Revealing the true utility of ctDNA

The most recent ctDNA paper verifies its utility as a “broadly applicable, sensitive, and specific biomarker” across “multiple different types of cancer” and useful in cancer research and medicine: Detection of Circulating Tumor DNA in Early- and Late-Stage Human Malignancies
 by Bettegowda et al Sci Transl Med 2014.

Bettegowda, Sausen, Leary & Kinde et al have analysed the largest set of cancer patients to date for detectable levels of ctDNA. 640 cancers from 18 different tumour types were analysed for the presence of ctDNA, which was detected in 82% of tumours outside the brain. Some tumours did not show such high detection rates: less than 50% in medulloblastomas, kidney, prostate, or thyroid; and under 10% in glioma. The numbers of individual cancers was quite variable (155 pancreatic vs 10 or fewer each of liver, neuroblastoma, SCLC, NSCLC, ovary, prostate, kidney and thyroid) but other datasets are on the way that will add to the data we have available. In the paper they also demonstrate that localised disease is slightly less likely to show detectable ctDNA compared to metastatic disease, and that detectable levels of ctDNA correlate with stage; no surprise but this is likely to have an impact on the use of ctDNA in screening.

ctDNA vs CTCs: In a recent NEJM paper Sarah Jane-Dawson et al showed a good correlation between levels of ctDNA and CTCs. ctDNA levels had greater dynamic range, and correlated better with tumour burden than CTCs, and ctDNA gave the earliest indication of response in 50% of the 19 patients studied. In the Bettegowda et al STM paper they found no cases where CTCs were found without ctDNA but did find the opposite: ctDNA was present without detectable CTCs, giving us the first proof that ctDNA is not coming directly from CTCs. But not giving us the definitive answer to where it comes from or how it gets there. Importantly; in comparison to normal cancer biomarkers like CA-125 which are expressed in asymptomatic patients and not in all cancer patients, patient specific somatic alterations in ctDNA appear to be detectable in a very large number of cancer patients.

What about lung cancer: in the Bettegowda et al paper lung cancer was obvious by its absence relative to its abundance. As a disease that is responsible for 15% of cancers and 35,000 cancer deaths in the UK, understanding the utility of ctDNA in the lung has got to be a high priority. Lung cancer was also one of the first cancers to have its genome sequenced and this paper showed the carcinogenic signatures of tobacco smoke. In the figure below I took incidence and mortality stats from Cancer Research UK's website and have highlighted only the cancers included in the paper, alongside lung cancer and "other". This shows quite graphically that the paper covers about half of cancers; that the impact of ctDNA in lung could be significant, and that more work will be required to look at the "other" cancers. Common driver mutations in the big four: lung, bowel, breast and prostate, are relativley clear and can be inculded in a capture or amplicon assay. ICGC is going to help get better data on the drivers important in "other" cancers.

Screening potential: I’m very bullish about how much impact ctDNA analysis is going to have in the clinic (possibly because I am a couple of steps removed and don’t see some of the challenges). To me the paper is a clear demonstration of how useful ctDNA might be and shows high detection rates in diseases like Pancreatic cancer. This particular disease is a case where early detection can have a meaningful impact on patients. The authors of the STM paper suggest that "evaluation of specific mutations in the primary tumour add both time and expense to patient management", I disagree; but it depends on how you find those patient specific mutations in the first place. Approaches that target common cancer drivers can quickly find patient specific tumour biomarkers without the need to look in the primary tumour first (e.g. Forshew et al STM).

So hopefully screening is not beyond the reach of ctDNA-based assays. But the use of ctDNA in treatment monitoring is likely to have an earlier impact, the community still needs to do the longer-term trials to see how getting information on progression several months earlier than imaging might impact patient outcomes. But as a relatively cheap and non-invasive test it's likely to have a significant impact on the number of patients we can monitor.

For me one of the most importantinf findings in the paper was detectable levels of ctDNA in almost half of patients with localised disease, i.e. the patients with the best outcomes. I'm confident improvements in the sensitivity of ctDNA-based assays will make tests better than the ones we're using today. It's only been in the last few years that we've really started to wake up to the possibility ctDNA analysis has; and not everyone believes it will have a significant impact. Only time will tell so you can be sure this is something I'll be covering again.