In a recent paper, (McNabb, 1978), we set up a method allowing to compute both the transient and steady-state exchange terms between the matrix and fractured regions of a naturally fractured porous medium using the continuous time random walk method (CTRW). In particular, the exchange coefficient alpha parametrizing the large-scale exchange term was computed on physical grounds using a time integration of the so-called time correlation function corresponding to the particle presence in the fractures. On the other hand, the large scale averaging theory (LSAT) as developed by Quintard and Whitaker (Quintard and Whitaker, 1996) gives another definition for this exchange coefficient alpha. It also provides a so-called 'closure problem' allowing to compute alpha from the solution of a well-defined steady state boundary value problem, to be solved over a representative volume of the high resolution fractured map. The goal of the present paper is to show analytically that both definitions coincide, yielding a unique and well defined value of the alpha coefficient. This provides an unification of two approaches whose respective backgrounds are very different at the first glance.