As our research team have and hope to continue working with Grail I am particularly interested in this (I have no financial ties to grail btw). The sensitivity is poor for stage 1-2 cancer and thus probably means it will fail in the big RCT. On the other hand we think it could help diagnose symptomatic patients especially those with vague symptoms. There are of course plenty of other companies snapping at Grails feet now, including a test developed by iur colleagues in oncology.
Helping diagnose symptomatic patients makes sense, given how good the test is at identifying cancer signal origin, but you may still end up with a frustrating proportion of false negatives. I didn't get into this in my post, but 15.8% of the true positives were stage III, and another 24.8% were stage IV. Given that overall sensitivity was so low (40.4%), there's a hint here that the test is also missing more advanced-stage cancers.
This is only a "hint", because the researchers don't disaggregate sensitivity by stage.
Here's where things get a little touchy.
You could argue that the researchers don't report this disaggregation because the numbers would be too small for meaningful interpretation (there were only 329 cancer diagnoses across the entire sample).
However, other small numbers do get reported (e.g., true positives for cancers that don't have recommended screenings, where the number of cases range from 1 to 14). An alternative, more cynically-minded argument is that Grail is carefully curating how they present their data because, as you said, they've got competitors.
Curation per se is not inappropriate – scientists routinely spin their data as positively as they can without misrepresenting it – but I've been struck by how tightly Grail controls a somewhat underreported, strictly positive presentation of the new findings. The only data is available on a slide deck, and the exact same deck was used in the new conference report, on Grail's website, and in a report to investors.
I'm not saying I buy into the cynical view. I'm just saying it's not unreasonable.
• Prenuvo offers a full-body MRI screening service that can detect 500 conditions.
• Radiologists say they usually only recommend MRI scans for patients who have a family history of cancer and only on a case-by-case basis.
• Although MRI scans are useful in detecting abnormalities in the body, they can also create anxiety in patients and sometimes lead to unnecessary procedures.
I would also question if your PCP's Clinic getting remunerations to advertise this service. If there was a need to it, they would have arrange for the labs.
I also question your choice of PCP. ;~) I would if it was mine.
Yes, at the time of that 2024 BMJ paper the sensitivity of the test had been consistently awful. It's not great now (40.4%), nor is the PPV.
The full-body MRI scan is an excellent analogy – I regret not including that in the post.
As for my PCP, it's a moot point. I've also gotten texts from them that they're switching to concierge service next year. I don't plan to stay on. I think it's reasonable to assume financial motives behind their texts pushing Galleri.
Grail and its researchers do at least acknowledge that Galleri has false positives and negatives. On the website, in the conference report, in a report to investors, and in media releases this is noted more than once. In contrast, my PCP's texts just mention an "early detection test" that "looks for a signal shared by 50+ cancers", which is true but fails to mention the known inaccuracies. IMO this misleading, if not deceptive.
Correct! "Hopefully they show this much persistence with other aspects of patient care."
When was the last time you got an email from your doctor telling you that eating highly processed food, smoking, drinking to excess, being a couch potato was not in your best interest!
I'm switching PCPs because mine is transitioning to concierge service, but I would've switched anyway because they don't seem sufficiently attentive to patient care. During my last physical, they didn't ask whether my diet, activity levels, etc. had changed since what they have on record from about 16 months ago.
My lifestyle is reasonably healthy, so I'm not worried about the oversight, but they ought to be asking every patient about these things!
(Or, to put it systemically rather than blaming them, we ought to have a health care system in which shortages of PCPs, insurance issues, administrative chores etc. don't result in such heavy caseloads that PCPs can't get to know their patients.)
Ken, we will have to see what data they put in the main paper (I assume one is on the way). I have only seen a conference abstract so far, which as you say, possibly paints a more optimistic take on the study.
I do know for a fact that we published sensitivity by stage, and cancer site in our paper, on MCED's in symptomatic patients.
Aside, from stage I and II which is not good, you'll see that the sensitivity for prostate cancer is terrible (1/11) - but that might actually be a good thing!
The other nuance to this story is that some of false positives were later diagnosed with cancer on 6-12 month follow-up. These were people who were MCED positive, but in whom the NHS could not find evidence of cancer at the time.
Those are important nuances. The 6-12 month follow-up data is good news. So, in a sense, is the prostate cancer sensitivity, although I get a little nervous, on purely statistical grounds, about making too much of accuracy data when case rates in the sample are so small.
As our research team have and hope to continue working with Grail I am particularly interested in this (I have no financial ties to grail btw). The sensitivity is poor for stage 1-2 cancer and thus probably means it will fail in the big RCT. On the other hand we think it could help diagnose symptomatic patients especially those with vague symptoms. There are of course plenty of other companies snapping at Grails feet now, including a test developed by iur colleagues in oncology.
