Asco 2018 – Grail makes a start in lung cancer

The first data emerge from the circulating cancer genome atlas trial.

Targeted therapies in first-line lung cancer have stolen the show so far at this year’s Asco, but the disease can only be treated once it is diagnosed. Grail, the Illumina spinout with a claim to unicorn status, is working towards doing this using a blood test, with data on its liquid biopsy technology presented as a late-breaker this morning.

The data are the first to emerge from the circulating cancer genome atlas (CCGA) study, which at 15,000 patients is the smaller of Grail’s two vast ongoing clinical trials. But the cut is a sub-analysis of a sub-study of CCGA, and while the results are the best Grail could reasonably hope for at this stage in the test’s development, they do not show much more than proof of principle.

The subset of CCGA patients that made up the Asco data included 878 untreated patients newly diagnosed with one of 20 different types of cancer and 749 healthy people. The sub-analysis concerned the 127 patients who had lung cancer and 580 control patients with similar demographics and history.

The version of Grail’s test used for the Asco analysis was necessarily a prototype – the point of CCGA is to develop a final version of the test suitable for commercialisation. This version ran three different assays on each patient’s blood: cell-free DNA and white blood cell-targeted sequencing of 507 oncogenes for single nucleotide variants; cell-free DNA and white blood cell whole genome sequencing for copy number variation; and cell-free DNA whole genome bisulfite sequencing for methylation.

DNA from white blood cells must be sequenced because mutations therein can interfere with the clinically relevant tumour DNA mutations (Early liquid biopsy data hint at Grail’s travails, June 5, 2017).

Bleeding edge

Knowing that the key to developing a successful liquid biopsy was always going to be specificity, Grail designed this analysis in an unusual way. Rather than simply coming up with figures for sensitivity and specificity, the company predefined what would count as a positive signal for cancer, setting the bar at 98% specificity.

At that level, sensitivity for late-stage cancer was 87-89% for the three tests – impressive. But the point of liquid biopsies is to pick up cancer at the earlier stages, when it can be more easily treated, and here the results were poorer, with specificity no better than 51%.

Sensitivity for lung cancer at 98% specificity
Stages I-IIIA (n=63*) Stages IIIB-IV   (n=54*)
Whole-genome bisulfite assay 41% 89%
    95% confidence interval 29-54% 77-96%
Whole-genome assay 38% 87%
    95% confidence interval 26-51% 75-95%
Targeted assay 51% 89%
    95% confidence interval 38-64% 77-96%
*For this analysis, 117 participants had evaluable data from all three assays and had informative staging. Source: company press release.

Cutting the data by histological subtypes found roughly similar sensitivity figures for adenocarcinoma, squamous cell and small cell lung cancer.

Sensitivities for lung cancer subtypes (stages I-IV) at 98% specificity
Adenocarcinoma (n=58) Squamous cell (n=28) Small cell (n=13)
Whole-genome bisulfite assay 62% 61% 85%
    95% confidence interval 48-75% 41-79% 55-98%
Whole-genome assay 59% 61% 77%
    95% confidence interval 45-71% 41-79% 46-95%
Targeted assay 60% 79% 92%
    95% confidence interval 47-73% 59-92% 64->99%
Source: company press release.

This level of cancer detection is not likely to make doctors throw their stethoscopes in the air with joy. Better data in larger populations will be necessary for Grail’s test to be approved and adopted, and while the company is working towards exactly this the Asco data make it plain that detection rates will have to be improved before the test is ready for prime time.

Even once CCGA concludes – it has a primary completion data of this September according to – a further two rounds of testing will be necessary before Grail expects to have a marketable product on its hands. And in parallel, the company is also working on the 120,000-patient Strive study in breast cancer; this study is about halfway enrolled.

Grail’s first marketed product is likely to be a blood test for nasopharyngeal cancer, though calling this a liquid biopsy is pushing the definition of the term. The test does not detect tumour DNA, but rather DNA from the Epstein-Barr virus, which causes the condition. Moreover it does not use Grail’s in-house technology, instead coming via the company’s takeout of Cirina a year ago (Viral success for Grail but not with core tech, August 10, 2017).

At least Grail has plenty of cash to further the development of its in-house tests: two weeks ago it reeled in a $300m series C round, meaning it has now raised $1.6bn in total. What valuation this confers on the company is not clear, but rumours of a $500m IPO in Hong Kong have swirled for some time.

The potential for a liquid biopsy for early-stage cancer is mind-boggling, but Grail still has a great deal to prove – and every extra million in venture cash increases the pressure to deliver.

Trial name Trial ID
CCGA NCT02889978
Strive NCT03085888

To contact the writer of this story email Elizabeth Cairns in London at or follow @LizVantage on Twitter

Share This Article