Check the scan dates before changing treatment
My August CT was read as progression and the plan was to add chemo. But the comparison scan was from before I started my new pill. With my oncologist's OK, I waited a month. The next scan showed no progression.
On August 18, two oncologists looked at my CT: my primary oncologist and a specialist at one of the top cancer centers in the country.
Same verdict. "Looks like progression. Your oral pill isn't working. We recommend adding chemo."
I'd been on that new pill for less than a month.
What bugged me
The scan they compared it to was from June 15, back when my old pill was already failing. I didn't start the new pill until July 20.
So how much of that growth happened on the old drug, before the new one was even in my body?
Here's the timeline as I saw it:
- June 15: comparison scan. My old pill was already failing.
- July 20: I started the new pill.
- August 18: CT read as progression, compared with June 15.
- September 22: the next CT.
How I used AI to check the timing
I gave AI my scans, the tumor measurements, and the start and stop dates for every medication. I had it estimate growth rates and back-calculate doubling times. Then I had it run thousands of simulations across the measurement error to see when that growth most likely happened, and update the odds that the new drug was working.
The answer: there was a high likelihood the growth happened in the roughly 5 weeks my old pill was failing, not after I started the new one.
That isn't proof. It was a better question to bring to my doctor.
The decision we made together
I took it to my oncologist and explained my reasoning. I asked to wait another month and scan again.
For me, the risk was worth it. If I was wrong, I didn't expect a month to change much. And after this pill stops working, there aren't many treatment options left for me besides chemo. My oncologist was comfortable with a one-month wait and a repeat scan.
On September 22, I had the CT. No progression. The new pill was working.
My care team is incredible. But they get 15 minutes with me. I get every waking hour to look into my own health. I show up prepared, we talk through probabilities and outcomes, and we decide together.
Questions to ask about any scan
If a scan comes back worse soon after a treatment change, these are the questions I'd ask:
- Which scan is this being compared to, and what's its date?
- When did my current treatment start? Was the comparison scan before or after?
- Could some of the change have happened before the new treatment had time to work?
- What would a short wait and a repeat scan risk for me? What would it gain?
- When would we scan again, and what result would change the plan?
- Could anything else explain what we're seeing?
Write the dates down before the appointment: treatment start and stop dates, scan dates, and what each scan was compared to. It's a small step, and it changed my plan.
How AI fits
AI didn't replace my doctors. It helped me walk in with a better question. If you try it, give it your real dates and measurements, and give it rules so you get odds and uncertainty instead of reassurance. Here are the 9 rules I use, and how to ask AI about your data instead of your symptoms.
The real next step is pairing physicians with AI, so these probabilities are already on the screen when they walk into the appointment.
Keep going
I'm not a doctor and this isn't medical advice. It's what happened to me. Please talk to your care team before you change anything.
What would you add?
Help make this page more useful.