Esmo 2022 – Galleri’s real-world exhibition disappoints

Grail finds the path for its liquid biopsy less clear than it might have hoped.

Full data from a study assessing real-world use of a version of Grail’s – or is it Illumina’s? – multi-cancer screening test has revealed respectable predictive values. But not as respectable as they used to be: an interim data cut, reported last summer, looked better. 

An updated version of the liquid biopsy, Galleri, is already on sale in the US as a lab-developed test, and the point of the Pathfinder trial is to convince doctors and payers of the test’s utility as a screen for a variety of cancers in seemingly healthy patients. The data might also form part of a regulatory submission. But the diminishing accuracy is a disappointment, though arguably Illumina’s investors have bigger problems to worry about.

The 6,600 participant-strong US-based Pathfinder study is the first time Galleri has been used in clinical practice to help guide diagnostic workups. The version of Galleri used in the study is an older one; the marketed version of the test, which sells for just under $1,000, has been tweaked to improve its performance. 

Galleri was used to detect cell-free tumour DNA in blood samples from the Pathfinder subjects, all of whom were aged over 50 with or without additional cancer risk factors. The subjects’ cancer status was then confirmed at one year. 

According to the final cut of Pathfinder, to be presented at Esmo on Sunday, Galleri detected a cancer signal in 92 participants with analysable samples. Cancer was confirmed in 35 of these, giving a positive predictive value (PPV) of 38.0%. This is a drop from the 44.6% figure from the interim cut reported at last year’s Asco

Pathfinder data – accuracy
  Initial test version Refined test version
(prespecified analysis)
True positives  35 25
True negatives  6,235 6,216
False positives  57 33
False negatives  86 95
Patients whose cancer origin was correctly identified 34 22
Sensitivity* 28.9% 20.8%
Specificity 99.1% 99.5%
Positive predictive value 38.0% 43.1%
Negative predictive value 98.6% 98.5%
Cancer origin prediction accuracy 97.1% 88.0%
*Vantage calculation. Source: Esmo abstract #903O. 

In a prespecified analysis, a refined version of the test was used to test participants’ banked specimens, without the results being given to doctors or used to guide further steps. The data here were slightly better, with the exception of when Galleri was used to detect the origin of the cancer. The abstract does not say which tumour types were identified – or misidentified.  

Accuracy figures were not the main thrust of the Pathfinder trial, however. Its primary endpoint actually concerns the number and types of subsequent tests required, and the time taken, to achieve a solid diagnosis in those in whom Galleri detected a cancer signal. Here, 24 of the true positives (73%) had a concrete diagnosis within three months. 

Pathfinder data – primary endpoint
  True positives False positives Total
N 35 57 92
Extent of diagnostic testing 33* 57 90
Proportion that had at least one imaging test 90.9% 93.0% 92.2%
Proportion that had at least one invasive procedure 81.8% 29.8% 48.9%
Median time to resolution (days) 57 162 79
*Excludes 2 true positives whose evaluation started before test results. Source: Esmo abstract #903O. 

When the interim cut of Pathfinder was released, Stifel analysts said an oncologist they had spoken to believed that the assay’s performance was in line with expectations. To what degree the fall in PPV might disappoint cancer specialists is not yet clear, though a disappointment it will surely be. 

And getting doctors on side is only part of the problem. Galleri has no reimbursement coverage; this almost never occurs ahead of FDA approval. When Grail filed its IPO in 2022, it said it intended to seek approval of the liquid biopsy in 2023. 

A lot has happened since then. Someone will have to grapple with the problems of regulation, reimbursement and uptake, but with Illumina’s purchase of Grail very much on the rocks it is uncertain who might will be.

Share This Article