Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • CoxPH assumption violation

    Hello everyone,
    I am new to this forum. I am a stata user but I don't have much background in statistics. I did CoxPH model to detect association between risk of hypertrophic cardiomyopathy (phenopos_fup) in different genotype groups(genotype_group) adjusted for age at 1st echocardiogram (age_1st echo) and sex. I check PH assumption after the test and found that age_1stecho violated the assumption. So, I re-ran the model by adding age_1stecho as tvc. But it didn't seem to fix the violation. In this case, how could I proceed? My main interest is risk of HCM in different genotype. Can I still interpret the effect size as it is in model 1 and ignore the PH assumption violation for age as it's not main effect I am interested in?

    Model 1
    . stcox i.sex age_1stecho i.genotype_group,nolog

    Failure _d: phenopos_fup==1
    Analysis time _t: time_to_phenopos

    Cox regression with no ties

    No. of subjects = 581 Number of obs = 581
    No. of failures = 105
    Time at risk = 7,065.6012
    LR chi2(4) = 77.15
    Log likelihood = -524.05486 Prob > chi2 = 0.0000

    -----------------------------------------------------------------------------------
    _t | Haz. ratio Std. err. z P>|z| [95% conf. interval]
    ------------------+----------------------------------------------------------------
    sex |
    Female | .6368161 .1330752 -2.16 0.031 .4228037 .959156
    age_1stecho | .8817264 .0198928 -5.58 0.000 .8435867 .9215904
    |
    genotype_group |
    Genotype - | .8720211 .2339626 -0.51 0.610 .5154069 1.475379
    Genotype unknown | .2573629 .0581113 -6.01 0.000 .1653289 .4006295
    -----------------------------------------------------------------------------------

    . estat phtest,detail

    Test of proportional-hazards assumption

    Time function: Analysis time
    --------------------------------------------------------
    | rho chi2 df Prob>chi2
    -------------+------------------------------------------
    1b.sex | . . 1 .
    2.sex | -0.14604 2.34 1 0.1260
    age_1stecho | 0.51468 16.61 1 0.0000
    1b.genotyp~p | . . 1 .
    2.genotype~p | 0.04664 0.22 1 0.6358
    3.genotype~p | 0.08506 0.76 1 0.3819
    -------------+------------------------------------------
    Global test | 18.81 4 0.0009
    --------------------------------------------------------

    Model 2
    . stcox i.sex age_1stecho i.genotype_group,tvc(age_1stecho) nolog

    Failure _d: phenopos_fup==1
    Analysis time _t: time_to_phenopos


    Cox regression with no ties

    No. of subjects = 581 Number of obs = 581
    No. of failures = 105
    Time at risk = 7,065.6012
    LR chi2(5) = 98.98
    Log likelihood = -513.13588 Prob > chi2 = 0.0000

    -----------------------------------------------------------------------------------
    _t | Haz. ratio Std. err. z P>|z| [95% conf. interval]
    ------------------+----------------------------------------------------------------
    main |
    sex |
    Female | .62324 .1298635 -2.27 0.023 .414278 .9376026
    age_1stecho | .6223885 .0538531 -5.48 0.000 .5253032 .7374169
    |
    genotype_group |
    Genotype - | .8949743 .2401153 -0.41 0.679 .52898 1.514195
    Genotype unknown | .2524777 .0571142 -6.08 0.000 .1620573 .3933485
    ------------------+----------------------------------------------------------------
    tvc |
    age_1stecho | 1.028269 .0067182 4.27 0.000 1.015185 1.041521
    -----------------------------------------------------------------------------------
    Note: Variables in tvc equation interacted with _t.



    Thanks
    Sandar


    .

  • #2
    Sorry, I tried to format the table previously and was not sucessful. I have now fixed tables, Could anyone help me whether I can interpret effect size for genotype group as in model 1 or whether I should make any adjustment?
    Mode; 1: CoxPH model for risk of HCM
    _t Haz. ratio Std. err. z P>|z| [95% conf. interval]
    sex
    Female .6368161 .1330752 -2.16 0.031 .4228037 .959156
    age_1stecho .8817264 .0198928 -5.58 0.000 .8435867 .9215904
    genotype_group
    Genotype - .8720211 .2339626 -0.51 0.610 .5154069 1.475379
    Genotype unknown .2573629 .0581113 -6.01 0.000 .1653289 .4006295
    Model 2 : CoxPH model for risk of HCM with tvc
    _t Haz. ratio Std. err. z P>|z| [95% conf. interval]
    main
    sex
    Female .62324 .1298635 -2.27 0.023 .414278 .9376026
    age_1stecho .6223885 .0538531 -5.48 0.000 .5253032 .7374169
    genotype_group
    Genotype - .8949743 .2401153 -0.41 0.679 .52898 1.514195
    Genotype unknown .2524777 .0571142 -6.08 0.000 .1620573 .3933485
    tvc
    age_1stecho 1.028269 .0067182 4.27 0.000 1.015185 1.041521

    Comment

    Working...
    X