Skip to main content
. 2023 Dec 6;229(5):1266–1276. doi: 10.1093/infdis/jiad558

Table 3.

Univariable and Multivariable Linear Regression Analysis for the Regression Coefficients of the Principal Components

Standardized β (95% CI) P Standardized aβ (95% CI) P PC Direction
CSF and plasma Aβ42 + CSF CCL2 (PC1 of C1 model P < .001, 39.2% explained variance)
Age, y .111 (−.035 to .256) .134 .048 (−.101 to .197) .527
CSF HIV RNA, cp/mL .203 (.039 to .420) .018 .162 (−.017 to .340) .075
Male sex −.193 (−.337 to −.050) .009 −.168 (−.314 to −.022) .023 Lower
NNRTI use .143 (−.001 to .288) .052 .097 (−.050 to .244) .093
Current ART, mo .093 (−.053 to .238) .211 .042 (−.105 to .188) .574
Positive urine screening −.129 (−.274 to .016) .081 Excludeda
SSRI use .172 (.028 to .316) .020 .144 (.001 to .287) .040 Higher
Duration of SSRIs, mo .180 (.036 to .324) .014 .050 (−.385 to .485) .227
SNRI use −.033 (−.672 to .427) .661
BDI-II, score −.066 (−.212 to .080) .460 −.047 (−.190 to .096) .507
Statin use .030 (−.292 to .443) .684
ACEI use .040 (−.280 to .489) .594
Tobacco smoking .153 (.009 to .298) .038 Excludeda
CSAR .132 (−.013 to .277) .074 Excludeda
CSF 8-oxo-dG + CSF total Tau + CSF sCD14 (PC1 of C2 model P < .001, 61.8% explained variance)
Age, y .163 (.018 to .307) .027 .097 (−.148 to .342) .264
CSF HIV RNA, cp/mL .307 (.143 to .471) <.001 .202 (.033 to .371) <.001 Higher
HIV infection, mo .182 (.038 to .326) .013 Excludeda
SSRI use −.171 (−.270 to −.072) .021 −.123 (−.231 to −.015) .042 Lower
Duration of SSRIs, mo −.123 (−.268 to .022) .096 −.086 (−.298 to .126) .579
SNRI use .072 (−.279 to .818) .334
BDI-II, score .213 (.069 to .357) .004 .205 (.122 to .288) .002 Higher
Statin use −.024 (−.427 to .308) .751
ACEI use −.023 (−.445 to .324) .758
Tobacco smoking −.162 (−.306 to −.017) .028 Excludeda
CSF total protein, mg/dL .540 (.360 to .720) <.001 .547 (.377 to .717) <.001 Higher
CSAR .183 (.040 to .327) .013 .149 (.043 to .252) .016 Higher
Plasma sCD14 + plasma sCD40L (PC2 of C2 model P = .006, 37.9% explained variance)
Age, y −.071 (−.216 to .075) .341 −.028 (−.173 to .116) .698
CD4+ T cell nadir, cells/µL .151 (.006 to .295) .041 Excludeda
CD4+ T cell count, cells/µL .130 (−.016 to .276) .080 Excludeda
Current ART, mo −.127 (−.272 to .018) .086 Excludeda
BMI .174 (.030 to .318) .018 .203 (.056 to .350) .007 Higher
SSRI use .146 (.001 to .290) .048 .149 (.006 to .292) .042 Higher
Duration of SSRIs, mo .101 (−.044 to .247) .172 −.009 (−.429 to .411) .966
SNRI use −.073 (−.825 to .273) .322
BDI-II, score .048 (−.100 to .195) .524 .075 (−.070 to .220) .309
Statin use .023 (−.310 to .425) .757
ACEI use −.056 (−.562 to .295) .539
Antimicrobial use −.181 (−.325 to −.037) .014 −.216 (−.360 to −.072) .004 Lower
Immunomodulator use −.132 (−.277 to .013) .075 Excludeda
CSF and plasma neopterin + CSF sTNFR-II (PC1 of C4 model P < .001, 46.5% explained variance)
Age, y .071 (−.047 to .189) .235 .091 (−.015 to .198) .092
PI use .113 (−.004 to .230) .058 Excludeda
NNRTI use −.124 (−.241 to −.007) .038 −.109 (−.221 to .004) .058
Current ART, mo .045 (−.073 to .163) .457 .041 (−.067 to .0150) .455
BMI .112 (−.006 to .229) .063 Excludeda
GDS .126 (.017 to .235) .024 .137 (.030 to .243) .012 Higher
SSRI use .033 (−.282 to .444) .659
SNRI use .237 (.124 to .350) <.001 .223 (.072 to .374) .004 Higher
Duration of SNRIs, mo .172 (.056 to .287) .004 .005 (−.145 to .156) .943
BDI-II, score .111 (.001 to .221) .049 .025 (−.083 to .133) .650
Statin use .045 (−.204 to .389) .540
ACEI use −.069 (−.456 to .164) .354
Antimicrobial use .114 (−.003 to .231) .057 Excludeda

Linear regression models for PC regression coefficients. The table shows the results for the 4 medication classes, for significant (P value < .10) variables in univariable analysis, and for variables that were used to adjust the multivariable models regardless of univariable significance (univariable nonsignificant variables are not reported). Variables with P values <.10 on univariable analysis were included as candidate covariates in multivariable analysis and the best-fitting final models were identified through Akaike Information Criterion. Age, the clinical indication, and the duration of every drug that resulted associated on univariable analysis were retained to adjust the final model, regardless of univariable significance. P values <0.05 are highlighted in bold.

Abbreviations: 8-oxo-dG, 8-oxo-2′-deoxyguanosine; ACEI, angiotensin-converting enzyme inhibitors; ART, antiretroviral therapy; aβ, adjusted β; Aβ42, fragment 1–42 of β amyloid; BDI-II, Beck Depression Inventory II; BMI, body mass index; CCL2, chemokine C-C motif ligand 2; cp, copies; CSAR, CSF to serum albumin ratio; CSF, cerebrospinal fluid; GDS, global deficit score; NNRTI, nonnucleoside reverse transcriptase inhibitors; PC, principal component; PI, protease inhibitors; sCD14, soluble cluster of differentiation 14; sCD40L, soluble ligand of cluster of differentiation 40; SNRI, serotonin and norepinephrine reuptake inhibitors; SSRI, selective serotonin reuptake inhibitors; sTNFR-II, soluble tumor necrosis factor receptor type II.

aExcluded variables by backward selection are still shown in each model.