to ymav4294: ct scan frequency and pfs results, page-2

i can try and give you my thoughts, but you must remember i have little knowledge in cancer, and methods used to detect the presence, or in this case re-appearance of cancer, so please take that into consideration. also, while i am aware of research statistics, and study designs, kaplan meier curves are not something i perform myself. i do have to interpret them in a lot of research i do, but i have never actually performed one myself...yet.


in this case, it really is a tricky thing to do, and you need to find a balance. You can have a CT scan, and show no progression, but progress the very next day, and this won't be picked up for another two months. increasing the CT scan frequency will increase the sensitivity of observing events. but, you need to take into account all the factors associated with CT scans. there is patient burden, radiation exposure, cost of performing the scan, and the precision to which you need to estimate progression. ultimately, the team in charge of the study need to decide what is most appropriate.

when it comes to ovarian cancer, you need to consider
what the usual rate of progression is for OSC, and how long people can go without progression. if patients can progress as early as 2 months, then there is no problem with this timeframe. but if almost everyone progresses by 4 months, then you want to be sampling much more frequently than 2 months. i don't know what the stats are for ovarian cancer, but i'm assuming the incidence of progression isn't large within the first two months, and judging from the phase 2 data, the fact that median survival is at 400 odd days, the 2 month time frame seems ok.

but, as i mentioned, we don't know where in those 2 months they progressed, so there is a loss of sensitivity there.

lets just assume 1 month is 30 days. if everyone in the OSC progressed on day 59, and everyone in the CVAC group actually progressed on day 61, this would show up as a difference of 2 months. but ultimately, you can't have everyone have a CT every day or every week. you have to have a trade off somewhere.

but not everyone progresses at the same time. so what you get is a survival curve. over time, this issue will matter less, as the incidence of progression will still be true at each measured time point, and what you hope to see, is that the incidence of progression is lower in treatment group, and that this incidence persists over time.

so what i mean here is:

at 0 months post treatment, the probablity of PFS is 1. that means, no one has progressed (which is a no brainer as you can't enter the study otherwise). now, from here on in, they are measuring the rate at which people progress.

at 2 months, you may have 2 in 100 people in the OSC group, and 2 in one 100 in the CVAC group progress. your probability of PFS at 2 months is 0.98 in both groups.

at 4 months, 4 people in the OSC group have progressed, and 3 people in the CVAC group. now, the probability in the OSC group at 4 months in 0.96 (4 in 100), while in the CVAC group it is 0.97 (3 in 100). you want this difference to continue to separate over time, and not rejoin. so for example, by 16 months, the probability in the OSC group might be 0.60 (40 in 100 have progressed), while in the CVAC group, it is 0.75 (25 in 100 have progressed). what you don't want to happen is for the curve to join back up again. so by 18 months, OSC is 0.58, and CVAC is 0.6. that would suggest that the treatment has lost its effect by 18 months (but was quite effective up until that point).

so in short, the CT scan data might effect the specificity to some degree, and will definitely have some impact on where exactly the median PFS would have been, but it is unavoidable. in the phase 2b trial, it shouldn't affect the survival curve too much. Also, i think they look at other definitions for progression, such as the CA125 serum marker. are they sampled at the same time point as the CT scan, or separate?

i hope i haven't confused you, or provided you with the wrong information. if you want a better answer, you would need to speak with someone more involved in it than me.