A Randomized Clinical Trial on the Impact of Individually Targeted Computerized Cognitive Training on Quality of Life Indicators in Adults With HIV-Associated Neurocognitive Disorder in the Southeastern United States

Abstract

HIV-associated neurocognitive disorder (HAND) is experienced by 30% to 50% of people living with HIV (PLWH), potentially impacting their quality of life (QoL). In the Training on Purpose (TOPS) Study, we investigated whether targeted cognitive training can improve QoL in PLWH with HAND. Using a two-group experimental design, we randomized 109 adults with HAND to either: 1) the Individualized-Targeted Cognitive Training group or 2) a no-contact control group. Those in the training group were assigned 10 hours of cognitive training per two selected cognitive domains (20 hours total) for which impairment was observed. Overall, two patterns emerged. First, significant improvements in measures of everyday cognitive complaints, depression, and mental health were consistently observed following the completion of many cognitive training protocols. Second, immediate and delayed spatial learning and memory training resulted in more significant indicators of QoL improvements compared to the other cognitive domain trainings. The findings suggest that some types of cognitive training may have advantages over others in improving aspects of QoL.

In 2018 in the United States, nearly 51% of people living with HIV (PLWH) were 50 and older, largely due to improved antiretroviral therapy that is extending the lifespan of PLWH to that of the general population (CDC, 2020). By 2030, this percentage is projected to increase to 70% (Wing, 2017). As PLWH age, they are more susceptible than people without HIV to comorbidities such as heart disease, diabetes, and neurovascular changes that may accentuate or accelerate cognitive aging, including HIV-associated neurocognitive disorder (HAND; Waldrop et al., 2021). Because 30% to 50% of PLWH experience HAND, the incidence and severity of HAND may increase in the coming years and diminish quality of life (QoL) (Small & Brew, 2018). In fact, PLWH with HAND have reported QoL issues associated with cognitive complaints, depression/mood problems (Waldrop et al., 2021), poor sleep quality (Campbell et al., 2020), and health-related QoL (Jones et al., 2019). Fortunately, cognitive remediation techniques, such as transcranial direct current stimulation or cognitive training to improve cognition, may improve QoL.

Cognitive training has been used extensively in older adults to protect and improve cognition (Waldrop et al., 2021). In the seminal clinical trial of cognitive training in community-dwelling older adults (without HIV), referred to as the Advanced Cognitive Training In Vital Elderly (ACTIVE) Study, Ball et al. (2002) randomized 2,802 participants into one of four groups: 1) speed of processing (SOP), 2) reasoning training, 3) memory training, or 4) no-contact control. Interestingly, reasoning training and, moreover, SOP training, uniquely enhanced QoL outcomes over 1 to 5 years after training. Benefits included: 1) protecting against depression (Wolinksy et al., 2009) and 2) improving health-related QoL (Wolinksky et al., 2006) and self-rated health (Wolinsky, Mahncke et al., 2010). A related study demonstrated that cognitive training may also improve sleep quality in older adults (Cody et al., 2020). These QoL outcomes are crucial areas in HIV that also require intervention (Vance, Fazeli, Azuero et al., 2018).

Cognitive training for PLWH has been associated with promising cognitive therapeutic effects. In a systematic review of 13 cognitive training studies in PLWH, Vance et al. (2019) observed that most studies found evidence of cognitive improvement after training, especially for domain-specific cognitive training. It remains unclear if the observed training effects are robust over time, as few of the reviewed studies reported follow-up beyond 6 months. Moreover, because the reviewed studies did not include QoL measures and indicators, it is unclear if the documented benefits of cognitive training on QoL observed in other populations transfer to PLWH.

Cognitive training capitalizes on inherent brain plasticity to favorably modulate brain function and morphology, resulting in cognitive improvement (Vance, Fazeli, Cheatwood et al., 2019). Studies suggest that SOP training may decrease reliance on frontally oriented activity by recruiting cognitive resources to more posterior brain regions; this in turn improves SOP which also improves indicators of everyday function and QoL (O’Brien et al., 2013). In one randomized controlled trial, participants in the cognitive training group (n = 34) showed improved sleep quality (i.e., sleep efficiency and sleep onset latency) compared to active control group participants (n = 17; O’Brien et al., 2013). The authors hypothesized that cognitive training improved neural function with an attendant improvement in overall sleep quality, possibly resulting from training-related cognitive exertion, which required more rest.

Given the potential benefits of cognitive training, the Training on Purpose Study (TOPS) was conducted to test the effects of cognitive training that targeted individual-specific cognitive deficits in adults 40 and older to reverse the diagnosis of HAND. Although the targeted training did not reverse HAND, improvement in cognitive function was observed in some domains (Vance et al., 2021). A secondary aim of TOPS, as reported in this article, was to examine whether targeted cognitive training could improve measures of QoL, similar to those described in the ACTIVE study.

Methods

Design Overview

The primary aim of the TOPS study was to reduce the severity of HAND and reverse the diagnosis using a framework to deliver individually targeted domain-specific cognitive training (Vance, Fazeli, Azuero et al., 2018). Employing a neuropsychological algorithm called the Frascati criteria, HAND is diagnosed using a norm-based (age/education) neurocognitive battery when cognitive impairment (<1 SD below the normative mean) is detected in two or more cognitive domains (i.e., attention, SOP, verbal learning and memory, delayed verbal memory, executive functioning, spatial learning and memory, delayed spatial memory, and spatial visualization; Blackstone et al., 2012; Vance, Cody et al., 2017).

In this two-group pre-/post-experimental design, participants 40 years and older with HAND were randomized to either: 1) an Individualized-Targeted Cognitive Training group or 2) a no-contact control group. Those in the training group were assigned 20 hours total of cognitive training targeting two selected cognitive domains (10 hours in each domain) in which impairments were detected. Because of the expense of training and observed attrition of the treatment group during the study, the randomization assignment was altered nearly halfway through the study from a 1:1 treatment allocation to slightly favor random assignment to the treatment group, resulting in a treatment allocation of 1:1.4 that favored the experimental condition. We maintained a stratified randomization method with permuted block and treatment allocation between African Americans/Caucasians and men/women.

