Table 2.
Mixed effect model of Log10 HIV‐1 DNA decay against the square root of time (days treated) for children on continued ART
| Fixed effects | Coefficient | SE | Df | t‐value | p(>|t|) |
|---|---|---|---|---|---|
| (Intercept) | 1.332 | 0.451 | 19.441 | 2.952 | 0.008 |
| Square root of days treated | −0.028 | 0.002 | 12.462 | −13.689 | <0.0001a |
| Baseline HIV‐1 DNA level | 0.799 | 0.115 | 20.852 | 6.955 | <0.0001 |
| Pre‐treatment HIV‐1 RNA level | −0.159 | 0.069 | 19.528 | −2.295 | 0.033 |
| Time interrupted | 0.000 | 0.000 | 24.036 | 1.029 | 0.314 |
| Study arm | |||||
| ART‐40W (vs. ART‐Def) | −0.406 | 0.195 | 22.631 | −2.085 | 0.049a |
| ART‐96W (vs. ART‐Def) | −0.532 | 0.204 | 21.993 | −2.606 | 0.016a |
| Part B ART‐40W (vs. ART‐Def) | 0.115 | 0.312 | 24.378 | 0.369 | 0.715 |
| Groups | Name | Variance | SD |
|---|---|---|---|
| For Random effects | |||
| Participant | (Intercept) | 0.078 | 0.279 |
| Participant | Square root of days treated | 0.000015 | 0.0039 |
| Residual | 0.145 | 0.380 | |
Significant p‐values (p < 0.05); CHER part A were randomized to delayed continuous therapy (ART‐Def) or time limited early elective therapy for 40 or 96 weeks respectively (ART‐40W or ART‐96W); Part B had a baseline CD4% <25% and received initial therapy for 40 weeks. The coefficients of the first four fixed effects factors represent the change in log10 (HIV‐1 DNA) for a one unit change in the factor adjusted for the other factors in the model. When assessing the association with HIV‐1 DNA decay and using the t‐statistic (a standardized difference) the days treated and baseline HIV‐1 DNA level have a strong effect on slowing decay, whereas pre‐treatment HIV‐1 RNA level has a small effect on increasing decay. The effect of time interrupted is not significant. With reference to both ART‐Def, ART‐96W and ART‐40W has a small effect, increasing decay, whereas the effect of Part B ART‐40W on decay was not significant. The difference in coefficients for ART‐40W versus 96W was not significant (p = 0.68), but the overall effect of treatment arm was significant (p = 0.037). The variance components for random intercepts, slopes and error are presented and in general, they are small and estimated with some uncertainty given the standard deviation of these estimates.