Cuffdiff find genes that are differentially spliced or switch promoters between conditions using the Jensen-Shannon distance metric. Previous versions of Cuffdiff tested for the signficance of observed shifts in relative isoform abundance using an analytic approximation of the variance of this metric for each gene. However, when few replicates are available or sequencing is shallow, this approximation can be poor. This release improves Cuffdiff's accuracy with a computational-derived estimate of the variance of the Jensen-Shannon metric by sampling in each gene. This improvement substantially reduces the false positive rate of Cuffdiff's tests.