The pooled specificity of GSTP1 was excellent (0. 89, 95% CI, 0.80–0.95) and much higher than the specificity of PSA. The specificity in each subgroup (stratified by sample type and methylation method) remained above 0.86. The sensitivity of GSTP1 was 0.63 (95% CI, 0.50–0.75) for samples collected before treatment and 0.40 (95% CI, 0.25–0.78) for samples collected after treatment; these sensitivities were not higher than the sensitivity of PSA screening. These results suggest that plasma, serum, or urine samples may complement PSA screening for prostate cancer diagnosis, although the positive link between GSTP1 methylation and pathological stage needs to be evaluated in more studies.
Collecting plasma/serum or urine samples is a non-invasive procedure, whereas invasive biopsy procedures may cause pain, anxiety, and increased medical costs. Urine samples were voided urine except the urine samples collected in the following studies: four were collected after a massage (Goessl et al, 2001b
; Rogers et al, 2006
; Woodson et al, 2008
; Payne et al, 2009
) and one after a biopsy (Rogers et al, 2006
). High specificity remained even after we excluded the studies with urine collection after a massage.
This study highlights several important issues. First, we identified and systemically evaluated the methylation test at the GSTP1 promoter as an important potential test to complement PSA screening. As a complement rather than a replacement for PSA is needed, a high specificity is more important than a high sensitivity. To combine the strengths of both tests, they should be used sequentially, not simultaneously. The PSA test will be initially used to screen out potential patients, and the GSTP1 methylation test will then be given to those patients who have elevated PSA levels. Only those who have elevated PSA levels, followed by positive results on the GSTP1 methylation test, will undergo further biopsies. With its high specificity, the methylation test will exclude patients unlikely to have PCa but have elevated PSA levels. Using the two tests sequentially will reduce the number of unnecessary biopsies considerably, compared with using the PSA test alone. Serial testing has been used clinically for embolism and diarrhea (Fekety, 1997
; Wells et al, 2001
Second, unlike previous studies and reviews, we rigorously evaluated the specificity of GSTP1 by excluding healthy controls. In epidemiological research, we use controls that represent the population from which the cases were derived. As described above, randomly selected healthy controls usually do not have elevated levels of PSA or abnormal urological symptoms; therefore, they cannot represent patients who have high levels of PSA and undergo biopsy tests. Third, no previous studies have systemically evaluated the diagnostic value of measuring GSTP1 promoter methylation in different types of body fluids for prostate cancer diagnosis. This study indicates that the use of plasma/serum or urine samples for prostate cancer diagnosis is an important, non-invasive procedure that can complement PSA screening and minimise unnecessary biopsies.
Future assays that measure DNA methylation at gene promoters need to be standardised, simplified, and evaluated with external quality assurance programmes. Quantitative methods, such as pyrosequencing (<200
bp) and MassArrays (<600
bp), which truly quantify all the DNA methylation in the CpG islands and measure levels of all CpG dinucleotides, are also high-throughput technologies. Therefore, pyrosequencing and MassArrays are considered to be more efficient for validating the DNA methylation of gene promoters.
Of note, indicates that the DNA methylation test in whole blood samples is less sensitive to prostate cancer stages than the same test done in plasma, serum, or urine (Roupret et al, 2008
). The methylated DNA in plasma, serum, or urine most likely derives from cancer cells whereas the methylated DNA detected in whole blood can be released from white blood cells as well. Therefore, the DNA methylation test in plasma, serum, or urine may be more accurate than the same test applied to whole blood in reflecting the severity of cancer stage. As most prostate cancer cases are detected at an early stage, the current PSA test does not predict specific prostate cancer stages. If GSTP1 methylation in plasma, serum, or urine samples is associated with pathological stage, this test will be even more appealing in addition to its high-specificity feature. More studies are warranted to confirm this finding.
The present study has several limitations. First, the validation assay of gene promoter methylation used in each study was different; some used N-MSP, others used Q-MSP, MSRE-Qpcr, or bisulfite sequencing, adding additional heterogeneity. Gonzalgo et al, 2004
(Gonzalgo et al, 2004
) commented that primers selected at different regions on the same CpG island may have different sensitivities and specificities. Fortunately, the studies included in this paper all chose primers targeting the CpG dinucleotides in the 5′ promoter region starting from −80 to 400
nt from its transcriptional start site; hence findings in these studies are still believed to be appropriate for our analyses, as DNA methylation mostly occurs at the 5′ promoter region. Nevertheless, the specificity of GSTP1 across different methylation methods was consistently higher than the specificity of PSA, even though a wide range of sensitivities for GSTP1 was noted. This indicates the robustness of specificity of the GSTP1 methylation assay. Furthermore, our study did not find an association between sensitivity or specificity and methylation method.
Second, only 15 studies that did not use healthy controls could be used for specificity calculations. However, some of these studies used more than one specimen and gave us additional statistical power. Nevertheless, the evidence is compelling, in that the majority of individual GSTP1 specificities calculated here were higher than 0.8, and the overall pooled specificity was 0.89.
Third, the sample collection time varied widely among the studies. As mentioned above, because of our concern over the influence of treatment, we limited our analysis to untreated cases when analysing the associations between gene methylation and pathological stage and other factors, thereby decreasing our statistical power.
In summary, we summarised primer sequences, nucleotide position, and PCR methods in each study and evaluated them using a meta-analysis, which is a unique approach compared with a traditional review without any statistical analyses. Our study represents a new trend in epidemiology: a cross-disciplinary approach between molecular biology and epidemiology. Measuring DNA methylation at gene promoters has the potential to provide a new generation of biomarkers for prostate cancer diagnosis. Future studies should focus on the following tasks: (1) standardising the primers and the PCR protocols for each target gene; (2) using plasma, serum, or urine samples; (3) using patients with negative biopsies as controls rather than randomly selected healthy controls; and (4) collecting samples from cases before biopsies or at least before treatment to improve sensitivity. These tasks will reduce the heterogeneity among studies, enabling us to conduct an accurate meta-analysis to find a complement for the PSA test. Finally, more studies are needed to examine the association between gene methylation status and the stage and prognosis of prostate cancer. This will help avoid unnecessary treatment of some localised prostate cancers, as prostate cancer therapies are associated with significant adverse effects that impact patients' health and quality of life.