Most people have heard of a differential diagnosis—the ranked list of possibilities a doctor holds in mind when figuring out what you have. It’s a useful framework because it forces you to hold multiple possibilities simultaneously, weigh them against evidence, and re-rank as new information arrives.
Two things before I get into specifics. If you know me, this post contains a lot of clinical detail about my situation, and I apologize if any of it catches you off guard. If you don’t know me, you can let the specifics of my particular case blur past—because the underlying idea, differential prognosis, is something you will need someday. Everyone does. The specifics are mine; the framework is yours.
I’ve been thinking about a related idea that, as far as I can tell, nobody has named: differential prognosis. Not “what do you have,” but “where is this going”—and critically, how that ranking should shift as new evidence arrives.
Here’s why it matters.
About 1 in 8 men will be diagnosed with prostate cancer in their lifetime. I’m one of them. Gate one.
When I was diagnosed at 52, that itself was unusual. Only about 2–3% of prostate cancer diagnoses happen at that age. Most men are in their 60s or 70s. Being in the left tail of that age distribution isn’t just a fun fact—it correlates with more aggressive biology. Gate two.
Post-surgery pathology came back with extracapsular extension, a positive margin, Gleason 3+4, Grade Group 2. None of those findings individually is rare, but the combination starts narrowing the population I actually belong to. Gate three.
Then my PSA never became undetectable after surgery—a distinction clinicians pay attention to, because it’s different from dropping to undetectable and later rising. It suggests something was there from the start. Gate four.
Then my PSA doubling time landed at roughly three months. Among men with biochemical recurrence (technically defined as PSA reaching 0.2—a threshold I hadn’t even crossed yet, but with a trajectory this clear, no one was waiting for a number), only around 10–15% have doubling times under six months. Three months puts me in the aggressive tail of an already-selected subgroup. Gate five.
The naïve way to interpret all of this—and I’ve encountered it from well-meaning people, and even from some clinicians—is to treat each of those gates as an independent coin flip. “Only X% of men with recurrence have a doubling time that fast, so the odds are still in your favor!” But that treats my trajectory as a series of fresh rolls of the dice. It isn’t.
What’s actually happening is that each gate is evidence of a latent variable—call it underlying disease aggressiveness, or biology, or whatever you want. That variable doesn’t change between gates. Each time I pass through a low-probability bad gate, I’m not just unlucky. I’m accumulating evidence that the hidden variable driving all of this is unfavorable. The gates are correlated through that hidden factor.
This is Bayesian updating, and it’s not complicated once you see it. Each gate is a filter. By the time you’ve passed through four or five of them on the unfavorable side, the population you actually belong to—statistically—is very different from the one your original diagnosis placed you in. Your prior for the next branch point being unfavorable should be substantially higher than the base rate would suggest.
That’s the differential prognosis. It’s a living, ranked list of possible futures that gets explicitly re-weighted at each gate. Not a static risk label you were handed at diagnosis and carry around forever.
And I think naming it matters, because the standard way prognosis gets communicated tends to be snapshot-based. You get staged, you get a risk stratification—CAPRA score, NCCN risk group, whatever—and that label tends to stick in people’s minds even as subsequent events should be revising it. “You’re intermediate risk” can persist psychologically long after the actual picture has shifted substantially toward unfavorable. Clinicians implicitly update their thinking, but patients often don’t. They anchor on the original framing.
The differential prognosis framework fights that anchoring by making the updating the point. You’re not revising a fixed label. You’re maintaining that ranked list and explicitly moving things up or down as you pass through each gate.
Where I think this has real power is in arguing against complacency at each new decision point. In my case, when it came time to decide on radiation and hormone therapy, the sequential history argued for treating aggressively—18 to 24 months of ADT rather than the minimum 6. My history is the prior, and the prior says: don’t bet on the favorable tail anymore.
There’s one nuance I want to be honest about. The framework is stronger at up-regulating concern than at down-regulating it. A negative PSMA-PET scan should move favorable trajectories back up the list. Good news is real and should count. For the framework to be complete—and not just a machine for generating anxiety—it has to account for that symmetry.
And there’s a ceiling effect. At some point you’ve already updated your prior substantially, and the next bad gate provides diminishing informational value. You’re already modeling aggressive disease; one more confirmatory data point doesn’t shift things as much as the first few did.
But I think most people facing a serious diagnosis are nowhere near that ceiling. They’re still anchored on their original risk label, still hearing base rates that no longer apply to them, still treating each new finding as a fresh surprise rather than as another piece of a pattern they should have seen forming.
The differential prognosis says: Pay attention to the pattern. Update the list. And make your decisions from where you actually are—not from where you started.
ɕ