Human Subject Protection

Ethical and human subjects approval for this study was granted by the University of Alabama at Birmingham’s Institutional Review Board (protocol number: IRB-F161122002). Participants signed an IRB-approved consent form. Data collection for this study began in July 2017 and ended in January 2019. The ClinicalTrials.gov number of this study is NCT03122288.

Participants

Flyers were posted at AIDS service organizations, an HIV/AIDS clinic, public businesses, and places of worship to recruit participants. As seen in the Consort Figure (Figure 1), 139 people were called and asked questions over the phone to determine eligibility criteria, including: (a) diagnosed with HIV for at least 1 year; (b) 40 years old or older; (c) not diagnosed with a severe neuromedical disorder (e.g., traumatic brain injury, Alzheimer’s disease); (d) not presently treated with radiation or chemotherapy; (e) fluent in English; (f) not legally deaf or blind; and (g) resides within 100 miles of the research center. Based on the phone screening, 135 eligible participants were consented at baseline and completed the pretest assessment. Using the Frascati criteria with the participants’ baseline cognitive performance, 109 (80.74%) participants met the HAND criteria and were randomized to either the control (n = 45) or active treatment (n = 64) group. Those who did not meet the criteria for HAND were withdrawn from the study. The observed HAND rate of 80% was much higher than the previously reported prevalence rates of 30% to 50% (Small & Brew, 2018). We posit that our recruitment materials, designed to target participants for a cognitive training program, may have attracted PLWH with cognitive symptoms. Compensation was provided for the pretest/post-test assessments ($75/assessment) and cognitive training sessions ($15/1-hour session of cognitive training).

Figure 1.

Tops Consort Diagram

Although the SAGER guidelines (Heidari et al., 2016) were largely followed, self-reported gender was controlled. In addition, the proportion of women in each group was approximately that of the HIV population in the United States; therefore, we have no reason to suspect that gender differentially affected the observed treatment effects.

Measures

Both pretest (baseline) and post-test assessment batteries required approximately 1.5 to 2 hours to complete. The neuropsychological tester was trained by the Principal Investigator (PI) (DEV) and was routinely monitored by the PI to assure fidelity to the testing protocol.

Sample Characteristics

Demographics/health information.

Self-reported demographic (e.g., gender, education, age) and health information (i.e., medical conditions such as diabetes, medications) was collected at baseline.

HIV health status.

At study onset, participants’ medical providers were contacted to verify the participants’ HIV status and acquire the following information: current and nadir CD4+ lymphocyte count, current viral load, prescribed antiretroviral therapy (ART) medication (yes/no), current ART regimen, and years diagnosed with HIV.

Cognitive domains and HAND diagnosis.

Employing normed-based cognitive performance tests at baseline (pretest) and post-test, eight cognitive domains were assessed. These gold standard cognitive tests are commonly utilized in neuroHIV research with excellent validity and reliability (Blackstone et al., 2017; Woods et al., 2008). We used t-scores of the cognitive performance tests to estimate clinical ratings for each cognitive domain, ranging from 1 (above average, that is, if the t-score is 55 or higher) to 9 (severe impairment, that is, if the t-score is 19 or lower). This cognitive performance battery comprised: (1) SOP (Stroop Color Naming Test, Trails A); (2) attention (Paced Auditory Serial Attention Test); (3) verbal learning and memory (Hopkins Verbal Learning Test–Revised); (4) delayed verbal learning and memory (Hopkins Verbal Learning Test/Delayed–Revised); (5) executive function (Stroop Interference, Trails B); (6) spatial learning and memory (Benton Visual Retention Test–Revised); (7) delayed spatial learning and memory (Benton Visual Retention Test Delayed–Revised); and (8) spatial visualization (WAIS IV Block Design). Based on baseline cognitive performance, cognitive deficits were identified and targeted for training purposes (see the following Procedures/Treatment section).

Using the Frascati criteria (Antinori et al., 2007), based on each participant’s cognitive performance test scores, a Global Clinical Rating score was calculated that ranged from 1 (above average) to 9 (severe impairment). Specifically, if cognitive impairment was detected (i.e., 5-9 range) in two or more cognitive domains, the Global Clinical Rating score reflected the higher level of impairment. Thus, a global deficit score of 5 or higher indicates HAND (Blackstone et al., 2017). In clinical practice, other factors, such as comorbidities or impairments in everyday functioning, are taken into consideration when diagnosing HAND; however, in our study, only cognitive performance was used to make this diagnosis (i.e., all participants met criteria for at least asymptomatic neurocognitive impairment), as cognition is the key component of HAND. (For details about HAND, the cognitive measures, and TOPS protocol procedures, see Vance, Fazeli, Azuero et al., 2018).

Outcomes: QoL Indicators

The following QoL indicators were measured because they represent broad domains of well-being and QoL.

Depression.

Depressive symptomology was measured by the Center for Epidemiologic Studies Depression Scale-Revised (CESD; Radloff, 1977). A score of 16 or greater indicates clinically relevant depressive symptomology. Cronbach’s α has been reported in prior research as 0.88.

Cognitive complaints.

The Cognitive Failures Questionnaire (CFQ) consists of 25 Likert-type items that document common everyday cognitive complaints. For example, perceived memory deficits are assessed by items probing forgetfulness (e.g., “Do you forget where you put something like a newspaper or book?”). Responses range from 0 (never) to 4 (very often). CFQ ranges from 0 to 100; the total score was used in the analyses. Higher scores indicate greater subjective cognitive complaints (Broadbent et al., 1982). Cronbach’s α has been reported as 0.89 in a sample of healthy older adults (Knight et al., 2004).

Sleep quality.

The Pittsburgh Sleep Quality Index (PSQI) consists of 19 items to assess participants’ sleep quality in the past month. The PSQI subscales include sleep efficiency, sleep medication use, sleepiness, sleep duration, sleep disturbance, sleep latency onset, and dysfunction due to sleepiness. Participants reported the extent to which various factors affect their sleep. Subscale scores were summed to create a global score ranging from 0 to 21; scores greater than 5 indicate poor sleep quality. The seven subscales have shown high internal consistency reliability, as evidenced by Buysse et al. (1989), while the entire instrument has shown high retest reliability (r = 0.87; Backhaus et al., 2002).

