**Figure 3: **Regression models in HIV-infected patients without signs and symptoms of opportunistic lung disease (OLD-negative). Linear regression
(A - top) to estimate changes in exhaled H_{2}O_{2} (y) as a function of log10-
transformed detectable HIV-1 viral load (x). Solid line: slope (β_{1}); long
dashed lines: 95% confidence intervals. Quantile regression (B-middle)
to estimate relationships between exhaled H_{2}O_{2} and log10-transformed
detectable HIV-1 viral load for the exhaled H_{2}O_{2} distribution from 0.2 to
0.8 quantiles. Solid line: slopes for quantiles (b1), connected with white
circles (non-significant) or black circles (significant at p<0.05) and with
error bars corresponding to 95% confidence intervals (bolded if p<0.05).
Linear regression slope β_{1} is shown as solid line with long dashed lines.
corresponding to 95% confidence intervals. Multiple linear regression
(C - bottom) in HAART-treated subgroup only, with exhaled H_{2}O_{2} as the
dependent variable and log10-transformed HIV-1 detectable viral load (A),
CD4 count (B), HIV-infection duration (C) and HAART duration (D) as the
independent variables. Predicted values of exhaled H_{2}O_{2} from the multiple
regression equation are graphed on the X-axis and the observed values of
exhaled H_{2}O_{2} are plotted on the Y-axis. |