Helping diagnose symptomatic patients makes sense, given how good the test is at identifying cancer signal origin, but you may still end up with a frustrating proportion of false negatives. I didn't get into this in my post, but 15.8% of the true positives were stage III, and another 24.8% were stage IV. Given that overall sensitivity was so low (40.4%), there's a hint here that the test is also missing more advanced-stage cancers.
This is only a "hint", because the researchers don't disaggregate sensitivity by stage.
Here's where things get a little touchy.
You could argue that the researchers don't report this disaggregation because the numbers would be too small for meaningful interpretation (there were only 329 cancer diagnoses across the entire sample).
However, other small numbers do get reported (e.g., true positives for cancers that don't have recommended screenings, where the number of cases range from 1 to 14). An alternative, more cynically-minded argument is that Grail is carefully curating how they present their data because, as you said, they've got competitors.
Curation per se is not inappropriate – scientists routinely spin their data as positively as they can without misrepresenting it – but I've been struck by how tightly Grail controls a somewhat underreported, strictly positive presentation of the new findings. The only data is available on a slide deck, and the exact same deck was used in the new conference report, on Grail's website, and in a report to investors.
I'm not saying I buy into the cynical view. I'm just saying it's not unreasonable.
Interesting, this company was mentioned in my August '24 edition, https://jkd0811.substack.com/p/its-saturday-at-the-petri-dish
Galleri promises to detect multiple cancers—but new
evidence casts doubt on this much hyped blood test
https://www.bmj.com/content/386/bmj.q1706
Also, I am going to put this on the same level as those whole body MRI.
from https://www.verywellhealth.com/prenuvo-full-body-mri-scan-7693049
• Prenuvo offers a full-body MRI screening service that can detect 500 conditions.
• Radiologists say they usually only recommend MRI scans for patients who have a family history of cancer and only on a case-by-case basis.
• Although MRI scans are useful in detecting abnormalities in the body, they can also create anxiety in patients and sometimes lead to unnecessary procedures.
I would also question if your PCP's Clinic getting remunerations to advertise this service. If there was a need to it, they would have arrange for the labs.
I also question your choice of PCP. ;~) I would if it was mine.
Yes, at the time of that 2024 BMJ paper the sensitivity of the test had been consistently awful. It's not great now (40.4%), nor is the PPV.
The full-body MRI scan is an excellent analogy – I regret not including that in the post.
As for my PCP, it's a moot point. I've also gotten texts from them that they're switching to concierge service next year. I don't plan to stay on. I think it's reasonable to assume financial motives behind their texts pushing Galleri.
Grail and its researchers do at least acknowledge that Galleri has false positives and negatives. On the website, in the conference report, in a report to investors, and in media releases this is noted more than once. In contrast, my PCP's texts just mention an "early detection test" that "looks for a signal shared by 50+ cancers", which is true but fails to mention the known inaccuracies. IMO this misleading, if not deceptive.
Correct! "Hopefully they show this much persistence with other aspects of patient care."
When was the last time you got an email from your doctor telling you that eating highly processed food, smoking, drinking to excess, being a couch potato was not in your best interest!
Lol, right.
I'm switching PCPs because mine is transitioning to concierge service, but I would've switched anyway because they don't seem sufficiently attentive to patient care. During my last physical, they didn't ask whether my diet, activity levels, etc. had changed since what they have on record from about 16 months ago.
My lifestyle is reasonably healthy, so I'm not worried about the oversight, but they ought to be asking every patient about these things!
(Or, to put it systemically rather than blaming them, we ought to have a health care system in which shortages of PCPs, insurance issues, administrative chores etc. don't result in such heavy caseloads that PCPs can't get to know their patients.)
Ken, we will have to see what data they put in the main paper (I assume one is on the way). I have only seen a conference abstract so far, which as you say, possibly paints a more optimistic take on the study.
I do know for a fact that we published sensitivity by stage, and cancer site in our paper, on MCED's in symptomatic patients.
https://pubmed.ncbi.nlm.nih.gov/37352875/ (Table 3)
Aside, from stage I and II which is not good, you'll see that the sensitivity for prostate cancer is terrible (1/11) - but that might actually be a good thing!
The other nuance to this story is that some of false positives were later diagnosed with cancer on 6-12 month follow-up. These were people who were MCED positive, but in whom the NHS could not find evidence of cancer at the time.
Those are important nuances. The 6-12 month follow-up data is good news. So, in a sense, is the prostate cancer sensitivity, although I get a little nervous, on purely statistical grounds, about making too much of accuracy data when case rates in the sample are so small.