2 large	CARB (g)	36.4
Bacon, regular	4 slices	FIBER (g)	15.9
Butter, regular	5 g	NET CARB (g)	20.4
Almond milk, unsweetened	11 fluid ounces	FAT (g)	140
Coffee	8 fluid ounces	PROTEIN (g)	108
Half and half	3 containers (3 T)	% CARB	7.3
		% FAT	68.2
Lunch		% PROTEIN	24.4
Canned tuna	5 ounces
Mayonnaise, regular	3 packages
Walden Farms Ranch dressing	50 grams
Tomato	5 cherry tomatoes
Iceburg lettuce	1/4 head wedge
Macadamia nuts	40 g
Dinner
Chicken breast	120 grams
Olive oil	2 tablespoons
Spinach, canned	8 ounces
Squash, frozen	1 cup
Butter, regular	10 g

Adherence Assessment

Capillary ketone assessments were completed using the Precision Xtra^© ketone monitor, a home ketone monitoring device with a 450 test memory, including date and time of assessment. All testing supplies were furnished. Ketone testing instructions were provided during the baseline visit. A member of the research staff reviewed all of the needed supplies, turned on the monitor, and, in a step-wise fashion, completed the capillary ketone testing procedure. Participants then repeated the process to complete a ketone assessment. The steps were repeated until the participant was able to complete ketone testing without questions or guidance from the research staff. During the study, capillary ketone sampling was prescribed as follows: 3 times per day prior to meals for weeks 1 and 2, prior to lunch on Wednesdays during weeks 3 to 12, a strategy to reduce participant burden.

Ketone assessments were recorded on a standardized paper log. Participants were phoned during week 1 to answer questions, provide clarification regarding ketone monitoring, and verify that they were able to complete ketone tests. Ketone values from the paper log were read to research staff during two telephone visits (week 4, week 8) during which weekly ketone testing was verified and the participant was asked about testing supplies. The second record of ketone scores assured data availability in case a paper log was misplaced or the ketone monitor malfunctioned or was lost. Participants were compensated for week 4 ($10) and week 8 ($10) telephone visits.

To assess reliability and validity of self-reported ketone values, ketone monitor data were downloaded and compared to recorded scores on paper logs at week 12. Scores from the paper log and values from the ketone monitor were individually totaled. Paired t-tests were conducted to compare means between self-report paper log and ketone monitor data. When the paper log was incomplete or not available, scores verbalized to research staff during telephone visits were substituted.

Demographics, Food Insecurity, and Mental Health Measures

Demographic questionnaire and health questionnaire.

A basic demographic questionnaire gathered information about age, gender, relationship status, sexual orientation, race/ethnicity, education level, and household income before taxes. Health data included participant’s perceived health status relative to the health of their peers (Table 2).

Table 2. Demographic variables.

Variable	KD (n = 7)	PCD (n = 7)
	Total/%	Total/%
Age* (years)	56.7 (5.7)	54.3 (4.7)
Gender
Male	3 (42.9%)	3 (42.9%)
Female	4 (57.1%)	4 (57.1%)
Race/ethnicity
African American	6 (85.6%)	6 (85.6%)
Caucasian	1 (14.3%)
Native American		1 (14.3%)
Not Hispanic or Latino	6 (85.7%)	5 (71.4%)
Marital Status
Single	3 (42.9%)	6 (85.7%)
Married	2 (28.6%)
Divorced		1 (14.3%)
Widowed	2 (28.6%)
Sexual Orientation
Heterosexual	5 (71.4%)	3 (42.9%)
Bisexual		1 (14.3%)
Homosexual		3 (42.9%)
Not Reported	2 (28.6%)
Education (years) Completed
Some High School (≤ 11th grade)	2 (28.6%)	2 (28.6%)
High School (12th grade)	1 (14.3%)	1 (14.3%)
Some College/Vocational	4 (57.1%)	4 (57.1%)
Household Income
$0 – $10,000	2 (28.6%)	3 (42.9%)
$10,001 – $20,000	1 (14.3%)	2 (28.6%)
$20,001 – 30,000	2 (28.6%)	1 (14.3%)
$30,001 – $40,000
$40,001 – 50,000
$50,001 – $60,000	1 (14.3%)
Not Reported	1 (14.3%)	1 (14.3%)
Perceived Health
Excellent	2 (28.6%)	1 (14.3%)
Very Good		1 (14.3%)
Good	4 (57.1%)	4 (57.1%)
Fair	1 (14.3%)	1 (14.3%)
Food Insecurity
Yes	2 (28.6%)	4 (57.1%)
No	5 (71.4%)	3 (42.9%)

Note:

^*Mean (SD)

Food insecurity.

