Low-income women are the group with the highest levels of obesity worldwide. In low-income settings, the use of predictive equations, which yields a measure of the individuals’ basal metabolic rate (BMR), is a feasible approach to estimate the individuals’ total energy expenditure (TEE), using the factorial method (calculated-TEE = BMR * Physical activity level), an important step of the obesity nutritional care. This study aimed to identify the predictive equation that, in conjunction with metabolic equivalents of tasks (MET) data from accelerometers, yields the calculated-TEE with better agreement compared to the TEE measured by doubly labelled water (TEE-DLW). Forty-five women aged 19-45 years, with excess weight; mothers of undernourished children were included. They received DLW to determine TEE (14 days), at the same time as they used triaxial accelerometers (7 days) to estimate their MET. The Bland-Altman method, paired-samples t-tests, concordance correlation coefficient (CCC), and root mean square error were used to assess the agreement. Maximum allowed differences were defined as 24%, based on the within-variance coefficient of the energy intake of the sample. Eleven equations were studied. The calculated-TEE obtained by 5 equations showed non-significant bias: Dietary Reference Intake (2005), FAO/WHO/UNU (2001), Harris-Benedict (1919), Henry-Rees (1991) and Schofield (1985). The mean percentage difference was, respectively: -1.5%, -0.8%, 2.2%, -2.2%, and 2.0%. Considering all parameters, FAO/WHO/UNU (2001) equation performed slightly better than the others; nevertheless, no equation in conjunction with the estimated-MET showed a calculated-TEE with its confidence intervals for the Bland-Altman limits of agreement inside the pre-defined acceptable range.