This contrasts with the conventionally used histopathological classification which highlighted a similar distribution of recurrence in high- and low-risk subgroups (Table 2). The integration of BRCA1 and TP73 markers into the panel of genes did not increase accuracy when either or both were considered in methylation status click here analysis (Table 4b). Table 4 Number of hypermethylated markers in recurrent lesions   Sensitivity (%) Specificity (%) Accuracy (%) (95% CI) (95% CI) (95% CI) a) FHIT, MLH1, ATM       ≥1 61.29 (43.82-76.27) 93.61 (82.84-97.81) 80.76 (72.02-89.52) ≥2 22.58 (11.40-39.81) 100 (92.44-100) 69.23 (58.99-79.47)

≥3 6.45 (1.79-20.72) JSH-23 cost 100 (92.44-100) 62.82 (52.09-73.55) b) FHIT, MLH1, ATM, TP73, BRCA1       ≥1 70.96 (53.41-83.90) 85.11 (72.31-92.59) 79.49 (70.53-88.45) ≥2 38.71 (23.73-56.18) 95.74 (85.75-98.83) 73.08

(63.24-82.92) ≥3 16.13 (7.09-32.63) 100 (92.44-100) 66.66 (56.21-77.13) NCT-501 in vitro ≥4 6.45 (1.79-20.72) 100 (92.44-100) 62.82 (52.09-73.55) ≥5 3.22 (0.57-16.19) 100 (92.44-100) 61.53 (50.74-72.34) c) FHIT, MLH1       ≥1 58.06 (40.77-73.58) 95.74 (85.75-98.83) 80.77 (72.02-89.52) ≥2 9.68 (3.35-24.90) 100 (92.44-100) 64.10 (53.45-74.75) Sensitivity, R patients who were correctly identified by the hypermethylated profile; Specificity, NR patients who were correctly identified by the hypermethylated profile; Accuracy, R patients, correctly identified by the hypermethylated profile, and NR patients, correctly identified by the hypermethylated profile, divided by the total

series; 95% CI, 95% confidence intervals. Unconditional logistic regression analysis was carried out to evaluate the capacity of MLH1, ATM and FHIT gene methylation to predict recurrence. FHIT and MLH1 proved to be independent variables with an RR of recurrence of 35.30 (95% CI 4.15-300.06, P = 0.001) and 17.68 (95% CI 1.91-163.54, next P = 0.011), respectively. CIMP analysis showed that hypermethylation of at least 1 of these gene promoters identified recurring adenomas with 58% sensitivity and 96% specificity (Table 4c). Methylation status was not related to age or grade of dysplasia. Conversely, a higher frequency of MLH1 hypermethylation was associated with site of lesion. In particular, a higher frequency of methylated MLH1 was observed in ascending with respect to descending lesions (71% and 29%, respectively, P = 0.07). Validation of MS-MLPA results Pyrosequencing measures the methylation level of single promoter CpG sites and is used to confirm the results from other analytical methods [23]. The average methylation percentage of the same CpG sites as those used for the MS-MLPA approach was considered for data analysis (data not shown). This approach was only utilized for MLH1 and ATM as reliable results were not obtained for FHIT. For this reason, FHIT was evaluated by immunohistochemistry.