Food insecurity is associated with greater fragility as well as slower speed of processing, reduced learning and memory, diminished motor function, and decreased overall cognition in older adult PLWH (Willig, Overton, & Saag, 2016). We assessed food insecurity via a 2-item survey: (a) The food I/we bought just didn’t last and I/we didn’t have money to get more (0 = never true, 1 = sometimes true, 1 = often true) and (b) I/we couldn’t afford to eat balanced meals (0 = never true, 1 = sometimes true, 1 = often true). A composite score of 1 or more indicated food insecurity and a score of 2 suggested more severe food insecurity. The internal validity of the scale, as determined by Cronbach’s alpha, was 0.94. Moreover, the instrument was validated in a study of adult PLWH with a sensitivity of 100% (95% CI 75, 100), specificity 78% (95% CI 61, 90), and a negative predictive value of 100% (95% CI 88, 100; Young, Jeganathan, Houtzager, Di Guilmi, & Purnomo, 2009).

Depression.

The Centers for Epidemiological Studies (CES)-Depression Scale was used to assess depression and depression symptoms, as these may negatively impact cognition (Radloff, 1977). Participants indicated on a 4-point Likert-type scale how often they had felt a certain way during the previous week. Scores were summed for a composite score (range = 0 – 60, with higher scores reflecting greater depressive symptomology). A summed composite score of 16 or higher indicated depressive symptoms. Cronbach’s alpha of the CES-D suggests excellent internal validity at 0.88 (Clark, Mahoney, Clark, & Eriksen, 2002).

Cognitive Measures

Cognitive testing was performed by research staff trained and validated in cognitive assessment by a licensed clinical psychologist. The cognitive battery consisted of five demographically normed tasks (Trails A [psychomotor speed], Trails B [executive functioning], Stroop [executive functioning/inhibition], Digit Symbol Substitution Test [speed of processing], the Hopkins Verbal Learning Test [verbal memory]) that were administered in a fixed order that took about 45 minutes to complete at baseline, week 12, and week 18. Tasks were divided into five cognitive domains to reduce the number of cognitive variables in the analysis and for clinical interpretation (Brooks, Strauss, Sherman, Iverson, & Slick, 2009; Lezak, Howieson, Loring, & Fischer, 2004; Reitan, 1992). The cognitive battery was completed at baseline, week 12, and week 18. A T-score was derived for each component test adjusting for age and years of education.

Stroop.

Executive control, selective attention, automaticity, speed of processing, and inhibition were measured by the Stroop Color-Word Naming Task (Stroop, 1935). The test presents a participant with a list of 100 color words (i.e., red, green, blue) that are printed in a color of ink incongruent with the word. For example, “red” may be printed in green ink or the word “green” may be printed in blue ink. During the test, the participant must state the color of ink, ignoring the mismatch. Score is determined by the number of correctly stated colors in 45 seconds. The Stroop Color-Word Test has demonstrated sensitivity to the effects of aging on cognitive performance across the life span (Jensen & Rohwer, 1966).

Trails A.

Psychomotor speed, visuospatial tracking, and attention were measured by Trails A, a paper/pencil test that requires an individual to connect randomly scattered numbers in sequence as quickly and accurately as possible. The score is determined by the amount of time (seconds) required to correctly connect the numbered sequence. This test has demonstrated good sensitivity to age-related declines in cognitive performance, including in older PLWH (Lezak et al., 2004).

Trails B.

Executive function (set switching, speed of processing, working memory) was assessed using Trails B, which is similar to Trails A in that it requires participants to sequentially connect randomly scattered letters and numbers on a paper in the correct order. The score is the amount of time required to complete the task. Trails B has demonstrated sensitivity to normative age-related cognitive decline (Lezak et al., 2004).

Hopkins Verbal Learning Test-Revised (HVLT-R).

The HVLT-R is a brief test of verbal learning and memory with 12 items organized into three semantic categories and administered over three consecutive learning trials (trial 1, trial, 2, trial 3). For our study, the total number of words correctly recalled from all three trials were summed to form a composite score in which higher scores represented better verbal memory function. Delayed recall (trial 4) was determined by the number of words a participant was able to recall from the original list of 12 semantically categorized words approximately 20 minutes later.

