A method of detecting a predisposition to, or the incidence of, cancer in a sample comprises detecting an epigenetic change in at least one gene selected from an NDRG4/NDRG2 subfamily gene, GATA4, OSMR, GATA5, SFRP1, ADAM23, JPH3, SFRP2, APC, MGMT, TFPI2, BNIP3, FOXE1, SYNE1, S0X17, PHACTR3 and JAM3, wherein detection of the epigenetic change is indicative of a predisposition to, or the incidence of, cancer. Also described are pharmacogenetic methods for determining suitable treatment regimens for cancer and methods for treating cancer patients, based around selection of the patients according to the methods of the invention. The present invention is also concerned with improved methods of collecting, processing and analyzing samples, in particular body fluid samples. These methods may be useful in diagnosing, staging or otherwise characterizing various diseases. The invention also relates to methods for identifying, diagnosing, staging or otherwise characterizing cancers, in particular gastrointestinal cancers such as colorectal cancers, gastric cancers and oesophageal cancers. The methods of the invention relate, inter alia, to isolating and analyzing the human DNA component from faecal samples and blood-based samples.