porc phreg output interpretation

[ANKH1]

Hi,

The output from proc phreg is something like this:

Parameter		DF	Parameter Estimate	SE	Chi-square	HR	Label
group	1	1	x.x	x.x		0.0567	group1

This means that group = 2 has less probability of experiencing an event? I did not specify the reference group in the proc, so group=1 should be the reference. Is this correct?

[PaigeMiller]

Show us the real output. Show us this table in its entirety. Use the "Insert Photos" icon to include a screen capture.

From what you have shown, we don't see group=2 and we don't see what the probability of experiencing an event is, we can't tell if it is lower or higher (and you can't just say "less probability of experiencing an event" — you need to tell us: less than what?).

[ANKH1]

Sorry, the HR=0.567. I just want to know how to interpret the HR. If the reference group was not assigned, I assume it is group 1?

[PaigeMiller]

Google is your friend

https://www.google.com/search?q=how+to+interpret+hazard+ratio&rlz=1C1GCEA_enUS855US855&oq=how+to+int...

I don't know what the control is in your study. You have provided such little discussion of the data and what your code is doing, I can't say if group=1 is the control or not. (But judging from the output you showed us, it doesn't look to me like group=1 is the control; of course I could comment further if you showed us this entire output table as I asked)

[PaigeMiller]

Please show us the ENTIRE table. Please show us your code. Please explain the data in this study.

[FreelanceReinh]

Hi @ANKH1,

According to the documentation, the default of the REF= option of the CLASS statement is LAST, which means: If the (formatted) values of CLASS variable group are '1' and '2', a hazard ratio of 0.0567 for group 1 indicates a considerably lower risk of experiencing the event in group 1 than in group 2 (e.g., event = infection, '1' = highly effective vaccine vs.  '2' = placebo).

[ANKH1]

Thank you!

[ANKH1]

This means that for this table group 1 is at lower risk of experiencing a first event? Below is also the code.

proc phreg data=ds;
class group;
model time2event*censor(1)=group/rl;
run;

[FreelanceReinh · Accepted solution]

The point estimate 0.75 of the hazard ratio indicates a slight tendency of a lower risk in group 1 than in the reference group 2, but the wide 95% confidence interval including 1.0 shows: At that confidence level it is well possible that the true hazard ratio is equal to 1 (or even greater than 1), i.e., it might be that the hazard rates in the two groups are the same (or the risk in group 1 is even higher) and the observed hazard ratio 0.75 is just due to random fluctuations (possibly in a small sample).

[ANKH1]

Yes, our sample size is very small. I just wanted to make sure that the interpretation of the reference group (group2) was correct. Thank you for pointing out to the documentation. I was under the impression that if you don't reference a group it takes the first "number", but I was clearly wrong. Thanks again!