A metamodel-based torque optimization strategy is proposed for the efficient design and electromagnetic performance optimization of outer rotor permanent magnet brushless DC motors (ORPMBLDCM), which is based on the dimensionality reduction and generalization principles in statistical modeling theory. The strategy combines Criteria Importance through Intercrieria Correlation and Technology for Order Preference by Similarity to an Ideal Solution (CRITIC-TOPSIS) theory. A parameterized two-dimensional finite element calculation model is established using a 10-pole 12-slot ORPMBLDCM as an example. The optimization objectives are set to be the low cogging torque, low torque ripple, and high average electromagnetic torque. The actual model is approximated using coefficient of prognosis and the constructed optimal prediction metamodel, based on variance-based sensitivity analysis. Multi-objective evolutionary algorithms are used to optimize the objectives and obtain the Pareto optimal solution set. Finally, the CRITIC-TOPSIS method is used to globally rank the feasible solution set and determine the optimal structural dimension parameters. The results of the finite element analysis demonstrate a 77.9\% reduction in cogging torque, a 18.5\% improvement in average electromagnetic torque, and a 14.5\% reduction in torque ripple. The effectiveness of torque optimization strategy using metamodel and CRITIC-TOPSIS method in improving the performance of ORPMBLDCM has been validated. The presented torque optimization strategy can guide the design of brushless DC motors.