Self-rated health.

Self-rated health was measured by a single item that asked participants to rate their health from 1 (excellent) to 5 (poor). This single item has been shown to be a valid measure of self-rated health (Wuorela et al., 2020).

Health-related quality of life (HRQoL).

The Medical Outcomes Study–HIV (MOS-HIV) was used to measure HRQoL. Adapted from a previous version of the Medical Outcomes Survey SF-36 Health-related QoL (HRQoL) Questionnaire, the MOS-HIV was used to capture HRQoL data in participants with HIV (Wu et al., 1997; Wu et al., 1991). This instrument is a commonly used and well-validated self-report questionnaire consisting of 35 items across several areas (Cooper et al., 2017). The present study analyzed the mental health and physical health composite scores (MHS and PHS). Values for each domain were converted to a 100-point scale, with higher values indicating better HRQoL. Cronbach’s α for the MHS composite ranges from 0.91 to 0.94, and the Cronbach’s α for the PHS composite ranges from 0.90 to 0.92 (Revicki et al., 1998).

Procedures/Treatment Design and Rationale

The target time period between pretest and posttest was 10 to 12 weeks. A cognitive trainer supervised the participants in the cognitive training group. Participants came to the research center where this cognitive trainer could assist them in logging onto the program, provide encouragement, and answer questions as needed. This helped ensure treatment fidelity. Participants in the cognitive training group received a total of 20 hours of training, specifically 10 hours in each of two cognitive domains in which impairment was detected during baseline testing.

Breaks were encouraged as needed because cognitive training can be monotonous and/or fatiguing. The cognitive training dose of 10 hours for each of the targeted cognitive domains was based on the meta-analysis of 52 cognitive training studies involving community-dwelling older adults; Lampit et al. (2014) observed a U-shape dosage-therapeutic response. Other studies support 10 hours as an optimal training dose for improving function in a particular cognitive domain (Ball et al., 2002; Vance, Fazeli, Cheatwood et al., 2019; Vance et al., 2012).

As participants may have impairments in multiple cognitive domains, 10 hours of training per impaired domain may not be feasible due to excessive participant burden. To address this, a cognitive training algorithm was created that theoretically would best target the cognitive domains needed to reverse a HAND diagnosis. Thus, the three-step Individualized-Targeted Cognitive Training Framework (Vance, Fazeli, Azuero et al., 2018) was created. In step one, if cognitive impairments were detected in either SOP and/or attention, participants were automatically assigned cognitive training in these domains. Based on the Salthouse’s Diminished SOP Theory of Aging (Salthouse, 1996) and the Wickens Applied Attention Theory (Wickens et al., 2013), SOP and attention are foundational cognitive domains on which other cognitive domains depend; therefore, improving functioning in these domains may transfer to improvements in other cognitive domains. In step two, if participants did not show impairments in SOP or attention, they were assigned cognitive training in the least compromised but still impaired (1 SD or below the demographically adjusted mean) cognitive domain. This approach was informed by the goal of reversing a HAND diagnosis, which is most probable when impairment is mild rather than severe. Mild impairment is more likely to indicate sufficient cognitive reserve in the affected domain to allow benefit from the targeted cognitive training. Finally, in step three, participants were directed to interact with the computerized modules/exercises that corresponded to the two cognitive domains selected from steps one and two. Training sessions usually lasted 1 hour, with no more than 2 hours of training at a time to prevent fatigue; however, we allowed participants to be somewhat flexible to accommodate travel arrangements and other scheduling conflicts. More information about the cognitive training modules for each domain can be found at https://www.brainhq.com and Vance, Fazeli, Azuero et al. (2018).

Data Analysis

Because of the exploratory nature of this small pilot study, the sample size was not powered for strict statistical inference (Leon et al., 2011). Instead, effect size cutoffs were applied for Cramer’s V (small ~0.1, medium ~0.3, large ~0.5, for cross-tabulations comparing two groups) and Cohen’s d (small ~0.2, medium ~0.5, large ~0.8; Cohen, 1988). Analyses were conducted using R software version 4.0.3 (R Core Team, 2020).

We first investigated the balance between the two groups’ baseline characteristics (e.g., gender, age, race, HIV markers [i.e., prescribed ART, years diagnosed/living with HIV, current CD4+ T lymphocyte count/mm³, nadir CD4+ T lymphocyte count/mm³]) using measures of effect size. These baseline characteristics were also examined for association with attrition from the study. Next, for each QoL indicator, we generated descriptive statistics at the preintervention and postintervention time-points by group (overall and matched by assigned training or hypothetical assigned training). Furthermore, linear mixed-effects models for repeated measures with a random effect for participant were used to approximate the between-group difference in change from preintervention to postintervention, using a time-by-group interaction coefficient. We then fit all models adjusted with pertinent covariates associated with missingness. This is an appropriate approach to handle Missing At Random (Allison, 2009; Groenwold et al., 2011). Because there were only two time-points, time was used as a categorical fixed-effect, and the model included a time-by-group interaction to estimate intervention effect. Also, in small samples, the simpler specification with random effect for subject prevents potential convergence problems while still accounting for covariance among repeated measurements. Because there were only two time-points, time was used as a categorical fixed-effect. Therefore, using a time-by-group interaction in the model, we computed means at pre- and post-times by group (i.e., we obtained four time-by-group cells), as opposed to slopes over a number of weeks or days. Thus, we were able to use Cohen’s d as a measure of effect between group means. Note that what is compared is the mean change between groups, which is estimated by the time-by-group interaction.

Following an intention-to-treat approach, the models were fitted with all available data, and relevant baseline characteristics associated with attrition were used as controlling covariates to address bias resulting from missing data (Groenwold et al., 2011). An effect size is a measure that estimates the magnitude of an association; we therefore used effect sizes to estimate the magnitude of in-sample association between dropout and each potential baseline covariate. The variance components from each model were used to estimate a pooled SD for each QoL measure, which allowed us to standardize the interaction coefficient and provided a measure of effect size (Cohen’s d). For significance tests of the interaction effect, we employed the Kenward-Roger approximation to degrees of freedom. As a sensitivity analysis, a completer-only analysis was conducted, including training participants who completed training and had complete postintervention data as well as control participants with postintervention data.

