Three-dimensional (3-D) measurements of flame stretch are experimentally challenging. In this paper, two-dimensional (2-D) and 3-D measurements of flame stretch and turbulence–flame interactions were examined using direct numerical simulation (DNS) data of turbulent premixed flames, and models to estimate 3-D statistics of flame stretch-related quantities by correcting 2-D measurements were developed. A variety of DNS cases were simulated, including three freely propagating planar flames without a mean shear and a slot-jet flame with a mean shear. The main findings are summarized as follows. First, the mean shear mainly influences the flame orientations. However, it does not change the flame stretch and turbulence–flame interactions qualitatively. The distributions of out-of-plane angle of all cases are nearly isotropic. Second, models were proposed to approximate the 3-D statistics of flame stretch-related quantities using 2-D measurements, the performance of which was verified by comparing modelled and actual 3-D surface averages and probability density functions of tangential strain rate, curvature and displacement velocity. Third, 2-D measurements of flame stretch capture properly the trends of the 3-D results, with flame surface area being produced in low curvature regions and destroyed in highly curved regions. However, the magnitude of flame stretch was under-estimated in 2-D measurements. Finally, 2-D and 3-D turbulence–flame interactions were examined. The flame normal vector is aligned with the most compressive strain rate in both 2-D and 3-D measurements. Meanwhile, the flame normal vector is misaligned (weakly aligned) with the most extensive strain rate in 3-D (2-D) measurements, highlighting the difference in 2-D and 3-D results of turbulence–flame interactions.