Statistical significance between placebo group and test group

[Confused101]

Hi. I need to work out whether or not the difference in number of deaths between a placebo group and a test group is statistically significant. I’d be grateful if someone was able to explain which test is best and how I should go about this. Thanks.

 

[Karabiner]

How were the data sampled (study design), how large are the respective groups,
and how many deaths were observed in each group?

With kind regards

Karabiner

 

[Confused101]

There were 2 randomised groups of 800. 5 deaths in the placebo group and 7 in the test group. Is that enough information?

 

[Karabiner]

If observation period(s) were the same for all participants, then you can use Fisher's exact test for
a 2x2 table (test/placebo versus dead /alive).

With kind regards

Karabiner

 

[hlsmith]

Is time to death important since those rates seem comparable.

 

[Karabiner]

I thought of the possibility that observation time was different between groups, and therefor survival anaysis should be used.
But maybe the information that it is a randomized design with a placebo control group suggests that they had controlled this.

 

[Confused101]

Is time to death important since those rates seem comparable.

Hi. No. They are very similar. Some people are arguing that because 7 is more than 5 then the trial drug increases the number of deaths. I can see that’s not the case but wanted to assure myself that this had a scientific basis. Thanks for taking an interest.

 

[Confused101]

I thought of the possibility that observation time was different between groups, and therefor survival anaysis should be used.
But maybe the information that it is a randomized design with a placebo control group suggests that they had controlled this.

It was a randomised double blind trial. Thanks.

 

[hlsmith]

Well I have an example I use that shows the 10-year mortality rate between two treatments is the same, but the rates across time are drastically different. Since the placebo group all dies within 2 years and the intervention group dies about 6-8 years post exposure. Thus a cross-sectional comparison misses this difference. How, perhaps your outcomes are immediate, which would make this a mote point. @Karabiner - it is hard to control for this even if you have discete time intervals for data collection without using survival analysis. However, you could add time to a multiple variable regression to try and remove its effect.

 

[Confused101]

Is time to death important since those rates seem comparable.

It’s an ongoing trial. I think I’m maybe missing the point though. Thanks for taking an interest.