Using baseline data of all individuals randomized for the study (N = 109), Pearson correlation coefficients between QoL indicators and cognitive domain scores were estimated and interpreted (small ~0.1, medium ~0.3, large ~0.5; Cohen, 1988). To evaluate fair comparisons between study groups across the various cognitive training domains, in the analysis only participants in the control group were assigned the same cognitive training assignment using the same algorithm had they been randomly assigned to the individualized-training group (Imbens & Rubin, 2015).

Results

Sample Characteristics

Table 1 shows baseline characteristics of the randomized participants (N = 109). On average, participants were 53.26 (SD = 6.7) years old, the majority were African American (84.4%), and about two-thirds were male (69.7%). Participants reported an average of 12.2 years of education, and their average annual income was $17,100. Concerning depressive symptomology, their average CESD score was 18.1; 65 (59.6%) had a CESD ≥16 indicative of clinically relevant depressive symptomology. Participants had an average of five (SD = 4.1) prescribed medications. The majority of participants were on ART (79.8%), had been diagnosed with HIV for an average of 16.6 years (SD = 8.1), had an average current CD4+ T lymphocyte of 670.6 (SD = 414.2.8) count/mm³, and their lowest average recorded CD4+ T lymphocyte was 360.7 (SD = 340.3) count/mm³. Using effect size, no major chance imbalance was detected between study groups, except for current CD4+ T lymphocyte. A few participants had large CD4+ T lymphocyte counts that skewed the right tail of the distribution, but after a log transformation to address this skewness, the between-group standardized difference decreased to d = 0.26 (a small effect size). Finally, the mean time between pretest and post-test was 85.25 days (SD = 48.69), with a nonsignificant difference between the training group (85.92 days) and no-contact control group (84.48 days; (p = 0.89).

Table 1.

Sample Characteristics (N = 109)

Variable	No-Contact Control Group (n = 45)			Individualized-Targeted Cognitive Training (n = 64)			Effect size
	n (%)	Mean (SD)	Median (IQR)†	n (%)	Mean (SD)	Median (IQR)†
Age		53.78 (7.17)	54 (50 - 58)		53.42 (6.42)	54 (49 - 58)	d=0.05
Race/ethnicity							V=0.09
African America	39 (86.67%)			53 (82.81%)
Caucasian	6 (13.33%)			10 (15.63%)
Other	0 (0.00%)			1 (1.56%)
Gender							V=0.06
Male	30 (66.67%)			46 (71.88%)
Female	15 (33.33%)			18 (28.13%)
Yearly household income ($10K)		1.71 (1.31)	1 (1 - 2)		1.70 (1.40)	1 (1 - 2)	d=0.01
Years of education		12.4 (2.30)	12 (11 - 13)		12.13 (2.39)	12 (11 - 13)	d=0.12
Alcohol use* (no. of drinks/day)		1.73 (0.81)	2 (1 - 2)		1.7 (0.95)	1 (1 - 2)	d=0.03
Tobacco use (cigarettes/day)		4.02 (6.49)	0 (0 - 5)		4.08 (5.00)	2 (0 - 8)	d=0.01
CES-Depression		17.24 (10.36)	18 (11 - 21)		18.73 (10.71)	17.50 (12 - 23)	d=0.14
Number of prescribed medications		5.02 (4.25)	4 (2 - 6)		4.97 (3.97)	5 (3 - 6.50)	d=0.01
Prescribed ART (yes)	33 (73.33%)			54 (84.38%)			V=0.13
Years diagnosed/living with HIV		16.33 (8.36)	16 (10 - 22)		16.81 (7.95)	18 (10 - 22)	d=0.06
Current CD4+ T lymphocyte count/mm3		752.98 (430.88)	671 (463 - 978)		610.82 (394.38)	548.50 (355 - 820)	d=0.35
Nadir CD4+ T lymphocyte count/mm3		400.34 (378.10)	308 (126 - 513)		332.52 (310.87)	270.50 (146 - 438)	d=0.2

Notes. d = Cohen’s d (small~0.2, medium~0.5, large~0.8); V = Cramer’s V (small~0.1, medium~0.3, large~0.5); ART = antiretroviral therapy; CES-Depression = Center for Epidemiological Studies Depression Scale; $10K = 10,000 dollars.

^*Alcohol use ranges from 1 (not applicable/don’t drink), 2 (one to two drinks), 3 (three to four drinks), 4 (five to six drinks), 5 (seven to nine drinks), and 6 (10 or more drinks);

^†No decimals were added to integer numbers.

Attrition

As seen in Table 2, when comparing participants who completed the study (i.e., completed the post-test assessment) and those who dropped from the study (n = 21), those who dropped were more likely to be assigned to cognitive training (n = 16, 76.2% vs n = 48, 54.5%, V=0.17); had a tendency to have fewer years of education (M[SD] = 11.4[2.6] vs 12.4[2.2], d = 0.46); had a tendency to be younger (M[SD] = 50.9[4.7] vs 54.2[6.97], d = 0.50); and had a tendency to report higher depressive symptomatology (CESD: M[SD] = 20.7[9.6] vs17.54[10.7], d = 0.31). The characteristics with the largest magnitude of association with attrition (years of education, age) were used as adjusting covariates in models estimating results shown in Tables 2 and 3.

Table 2.

