This study aims to improve a variation of the differential scanning calorimetry (DSC) technique for measuring boiling points of pure compounds at low pressures. Using a well-known n-paraffin (n-hexadecane), experimental boiling points at a pressure of 3.47 kPa with u(P) = 0.07 kPa were obtained by using a variation of the DSC technique, which consists of placing samples inside hermetically sealed aluminum crucibles, with a pinhole (diameter of 0.8 mm) made on the lid and a tungsten carbide ball with a diameter of 1.0 mm over it. Experiments were configured at nine different combinations of heating rates (K.min(-1)) and sample sizes (mg) following a full factorial design (22 trials plus a star configuration and three central points). Individual and combined effects of these two independent variables on the difference between experimental and estimated boiling points (NIST Thermo Data Engine v. 5.0 - Aspen Plus v. 8.4) were investigated. The results obtained in this work reveal that although both factors affect individually the accuracy of this variation of the DSC technique, the effect of heating rate is the most important. An optimized region of combinations of heating rate and sample size for determining boiling points of pure compounds at low pressures was obtained using the response-surface methodology (RSM). Within this optimized region, a selected condition, combining a heating rate of 24.52 K.min(-1) and a sample size of (4.6 +/- 0.5) mg, was tested for six different compounds (92.094-302.37 g mol(-1)) comprising four fatty compounds (tributyrin, monocaprylin, octanoic acid and 1-octadecanol), glycerol and n-octadecane, besides n-hexadecane. This condition was also successfully applied for obtaining boiling points of n-hexadecane at pressures up to 18.66 kPa with u(P) = 0.18 kPa. The optimized region obtained in this work, in terms of heating rates and sample sizes, is specific for the crucible configuration tested. (C) 2016 Elsevier Ltd. All rights reserved.