For example, age analyzed with 4 other variables then analyzed with 8 variables the second time. Can the AUC and p value for the age change in both analysis situations?

Depends, If you're computing the ROC curve + AUC for a single variable (e.g., age) alone, then adding other variables to the model does not change the ROC/AUC for age alone since you're not conditioning on the others.

"age analyzed with 4 other variables then with 8 variables" - if you refer to multivariable models where age is included as one of several predictors, then 100% yes, AUC and the p-value associated with age can absolutely change

Thank you Marius

It is neither regression, univariate nor multivariate analysis, and we did not search for Model

We do ROC curve to obain separate evaluation for each variable.

Age was analysed as other variables, example (first time; age, weight and hight, the second time age, weight, hight, respiratory rate and pulse)

Ah that clarifies things, then no, the AUC & p-value for age should not change

Jamil Obaid

If you're using a univariate ROC curve, where you test just one variable against a binary outcome, then:

AUC (Area Under Curve) and p-value should stay the same, no matter what other variables you're analyzing elsewhere.

That’s because you're only assessing the discriminative power of that single variable independently.

But if you're using multivariate models like logistic regression, and you compute ROC curves and AUCs based on predicted probabilities from the full model (including multiple variables), then:

The AUC for each model can change depending on which other variables are included.

The p-value for a variable can also change, since it depends on the unique contribution of that variable given the others (i.e. adjusted effect).

Thank you Dr.Anuraj,

I used ROC curve alone without regression analysis. I got different results.

You said that the results must be the same, that's true. I will try to replace the softwar.