Attrition Characteristics of the Sample (N = 109)

Variable	Stayed to Post-Test (n = 88)			Dropped Out (n = 21)			Effect size
	n (%)	Mean (SD)	Median (IQR)†	n (%)	Mean (SD)	Median (IQR)†
Group assignment							V=0.17
Cognitive training	48 (54.55%)			16 (76.19%)
Control	40 (45.45%)			5 (23.81%)
Age		54.2 (6.97)	55 (50 - 58)		50.9 (4.72)	51 (47 - 54)	d=0.5
Gender							V=0.07
Male	16 (76.19%)			60 (68.18%)
Female	5 (23.81%)			28 (31.82%)
Race/ethnicity							V=0.08
African America	17 (80.95%)			75 (85.23%)
Caucasian	4 (19.05%)			12 (13.64%)
Other	0 (0.00%)			1 (1.14%)
Education (years)		12.44 (2.24)	12 (12 - 13)		11.38 (2.64)	11 (9 - 12)	d=0.46
Household income ($10K)		1.76 (1.46)	1 (1 - 2)		1.48 (0.75)	1 (1 - 2)	d=0.21
Years diagnosed with HIV		17.03 (8.08)	17.50 (10 - 23)		14.86 (8.07)	16 (9 - 21)	d=0.27
Current CD4+ T lymphocyte count/mm3		691.52 (435.84)	655.50 (392 - 914)		584.95 (303.77)	554 (434 - 784)	d=0.26
Nadir CD4+ T lymphocyte count/mm3		359.61 (352.95)	268 (141 - 473)		364.95 (291.20)	323 (161 - 507)	d=0.02
Number of prescribed medications		5.10 (4.22)	5 (3 - 6)		4.52 (3.40)	4 (2 - 6)	d=0.14
Prescribed ART (yes)	18 (85.71%)			69 (78.41%)			V=0.07
CES-Depression		17.5 (10.72)	17 (10.5 - 22)		20.71 (9.59)	19 (15 - 23)	d=0.31
Global Cognitive Score*		7.34 (1.45)	8 (6 - 9)		7.19 (1.47)	7 (6 - 8)	d=0.1
Alcohol use** (no. of drinks)		1.82 (0.93)	2 (1 - 2)		1.29 (0.56)	1 (1 - 1)	d=0.61
Tobacco use (cigarettes/day)		3.72 (5.50)	0 (0 - 6.50)		5.48 (6.11)	4 (0 - 10)	d=0.31

Notes. d = Cohen’s d (small~0.2, medium~0.5, large~0.8); V = Cramer’s V (small~0.1, medium~0.3, large~0.5); ART = antiretroviral therapy; CES-Depression = Center for Epidemiological Studies Depression Scale; $10K = 10,000 dollars.

^*Alcohol use ranges from 1 (not applicable/don’t drink), 2 (one to two drinks), 3 (three to four drinks), 4 (five to six drinks), 5 (seven to nine drinks), and 6 (10 or more drinks);

^†No decimals were added to integer numbers.

Table 3.

Quality of Life Baseline and Post-Test Scores by Treatment Group (N = 109)

QoL Indicators	No-Contact Control Group		Individualized-Targeted Cognitive Training		Difference in Change
	Baseline (n = 45)	Post-test (n = 40)	Baseline (n = 64)	Post-test (n = 48)	Estimate (SE)**	p-value	Effect Size, Cohen’s d
	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)
CES-Depression*	17.2 (10.4)	16.6 (10.6)	18.7 (10.7)	16.4 (11.4)	−2.41 (1.78)	0.18	−0.23
Cognitive Failures Questionnaire*	41 (19.8)	38.4 (19.2)	41.5 (23.9)	38.6 (21.1)	−2.75 (3.36)	0.415	0.13
Pittsburgh Sleep Quality Index*	9.7 (4.3)	10.2 (5.1)	9.5 (4.8)	9.3 (4.8)	−0.71 (0.77)	0.36	−0.15
Self-rated Health*	2.7 (1.1)	2.2 (1)	2.6 (1.2)	2.4 (1.1)	0.12 (0.21)	0.562	0.11
MOS-HIV – Mental Health Composite	43.9 (10.9)	49.7 (11.1)	42.7 (12.1)	49 (11.4)	1.44 (1.77)	0.418	0.13
MOS-HIV – Physical Health Composite	43.4 (11.5)	47.3 (12.6)	44.7 (9.7)	47.2 (12.9)	−1.21 (1.92)	0.532	−0.11
	Completer-only Analysis
	Baseline (n = 40)	Post-test (n = 40)	Baseline (n = 41)	Post-test (n = 41)
	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)
CES-Depression*	16.1 (9.1)	16.6 (10.6)	20 (11.6)	17.4 (11.7)	−3.04 (1.91)	0.115	−0.29
Cognitive Failures Questionnaire*	39 (18.3)	38.4 (19.2)	44.1 (24.9)	40.7 (21.3)	−2.79 (3.62)	0.443	−0.13
Pittsburgh Sleep Quality Index*	9.7 (4.5)	10.2 (5.1)	9.7 (5.2)	9.4 (5.2)	−0.77 (0.81)	0.349	−0.16
Self-rated Health*	2.6 (1)	2.2 (1)	2.7 (1.1)	2.4 (1.1)	0.08 (0.23)	0.711	0.08
MOS-HIV – Mental Health Composite	45 (9.4)	49.7 (11.1)	41.4 (13.2)	48.1 (11.4)	1.92 (1.88)	0.311	0.18
MOS-HIV – Physical Health Composite	43.9 (11.8)	47.3 (12.6)	44.9 (10.6)	47.3 (12.6)	−1.01 (2)	0.616	−0.08

Notes. Cohen’s d = small ~0.1, medium ~0.3, large ~0.5;

^*Higher scores indicate poorer quality of life.

^**Group by time interaction coefficient in linear mixed-effect model adjusted for baseline age, alcohol use, and education.

Effects of Training on QoL Indicators

Table 3 displays descriptive statistics of training effect on QoL indicators regardless of training received (or hypothetically assigned for control participants). In other words, we compared the full training group to the no-contact control group. On average, compared to the no-contact control group, a small beneficial effect was observed for reducing CES-Depression scores (d = −0.23). Effects for other measures were of trivial magnitude. Table 3 also shows results of the completer-only analysis, which indicated similar conclusions compared to analysis with all available data, although there was some variability in the estimates, as expected from changes in small samples.

Table 4 presents measures of training effect on QoL indicators, comparing participants trained in a specific domain vs their respective matched control participants. With their control counterparts as a comparison, participants who received SOP training (n = 22) demonstrated, on average, small improvements in sleep (PSQI, d = −0.22), and differences of trivial magnitude in the other measures.

Table 4.

