Dynamic immune markers predict HIV acquisition and augment associations with sociobehavioral factors for HIV exposure

Protocol for evaluating mechanistic pathways associated with HIV acquisition via nested Least Absolute Shrinkage and Selective Operator analysis

“…In brief, of 2,685 MSM and TW who did not have HIV but were considered at high risk for exposure, 2,109 entered a longitudinal cohort and were tested monthly by point-of-care third-generation HIV immunoassay and for HIV-1 RNA by pooled NAAT test if negative. 10 , 11 The case-control sample was created by identifying from Sabes, 90 case participants who were diagnosed with incident HIV, either while HIV-seronegative with detection of HIV-1 RNA by nucleic acid amplification test (NAAT) or having had a negative HIV Ag/Ab test within 3 months prior, 1 and by identifying 3 matched controls per case who did not acquire HIV during the study (N = 270). Two controls were matched by time under observation, meaning they were followed for at least the same number of months as the case and had available samples at matched study visits.…”

“…The examples below demonstrate application of this protocol to data from the Sabes study cohort. Rather than repeat the published results of the effect sizes (ORs) from the models as presented in Table 3 of the referenced manuscript, 1 our results tables below focus on whether each selected variable was retained in the base or nested models, and if so, whether the coefficient was significant and if it was associated with effect modification in the final nested model.…”

“…The odds ratios (ORs) resulting from the conditional logistic regression conducted subsequent to LASSO step and associated p values are presented in primary manuscript. 1 Throughout our study, when running LASSO, we selected ‘optimal’ using a 10-fold cross-validation, by choosing the value which minimized test prediction error for a given dataset. That value is referred to in the literature as lambda.min .…”

“…For complete details on the use and execution of this protocol, please refer to Bender Ignacio et al. 1 …”

Data_Sheet_2.pdf

“…They selected 13 cytokines and used logistic regression and principal component analysis (PCA) in their statistical analysis. On the other hand, Ngcobo et al (20), in their study examining whether preinfection plasma cytokine concentrations predicted the rate of HIV disease progression in the same study cohort, considered all 48 cytokines. They used linear regression to assess the impact of each cytokine on viral load (VL) and the CD4:CD8 ratio in both bivariate and multivariable models, adjusting for age, contraceptive use, HSV-2 status at baseline, study site, and study arm at randomization.…”