Using autodiff types for ODEint solvers

I am exploring whether it is possible to compute first order (or even higher order) sensitivities of ODE solutions using Boost.ODEInt by giving the integrator an automatic differentiation type (like autodiff::dual or similar) as the state type.

My questions are:

- Is it feasible to use generic autodiff types in adaptive Runge-Kutta solvers or in stiff solvers provided by ODEInt to compute sensitivities directly?

- Are there recommended practices or caveats when doing so?

Any guidance, examples, or references would be greatly appreciated.

Thank you very much!