The HVLT-R has demonstrated retest reliability (Cronbach’s alpha: verbal memory = .78; recall = .74). Construct validity for verbal learning and recall has good diagnostic accuracy for detection of cognitive impairment across cognitive disorders and cultures, including HIV-associated cognitive impairment (Woods et al., 2005).

Digit Symbol Substitution Test (DSST).

The DSST is a paper/pencil assessment of psychomotor speed. Psychomotor speed is used to assess changes in cognition across aging populations because psychomotor slowing reflects fundamental aspects of brain function (i.e., overall efficiency of processing). Participants are provided a key grid of numbers and matching symbols and a test section with numbers and empty boxes and instructed to fill as many empty boxes with a symbol matching each number. The score is the number of symbols correctly matched in 90 seconds. Higher scores are indicative of greater cognitive performance. Reliability estimates between .89 and .96 have been observed in older, cognitively impaired populations (Rosano, Newman, Katz, Hirsch, & Kuller, 2008).

Biomarker and Anthropometric Assessments

All serum biomarker testing was performed after a 12-hour fast. Lipid and metabolic panels were assessed at baseline and week 12. Testing was completed at the UAB Clinical Research Unit, which is part of the UAB Center for Clinical and Translation Research and is staffed by registered nurses and registered dietitians trained to implement nutrition-based research protocols.

Concentrations of glucose, insulin, and C-peptide were analyzed in the UAB Diabetes Research Center Metabolic Core. Glucose was measured in 3 µL of sera using the glucose oxidase method on a Stanbio Sirrus analyzer (Stanbio Laboratory, Boerne, TX). Insulin was assayed in 50-µL aliquots using immunofluorescence on a TOSOH AIA-II analyzer (TOSOH Corp, S. San Francisco, CA). C-peptide was assayed in 20-µL aliquots using the TOSOH analyzer. Homeostatic model assessment (HOMA) was calculated using the approach described by Matthews et al. (1985).

Total, high density cholesterol (HDL), and triglycerides were measured using SIRRUS analyzer (Stanbio Laboratory, Boerne, TX). Low-density lipoprotein was calculated using the method of Friedewald, Levy, and Fredrickson (1972).

Height, weight, body mass index (BMI), and waist and hip circumferences were assessed at baseline and week 12. Height (nearest 0.1 cm), waist circumference (nearest 0.1 cm), and hip circumference (nearest 0.1 cm) were measured with a Gulick tape measure. Weight (nearest .01 kg) was assessed using a Tanita body composition analyzer BC418 (Tanita Corp of America, Arlington Heights, IL).

Statistical Analysis

Data were analyzed using SPSS 25.0. Variables known to deviate from a normal distribution were log10 transformed prior to statistical analyses. Cognitive data were adjusted using nationally representative means for age and level of education (i.e., transformed into t-scores) congruent with standard psychology practice. Higher scores for these measures indicated more optimal cognitive function. All statistical analyses were two-tailed using a type one error rate of 0.05. No alpha corrections were performed due to the small sample size. Basic descriptive statistics were used to describe sample characteristics by group assignment. Independent t-tests were conducted at baseline to assess for between-group differences on continuous descriptive (e.g., years of education), metabolic (e.g., weight, total cholesterol), and cognitive variables (e.g., Trails A and B, Stroop; Table 3). Analysis of covariance was used to isolate between-group effects of the intervention on primary outcome measures. Specifically, for cognitive measures, baseline age- and education-adjusted normed scores were entered as covariates and week 12 age- and education-normed cognitive outcomes were the dependent variables. To determine persistence of any between group diet effects after a 6-week washout period, week 18 age- and education-adjusted cognitive data were entered as dependent variables and age- and education-adjusted cognitive baseline data served as independent variables. Cardiometabolic data were assessed in a similar fashion: week 12 cardiometabolic outcomes served as dependent variables while baseline measures adjusting for total body fat mass served as covariates (Table 4).

Table 3. Cognitive Assessment Score at Baseline, Week 12, and Week 18 Adjusted for Age and Education (N = 14).