Individualized Trainings on Quality of Life Indicators (N = 109)

	Cognitive Domain Training
	SOP	Attention	Verbal Learning and Memory	Delayed Verbal Memory	Executive Functioning	Spatial Learning and Memory	Delayed Spatial Memory	Spatial Visualization
N Training at baseline	29	21	21	17	5	11	11	13
N Training at post-test	22	16	15	13	4	8	8	10
N Control at baseline	20	18	11	13	6	9	5	8
N Control at post-test	19	17	9	10	5	8	5	7
QoL Indicators	Cohen’s d	Cohen’s d	Cohen’s d	Cohen’s d	Cohen’s d	Cohen’s d	Cohen’s d	Cohen’s d
	(p-value)	(p-value)	(p-value)	(p-value)	(p-value)	(p-value)	(p-value)	(p-value)
CES-Depression	−0.05	−0.29	−0.34	−0.27	0.5	−0.52	−0.14	−0.66
	(0.849)	(0.328)	(0.247)	(0.491)	(0.189)	(0.282)	(0.751)	(0.079)
Cognitive Failures Questionnaire	−0.1	0.18	−0.21	−0.27	−0.22	−0.34	−0.35	0.03
	(0.704)	(0.523)	(0.418)	(0.343)	(0.282)	(0.375)	(0.217)	(0.94)
Pittsburgh Sleep Quality Index	−0.22	−0.49	0.36	0.21	−0.09	−0.54	−0.76	0.15
	(0.326)	(0.061)	(0.293)	(0.587)	(0.8)	(0.121)	(0.091)	(0.681)
Self-rated Health	−0.13	0.31	0.19	0.3	0.29	−0.13	−0.42	0.62
	(0.625)	(0.383)	(0.691)	(0.372)	(0.631)	(0.734)	(0.086)	(0.227)
MOS-HIV – Mental Health Composite	−0.08	0.19	0.32	0.29	−0.39	0.39	0.34	−0.12
	(0.773)	(0.511)	(0.22)	(0.364)	(0.255)	(0.416)	(0.38)	(0.719)
MOS-HIV – Physical Health Composite	−0.08	−0.23	−0.18	−0.18	−0.04	−0.14	0.33	−0.14
	(0.746)	(0.38)	(0.634)	(0.671)	(0.933)	(0.718)	(0.31)	(0.699)

Notes. Cohen’s d = small ~0.1, medium ~0.3, large ~0.5.

Participants who received Attention training (n = 16) showed small to medium improvements on CES-Depression (d = −0.29) and Pittsburgh Sleep Quality Index (d = −0.49); however, they showed small to moderate detrimental effects on self-rated health (d = 0.31) and MOS-HIV–Physical Health Composite (d = −0.23; Table 4).

Participants who received Verbal Learning and Memory training (n = 15) showed small to moderate beneficial effects on CES-Depression, the Cognitive Failures Questionnaire, and the MOS-HIV–Mental Health Composite (|d| ranging from 0.21 to 0.34); however, they showed small to moderate detrimental effects on Pittsburgh Sleep Quality Index (d = 0.32; Table 4).

Participants who received Delayed Verbal Memory training (n = 13) showed small to medium improvements on CES-Depression, the Cognitive Failures Questionnaire, MOS-HIV–Mental Health Composite (|d| ranging from 0.27 to 0.29); however, they showed small to moderate detrimental effects on Pittsburgh Sleep Quality Index (d = 0.21) and self-rated health (d = 0.3; Table 4).

Participants who received Executive Functioning training (n = 4) showed high variation on effects, as expected from the very small sample size for this subgroup, and therefore these estimates should be interpreted with caution. These participants showed small improvements on the Cognitive Failures Questionnaire (d = −0.22); however, on average they also showed small to large detrimental effects on CES-Depression, self-rated health, and MOS-HIV–Mental Health Composite (|d| ranging from 0.29 to 0.5; Table 4).

Participants who received Spatial Learning and Memory training (n = 8) also had, on average, small to large improvements on CES-Depression, Cognitive Failures Questionnaire, Pittsburgh Sleep Quality Index, and MOS-HIV–Mental Health Composite (|d| ranging from 0.34 to 0.52; Table 4).

Participants who received Delayed Spatial Memory training (n = 8) showed moderate to large improvement on all measures (i.e., Cognitive Failures Questionnaire, Pittsburgh Sleep Quality Index, and MOS-HIV–Mental Health Composite and MOS-HIV Physical Health) except CES-Depression, with |d| ranging from 0.33 to 0.76 (Table 4).

Participants who received Spatial Visualization training (n = 10) showed high variation on effects compared to those in the control group that would have been assigned this training. This subgroup had an improvement on CES-Depression of medium magnitude (d = −0.66); however, the subgroup also showed detrimental effects of medium magnitude on self-rated health (d = 0.62; Table 4).

Association Between QoL Indicators and Baseline Cognitive Domain Scores

Table 5 presents sample correlation coefficients between QoL Indicators and baseline Cognitive Domain Scores using data from the N = 109 individuals randomized for the study. All correlations were of small magnitude (|r|<0.2). These results suggested that the QoL indicators and baseline Cognitive Domain Scores were not strongly related constructs in the sample.

Table 5.

Associations of Baseline Cognitive Domain on Quality of Life Indicators (N = 109)

