Lung cancer is one of malignancies liquid biopsy developers are currently most closely focused on, with at least seven companies applying blood testing tech to the disease. Amid the herd Veracyte is investigating a very different type of technology, NasaRisk, which detects genomic changes to epithelial cells in the nose.
“From the nasal airway, can we get similar or better classification of patients pre-work-up for lung cancer?” asks Veracyte’s chief executive, Bonnie Anderson. The answer appears at first glance to be yes – but a close read of the results shows that the way the company has presented them is not the full story.
Data unveiled at the Chest meeting last month, from a trial in which the nasal test was used to stratify 261 current or former smokers with suspicious lung nodules, showed remarkable accuracy, with figures for sensitivity and specificity greater than those claimed for liquid and tissue biopsies.
Among patients whose nodules were subsequently determined to be benign, NasaRisk classified over 40% as low risk for cancer with a sensitivity of over 95%. In the patients whose nodules were malignant the test classified over 40% as high-risk for cancer, with a specificity of over 94%.
But this is statistical sophistry, because the company has applied two different cutoffs to the same dataset, rather than a more accepted single cutoff. For the low-risk group the researchers pegged sensitivity at 96.6%, with specificity of only 45.6%. For the high-risk group it was the specificity that was prespecified, at 94.7%, at which point sensitivity shrank to 50%.
There is no reason all the other test developers could not take the same approach and make very similar claims. In any case, for all noninvasive cancer tests so far, a high prespecified sensitivity has come with a trade-off in the shape of middling specificity, and vice versa.
The diagnostics specialists Vantage spoke to about the way Veracyte has presented its data urged caution. One said that it was disappointing that Veracyte had not documented exactly how the specificity and sensitivity calculations were defined, and said this would likely be challenged if presented in a peer-review publication.
Another expert had no doubt that the test has some value, and might push the diagnostic paradigm forward, but pointed out that the use of two cutoffs seemed a “statistical trick, and was based on a post-final diagnosis bias analysis. Why not simply comment on the sensitivity and specificity that they have now, and try to ameliorate the current test algorithm?”
Fixing either the sensitivity or specificity is not uncommon. Grail, for instance, reported data on its liquid biopsy in which specificity was pegged at 99%, and gave the corresponding sensitivities for various tumour types – in lung cancer the figure was 59% (Grail reports liquid biopsy data as Guardant heads to market, June 3, 2019).
|Selected noninvasive lung cancer tests|
|Guardant Health||Guardant360||Blood test - ctDNA||Nile: 282 patients with metastatic NSCLC||As good as tissue biopsy|
|Grail||Unnamed||Blood test - ctDNA||CCGA: 15,000 patients with various cancers including lung||Sensitivity of 59% at a pre-set specificity of 99%|
|Thermo Fisher Scientific||Oncomine Lung||Blood test - ctDNA||94 NSCLC patients||88% concordance between the tissue and plasma samples|
|Foundation Medicine (Roche)||FoundationOne Liquid||Blood test - ctDNA||BFast: 2,219 patients treated with Roche’s Alecensa||ORR with Alecensa of 87% – as good as with tissue biopsy|
|Thrive Earlier Detection||CancerSeek||Blood test - ctDNA and protein biomarkers||1,005 patients with various cancers including lung||Approx 59% sensitivity; 99% specificity|
|Laboratory for Advanced Medicine||IvyGene Core||Blood test - methylation of ctDNA||No details given||Approx sensitivity 84%; approx specificity 90%*|
|Oncimmune||EarlyCDT-Lung||Blood test - immune gene expression||ECLS: 12,000-patient trial in the screening setting||33% sensitivity; 90% specificity|
|Owlstone Medical||ReCiva||Breath test||Lucid: 520 patients with suspected lung cancer||Data possible early next year|
|Veracyte||Unnamed||Nasal swab||261 current or former smokers with suspicious lung nodules||Sensitivity and specificity vary depending on cutoff used|
|ctDNA=circulating tumour DNA. CTCs=circulating tumour cells. ORR=objective response rate. *Figures are averages for breast, colon, liver and lung cancers.|
Veracyte aims for the test to be used to triage patients rather than as a true diagnostic. Each year around 750,000 patients with suspicious nodules must undergo expensive, invasive further testing, 550,000 of whom are found to have benign nodules. NasaRisk could, Ms Anderson argues, classify around 200,000 of those as low-risk “with a very high sensitivity”.
It is also capable of classifying about 100,000 of the 200,000 malignant patients as high-risk “with over a 94% specificity” she says. Of course, this means that over half the 750,000 indeterminate patients will remain indeterminate. Ms Anderson says such performance is still good enough.
Accuracy questions aside, the test is based on genuinely innovative science. Smoking is known to cause genomic damage to epithelial cells in the lung, but also damages the nasal epithelium, allowing a much less invasive sampling procedure, in which a brush resembling a mascara wand is inserted into the nose. This approach has the tacit approval of one of the major medtechs: it is part of a deal Veracyte signed in January with Johnson & Johnson for the development and sale of novel diagnostic tests to detect early-stage lung cancer.
Veracyte intends to launch NasaRisk by early 2021 as a lab-developed test performed only at Veracyte’s Clia-certified labs. According to Leerink analysts, Veracyte’s management has indicated that the average price of the test in the US could be around $2,500.
When the test is marketed it will go up against liquid biopsies such as Guardant360, already sold in the US on a lab-developed basis. Guardant’s sales topped $60m in the third quarter, and it is on track for 2019 revenues of $205m, according to Leerink.
But profitability is not in Guardant’s immediate future, the analysts believe, whereas Ms Anderson says Veracyte will break even by the end of this year. The group will want NasaRisk to add to its revenues, but its uptake will depend in part on whether doctors are convinced by the Chest data.
|Veracyte's test revenues|
|Global sales ($m)|
|Afirma (thyroid cancer)||105||120||127||130||134||139||+6%|
|Percepta (lung cancer)||4||9||18||32||44||56||+67%|
|Envisia (idiopathic pulmonary fibrosis)||1||3||6||12||18||23||+91%|
|Total company revenues||120||134||153||175||198||220||+13%|