Hazard Ratio for different time intervals

[kc]

Hello, I am trying to see the impact of treatment group on an event using the repeat event analysis by creating a counting dataset. But, I need the HR's in different time intervals (3 intervals) for the treatment group. After creating the start and end time variables in the dataset, I used the ‘end’ time variable to create ‘Time_Interval’ (code below). The issue is, although the HR’s for time interval 2 and 3 make sense clinically, HR for interval 1 is way off base. My question is whether the ‘Time_Interval’ calculation should be based on the ‘end’ time variable in the first place? If not, can someone point me in the right direction of how to accomplish this task? Thanks! Code: if 0 <= end <= 30 then Time_Interval='0-30D'; if 30 < end <= 365 then Time_Interval='30D-1Y'; if 365 < end <= 1825 then Time_Interval='1-5Y'; proc phreg data=counting_dataset covs(aggregate) covm; class group (ref='Trt2') Time_ Interval; model (start, end)*event_flag(0 2) = group Time_ Interval group*Time_ Interval / ties=breslow; id Subject; hazardratio 'HR Trt1 vs. Trt2' group ; run;

[OsoGris · Accepted solution]

Please open a tech support ticket and I can send you some materials on hazard ratios over time.   By the way, if you are using the model (t1, t2) counting process syntax it would be appropriate to use t2 in any programming.