Lauren Corona

For those of you who can't fathom how there could possibly be a downside to buying your own full body MRI…. Rough math, (my own estimates, my arguments, AI tightened the post for me), per 10M unselected scans: 3.5M false alarms 300K invasive procedures 15K major complications 100+ procedural deaths Lives saved? No randomized trial has ever shown one. Some almost certainly are saved. Nobody can tell you how many, and whatever the number is, it gets diluted across 10M people who were never going to have the disease. Every one of those harms can happen even with perfect execution by the doctors involved. The renal biopsy goes exactly right and the patient bleeds, because you put a needle through a vascular organ. The prostate biopsy goes exactly right and the patient dies of septic shock at home, because the needle crossed the rectal wall. We have spent twenty years driving those rates down with transperineal approaches, targeted prophylaxis, MRI-first pathways. The rates are lower. They have a floor though! Multiply a floor by 300K procedures and you get innocent people dying. The false alarms are just arithmetic. Screen a population where almost nobody has the disease and most positives will be false regardless. A better scanner finds more incidentalomas. And the overdiagnosis is just biology. The mass is seen on MRI, the radiology read is right, the surgery is flawless, the pathology confirms cancer, and the man still loses a kidney that was going to sit there quietly until he died at 89 of something else. He calls himself a survivor. He tells his friends. They book scans too Aim the same MRI magnet at high pre-test probability (people at high risk for cancer) and all three numbers get waaaay better. More lives saved but with much much less harm. Bayes doesn't care how good your scanner is. In this hypothetical, would you accept 100 procedural deaths in healthy people for an unknown number of lives saved?