Dear listers,
In fitting a multivariable model for survival after treatment for
cancer I get that radiotherapy is beneficial and that multiple lesions
are detrimental (surprise..) Also the proportional hazard assumption
is not fulfilled for radiotherapy. However, after defining an
interaction between the two I get that the interaction is siginificant
and that the PH assumption on the model is fulfilled. In trying to
interpret this I would greatly appreciate it if someone could comment:
gen interact=rt*multiple
local g rt multiple
local u `g' interact
stcox `g'
eststo a
stcox `u' - all factors significant
------------------------------------------------------------------------------
_t | Haz. Ratio Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
rt | .5674932 .10614 -3.03 0.002 .3933306 .8187732
multiple | 7.46303 2.458919 6.10 0.000 3.912564 14.23538
interact | .1807812 .0658104 -4.70 0.000 .0885702 .3689937
lrtest a -->significant.
Analyzing the interaction further::
lincom rt+interact,eform
_t | exp(b) Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
(1) | .1025921 .0324932 -7.19 0.000 .0551466 .1908572
lincom multiple+interact,eform
_t | exp(b) Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
(1) | 1.349175 .2142533 1.89 0.059 .9883139 1.841797
I interpret this to mean that rt decreases the hazard of death within
one unit of time by 90% if there are multiple metastases, whilst only
decreasing the hazard of death by 45% when there isnt multiplicity.
Conversely, that the HR is 7.46 for patients with multiplicity without
rt, but that multiplicity is only border-zone sigificant if the
patient receives rt.
Am I right in my interpretation?
regards,
M
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
