As many of us do, I spend time looking at various phase 3 trials on prostate cancer. As I’m also sure most do, I focus on the median life expectancy for those on a particular treatment vs those on a placebo. “Median Survival l” as I understand is the date that half the patients are alive and the other half are dead.
As I began to think about it, I realized that median life expectancy does not tell the whole picture. Take for example the IMPACT phase 3 study on Provenge. The trial showed an increase in life expectancy of only 4.1 months. One might look at that and determine that the cost in trouble don’t justify the use of such an expensive drug. But let’s look at it a different way. Assume that people on the placebo lived a median of 30 months And the people on Provenge lived a median of 34 months. You might conclude that the 4 months weren’t much of a difference. But now let’s look at it a different way. Assume that of the 500 people on the placebo arm, 250 died between by month 30 but 125 of those were dead by month 10. And the other 250 on the placebo arm were all dead by month 36. 30 months would be the “median survival” of the placebo arm being the time half were dead and half were alive.
Now assume that of the 500 men on Provenge, 250 died between months 31 and month 34. but (and here’s the kicker) of the other 250, 125 men died between month 36 and month 44 but the other 125 all lived past Month 72 and were considered in long term remission or even cured. The median survival for the Provenge arm would be at 38 months, the date half those on that arm were dead.
So a prospective user of Provenge who passed on that treatment because they thought it only extended life 4 months, probably didn’t realize they’d have a 25% chance of extending their life 6 years or more. And they also didn’t realize that that without Provenge, they had a 25% chance of being dead in less than a year while no one in the Provenge arm died sooner than 3 years
I realize these numbers are somewhat absurd but I use them to illustrate just how misleading @median survival” can be . It doesn’t seem to tell the whole picture. 25% of these men on Provenge lived more than 72 months in my example and 25% of the men on a placebo died in less than a year and none of that is learned by looking at just the median survival. I would guess these outlier survival and deaths scenarios could cut both ways and may exist in many treatments
For those here that are smarter than I (and there are many), please tell me if my thinking is skewed and if not, why do these these studies not show something like a colored plot mapping (with different colors for placebo and Provenge ) of all those treated on both arms to give a more complete picture of the results? Or is that info readily available and if so why do we never hear it quoted ? Even a average life expectancy would be helpful as additional info.
Schwah
Written by
Schwah
To view profiles and participate in discussions please or .
That's why it might be time for you to learn to interpret Kaplan-Meier diagrams. They show what percentage of the patients are still alive every few months for both the treatment and the placebo groups. Basically, if the treatment curve is above the placebo curve, that's a good thing since it means more people are alive. A lot of people talk about the gap between the two curves, which tells you how much better one is than the other, and how the gap changes over time.
"Assume that of the 500 people on the placebo arm, 250 died by month 30... and the other 250 on the placebo arm were all dead by month 36. Now assume that of the 500 men on Provenge, 125 all lived past Month 72."
I don't understand the basis for assuming anything like that could or would occur... every single man on placebo who managed to survive to 30 months somehow dies in the next six months? A full 50% of the placebo arm dies within that particular time window? I don't think that ever really happens.
It would seem to make more sense to just look at the published survival curves. I completely agree with you that median survival tells an incomplete story. These graphs show a better one... namely that men in both arms are dropping like flies in the first few years, and that some in both arms survive much longer, but by that time there are so few survivors in either arm that the study has become under-powered and it is hard to reach conclusions about the few surviving men in either arm.
If in fact you could show me a curve where every single man in one arm is actually dead by month 36 and 25% of men in the other arm are still alive after month 72, trust me, I am ALL IN on whatever the long-surviving arm is doing!
But looking at curves like the one in the IMPACT study (linked below) shows that not to be the case and is exactly why I do NOT focus on the median life expectancy for those on a particular treatment vs those on a placebo. I can see that there are earlier-diers and later-diers in BOTH arms, and I have no idea which I would be or why... so the graphic illustration in these charts of men dropping like flies at "roughly" the same rate in both arms, month after month, makes me not care too much about trying to figure out if I will one of the lucky who benefit or one of the unlucky who don't.
If the costs and side effects of a particular treatment are reasonable, and/or it relieves symptoms, I'm sure I will give it a try whether it shows a "median survival advantage" of a measly two months or of a full two years.
Figure 2. Kaplan–Meier Estimates of Overall Survival. [scroll down to the "Results" section of the paper to find fig.2, click on it to enlarge]
Yes I did not know that these survival charts even existed. That’s why I was confused. Filling in the blanks in my (and others) ignorance is a huge advantage of this site. Trust me when I say there are far more like me (pretty ignorant) than like you guys who usually pipe in and that’s awesome. Thank you and all those that take the time to share their knowledge. That’s what makes this site so incredible valuable.
Well, I kind of discovered those graphs by accident myself, and they make more intuitive sense to me (as a reflection of "what's really going on" with the men being studied) than the lingo of the statisticians, which I confess to barely understanding.
Excellent explanation by noahware. One simplistic view is those who survive initial 2 or 3 years are likely to survive longer because most aggressive variants kill men in first 2 or 3 years, In brief . .Longer one survives. .longer one survives !
When we lump all PCa variants together ..we do not get the right picture...Only subset wise curves can provide accurate estimates.
What it shows is that the probability of survival follows the same general pattern with or without Provenge. You can clearly see what percent survived at all time periods. Survival dropped off more quickly without Provenge. The difference between the two curves is, I think, what you are talking about. As you see, after about a year, the two curves are consistently about the same distance apart.
