<p>This paper proposes a model predictive path integral (MPPI) control framework with a parameterized-policy for quadrotor path-following. Although MPPI control is attractive for nonlinear, nonconvex problems, its direct application to path-following is hindered by a large control-input search space and sensitivity to model imperfections. To address these issues, the proposed method embeds a pursuit-type guidance law with virtual-target placement as a low-dimensional, physically interpretable policy. Rather than optimizing full three-axis accelerations, the controller samples and adapts a few guidance parameters (e.g., guidance gain and axial/forward acceleration), thereby improving sampling efficiency and computational tractability. To handle model uncertainty and wind disturbances, a nonlinear disturbance observer (NDO) is integrated to provide online disturbance compensation. The resulting formulation couples along-path progress with bounded cross-track error through a problem-specific cost and yields smooth, stable control suited to typical onboard resource constraints. Numerical simulations on piecewise-linear reference paths under no-wind and wind conditions indicate consistently lower path-following error, smoother control inputs, and markedly higher sample efficiency than both guidance-only baselines and standard MPPI control with full-dimensional inputs. These results support the practicality of the proposed approach for robust, energy-aware quadrotor guidance.</p>