Figure 1.
Prevalence of transmitted surveillance drug resistance mutations (SDRMs) over time.
Sampling years 1998–2002 were grouped into a single level (‘pre-2003’) to adjust for small sample sizes (, and
, respectively). Sample sizes for the bins are displayed below the
-axis. Dashes (−) indicate the fraction of naïve sequences in a given sampling period that contain at least one SDRM (right
-axis). The trend in mean number of SDRMs per naïve sequence (left
-axis) over time is displayed as a solid line annotated with the actual values. The
quantile per time point is depicted by a lighter line bounding a shaded region to illustrate the trend in the distributions (all medians were zero).
Figure 2.
Bayesian network depicting conditional dependencies among surveillance drug resistance mutations (SDRMs) in ARV-naïve sequences.
Each labeled node corresponds to a position in protease [rectangular] or RT (rounded). Nodes without dependencies (L23, L24, V32, I47, I50, F53, G73, L76, K65, V75, F77, Y115, F116, L100, V106, V179, Y181, and Y188) were omitted from the graph for clarity. Connections between nodes (edges) are labeled with the log odds ratio of the
contingency table for the presence/absence of SDRMs at the respective sites, e.g., an SDRM at position PR-D30 is estimated to be
times more likely to be present in an ARV-naïve sequence that contains an SDRM at position PR-N88. Line widths for edges are also drawn in proportion to the log
odds ratios. Inf = infinity, i.e., unable to estimate because of zero count(s) in the contingency table.
Table 1.
Number and percentage prevalence of putatively transmitted surveillance drug resistance mutations (SDRMs) by residue position in antiretroviral (ARV)-naïve sequences.
Figure 3.
CD4 cell count (raw) plotted against the sum of Stanford scores for nucleoside and non-nucleoside reverse transcriptase inhibitors ( and
, respectively).
The upper limit of CD4 ( cells/mL) was truncated to
cells/mL (omitting 19 outliers) to emphasize the overall trend. Because the predicted baseline CD4 tended to decline with both
and
, we combined the scores into a single ordinal variable to facilitate interpretation. The linear model prediction is displayed as a solid line (generated by fitting a smoothing spline to the predicted values with smoothing parameter
), with
confidence intervals displayed as dashed lines.
Figure 4.
Three-dimensional scatterplot of -transformed plasma viral load (pVL) as a function of Stanford scores for NRTI and NNRTIs (
and
).
Overall, higher was associated with higher pVL, while
was inversely associated with pVL. The range of
pVL (
) was truncated to emphasize the overall trend (omitting 73 outliers). The linear model prediction is displayed as a wireframe surface (generated by local polynomial regression with smoothing parameter
).
Table 2.
Coefficient estimates from Akaike Information Criterion (AIC)-selected linear models of baseline and plasma HIV RNA on Stanford scores by drug class and demographic and risk factors.
Figure 5.
Box-and-whisker plots illustrating (A) the effects of SDRMs at positions D67 and K219 on CD4 cell count and (B) the effects of SDRMs at positions M184 and K103 on plasma viral load (pVL).
‘Wildtype’ denotes sequences lacking SDRMs at both positions, irrespective of whether any other SDRMs were present at other positions in protease or RT. Solid lines indicate the group median and open circles denote outliers that fall outside the region defined by 1.5 times the interquartile range. Plot (A) was rescaled with respect to untransformed CD4 counts to emphasize differences among groups, trimming 4 outliers from the ‘wildtype’ group (1602, 1626, 1838, and 2073 cells/mL) from the plot region. Sample sizes per group are annotated on each plot.
Table 3.
Coefficient estimates from AIC-selected linear models of baseline and
plasma HIV RNA on position-specific SDRMs and demographic and risk factors.
Table 4.
Unadjusted and stabilized weight-adjusted linear models of multinomial exposure variables (M184X and (K103X or K219X)) and
(D67X and K219X) on
plasma HIV RNA and
, respectively.