The line-level alternative requires that impedance matching be considered. More often than not preamps providing two sets of outputs do not separately buffer those outputs (i.e., they are simply connected together inside the rear panel, and driven by the same output stage). In that situation the preamp's output stage will see a load impedance equal to the parallel combination of the input impedance of the power amp and the sub, which will be significantly lower than either of those impedances individually.
That is most likely to be a problem if the preamp is tube-based, although it can sometimes also be a problem with solid state preamps. The value of the parallel combination of the two input impedances is equal to the product (multiplication) of the two numbers divided by their sum. For instance, if the input impedance of the power amp is 50K, and the input impedance of the sub is 20K, their combined input impedance is about 14.3K. That number should ideally be 10x or more higher than the output impedance of the preamp, at the frequency for which that output impedance is highest (which is often 20Hz). If only a nominal output impedance is known (presumably at mid-frequencies), I suggest using a factor of 50x.
Also, if the length of either or both of the interconnect cables is long, and the output impedance of the preamp is high, using low capacitance cables becomes increasingly important. The capacitance of the cables to the sub will affect the signals going to the main amp, as well as the signals going to the sub. A slight but perceptible high frequency rolloff might occur if the total capacitance of the two cables is high and the preamp's output impedance is high. Again, this assumes that the preamp's two outputs are not individually buffered.
Regards,
-- Al