Table 2.
Summary of the analyses of the variation of HIV-1 CVL RNA among treatments: test statistics and p-values, and treatment effect coefficients (standard error estimates in parentheses: copies/mL). Phase I Carraguard trial, Chiang Rai, Thailand.
| Model | Test statistic (P-value) |
Gaussian coefficients | Logistic coefficients | ||
|---|---|---|---|---|---|
| Placebo gel | Carraguard | Placebo gel | Carraguard | ||
| Nonparametric analyses: Friedman test | |||||
| NQ and ND combined | 10.7 ( .005) | NA | NA | NA | NA |
| NQ and ND separate | 15.2 (<.001) | NA | NA | NA | NA |
| Mixed models using day seven only | |||||
| Equal variances | 14.1 (0.001) | −0.439 (0.148) | −0.543 (0.150) | NA | NA |
| Unequal variances | 14.2 (0.003) | −0.455 (0.153) | −0.561 (0.157) | NA | NA |
| Mixed models using both days zero and seven | |||||
| Equal variances | 7.6 (0.02) | −0.526 (0.191) | −0.303 (0.193) | NA | NA |
| Unequal variances | 7.6 (0.06) | −0.532 (0.194) | −0.309 (0.196) | NA | NA |
| Mixture models using the likelihood L2 and data from day seven only | |||||
| Equal variances | 4.6 (0.34) | 0.036 (0.167) | −0.195 (0.171) | −0.553 (0.387) | −0.522 (0.388) |
| Unequal variances | 5.8 (0.32) | 0.036 (0.167) | −0.195 (0.171) | −0.653 (0.387) | −0.522 (0.388) |
| Mixture models using the likelihood L2 and data both days zero and seven | |||||
| Equal variances | 4.30 (0.37) | 0.036 (0.171) | −0.195 (0.177) | −0.553 (0.362) | −0.522 (0.356) |
| Unequal variances | 6.86 (0.23) | 0.036 (0.148) | −0.195 (0.152) | −0.554 (0.363) | −0.521 (0.358) |
| Generalized estimating equation logistic regression models for a meaningful decrease | |||||
| NQ and ND combined | 2.2 (0.34) | NA | NA | −0.345 (0.417) | − 0.550 (0.410) |
| NQ and ND separate | 1.0 (0.62) | NA | NA | −0.152 (0.389) | − 0.338 (0.387) |
NQ: sample detectable but not quantifiable; ND: sample not detectable;
NA: not applicable.
Test statistics:
Friedman test: chi-square with 2 degrees of freedom (df).
Mixed models and mixture models: likelihood ratio chi-square with 2 and 3 df for equal and unequal variances, respectively.
Generalized estimating equation logistic regression models: Wald chi-square with 2 df.