Evaluation of a frequency lowering hearing instrument algorithm using a non-inferiority test

The primary goal of hearing instrument verification is to demonstrate an improvement on a relevant outcome. It is imprudent to implement an algorithm that improves one outcome while simultaneously degrading another. A traditional test typically uses a superiority hypothesis – H0: New = Conventional and H1: New ≠ Conventional. The absence of statistical significance may be interpreted incorrectly as an absence of clinically relevant differences. An alternative is to start the test with a non-inferiority hypothesis – H0: New < Conventional and H1:    New ≥ Conventional. Cross-over designs are often employed because treatment differences are frequently measured within a subject rather than between subjects. Each test period should be long enough for the subject to become acclimatized to each processing change. With these conditions, it is possible to estimate, with the same test, the overall effect of the developed feature and also the period effect. The method of using a cross-over design with a non-inferiority analysis was applied in the testing of a new frequency lowering algorithm. Improved high-frequency functional gain and fricative discrimination was observed. Significant non-inferior SSQ scores between the processing on and off were seen while no period effect was found. These results provide a good approximation of ‘real world’ acceptance.