You might imagine that the gap seems to be a little wider at about 36 months. This is problematic because the sample sizes get very tiny the farther out you look (# of men still in study at all time periods is at the bottom). The authors looked at the deaths that occurred after data collection was frozen at 321 events and the study ended, adding a few months of additional f/u. This added 18 additional deaths observed between the data cutoff and study completion dates, with a median of 36.5 months of follow-up. However, the hazard ratio did not change appreciably.
Thank you TA. That now makes more sense. I know that they later looked at the data stratified by different starting PSA levels. And they apparently found that every lower level showed a greater survival benefit all the way down to the lowest level they looked at of <22 PSA which showed a survival benefit of 14 months vs the 4.1 months for the overall differential. I do not see that information on your link. Must have been done later I assume.
Glad you are thinking more deeply Schwah. If you had to pay for Provenge out of pocket it could be very difficult to decide. But the side effects other than treatment inconvenience is practically zero. T_A s 2nd reference is useful on real world benefits especially for those with lower PSA at the start (<5.3). Another visual tool is a “waterfall” plot, where each individual patient is shown when they go “over the falls” ( die or other outcome). Always makes me appreciate how much is put on the line by those in trials. We are in their debt.
Beyond that if you drill down into that study, those men who were treated with Provenge who had a PSA of <5 and were asymptomatic or minimally symptomatic had an increased survival benefit of 14 months! That’s pretty damn significant.
I actually read that the 14 month survival benefit over the control group (was actually 13 months) in the impact trial was for the tranche of men who were under 22 PSA. Not under 5. The group under 22 PSA was the lowest they looked at. I think the under 5 PSA (Actually under 5.27) you might be thinking about was referenced in the PROCEED Registry – Baseline PSA and Median Survival<5.27 ng/mL – 47.7 months
Interesting, I was on a drug trial with xtandi. After 4 years it stopped working. My research doc told me that 4 months was the median usefulness for patients. Xtandi stopped working after 4 months.
For a better understanding of the survival benefit, or lack thereof, of a treatment intervention, one would have to wait until everyone in the trial had died.
At the point where half have died, at least the weary team can publish something useful - the median - although it can't be used to predict the fate of the half still living.
& if we outlive those guys, we may get to read the final paper.
Yes, you've got the general gist of the problem with median survival figures. As others have pointed out, the survival curves give more information than a simple median figure. A good example IIRC is Atezolizumab. It provides a very small advantage in median survival, but if you go past the median you reach a point where the curves diverge greatly. Basically it does nothing for about 80% of people (myself included, sadly), but for the other 20% it can provide years of additional survival.
Here's another hypotetical case I like to ponder: What if a new drug was very toxic and killed 50% of patients in three months, but cured the other 50%? Very likely it would be outperformed by the placebo group, and the drug would not be approved, but imagine how many cancer patients would be willing to take that kind of risk for a possible cure! It's not entirely hypothetical either, there are a small number of members here that are doing very well years after being part of clinical trials that ultimately were unsuccessful.
And finally, there always seems to be a lucky handful of subjects in the placebo group that outlive most of the people in the study, even if the drug is wildly successful. For example, the STAMPEDE trial showed that early use of Abiraterone is highly beneficial, though some subject on Lupron only still hadn't progressed after 5 years. And of course, I seem to be the odd one that took Abiraterone, got to undetectable PSA, then developed neuroendocrine cancer as a result. Apparently my cancer didn't read the study.
Actually, not that sad. While I've had a couple treatments that didn't work, my overall success rate is far higher than the statistics would suggest. Also, given the number of neuroendocrine patients that have died since my NEPCa diagnosis I'd guess I've already beaten the survival median, which rumor has it is about one year.
Currently Folfiri, commonly used for colon cancer but a somewhat decent second line for any kind of neuroendcrine tumors. My scan a week ago showed impressive shrinkage below the diaphram, but above it in the lungs there's some new lesions. However, I just had a Covid shot which is known to cause swelling of lymph nodes so going to rescan in about 6 weeks (maybe 8 to get further away from the second Covid shot) and see if it's still a thing.
At NECPa diagnosis I had 6 cycles of Carboplatin+Etoposide+Atezolizumab (standard treatment for small cell lung cancer, which is a form of neuroendcrine tumor), then switched to Atezolizumab only for maintenance. I had good initial results which didn't last. Then cabazitaxel which knocked my slowly rising PSA back down but did nothing for the NEPCa.
Also had docetaxel at original prostate cancer diagnosis. Not sure how many patients live long enough to have been on four different types of chemo, now totaling 25 infusions. My fate seems to be to try to take as much chemo as possible, but I still have a ways to go as some people have had well over 100 cycles!
QOL is tolerable. I don't have the aerobic capacity for sustained running, but am still able to drive myself to some appointments, grill burgers outside, walk in the woods, etc. Next week I'm planning to be in the garage rotating tires and doing an oil change on our F150. The first week of each cycle sorta sucks but the second week I'm able bodied enough that I can find ways to enjoy life. Now that I'm vaccinated I'm planning to drive to visit family in a few weeks (about a three hour drive one-way, I'll stay the night cause six hours of driving in a day means no energy for socializing).
IMHO QOL is a use it or lose is proposition, though I'm also seeing some evidence that if you really get out of shape it's possible to regain some of your fitness while still on chemo.
Content on HealthUnlocked does not replace the relationship between you and doctors or other healthcare professionals nor the advice you receive from them.
Never delay seeking advice or dialling emergency services because of something that you have read on HealthUnlocked.
Treatment naïve and clinical trials
Chemo or clinical trial?
Clinical Trial vs. Zytega
