To obtain protection under the UPOV Convention, a new plant variety must satisfy three criteria: distinctness, uniformity and stability. New varieties under assessment are usually compared to existing varieties and other candidates based on a range of characteristics measured in field trials over two or three seasons. A widely adopted test used for assessing uniformity based on quantitative characteristics is known as the Combined-Over-Years-Uniformity method (COYU). With this method, any dependency between uniformity, represented by standard deviations, and level of expression, represented by means, is accounted for using an adjustment based on moving averages. We show that the moving average adjustment method currently used in COYU leads to a higher rejection rate of uniform varieties than intended. We investigate alternative methods of adjustment, finding that cubic smoothing splines with four degrees of freedom perform well. This approach is flexible enough to fit relationships between uniformity and level of expression seen in practice, whilst introducing only minimal bias.