If you want to measure the actual die temperature precisely you may be able to use an internal diode anyway. Such parasitic diodes usually exist, for example from an input to V- on a bipolar chip. You could calibrate the diode in an oven or environmental chamber on a particular chip (no power). To measure the actual temperature you would remove the power and quickly (before the temperature changes much) check the diode.
Alternately, a small very fine wire bead or fine ribbon thermocouple to the leadframe will measure some approximate temperature, and an IR thermometer looking at the encapsulation an approximation of the package temperature. If you use a thermocouple, double back the wires and heatsink to other parts of the leadframe to prevent excessive errors due to heat loss down the wires (while maintaining electrical isolation).