Sodium bisulfite-induced cytosine deamination/PCR (CD-PCR) is currently the most sensitive and robust method to determine the methylation status of all cytosines in a specific DNA sequence. The CDPCR products are directly sequenced with Thermosequenase and capillary electrophoresis; peak areas are then used to determine the mole fraction of methylated cytosines at each site in a single analysis. Here we show that, if the original DNA sample is a mixture of methylated and unmethylated DNA, conventional CDPCR discriminates against the sequence originating from the methylated DNA; CDPCR product does not accurately represent the methylation status of the original DNA sample. While CDPCR bias can lead to serious errors when determining methylation levels, the addition of betaine (N,N,N-trimethylglycine) to the PCR reaction buffer reduces this bias to less than 10%.