Can you post the model you have? It's hard to see why this should be a problem without some idea as to what your model looks like.
Most VCO models tend to work by converting the input voltage to a frequency (gain*V(in)+fc) and then integrating that (using idtmod), and then using this as the phase for sin() or cos(). If modelled like this, negative input voltages wouldn't be a problem - provided that the overall frequency remains positive.
In other words:
Code:inst_freq = gain*V(vin)+fc;
phase = idtmod(inst_freq,0,1);
V(vout) <+ amp*sin(2*`PI*phase);
Regards,
Andrew.