Research published in Nature Communications demonstrates a test for the detection of cancer that requires 10 minutes or less to analyse.
Prof Paul Pharoah, Professor of Cancer Epidemiology at the University of Cambridge, said:
“This paper is too preliminary to be exciting.
“In short they have an interesting technology that can be used to determine patterns of methylation of DNA. And they have evidence that those patterns differ in cancer tissue compared to normal tissue.
“It is already known that the DNA methylation pattern in cancer tissues differ from those of normal tissues, but the assays to show this are not easy to perform as a simple test. What they have shown is that their test can distinguish the DNA methylation in cancer tissues and normal tissues. They compared 72 samples from several cancers with 31 normal tissue samples to show this.
“However, there are many issues with this study that limit its interpretation and it is clearly too preliminary to state that it could be a game changer. Well, anything could be a game changer but most developments probably are not!
“However, the evaluation of the accuracy of this test as described in the paper is very weak. Firstly they describe the area under the roc curve as the specificity “the ROC curve (Fig. 3c) for the range of tissue samples tested shows high-specificity for cancer detection (AUC = 0.909)”. The area under the ROC curve and the specificity are completely different measures so this makes no sense.
“And, the positive and negative predictive values quoted are entirely meaningless as these metrics are not just properties of the test, but also a property of the proportion of cancer tisses tested out of all tissues tested (which they have determined).
“… with high positive (PPV) and negative (NPV) predictive values (Table-Fig. 3c, PPV = 91.78%, NPV = 83.33%”
“Just to illustrate if all the samples tested were cancer tissues the positive predictive value would be 100% and the negative predictive value would be 0%.
“They also compared circulating DNA in samples from cancer patients with samples from so-called healthy controls. The details of these samples are in supplementary material that I do not have access to. Again the presentation of the findings of these analyses are sub-optimal.
“In general, there have been many novel tests that can be used to differentiate cases diagnosed by other means from controls, but their performance in other settings is often much poorer.
“So, the test is promising, but it really needs to be applied from some carefully collected and characterised samples in order to be able to judge its potential usefulness as a diagnostic test. As it stands it is just one more technological innovation that may or may not be useful in the clinical setting.”
‘Epigenetically reprogrammed methylation landscape drives the DNA self-assembly and serves as a universal cancer biomarker’ by Abu Ali Ibn Sina et al. was published in Nature Communications at 4pm UK time on Tuesday 4 December.
None to declare.