WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
Building, troubleshooting and testing of these amplifiers should only be
performed by someone who is thoroughly familiar with
the safety precautions around high voltages.
The SPICE simulation will only be as good as the models for the valves are, and the model for the transformer. Apparently some people have had success with LTSpice/SwitcherCad (which is free), and possibly CircuitMaker, though I personally haven't bothered to learn how to use either.
There was a time when all these things were designed without the aid of computers, and many people still do. You will likely get better results by having a look at the curves, and using the general principles for minimising distortion such as choosing an appropriate operating point. Note that you can approximate the distortion by preparing composite curves after choosing the operating point from various formulae, but this can be rather tedious.
Choosing an appropriate operating point and load impedance, then fine tuning it in an actual prototype (you could build a notch filter, or use the computer sound card for distortion measurement) will be far more accurate than the computerised simulation, and may well give a better result.
Actually, I think I may have misunderstood your question.
If you already have a SPICE model, you could try applying a signal to the input, say a 1KHz sine wave, then monitor the output (which will of course be loaded).
Performing a Fourier transform on the output will give the distortion spectrum of the amplifier, THD can be calaculated as the geometric mean of the individual harmonics. Of course, the accuracy of this is still entirely dependent on how accurate the models are for your valves, transformers and other parts.
As stated above modelling programs are only as good as the models used and even then don't take into consideration device to device variations.
Modelling is good for getting operating points right, it can be good for frequency response PREDICTIONS, and it will generally let you know if there is a stabilty problem ( except for tube parasitics).
It a complete waste of time for any sort of accurate prediction of distortion (except gross stuff like clipping) - the models are just not good enough.
The best model of a tube is the glowing thing sitting on the bench in front of you.
Use modelling by all means for initial design BUT at some point you'll have to "bite the bullet" and just build the thing. Then you can finese it to your hearts content using Signal Generators, Oscilloscopes, Distortion Analysers (if your rich) BUT MOST OF ALL play music through it and use the cheapest and best test kit you have (YOUR EARS). Besides that if anyone has been able to exactly correlate distortion behaviour against listening pleasure I've yet to see his paper. A bit of second harmonic with a bit less third harmonic and preferrably no higher order harmonics is about as close as any has got.