	Ketogenic Diet (n = 7)			Patient Choice Diet (n = 7)
	Baseline M (SD)	12-Week M (SD)	Posttest Mean (SD)	Baseline Mean (SD)	12-Week Mean (SD)	Posttest Mean (SD)
DSST	46.1 (9.3)	46.7 (8.1)	52.0 (10.8)	46.0 (7.5)	46.1 (9.5)	44.7 (10.2)
Trails A	41.6 (17.0)	44.6 (17.6)*	40.1 (16.2)	50.0 (7.7)	44.3 (8.3)	44.1 (11.9)
Trails B	38.4 (18.1)	49.7 (13.1)*	45.9 (15.2)	48.4 (8.4)	44.7 (10.9)	46.6 (7.5)
HVLT Total	29.4 (9.8)	35.3 (9.6)	39.6 (9.1)	31.1 (8.6)	34.9 (12.6)	36.6 (11.7)
HVLT Delay	27.1 (11.1)	33.7 (13.9)	34.9 (8.4)	26.1 (8.0)	28.0 (9.0)	32.6 (12.0)
Stroop	36.3 (6.8)	40.0 (8.6)	39.1 (8.1)	36.0 (6.0)	39.4 (6.2)	40.1 (6.3)

Note: SD = standard deviation;

^*p <0.05; DSST: Digit Symbol Substitution Test; HVLT: Hopkins Verbal Learning Test

Table 4. Body composition and blood-derived data (N = 14) Means/SD.

	KD Group		PCD Group
Parameter	Baseline	12 Weeks	Baseline	12 Weeks
Anthropometric Measures
Weight (kg)	83.3 (13.9)	82.79 (12.96)	92.97 (20.6)	90.1 (21.9)
Waist circumference (cm)	103.0 (14.7)	101.5 (13.1)	93.6 (47.3)	106.2 (16.0)
Hip circumference (cm)	107.6 (11.8)	110.97 (14.7)	96.8 (48.25)	111.2 (19.6)
Body fat, %	32.96 (9.2)	28.4 (19.2)	34.96 (13.2)	34.4 (12.4)
Lipid Profile
Total cholesterol (mg/dL)	183.1 (39.2)	191.0 (31.8)	196.3 (43.4)	188.7 (23.9)
Triglycerides (mg/dL)	97.7 (73.3)	87.6 (58.1)	105.4 (49.05)	124.0 (61.0)
LDL (mg/dL)	93.8 (28.2)	102.5 (23.7)	107.8 (34.5)	91.8 (17.8)
HDL (mg/dL)	87.6 (58.1)	71.0 (20.3)	67.4 (22.9)	72.1 (26.1)
Metabolic Indicators
Insulin (μU/mL)	9.1 (3.6)	12.4 (4.8)	12.3 (9.7)	16.8 (8.1)
Glucose (mg/dL)	97.0 (4.4)	98.1 (6.3)	101.4 (11.5)	100.6 (16.6)
C-peptide (μg/L)	2.1 (0.6)	2.5 (0.7)	2.6 (0.9)	3.3 (1.1)
HOMA-IR	2.2 (0.9)	3.0 (1.4)	3.15 (2.7)	4.4 (2.8)

Results

Fourteen participants completed the study from April 2016 to June 2017. As seen in Table 2, intervention and control group composition were similar. Most participants had at least a high school education (71%) and more than 50% had attended some college. Based on reported income, poverty was observed across both groups. Although not statistically significant, lower levels of depression were observed in the KD group compared to the PCD group (M = 12.3, M = 22.6, respectively; CES-D score ≥ 16 indicates depression). However, a wide range was noted across both groups (Table 3). Unfortunately, depression is common in PLWH with evidence suggesting depression estimates range from 36% to greater than 50% depending on HIV subpopulation, diagnostic criteria, and/or assessment techniques (Nanni, Caruso, Mitchell, Meggiolaro, & Grassi, 2014). The groups were similar at baseline for cognitive function and cognitive function varied broadly in both groups, which was anticipated given the well-documented variation in cognition in older adult PLWH (Vance, Fazeli, Ball, Slater, & Ross, 2014).

Aims

Aim 1: Assess feasibility of a KD protocol in a cognitively impaired population of adult PLWH.

We considered intervention fidelity as part of the feasibility assessment. To determine KD adherence in the intervention group and to establish that individuals in the control group were not independently consuming a KD, capillary ketone monitoring was prescribed. Of the 14 completers, five participants (KD: n = 4, PCD: n = 1) routinely assessed capillary ketones as instructed. This determination was based on comparisons between data stored in the internal memory of participant ketone monitors and ketone paper logs. Of those five participants, scores recorded on the ketone paper log were consistent with date, time, and values retrieved from monitor internal memory. Seven participants (50%) returned ketone logs with documented ketone values that were inconsistent with the monitor’s internal memory. One PCD participant reported that the monitor and log were misplaced during a move at week 9. Another PCD participant recorded “error” on the telephone log across the 12 weeks. Neither of these participants were able to be reached for telephone visits during weeks 4 and 8, but did return for the week 12 onsite visit. Of the participants on the KD who maintained accurate logs, all four achieved mild ketosis (> 0.5 mmol/L) and the sole participant from the PCD who consistently assessed capillary ketone values did not achieve ketosis.