	Baseline Cognitive Domain Scores
QoL Indicators	Global Clinical Rating Score	SOP	Attention	Verbal Learning and Memory	Delayed Verbal Memory	Executive Functioning	Spatial Learning and Memory	Delayed Spatial Memory	Spatial Visualization
	r	r	r	r	r	r	r	r	r
	(p-value)	(p-value)	(p-value)	(p-value)	(p-value)	(p-value)	(p-value)	(p-value)	(p-value)
CES-Depression	0.10	0.11	0.20	0.15	0.07	−0.08	0.04	−0.01	0.10
	(0.312)	(0.234)	(0.041)	(0.132)	(0.453)	(0.392)	(0.707)	(0.936)	(0.295)
Cognitive Failures Questionnaire	0.02	−0.05	0.12	0.13	−0.06	−0.12	−0.04	−0.11	0.00
	(0.845)	(0.592)	(0.196)	(0.19)	(0.54)	(0.221)	(0.677)	(0.269)	(1)
Pittsburgh Sleep Quality Index	−0.01	0.10	0.05	−0.05	0.00	−0.06	−0.09	−0.11	0.01
	(0.887)	(0.29)	(0.605)	(0.639)	(0.992)	(0.568)	(0.347)	(0.267)	(0.91)
Self-rated Health	0.04	0.07	0.10	0.12	0.00	−0.10	−0.03	−0.15	−0.03
	(0.664)	(0.456)	(0.324)	(0.227)	(0.973)	(0.322)	(0.747)	(0.132)	(0.762)
MOS-HIV – Mental Health Composite	−0.14	−0.12	−0.20	−0.15	−0.05	0.08	−0.03	−0.03	0.01
	(0.152)	(0.213)	(0.041)	(0.13)	(0.625)	(0.429)	(0.76)	(0.764)	(0.954)

Note. r = Pearson’s r (small ~0.1, medium ~0.3, large ~0.5).

Discussion

This study was innovative in two ways. First, we examined if individualized cognitive training could improve overall QoL outcomes in PLWH, thus advancing the symptom science in this area. Second, we examined whether specific types of cognitive training showed a differential impact on such QoL indicators. Overall, the results were mixed regarding the effects of cognitive training on producing improvements in QoL indicators, which consisted of depressive symptomology, everyday cognitive complaints, sleep quality, self-rated health, and HRQoL. Compared to the no-contact control group, those assigned to an individualized cognitive training group experienced some benefit on measures of depression, mental health, sleep, and cognitive failures. Additionally, a mixture of improvements and declines was evidenced, with 20 analyses revealing QoL improvements and nine analyses revealing QoL declines.

Across the individual cognitive training protocols, four patterns emerged. First, improvements in everyday cognitive complaints were observed in five out of the eight cognitive training protocols. Second, improvements in depression symptomology were observed in five out of the eight cognitive training protocols. Third, improvements on mental health (measured by the MOS-HIV) were observed in four of the cognitive training protocols; however, one cognitive training protocol, executive functioning training, produced a detrimental effect on this indicator. Fourth, spatial learning and memory training and delayed spatial learning and memory training resulted in noticeably more improvements in QoL indicators than any of the other cognitive training protocols. Of note, detrimental effects on the QoL indicators were not consistently observed across the cognitive training protocols except for self-rated health, but in these four cases, the effects were small to moderate. Collectively, these findings suggest that these observed patterns, especially for spatial learning and memory training and delayed spatial learning and memory training, may be particularly robust for improving QoL in PLWH.

Compared to cognitive outcomes in the HIV cognitive training literature, the impacts of cognitive training on QoL outcomes have been understudied in PLWH (Vance et al., 2019). Although there are many avenues of discussion for the current study, we focus on the findings that can be compared with previous cognitive training studies using similar QoL outcomes in older adult samples. Starting with everyday cognitive complaints, Canevelli et al. (2013) found six studies testing nonpharmacological cognitive interventions in older adults who self-reported cognitive complaints. Only one study reported an improvement in subjective memory functioning in the experimental group compared to the control group (Valentijn et al., 2005). This experimental group used a variety of approaches to target memory improvement; in contrast, our study found improvements in cognitive complaints following memory-focused training (i.e., verbal memory, delayed verbal memory, spatial memory, delayed spatial memory) as well as following non–memory-focused training (i.e., executive function, spatial visualization); this suggests that several types of cognitive training may benefit perception of memory performance and perhaps other types of cognitive performance.

In terms of cognitive training and its benefit for improving depressive symptomology, using the data from the ACTIVE study, Wolinsky et al. (2009) reported that SOP training reduced the risk of clinically worsening depressive symptoms 1 and 5 years out from training and development of suspected clinical depression at 1 year. Unfortunately, in our study, a medium training effect was seen for SOP on mental health (measured by the MOS-HIV) but not for depressive symptomology (measured by the CESD). More research is needed into the specific training protocols that are sensitive to clinical depression and if results may differ by the clinical sample of interest. Additionally, the effects of antidepressant medications should be analyzed, as previous meta-analyses have shown significant effects of cognitive training on depressed mood in those with major depressive disorder and mild cognitive impairment (Hill et al., 2017; Motter et al., 2016).

Findings are also mixed for HRQoL. Eeva-Lissa Kallio et al. (2018) did not find a significant training effect on HRQoL in a sample of older adults with dementia when using a pencil and paper executive function training protocol over a 9-month period. In contrast, Wolinsky et al. (2006) found that those who were administered SOP training experienced less extensive HRQoL decline 5 years after baseline, compared to those in the memory training and reasoning training groups in the ACTIVE trial. In the current study, while the findings were mixed across the two HRQoL composites, improved scores on the mental health composite score were seen for the following training protocols: 1) spatial learning and memory, 2) delayed spatial learning and memory, 3) verbal learning and memory, and 4) delayed verbal learning and memory. In contrast, decrements were seen for executive function training on the mental health composite score and for attention training on the physical health composite score. It is unclear why these aberrant findings occurred, but two ideas emerge. First, with the number of statistical comparisons, this could have been a statistical artifact and spurious. Second, improvement in executive function and, in particular, attention could have resulted in greater awareness of one’s mental and physical functioning, which contributed to a negative evaluation of one’s circumstances and HRQoL. What remains unknown is what specific factor regarding cognitive training may not be beneficial for the way health affects an individual in completing their daily activities (e.g., responsibilities at a job, completing housework or schoolwork).

Implications for Practice

Combined with the cognitive aging and neuroHIV literature, implications for clinical practice are advanced. First, as many PLWH age and experience declines in cognitive and everyday functioning, some evidence suggests that certain types of cognitive training, such as SOP, may improve QoL indicators (Wolinsky, Vander Weg, et al., 2010). Unfortunately, not all types of cognitive domain training are comparable, and their effects on cognition may vary depending on the domain of training, as was observed in the current study. In fact, some cognitive training produces more robust cognitive effects in certain cognitive domains than others; likewise, it is not surprising to observe that there were various effects on QoL, as well. Thus, as clinicians make recommendations about cognitive training to their patients with HIV, clear expectations about the benefits and limitations of such training should be articulated.

