Analysis of methylated DNA, which refers to 5-methycytosine (5mC) versus cytosine (C) at specific loci in genomic DNA (gDNA), has received increased attention in epigenomics, particularly in the area of cancer biomarkers. Many different methods for analysis of methylated DNA rely on initial reaction of gDNA with concentrated acidic sodium bisulfite to quantitatively convert C to uracil (U) via sulfonation of denatured, single-stranded gDNA under conditions where 5mC is resistant to analogous sulfonation leading to thymine (T). These methods typically employ polymerase chain reaction (PCR) amplification after bisulfite conversion, thereby leading to readily detectable amounts of amplicons where T and C are measured as surrogates for C and 5mC in the original unconverted gDNA. However, incomplete bisulfite conversion of C in gDNA has been reported to be a common source of error in analysis of methylated DNA. Incomplete conversion can be revealed during the course of bisulfite sequencing, which is the generally accepted "gold standard" for analysis of methylated DNA. Previous bisulfite sequencing investigations of conventional predenaturation of gDNA with NaOH followed by the use of bisulfite containing added urea to maintain denaturation and thus mitigate incomplete conversion of C have been reported to give conflicting results. The current study describes a new approach where conventional predenaturation of gDNA with NaOH is instead achieved with formamide and maintains denaturation during subsequent sample handling and sulfonation. This formamide-based method was applied to 46 formalin-fixed/paraffin-embedded (FFPE) biopsy tissue specimens from well-characterized patients with primary prostate cancer. These specimens were representative of difficult-to-analyze samples due to the chemically compromised nature of the gDNA, which was recovered by modifying the protocol for a commercially available total RNA/DNA extraction kit (RecoverALL). An additional novel aspect of this study was analysis of CpG-rich promoter regions of two prostate cancer-related genes: glutathione S-transferase pi (GSTPi) and retinoic acid receptor beta2 (RARbeta2). High-quality bisulfite sequencing results were obtained for both genes in 43 of 46 (93%) specimens. Detection of methylated GSTPi and RARbeta2 genes was significantly associated with primary prostate cancer as compared with the benign prostate (Fisher's exact test, P < 0.001). The sensitivity and specificity of detection of methylated GSTPi and RARbeta2 genes were 86% and 100% and 91% and 100%, respectively. Moreover, the presence of either methylated gene was detected in primary prostate cancer with sensitivity and specificity of 100% and 100%, respectively. The results demonstrated a high degree of reliability of formamide-based denaturation and bisulfite conversion that should extend, generally, to FFPE and other types of samples intended for any analytical method predicated on bisulfite conversion. This pilot study also demonstrated the efficacy of determining methylation of these two genes with high sensitivity and specificity in FFPE biopsy tissue specimens. Moreover, the results showed a highly significant association of methylated GSTPi and RARbeta2 genes with primary prostate cancer. Finally, this improved procedure for determining these two methylated genes may allow the detection of prostate cancer cells in core biopsy specimens with insufficient numbers of cells and poor morphology.