Aim 2: Explore preliminary cognitive effects of 12-week KD intervention and investigate if cognitive changes observed at week 12 were sustained after a 6-week washout period.

Cognitive data are provided in Table 3. These normed data were adjusted for age and education based on nationally representative data (T-scores), as is routine in neuropsychological measurement. The KD and PCD groups were similar at baseline across all cognitive domains. At week 12, controlling for baseline scores, between group analyses indicated that the KD group performed significantly better on Trails A and Trails B assessments (p = 0.035 and 0.039, respectively; Figure 1) compared to the PCD group. No other between group differences were observed at week 12. No between-group differences were observed at week 18, suggesting the cognitive gains observed in the KD group at week 12 were not sustained after resuming their usual diet.

Aim 3: Examine effects of 12-week KD on rudimentary body composition assessments and cardiometabolic biomarkers in older adult PLWH.

Table 4 shows values for anthropometric measures, serum biomarkers, and metabolic indicators. Overall, the groups exhibited similar body characteristics. This was expected as the diet was intended to maintain weight. Similar lipid profiles and levels of insulin resistance as determined by HOMA-IR were also noted. In short, being overweight (BMI 25.0 kg/m² – 29.9 kg/m²) and obese (30.0 kg/m² – 34.9 kg/m²) as well as insulin resistance was prevalent across both groups at baseline and week 12

Anecdotal Findings

No serious adverse effects were reported. Lipid profiles and weight were relatively unchanged. This was noteworthy given the increased percentage of daily fat consumed in the KD group. One participant was hospitalized for 23 hours during the study for an unrelated illness. Only one participant reported a problem with food delivery; food items, even though packaged in commercial grade coolers, had warmed during the day because the participant was not at home at the time of delivery. An exit interview questionnaire indicated that KD group participants were generally satisfied with the food delivery process, clarity and ease of the food preparation, and palatability of the food. Participants also relayed general satiety with the quantity of food.

Discussion

The purpose of our study was to examine the efficacy and feasibility of a 12-week KD intervention and translate it to a new clinical population of older adults with HIV-associated neurocognitive impairment, a population with well-documented premature cognitive decline. After consumption of a KD for 12 weeks, the intervention group exhibited significant cognitive gains in the domains of executive function, speed of processing, attention, and visuospatial tracking, whereas the PCD group either remained unchanged or continued to decline across the 12-week intervention.

A number of mechanisms may be considered with respect to the significant cognitive gains noted in the KD group. An emerging body of evidence suggests pathways involving glucose hypometabolism and neuroinflammation play a key pathophysiologic role in promoting cognitive impairment. A hallmark feature of glucose restriction as observed during consumption of a KD is the heightened rate of fatty acid oxidation by the liver, which, in turn, promotes a rise in ketone bodies. Neuronal cells are well-equipped to use ketones for energy and neuronal ketone metabolism reduces production of reactive oxygen species (Hartman, 2012). These physiologic adaptations may partially explain the subsequent lower levels of inflammation, decreased inflammatory mediator signaling (i.e., TNFα and interleukins), and more stable synaptic function observed in persons consuming a KD (Branco et al., 2016). As such, it is plausible that these mechanisms translate to improved cognitive function, particularly in the context of cognitive disorders in which maladaptive energy metabolism in the brain likely plays a role (e.g., Alzheimer’s, Parkinson’s, HAND, mild neurocognitive impairment of aging; Daulatzai, 2017). In the context of HIV, the reduction or improvement of premature cognitive aging using non-pharmacologic approaches is particularly important given the excessive polypharmacy burden in adults aging with HIV, independent of cART.

It was not possible for our participants to be blinded to group assignment. Thus, protocol adherence may have been influenced by group assignment. It is also possible that capillary ketone assessment was too complex for some cognitively impaired older adults. Urinary ketone assessment has been used in other low carbohydrate diet studies with moderate success. Taylor, Sullivan, Mahnken, Burns, and Swerdlow (2018) observed urine ketone responses that paralleled serum ketone assessments in a 16-week study of patients with Alzheimer’s disease. Although urinary reagent strip ketone assessment is less precise in comparison to capillary ketone evaluation, urinary ketone detection requires fewer steps and may be more feasible (Taboulet et al., 2007). However, adherence to any home ketone assessment method may be challenging in a vulnerable, cognitively impaired population with pre-existing complex treatment regimens. Thus, researchers may need to consider the feasibility of more frequent laboratory serum ketone sampling.