Second, as cognitive training may be used to reduce anxiety or concern about HAND and cognitive aging, clinicians can recommend cognitive training to patients concerned about cognitive loss as they age. In general, cognitive training itself is a safe approach that shows promise for improving both cognitive function and everyday functioning, although the impact on QoL was relatively mixed in our study. In the TOPS study, participants with HAND were informed that they performed at a suboptimal neurocognitive level that met the criteria of HAND. At post-test, these responses were qualitatively assessed to determine how participants interpreted that information. As documented elsewhere, several themes emerged, including: Anxiety, Sadness, Unexpected, Concerned, and Not Concerned/No Reaction. Albeit, other more positive themes emerged including: Confirmation (acknowledging they have a cognitive impairment), Knowledge Seeking, and Desire to Improve (please refer to Vance, Jensen et al., 2019). Although these participants expressed trepidation and ambivalence about having HAND, they voiced that they also wanted tools on how to address this diagnosis and protect their cognitive ability as they age. Cognitive training could be such a tool to empower them, reduce anxiety, and improve QoL. Unfortunately, given the time requirements to administer cognitive training, its use in a typical clinical environment is limited due to the short visits. For that reason, referrals of cognitive training are recommended for rehabilitation clinics or home use where patients can engage with the software for the hours they need.

Implications for Research

Two research implications are noted. First, this study tested the efficacy of a framework designed to select what type of cognitive training should be assigned to participants to reverse a diagnosis of HAND. In that initial study, although cognition was improved, the diagnosis of HAND was unaltered (Vance, Fazeli, Azuero, Wadley et al., 2021). That previous framework implemented two strategies: 1) SOP and attention training would be assigned to participants if any impairment, no matter how slight, was observed in these domains; and 2) targeting mild cognitive deficits vs severe cognitive deficits for cognitive training in order to no longer meet the criteria for HAND. After further consideration, targeting the most compromised cognitive domains for training may have been a more straightforward and face valid approach for improving cognition. Thus, such an approach may have had more impact on improving QoL, as well.

Second, in the TOPS study, the cognitive training intervention was able to improve certain aspects of subjective and performance-based everyday functioning (Vance, Fazeli, Azuero, Frank et al., 2021). If actual or perceived everyday functioning is improved as a result of cognitive training, it is likely this alone could improve QoL, beyond cognitive improvement itself. Future analyses should examine the mediation and moderation effects between cognitive training improvement and QoL indicators.

Strengths and Limitations

Four methodological strengths are noted. First, standardized norm-based cognitive measures were utilized to quantify cognitive function and determine HAND; by using such established population-based measures, the results are generalizable across the literature. Second, race and gender were controlled methodologically by using a block randomization procedure. Third, standardized cognitive training modules were used, which ensures consistency of the cognitive therapeutic approach across the cognitive domains targeted for improvement. And fourth, a variety of measures of QoL were used, which is important given that QoL is a multifaceted concept (Meeberg, 1993).

Related to the methodological strengths, methodological limitations are also noted. First, although a range of QoL indicators were included in this study, they were not exhaustive. Clearly, other indicators of QoL could be included, such as measures of leisure time activities and other measures of mood (e.g., happiness, feelings of security/safety). In fact, a qualitative assessment of how cognitive training may or may not have improved or diminished certain aspects of their lives would be a novel way to measure QoL. Second, although standardized cognitive measures were used to quantify function in specific domains, we acknowledge there is no true measure of cognition. Despite years of skilled and expert development to measure cognitive function in one domain and minimize reliance on cognitive function in other domains, all cognitive tests experience some spillover effect. For example, even in a measure of spatial memory, attention skills and memory of the verbal instructions are still being utilized in the measurement process. Third, and similarly, there is spillover from the cognitive training; in other words, despite careful planning, there is no “pure” cognitive domain training. Fourth, although the study design provided a direct pre-post comparison with a control group, it is possible some improvements in QoL may not be observed directly and may emerge over time, which would only be captured in a longitudinal design. Fifth, a key limitation of this study is the lack of inferential conclusions that can be drawn for between-group comparisons. As a pilot project, the sample size was determined on the basis of feasibility and, as with any small sample, a large degree of uncertainty on any estimates should be expected. Therefore, conclusions regarding between-group comparisons apply to the sample only, and no statements about their generalizability can be made. Findings should be interpreted with caution and considered exploratory rather than confirmatory. Sixth, we changed the treatment allocation ratio during the study to favor more assignment to the active treatment condition in order to maximize resources to provide more opportunity to test the feasibility of this condition. Fortunately, this did not skew the randomization of gender and race to the control vs the experimental condition. Finally, the overall attrition rate of 19.26% may have introduced threats to external validity in that the findings may not be as applicable to those who are younger, less educated, and with higher depressive symptomatology. But it should be mentioned that this attrition rate is in range with other studies of cognitive training in PLWH, with some reporting attrition rates as high as 35% (Vance, Fazeli et al., 2019).

Conclusion

This study was among the first to take an individualized-computerized cognitive training approach to improving QoL indicators in a sample of PLWH with HAND. Although the pattern of findings is mixed, consistent across cognitive training protocols is early evidence of a potential benefit to depressive symptomology and mental health indicators. The study helps to lay the groundwork for future cognitive training studies that are adaptable to an individual’s health needs and not a one-size fits all approach to cognitive rehabilitation.

Key Considerations.

• Nearly half of people living with HIV (PLWH) have HIV-associated neurocognitive disorder (HAND), and their cognitive impairment may impact their quality of life (QoL) as indicated by sleep quality, depression, mental health, and cognitive complaints.

• Although findings have been mixed, cognitive training that targets individual-specific cognitive deficits may improve QoL of persons with HAND.

• Cognitive training focusing on spatial learning/memory-type training protocols may have an advantage over other cognitive training protocols in improving certain QoL indicators.

• Given the time required to conduct cognitive training, its adoption in a typical fast-paced clinical setting is limited; therefore, referral of such cognitive training in a home-based or rehabilitation setting is more appropriate.