About this file

This refers to additional details from data analysis required from a reviewer for an apreptant trial for preventing postoperative nausea and vomiting (PONV), ClinicalTrials.gov register NCT02357693 (url https://clinicaltrials.gov/ct2/results?term=NCT02357693&Search=Search ). This document can be used for analysis verification.

Dummy data for replication purposes

gcount=c(rep("Hysterectomy",24),rep("Neprhectomy",4),rep("Hemicolectomy",3),"Gastrectomy",rep("Hysterectomy",25),rep("Neprhectomy",3),rep("Hemicolectomy",4),rep("Gastrectomy",2))
group=c(rep("Control",32),rep("Treatment",34))
doseep=c(1,rep(2,22),rep(0,9),1,1.5,rep(2,21),rep(0,34-23))
doseit=c(rep(60,1),rep(80,4),rep(100,3),rep(120,1),rep(0,32-(1+4+3+1)),rep(60,4),rep(80,7),rep(100,0),rep(120,0),rep(0,34-(4+7)))

Chi-squared test or Fisher exact test?

Although both tests were originally developed for 2x2 contigency tables, both tests are suitable for any bidimensional table (for example, 5 x 4 table) (read Mehta, Cyrus R., and Nitin R. Patel. “A hybrid algorithm for Fisher’s exact test in unordered rxc contingency tables.” Communications in Statistics-Theory and Methods 15.2 (1986): 387-403.). Both tests are limited. Fisher exact test is limited by computational power, because when a sample is big, factorials get too big (although new algorithms have overcome this limitation).

With large samples, a chi-squared test can be used. However, the significance value it provides is only an approximation, because the sampling distribution of the test statistic that is calculated is only approximately equal to the theoretical chi-squared distribution. This approximation is inadequate when sample sizes are small, or the data are very unequally distributed among the cells of the table, resulting in the cell counts predicted on the null hypothesis (the “expected values”) being low. The usual rule of thumb for deciding whether the chi-squared approximation is good enough is that the chi-squared test is not suitable when the expected values in any of the cells of a contingency table are below 5, or below 10 when there is only one degree of freedom, as recommended by Cochran (Cochran, William G. “The ??2 test of goodness of fit.” The Annals of Mathematical Statistics (1952): 315-345.)

Surgery distribution comparison

table(gcount,group)
##                group
## gcount          Control Treatment
##   Gastrectomy         1         2
##   Hemicolectomy       3         4
##   Hysterectomy       24        25
##   Neprhectomy         4         3
chisq.test(group,gcount)
## Warning in chisq.test(group, gcount): Chi-squared approximation may be
## incorrect
## 
##  Pearson's Chi-squared test
## 
## data:  group and gcount
## X-squared = 0.57938, df = 3, p-value = 0.9011
fisher.test(group,gcount)
## 
##  Fisher's Exact Test for Count Data
## 
## data:  group and gcount
## p-value = 1
## alternative hypothesis: two.sided

Epidural morphine

Epidural morphine dose is not a cathegorical but an numerica variable. As such, group comparison could use either T-Test, ANOVA or Wilcoxon-Mann-Whitney for comparison. In this sample, parametric tests’ requirements are not fulfilled, therefore Wilcoxon-Mann-Whitney test was used.

table(doseep,group)
##       group
## doseep Control Treatment
##    0         9        11
##    1         1         1
##    1.5       0         1
##    2        22        21
wilcox.test(doseep~group)
## Warning in wilcox.test.default(x = c(1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, :
## cannot compute exact p-value with ties
## 
##  Wilcoxon rank sum test with continuity correction
## 
## data:  doseep by group
## W = 578, p-value = 0.6063
## alternative hypothesis: true location shift is not equal to 0

Intrathecal morphine

Intrathecal morphine dose is not a cathegorical but an numerica variable. As such, group comparison could use either T-Test, ANOVA or Wilcoxon-Mann-Whitney for comparison. In this sample, parametric tests’ requirements are not fulfilled, therefore Wilcoxon-Mann-Whitney test was used.

table(doseit,group)
##       group
## doseit Control Treatment
##    0        23        23
##    60        1         4
##    80        4         7
##    100       3         0
##    120       1         0
wilcox.test(doseit~group)
## Warning in wilcox.test.default(x = c(60, 80, 80, 80, 80, 100, 100, 100, :
## cannot compute exact p-value with ties
## 
##  Wilcoxon rank sum test with continuity correction
## 
## data:  doseit by group
## W = 547.5, p-value = 0.9621
## alternative hypothesis: true location shift is not equal to 0