190 Chapter 7 (45). However, alternative results have been described for methylation of the SEPT9 gene in plasma samples of a small group of 11 colorectal cancer patients. The highest SEPT9 methylation levels were found at midnight, detecting all (pre)cancers included in the study, as compared to 77.7% of the cases during the other time points. Methylation levels of CDO1 and SOX17 reached moderate to good reproducibility (0.74 and 0.57), while TAC1 showed poor reproducibility (0.14). The reproducibility of the markers seems to reflect the level of DNA methylation detected. TAC1 with the lowest ICC showed the lowest methylation levels of the three markers studied. Comparable variation within subjects has been observed for the mutant allele concentration of tumor-specific mutations in KRAS and P53 in the plasma of non-progressive lung cancer patients (22). From a patient monitoring perspective, the between-subject variation observed in this study implies that evaluating DNA methylation levels within individual patients, using longitudinal testing, may be more useful than using a dichotomous population-based threshold. Also, contrary to the moment of sampling, additional value was observed with collecting multiple urine samples for markers with the highest within-subject variability. This suggests that detecting lung cancer in urine will become more likely when multiple urine samples are being collected. Collecting urine at multiple time points has also been proposed by Liu et al. (10) who published the proof-ofconcept study for lung cancer detection in urine by the analysis of methylated DNA. A previous study by Hubers et al. (46) indicated that prolonged sampling increased the sensitivity of lung cancer detection by methylation analysis in sputum, with a slight decrease in specificity. Other options to improve test accuracy would be to increase urine volume, as shown for bladder cancer detection in urine (47), or to pool several urine samples before DNA isolation, as suggested for gene polymorphism analysis (43). The significant association found between the methylation levels of each marker and the cfDNA concentration of the urine sample indicates that adjusting the threshold of the reference gene, used for normalization and for excluding samples with insufficient DNA quality or quantity, could also increase the test reproducibility and accuracy. The current study has several limitations. Due to the substantial biological variance observed, our sample size may have been too small to accurately address systematic changes of cfDNA concentration and methylation levels in time. Nonetheless, the total of 138 urine samples included is similar or even higher as compared to previous studies assessing the biological variation of cfDNA in plasma (22-24). Moreover, because this study only included patients with active disease, further studies are warranted to examine the biological variation of cfDNA in patients with early-stage or non-progressive disease. Apart from that, the variability of the current marker selection in the urine of healthy controls and individuals at risk for lung cancer (e.g. heavy smokers or patients
RkJQdWJsaXNoZXIy MjY0ODMw