Researchers have documented the unique challenges of recruiting adults into diet studies in general (Piantadosi et al., 2015) and particularly KD intervention studies (Taylor et al., 2018). Public perception of the KD may be negative because of diet limitations. However, given the extensive polypharmacy burden observed in adult PLWH, non-pharmacological approaches to symptom management are crucial. Thus, developing recruitment strategies that highlight potential functional and quality of life gains are critical. Exclusion of older adult PLWH with diabetes was also a significant barrier to adequate enrollment in our study: Pre-existing diabetes was the second most common co-morbid factor in exclusion (10%). Inefficient energy metabolism beyond the brain may also promote systemic inflammation, furthering cognitive decline. Given the high prevalence of cardiometabolic disease in adult PLWH earlier in the lifespan (Willig et al., 2015), researchers should consider including older adults living with HIV and diabetes. Moreover, it is plausible that the KD may provide even greater cognitive gains in older adults living with HIV and diabetes given the adverse metabolic effects of cART and the metabolic alterations observed in the brain soon after HIV infection.

The lack of cardiometabolic improvement in our intervention group was somewhat unexpected despite eucaloric KD conditions. Mild weight loss, more optimal body composition (e.g., smaller waist circumference), and an improved lipid profile are commonly reported in similar, weight-maintenance ketogenic studies. However, it is plausible that persistent chronic levels of inflammation observed in PLWH and/or cART required to suppress viral load confounded these cardiometabolic indicators. The study was underpowered to detect small but clinically meaningful changes in the lipid profile, heightened insulin resistance, and greater abdominal obesity, due to the exclusion of participants with pre-existing metabolic disease and the small sample size; these have all been reported in the context of cART (Nuvoli et al., 2018).

Our findings were consistent with other KD studies of similar cognitively-impaired, aging populations. In a single-arm, 12-week KD intervention pilot study of 15 participants with Alzheimer’s disease, Taylor et al. (2018) observed significant cognitive improvements as determined by the Alzheimer’s disease Assessment Scale-cognitive subscale that assesses language, praxis, and orientation. Similar findings were reported after consumption of a KD for 8 weeks in cognitively-impaired patients with Parkinson’s disease (Phillips, Murtagh, Gilbertson, Asztely, & Lynch, 2018), and Krikorian et al. (2012) found improved verbal memory performance in participants randomized to a very low-carbohydrate group compared to participants on a high carbohydrate diet.

Limitations

Due to the small sample size, caution should be applied to interpretation of our cognitive performance data. Our sample was also relatively homogenous, limiting generalizability of our findings. Lastly, home ketone monitors determined intervention adherence; however, few participants completed the ketone assessments as prescribed and thus the degree of diet adherence is unknown. However, we were encouraged that statistically significant improvements in measures of executive function and visuospatial tracking were observed, suggesting interventions that target brain metabolism, such as the KD, may benefit older PLWH. How cognitive gains translate to indicators of successful aging (e.g., daily function) remain unknown. Our findings support tolerability and palatability of the KD in older PLWH, but safety was not conclusive given the small sample size, warranting further research.

Conclusion

A growing body of evidence suggests that brain hypometabolism plays a primary role in HIV-associated neurocognitive impairment pathogenesis (Lentz et al., 2011; Vázquez-Santiago et al., 2014). Brain imaging studies indicate ketones provide a better energy source for hypometabolic neurons in comparison to glucose (Croteau et al., 2018). Our findings suggest that (a) a KD may increase the brain’s energy supply, promoting positive downstream effects on cognitive function in older PLWH, and (b) non-pharmacologic approaches to treat chronic co-morbid conditions (such as the KD) decrease the risks of polypharmacy and uniquely benefit this vulnerable population.

Figure 2.

Change in cognitive outcomes by group

*p ≤ 0.05; Trails A: Psychomotor speed; Trails B: Executive function

Key Considerations.

• By improving brain metabolism, the KD may partially reverse some age-related cognitive losses. Future research may consider whether such cognitive gains are associated with everyday function and quality of life.

• In real world living conditions, simple, easy-to-follow nutrition counseling with a special emphasis on nutrition label interpretation is critical.

• The KD safely improves the brain’s energy supply without exacerbating polypharmacy concerns such as the risk of drug-to-drug interactions and drug-related adverse reactions.