# Cox Proportional Hazards Model

#### J_Heart

##### New Member
For a class I am taking, I need to understand some of the basics for what and how some stats are utilized in studies. I am currently looking at the following article: https://www.nejm.org/doi/full/10.10...=ori:rid:crossref.org&rfr_dat=cr_pub 0pubmed. I was specifically curious about their use of the Cox Proportional Hazards Model in their analysis. From what I could gather, the use of this model is to make hazard ratios usable for time-varying outcomes (i.e. time to an event). Without it, the hazard ratios would not be able to provide information to the appropriate risk associated with the use of the drug. I am just curious if I am looking at this in the correct way and not missing too many details. Any help would be appreciated!
@hlsmith

#### hlsmith

##### Less is more. Stay pure. Stay poor.
@J_Heart

Well, close. At a first glance Cox proportional hazards (PHREG) models seem simple like logistic regression - but they also control for time-to-event. Which this can reveal differences between treatments that are not discernible by just examining outcomes at a cross-section of time (e.g., 6-months). So this is fairly clear via the produced figures that go along with them (e.g., survival curves, etc.). The big assumption of PHREG models is proportional hazards, which is to my understanding that the slopes of the survival or hazards are proportional to themselves across time.

What you need to watch out for is what you labelled as "time-varying effects", which are something different. Time-varying effects are when an exposure changes or its effect changes with time. For example, my sex won't change during the study, it is fixed; however, my treatment may change, so it can be time-varying. These changes can be controlled for in PHREG models, but are not what they are addressing in general. The are addressing that hazards for the outcome need to be modeled while keeping time in mind, meaning they are modeling the rate of the outcome during the study period. I hope this helps - and note that "time varying" is a particular term that isn't necessarily exclusive to PHREG models. Thanks.