Table 2.
Improvement in nausea and vomiting: primary and secondary outcomes and sensitivity analysis
| Outcomes | Aprepitant | Placebo | Relative risk ratio or adjusted mean changes from baseline (95% CI) Aprepitant vs. placebo† | P† |
|---|---|---|---|---|
| Primary analysis | ||||
| Improvement in nausea (ITT)*‡ | ||||
| No. of patients randomized | 63 | 63 | ||
| Improved nausea, No. (%) | 29 (46%) | 25 (40%) | 1.2 (0.8, 1.7) | 0.43 |
| Sensitivity analysis | ||||
| No. evaluable patients | 59 | 63 | ||
| Improvement in components of composite primary outcome, No. (%) | ||||
| 1) Mean change of 28-day VAS versus baseline VAS ≤ −25 mm | 25 (42%) | 20 (32%) | 1.3 (0.8, 2.1) | 0.24 |
| 2) Mean 28-day VAS < 25 mm | 26 (44%) | 16 (25%) | 1.7 (1.1, 2.8) | 0.02 |
| 3) Both mean change in VAS ≤ −25 mm and mean 28-day VAS < 25 mm | 22 (37%) | 11 (17%) | 2.1 (1.1, 4.1) | 0.01 |
| Secondary outcomes of nausea/vomiting | ||||
| Measured change (mm) 28-day VAS versus baseline VAS, mean change (95% CI)§ | −21.4 (−27.0, −15.8) | −14.9 (−20.3, −9.5) | −6.5 (−14.3, 1.3) | 0.10 |
| Repeated measures of daily VAS¶ | ||||
| No. of measures of VAS change | 1701 | 1833 | ||
| Change (mm) 28-day VAS versus baseline VAS, adjusted mean change (95% CI) | −21.8 (−28.2, −15.1) | −14.9 (−19.3, −15.3) | −6.9 (−14.7, 0.95) | 0.09 |
| PAGI-SYM Severity index (0=none to 5=very severe) | ||||
| Nausea severity | −1.8 (1.5) | −1.0 (1.4) | −0.7 (−1.3, −0.2) | 0.005 |
| Vomiting severity | −1.6 (1.7) | −0.5 (1.4) | −0.7 (−1.1, −0.3) | 0.001 |
| Retching severity | −1.7 (1.6) | −0.8 (1.4) | −0.8 (−2.3, −0.3) | 0.003 |
| Daily Diary (GCSI-DD) symptom severities (0=none to 4=very severe) | ||||
| Nausea severity | −0.8 (1.0) | −0.5 (0.7) | −0.3 (−0.6, 0.0) | 0.06 |
| Vomiting severity | −0.4 (0.8) | −0.2 (0.5) | −0.1 (−0.3, 0.1) | 0.24 |
| Daily Nausea (hours) | −2.5 (3.2) | −1.2 (4.3) | −1.5 (−2.8, −0.1) | 0.03 |
| Percent of nausea-free days during 28-day follow-up (mean, 95% CI) | 21.8% (14.0, 29.6%) | 11.3% (5.5, 17.1%) | 10.5 (0.9, 20.2) | 0.03 |
| Imputation analyses of primary outcome|| | ||||
| Best case: missing outcome=improve | 33 (52%) | 25 (40%) | ||
| Improved nausea, No. (%) | 33 (52%) | 25 (40%) | 1.3 (0.9, 1.9) | 0.13 |
| Worst case: missing outcome=not improve | ||||
| Improved nausea, No. (%) | 29 (46%) | 25 (40%) | 1.2 (0.8, 1.7) | 0.43 |
| Multiple imputation of primary outcome | ||||
| Improved nausea, % (range) | 50% (46% – 52%) | 40% (40% – 40%) | 1.2 (0.8, 1.9) | 0.28 |
The primary outcome of Improvement in nausea is a binary composite outcome defined as either 1) an improvement in the mean of available nausea visual analog scale (VAS) scores over the 28-day treatment period compared to the means of VAS during the 7-day baseline (BL) period being ≤ −25 mm, or 2) the mean VAS after 28-days of treatment was < 25 mm. The number of 28-day treatment VAS available per subject ranged from 14 to 41 (median=28.5, IQR:27,33; PAprepitant Vs Placebo=0.76).
Baseline VAS denotes the mean of VAS scores over the 7-day pre-randomization period.
28-day VAS denotes the mean of all available VAS scores over the 28-day treatment period.
P-values, relative risk ratios, and 95% confidence limits (CI) for the primary ITT were calculated using the Cochran-Mantel-Haenszel chi-square test, stratified by clinic. Mean adjusted change from baseline, 95% confidence limits (CI), and P-values were calculated using ANCOVA, regressing change from baseline to 28 days on treatment group and baseline value of the secondary outcome.
There were 4 subjects missing all 28-day VAS scores; the primary outcome for these subjects was imputed to no improvement for the intention-to-treat (ITT) analyses.
Improvement indicated by a decrease in the change from the average 7-day baseline VAS of the average available VAS at 28-days of treatment as a percent of the baseline VAS.
Adjusted mean changes from baseline for each treatment group, and adjusted mean change from baseline between treatment groups, P-values, and 95% CI were determined from multiple linear regression of daily change of the VAS during follow-up in relation to the mean 7-day baseline VAS, with an indicator for treatment group. A GEE random effects model with a robust variance estimate was used to account for repeated subject measures.
No. patients in imputation sensitivity analyses: 63 (aprepitant), 63 (placebo).
Observations with missing 28-day mean VAS score were imputed using 500 datasets. Imputation model included the following baseline variables: treatment group indicator, baseline VAS, clinic, delayed gastric retention, diabetes status, age and sex.