Erratic breathing at rest and during exercise has been described in patients with dysfunctional breathing (DB). Sample entropy (SampEn) and approximate entropy (ApEn) were developed to assess the predictability of parameters in time series data. However, SampEn and ApEn were developed for data without trends. Their potential role in interpreting datasets during cardiopulmonary exercise testing (CPET) remains uncertain. We used simulations based on real-life exercise data to evaluate the effects of trends, varying numbers of analyzed respiratory cycles, and different coefficients used to calculate the tolerance interval on ApEn and SampEn measurements. We tested the LOESS_0.75 method to correct the trend in the calculation of SampEn and ApEn. Trends led to a significant underestimation of ApEn and SampEn values. The LOESS_0.75 method yielded values very close to the predicted ApEn and SampEn. Modifying the number of respiratory cycles and the coefficient used to calculate the tolerance interval had major implications on ApEn and SampEn. In conclusion, the residuals from the LOESS_0.75 method can provide a corrected estimate of ApEn and SampEn in exercise data. The number of respiratory cycles analyzed and the coefficient used to calculate SampEn or ApEn should always be reported.