The frequency of Earth-like planets in the habitable zone of Sun-like stars (ηEarth) is a fundamental input in estimating the occurrence rate of life resembling that on Earth, and therefore an important parameter for designing future direct imaging missions. ηEarth is currently best estimated by the Kepler transit survey, but Earth-analog systems with long periods and shallow transits are on the edge of the survey’s sensitivity. The Roman Galactic Exoplanet Survey will be able to detect Earth-analog systems through microlensing, but similarly they will be on the boundary of its sensitivity due to low-mass ratios and small projected separations. We perform simulations of the Roman Galactic Exoplanet Survey to estimate its sensitivity to these systems. Roman’s ability to estimate ηEarth will hinge on the extrapolation from systems with larger mass-ratios and wider projected separations. Finally, we consider the possibility of improving the fidelity of estimates of ηEarth by interpolating between the results from Roman and Kepler by adopting a mass-radius relation for Earth-like planets.