Characteristics of the Study Cohort by HIV Status
Table 1 summarizes the characteristics of the study cohort by HIV serostatus. Although there were no significant group differences in age, PLHIV had less formal education and were more likely to be male and identified as racial/ethnic minorities than the HIV− group. PLHIV exhibited increased neurocognitive deficits and had a higher prevalence of MetS, particularly elevated triglycerides, reduced HDL cholesterol, and diabetes, but did not significantly differ in BMI compared with the HIV− group. PLHIV had increased rates of current and lifetime major depressive disorder and also lifetime substance use disorder when compared with the HIV− group. Most PLHIV had AIDS, were on ART, and were HIV suppressed (Table 1).
Association Between MetS and Neurocognitive Function in the Overall Cohort
Univariable analyses aimed at selecting significant covariates showed that among the variables that were different by HIV status at P < 0.10 (Table 1), only WRAT-4 Reading SS was significantly associated with GDS (P< 0.01). A multivariable model in the overall sample on GDS, with HIV status, MetS, and their interaction as predictors, and adjusting for WRAT-4 Reading SS, found a borderline significant HIV by MetS interaction (P = 0.07). The F-test for the overall regression model was statistically significant: F(4, 197) = 6.17, P < 0.001, Adjusted R2 = 0.10. Follow-up multivariable linear regression analyses of GDS (stratified by HIV status) with MetS and WRAT-4 Reading SS as predictors found that MetS was significantly associated with GDS among PLHIV (Estimate = 0.22, SE = 0.10, P = 0.03) but not in the HIV− group (Estimate = −0.01, SE = 0.09, P = 0.93). Figure 1 shows the least squares means for the association between MetS and GDS based on these multivariable models.
Least mean squares (SE) based on the results from separate multivariable linear regression models on GDS by HIV status group investigating the association of MetS with GDS, after accounting for WRAT-4 Reading SS. *P < 0.05.
Association Between MetS and Neurocognitive Function in PLHIV
Among PLHIV, univariable analyses to identify significant covariates showed that nadir CD4 (P = 0.03) and WRAT-4 Reading SS (P < 0.001) were significantly associated with GDS. A multivariable model on GDS with nadir CD4, WRAT-4 Reading SS, and MetS as predictors showed that MetS continued to be significantly associated with GDS (P < 0.05), after consideration of the effect of WRAT-4 Reading SS (P < 0.01), and nadir CD4 (P = 0.08). The F-test for the model was significant: F(3,104) = 5.93, P < 0.001, Adjusted R2 = 0.12. Similar separate models among PLHIV on ART (n = 104) and among PLHIV on ART with undetectable plasma viral loads (n = 88) yielded comparable findings.
Next, we ran a series of independent sample t tests investigating the association between MetS and neurocognitive DDSs. Compared with PLHIV without MetS, PLHIV with MetS had lower scores in the domains of learning (P = 0.04) and fine motor skills (P = 0.04) and had borderline lower scores in executive functioning (P = 0.09). No other group differences were significant (P's > 0.11). Figure 2 shows Cohen's deffect sizes on the association of MetS with each of the neurocognitive domains among PLHIV. Multivariable analyses on neurocognitive DDSs (adjusting for nadir CD4 and WRAT-4 Reading) showed that MetS remained significantly associated with worse fine motor skills (Estimate = 0.14, SE = 0.07, P < 0.05) and borderline associated with learning (Estimate = 0.18, SE = 0.09, P = 0.07) but not with executive function (Estimate = 0.12, SE = 0.08, P = 0.12) or other neurocognitive domains (verbal fluency: Estimate = 0.05, SE = 0.06, P = 0.44; speed of information processing: Estimate = 0.08, SE = 0.06, P = 0.19; recall: Estimate = 0.11, SE = 0.11, P = 0.32; working memory: Estimate = 0.11, SE = 0.08, P = 0.16).
Cohen's d effect sizes on the association between MetS and neurocognitive DDSs among PLHIV. Exec, executive functioning; SIP, the speed of information processing; learn, verbal and visual learning; recall, delayed recall; motor, fine motor skills.
A series of independent sample t tests, assessing differences by each component of the MetS on GDS within PLHIV, showed that elevated triglycerides (P = 0.04, Cohen's d = 0.53) and diabetes (P = 0.02, Cohen's d = 0.52) were significantly associated with worse neurocognitive performance, whereas other components of MetS were not (P's >0.10; Figure 3). Of note, the Cohen's d effect size for the association of MetS (as a whole) and GDS (without adjustment for other covariates) was comparable with that of triglycerides and diabetes (0.53). Prior findings among PLHIV showed that central obesity was associated with neurocognitive function after considering the effect of overall BMI. Thus, we performed a multivariable model on GDS with elevated waist circumference and BMI as predictor variables, and found that waist circumference remained nonsignificant as a predictor of GDS (P = 0.50). A similar model with waist circumference as a continuous variable showed comparable findings. Separate multivariable models on GDS with elevated triglycerides and diabetes as predictors and adjusting for nadir CD4 and WRAT-4 Reading showed that diabetes (Estimate = 0.28, SE = 0.11, P = 0.02) remained significantly associated with GDS, whereas elevated triglycerides was unrelated (Estimate = 0.12, SE = 0.10, P = 0.23).
J Acquir Immune Defic Syndr. 2019;81(1):95-101. © 2019 Lippincott Williams & Wilkins