While there are big problems with this at present, & with "peace of mind" ads like @ezrainc - which is why we're seeing such strong reactions - I *do* think the future of cancer diagnosis @EricTopol could look something like this. (Bear with me.) At least 5 things are needed: /1
#1 First, the imaging devices would need to be so inexpensive that they could be in homes/non-medical offices, so people could be scanned frequently enough that biologically aggressive lesions are picked up. Yearly scans are too infrequent and will miss many interval lesions. /2
For instance, one of my patients with a cancer predisposition syndrome recently had a surveillance scan that was stone-cold normal; just 2 months after that had widely metastatic disease. In order to be that ubiquitous, the devices would need to be very inexpensive./3
If all an imaging technology can ever find is indolent diseases (eg localized low-grade prostate cancer) that will never cause trouble or that could be detected by something simpler like a periodic physical exam, benefit from imaging will be difficult or impossible to show./4
Need #2 Frequent scans of the general population would of course mandate very high quality automated interpretation – there could never be enough radiologists in the world to read all the resulting scans. This seems the easiest hurdle to surmount given AI developments./5
Need #3 Improved understanding of imaging signal characteristics such that benign or indolent lesions are definitively recognized for what they are >99% of the time, sparing patients unnecessary invasive procedures and accompanying worry. This work is ongoing but incomplete. /6
Need #3 is the greatest worry of physicians, because we've all seen patients harmed by biopsies (one of my own pts just suffered a pneumothorax from biopsy of a nodule that turned out to be benign). I think that experience drives most of the comments on @awilkinson thread./7
We know from prior studies that those benign/indolent lesions are incredibly common - in fact the majority of individuals have them. Many are easily recognized for what they are but some are genuinely difficult to distinguish at present. /8
Need #4: Imaging would still need to be the best way to detect cancer! Perhaps imaging ultimately will turn out to be inferior to some other technique (eg if there were a cell free/circulating tumor DNA assay with much improved sensitivity compared to currently available tests)/9
If that were the case, then a frequent blood test looking for a signature of emerging malignancy would be better than serial imaging. But already we know that clonal hematopoiesis (ubiquitous) confounds any simple link between mutation detection + cancer. /10
Need #5 The fifth necessity for this to be an effective approach is that the status quo would need to persist such that treatment of an “early” or localized lesion would remain more effective than treatment of a later-detected lesion./11
We can imagine therapies so effective that they routinely cure metastatic disease (some already are in development or approved, eg CAR-T cells or immunotherapies). If treatment were so effective, then it wouldn't matter how early cancer is detected./12
But for now, most cancers remain easier to treat early on. While that is less true for hematological malignancies, even imatinib works much better in chronic-phase early CML compared to blast crisis when additional mutations and clonal complexity have emerged./13
We all know the phenotype of overly hyped tech and silly or fraudulent startups, so caution is important. But we also don’t want to toss out emerging technologies that may someday provide real benefit just because they look similar to other techniques that have failed./14
Finally, as someone who personally spent a dark week at age 34 awaiting a biopsy and histology result after being told he had likely metastatic incurable malignancy based on what turned out (thank God!) to be a false-positive image, this all hits close to home./15End
