Programming the statistical procedures from SAS

Conditional Logistic regression, proc phreg, HR goes to extreme

Occasional Contributor
Posts: 10

Conditional Logistic regression, proc phreg, HR goes to extreme

I am using proc phreg to do conditional Logistic regression. It is case-controlled study. We have cases and controls. Our event is Stent Thrombosis. Some patient stopped the standard drug when the Event occured while some did not stop the drug. What we want to see if off-drug is cirital for the Event, i.e. high HR. We have created dummy variables. It is for different interval that patients were off the drug. We are doing subgroup analysis. For example, age < 65 or age > 65. Using the model below, for age < 65, it gets reasonable results. However, for age> 65 group, we get either very larger HR or very small HR. The model can converge and the procedure seems running well.  We have tried to make our time interval larger so that we can have more patients in each categories, such as only  < 7 days and > 7 days. But the result  did not change much.

Can we do something to make it better?

Thanks  a lot!!


if 0 < days_off_plavix < = 7     then offplavix_st7 = 1;

else                                  offplavix_st7 = 0;


title "Subgroup analysis of Risk for Discontinuing Clopidogrel at Various Interval on All

title2 "For subgroup factor, &subvar = &subgrp";


proc phreg data = plvx_st1  covs nosummary ;

      where &subvar = &subgrp;

      model time * case(0) = no_asa offplavix_st7 offplavix_st14 offplavix_st30 

                         offplavix_st180 offplavix_st365 offplavix_st366 /rl ties = discrete;

      strata setid;


offplavix_st7    1  33.82858    0.86602  0.000  1525.8330     <.0001 4.915E14 9.003E13 2.684E15

offplavix_st14   1   18.70692    0.88986  0.000   441.9417     <.0001 1.3314E8 23273673 7.6166E8

Ask a Question
Discussion stats
  • 0 replies
  • 1 in conversation