Path-integral derivation of the anomaly for the Hermitian equivalent of the complex PT-symmetric quartic Hamiltonian

It can be shown using operator techniques that the non-Hermitian PT-symmetric quantum mechanical Hamiltonian with a 'wrong-sign' quartic potential -gx4 is equivalent to a Hermitian Hamiltonian with a positive quartic potential together with a linear term. A naieve derivation of the same result in the path-integral approach misses this linear term. In a recent paper by Bender et al. it was pointed out that this term was in the nature of a parity anomaly and a more careful, discretized treatment of the path integral appeared to reproduce it successfully. However, on re-examination of this derivation we find that a yet more careful treatment is necessary, keeping terms that were ignored in that paper. An alternative, much simpler derivation is given using the additional potential that has been shown to appear whenever a change of variables to curvilinear coordinates is made in a functional integral