Statistiek 9: cox regression

Question 1

verschil Kaplan Meier en Cox regression

Answer 1

KM: survival curves, compare survival curves by means of the log-rank test.

CR: quantiy differnce in survival, assess independent effect of multiple predictors

Question 2

hazard function

Answer 2

hazard rate h(t) = risk of dying after a certain time (t), given the patient has not died until that time.

interpretaion: disk of dying at time t (instanteous)

Question 3

formule cox regression:

Answer 3

h(t) = h0(t) * e (B1X1+B2X2+…. + BnXn).

h(t) = baseline hazard
Wanneer je ln van e neemt dat je een lineare relatie krijgt.

Question 4

HR

Answer 4

h1(t)/h2(t)

how much higher or lower is the instantaneous risk of dyung for one group compared with antoher group

Question 5

Coefficient to HR: 0, > 0, < 0

Answer 5

HR 1 (no effect), > 1 (higher risk, lower survival), < 1 (lower risk, higher survival)

Question 6

Output to meaning

B, bv 0.377
Exp (B) 1.459

Answer 6

B = male gender gives higher hazard rate

Exp (B): male gender raises mortality by 46%

Question 7

Assumptions specific for cox regression:

Answer 7

proportionality, always to the same extent

covariates linearly related tot the log

Question 8

Grafische proportianlieit inschatten:

Answer 8

Kaplan meier curve bekijken: lopen ze ongeveer paralell? ze mogen niet kruisen

Question 9

Log-minus-log plot?

Answer 9

verschil tussen both log minus log curves = b*covariate.

Als PH assumpotion is aangenomen zullen de curves over elkaar lopen.

Question 10

Hazard ratio
Hazard rates
Cumulative Survival Probability
Proportional hazards

Answer 10

• presents how much higher or lower the instant.. risk of dyng is for 1 person compared to another
• risk of dygin tiwht short time interval after a certain time t, given that the patient is alive
• proportion still alive at time t’
• • the proportion of the hazards of two persons, is constant over time

Question 11

assumptions of cox regression models

Answer 11

Also for KM: non informative censoring, prognosis early and late included is equal

specific voor CR: proportionality always to the same extent, covariates linearly realte to the log of the hazard rate

Question 12

3 manieren voor proportionaliteit inschatten

Answer 12

graphisch inschatten, goodness of fit, time-dependent covariates.

Question 13

wat te doen als proportional hazard assumption niet gehaald wordt

Answer 13

stratified cox model

time varying covariatesd in je model