Thermogravimetric determination of l-ascorbic acid in non-effervescent formulations using multiple linear regression with temperature selection by the successive projections algorithm

Changing experimental thermogravimetric conditions, such as heating rate and sample mass, may not be sufficient to eliminate the effect of interactions in pharmaceutical analysis. This motivates the investigation of a chemometric approach to determine active drugs in pharmaceutical formulations. The use of multiple linear regression (MLR) with temperatures selected by the successive projections algorithm (SPA) for determination of l-ascorbic acid (AA) in simulated non-effervescent formulations using microcrystalline cellulose as excipient is evaluated. For comparison, two other multivariate calibration methods, MLR with temperatures selected by genetic algorithm (GA) and partial least squares (PLS) using the entire range of temperatures were chosen. MLR–SPA provided the best predictions, in agreement with the expected AA concentration (correlation of 0.991 and root-mean-square error of 0.8%, m/m in the range 61.3–74.9%, m/m). No significant differences were found between the MLR–SPA values and those from iodimetric titration, according to t-test at 95% confidence level.

► Physical–chemical interactions may interfere in TGA analysis of pharmaceuticals.
► Chemometry was used in ascorbic acid/methylcellulose/water mixture TGA analysis.
► Multiple linear regression with temperature selection by SPA showed to be helpful.
► Results agreed with those from official method of analysis for ascorbic acid.
► This points out for a usefulness of such strategy in TGA pharmaceutical analysis.