In , demographic, clinical, and metabolic characteristics are presented for all 130 participants in the substudy who had no confounding neurologic history or conditions. This substudy group differed at first CHARTER visit in some demographic and HIV-related variables from the remainder of the CHARTER participants who were also free of serious confounding neurologic conditions (n = 1,195). However, their metabolic profiles were similar to those of the CHARTER cohort except for lower BMI and higher triglycerides. The metabolic substudy group also differed from all US HIV-infected patients based on the CDC 2005 HIV/AIDS surveillance report.17
For example, our study subjects were on average 9 years older (mean age of 46 vs 37 years) and more likely to be Caucasian (57% vs 31%) and male (87% vs 74%).
Demographic and clinical characteristics for the CHARTER metabolic substudy group combined and stratified by cognitive impairment
The majority of participants had experienced CART-induced immune reconstitution based on having ARV therapy at the time of their study visit (82%), high current CD4 counts (median = 501 cells/mm3) compared to either much lower nadir CD4 counts (median = 120 cells/mm3) or AIDS diagnoses (70%), and median plasma HIV concentrations below the limit of detection (1.7 log copies/mL). The average duration of known HIV infection was 13 years.
Cognitive impairment (global rating ≥5) was common (52/130 = 40%). When demographic and clinical characteristics were compared between cognitively impaired and unimpaired subjects (), impaired subjects were older (age 48.3 vs 44.9 years, p = 0.02), had longer self-reported durations of HIV infection (14.5 vs 12.0 years, p = 0.04), had larger waist circumferences (WC) (99 vs 88 cm, p = 0.0005), and were more likely to have been diagnosed with diabetes mellitus (DM II) (15% vs 3%, p = 0.007). Impaired and unimpaired participants were similar in other demographics and HIV disease indicators including current and nadir CD4 count, plasma and CSF viral loads, and ARV exposure.
Levels of most metabolic variables were also similar for the 2 groups, but several were nonsignificantly higher in the impaired group. DM II was self-reported by 10 of 130 (8%) subjects. DM II patients were more likely to be cognitively impaired than nondiabetic patients (80% [8/10] vs 37% [44/120], OR = 6.9, p = 0.01). Similarly, cognitive GDS were higher in diabetic than nondiabetic patients (median GDS = 0.60 vs 0.13, p = 0.0012). These differences appear unrelated to CART because diabetic and nondiabetic participants were similar for proportions on antiretroviral therapy currently, time on their current regimen, and class of CART drugs taken (data not shown, p > 0.18 for all comparisons).
To examine the association between diabetes and NCI further, we assessed 1,325 participants in the entire CHARTER cohort who had self-reported diabetes and had no conditions that confounded their cognitive assessment. In these participants, prevalence of NCI by global rating was similar in diabetic and nondiabetic patients (55/115 [48%] vs 563/1,210 [47%], p
= 0.79). Because the effects of diabetes on NCI had been demonstrated in older HIV-infected patients in a prior study,12
we examined effects of diabetes in patients aged >55 years (). Prevalence of NCI was significantly higher in these older HIV+ diabetic patients as diagnosed by GDS of >0.5 (11/21 [52%] vs 29/97 [30%], p
= 0.05), but not when diagnosed by global ratings (13/21 [62%] vs 43/97 [44%], p
= 0.14). Thus, if diabetes contributes to NCI, it may do so only in older patients.
Relationship of diabetes and NCI (by GDS and global rating) in the CHARTER cohort and in those aged 55+ (excluding cognitively impaired group)
Multivariate logistic regression analysis was performed using the 90 substudy participants with complete data for 17 of the 26 variables that are marked with superscript b in . As the measure of obesity in our first model, we used BMI instead of WC because only 55 subjects had WC measurements and WC is highly correlated with BMI (r = 0.83). This model which used AIC as a selection criterion contained age (OR = 1.06/per year, p = 0.027) and BMI (OR = 1.12/kg/m2, p = 0.039) as predictors of NCI (). Diabetes appeared to increase the goodness of fit for the model, but its individual effect did not reach statistical significance (p = 0.12). However, the overall fit of this model based on the AIC was better with diabetes included than in competing models.
Multivariate regression model selected based on AIC to model NCI as a function of demographic, medical, and metabolic predictors of interest including BMI (n = 90)
Finally, we examined WC in a second model of NCI based on the 55 patients with measurements of this and all other variables (). BMI (OR = 0.69, p = 0.038) and WC (OR = 1.34 per cm, p = 0.001) both entered the model with the best goodness of fit along with diagnosis of AIDS (OR = 49.6, p = 0.027), diabetes (OR = 17.6, p = 0.07), and triglycerides (OR = 0.32, p = 0.09) (). In this model both BMI and triglycerides have ORs less than 1.0 and thus appear protective rather than predisposing, an unexpected finding. The C-statistic (area under the curve of the receiver operating characteristic curve) for this model (0.893) was greater than that of the first model (0.722).
Multivariate regression model selected based on AIC to model NCI as a function of demographic, medical, and metabolic predictors of interest including BMI and average mid waist circumference (n = 55)
Reversal of the effect of BMI when WC enters the model suggests that WC is the component of obesity that is most correlated with NCI and that BMI may correlate only because it is a marker of WC. As expected, BMI and WC were highly correlated (r = 0.83). The impaired group had larger WC than the unimpaired group at any level of BMI, consistent with findings from the multivariate models.