Table 2.

Association of immunological parameters with CD4 T-cell count.

Variable	Lasso estimate	Post-lasso estimate	p-value
CD4+Ki67+	−0.180	−0.265	< 0.0001
CD4+CM HLA-DR+	−0.033	−0.156	< 0.0001
CD4+CD25-CD127-	−0.095	−0.117	< 0.0001
Treg count	0.069	0.111	< 0.0001
CD8+CD127+	0.032	0.063	< 0.0001
Treg EM	−0.026	−0.057	< 0.0001
Viral load	−0.007	−0.046	0.0002
Non-suppr Treg	−0.003	−0.044	0.002
CD4+HLA-DR+	−0.059	−0.025	0.133

Standardized regression parameter estimates for the selected set of covariates by an unregularized conventional GLM (post-LASSO) using the R-function glm are shown in the middle two columns. In the right column the calculated p-value is shown. However, it has to be mentioned that the p-values and significance statements obtained from the post-LASSO analysis should be regarded with caution. In general, these values tend to underestimate the overall level of uncertainty in the estimates, as the uncertainty from the preceding LASSO procedure is ignored. Following covariates were included into the LASSO analysis but not selected and, hence, are not displayed in the table: Age, CD8⁺HLA-DR⁺, CD8⁺Ki-67⁺, T_REG frequency, T_REG memory subsets other than EM, T_REGKi-67⁺, T_REG EmraKi-67⁺, CD4⁺IL-10, and T_REG IL-10 production in response to PMA/Ionomycin, CD4⁺CD127⁺, CD4⁺ memory subsets CD127⁺, HLA-DR⁺ and Ki-67⁺, suppressive-activated T_REG, suppressive-resting T_REG, plasma IL-7, and sCD127.Selected regression parameter estimates with LASSO approach implemented in the R package glmnet are shown in the left column in HIV-infected children (n = 26) with available data for all regarded parameters.