Question

This published table summarizes the clinical trial results for a colon cancer drug, panitumumab (Vectibix™). The experimental group is summarized in the middle column, described as P + BSC, i.e., experimental drug (panitumab) plus best supportive care (BSC). The far right column is the control group, who received the best supportive care (BSC) without the experimental drug.

The drug study data in Table 3 is “statistically” significant. What does that term mean, precisely?

Consider the difference between the medians for the two groups (this difference would normally be the relevant comparison for such drugs); is the difference between the medians “clinically significant?” Why or why not? How does your answer here differ from the answer about statistical significance?

	Expiremental	Control
Median	56	51
95% ci	55-59	50-54
Mean (sd)	96.4 (5.3)	59.7 (3.75)
Min, max	0, 357	0, 337

Answer 1

The drug study data in Table 3 is “statistically” significant. What does that term mean, precisely?

The median is used instead of mean because the standard deviation of the experimental group (5.3 SD) is 1.5 times higher than the control group (3.75 SD). One can conclude that there is an extreme value(s) in the experimental group causing the data to skew.

The mean would not be useful to describe the results of the clinical trials because the mean follows the direction of the skew and becomes a higher value. The median will be the central tendency of the data because the smaller the standard deviation the more accurate the results.

Consider the difference between the medians for the two groups (this difference would normally be the relevant comparison for such drugs); is the difference between the medians “clinically significant?” Why or why not? How does your answer here differ from the answer about statistical significance?

I think that the difference between the medians “clinically significant". The sponsor of the drug would want the FDA to focus on the mean because it looks more favorable to the drug company. The mean indicates the experimental group has a rate of 96.4% and in the control group the rate was 59.7%.

In my opinion, I believe the median is more accurate as the median truly displays the central numbers without the data being skewed by one extreme number. The mean is too easily affected by extreme numbers.