If there exists 异方差 based on a gender for example, this means the data is generated from two different distribution, hence the model will have two different regression lines.
The testing procedure will be ,
1) run a model without the gender
2) run a model with the gender but interacting with other co-variates.
3) compare the 2loglike value as (likelihood ration test)
In SAS you may do as
proc logistics data=<dsn>;
model y= x;
run;
proc logistics data=<dsn>;
class gender;
model y= gender gender*x;
run;