Actually, you can as long as you're using standard NAT ("ip nat inside source static") and not using NVI ("ip nat source static") for your forwards. You apply the route-map at the end of the static NAT statement to indicate which interface it should apply to. So, if you have something like this:
ip access-list extended ACL_NAT
permit ip 192.168.0.0 255.255.255.0 any
!
route-map RM_NAT_ISP1
match ip address ACL_NAT
match interface GigabitEthernet0/1
!
route-map RM_NAT_ISP2
match ip address ACL_NAT
match interface GigabitEthernet0/2
Using port 80/tcp as an example, you can do this:
ip nat inside source static tcp x.x.x.x 80 y.y.y.y 80 route-map RM_NAT_ISP1
ip nat inside source static tcp x.x.x.x 80 z.z.z.z 80 route-map RM_NAT_ISP2
Just substitute x.x.x.x with the LAN address of the machine you're forwarding to, y.y.y.y with the WAN address you're forwarding from on ISP1 and z.z.z.z with the WAN address of the ISP you're forwarding from on ISP2. The static NAT will be conditional on the route-map at this point.
This works with TCP, UDP and IP forwarding but does require that you use an IPv4 address for your WAN address. For whatever reason, it doesn't work if you use an interface... so if you're using dynamic addresses, this will be more complicated.
