actually it is expected that all people eventually die, therefore achieve an event if they are sufficiently followed. One could even consider that if all of them do not die over long period of time (let's say 100 years), they are in some way inadequately followed/lost to follow up. Of course, this is typically not seen in human studies that are much shorter, and survival analysis usually works with censored data. Same can be considered in animal studies where all animals could be ultimately sacrificed - therefore also achieve an event. However methods of survival analysis are also sensitive to timing of events, and if deaths in one group tend to happen later than in other, analysis can yield the significant result irrespective if all or only some of patients die during whole study period. Whether or not "all subjects dying" implies specific conclusion when interpreting results can vary depending on the setting that you investigate (Not expected for all subjects to die during short follow up - unusually high mortality?; Expected for all subjects to die due to hazardous setting - adequate follow-up/no specific conclusion?).
In case all subjects die, T-test/Mann Whitney U test and other tests for non-censored data (regression analyses other than Cox regression) can be used to compare survival times, therefore enabling more extensive analyses.
Thank you Dr. Lucijanic for your quick response. I agree with you " Not expected for all subjects to die during short follow up - unusually high mortality ", but what if event is progression instead of death.
How can we quote median survival and curve by using Mann Whitney U test?
Then this would probably mean unusually high progression rate I guess. I meant to suggest that If you handle completed observations, you are not limited with methods for censored data and you are allowed to use other tests to compare survival between subgroups in your sample (like you would compare height of patients or hemoglobin level, median survival time would be median of all observations), you would still need to plot the survival curve to show how your subjects went over time as this is what readers would probably expect to see.
.... all study subjects got event before last follow-up, then how ...Is it getting event before lost for followup or before end of the study. In survival studies lost for followup has different meaning than end of study. If all got event before end of study then the survival estimate comes to zero at particular time (which is before end of study). As such there will not be any difference in the interpretation.