Could my psa variations be a result of lab errors ?

[diller]

See table at end of this posting for data being referred to here.

PSA has been going up for 2 1/2 years, starting a few years after inital IMRT

(no surgery was done)

Before lab of 12/30/2016 as shown below, the increases had been at a regular

rate, going up either 0.2 or 0.3 every 5 weeks when the lab was done. (am not concerned with psadt (psa doubling)here for this discussion)

And it always used to increase with each lab.

However, since 12/30/2016, some of the labs have either stayed the same

as pervious ones or even gone down a bit, and then the lab results

from a subsequent test after that shows the value of where it would have been

had it kept going up at the rate it had in the past.

And this pattern seems to be repeated - see chart below.

I've read about psa variance and find just a relatively few references online,

some relating to that psa can vary even within a day, but also a week or time of day

or season of the year, etc - according to some studies. Some articles seemed

to mention it can vary those ways by 20 to 30%.

Am also aware that there are 2 calibration standards used by different lab equipment manufaturers -

and as a result, the psa results stated can vary up to 20 % on the same blood sample -

, due only to those 2 different standards, and thats why I think its another reason

why people are recommended to use the same lab - assuming of course a given lab

uses just one manufacturer equipment/calibration standard for doing the psa testing,

Each lab I have done always says it was done using the Roche ECLIA

methodology -

and I was told by Roche that the calibration standard used by their machines

is the WHO standard, so am assuming that the same calibration standard has

been used for all my tests for many years.

The psa tests were not ultrasensitive ones.

I don't think these variances are psa bounces in context of post IMRT bounces -

since the IMRT was completed 5 1/2 years before these patterns were seen.

And since before the dates below, the psa kept going up once it started

back up. I do recall several bounces in the first 2 years after IMRT.

I've had no adt, no other rx or treatment that might affect psa, no new rx than the 2

have been taking for a long time, no change in rx dosage, no changes in diet

and before psa tests, as for many years, no ejaculation or exercise bike usage for several

days before the test. And I always go at the same time of day to the lab.

===> Thus am wondering, could the pattern I'm seeing of psa being the same or lower

on subsequent tests, before returning to where it would have been given

previous trends -- could that be due to some problems at the lab ?

Or is it really related to some natural and expected psa variations,

even weeks apart and even though before 12/2016 labs, there had never

been such variations, it had always just gone up each time.

Or perhaps this is a known pattern when one is recurrent and it has possibly spread outside of the prostate ?

Here is the charting for the last year (again, not concerned here with actual psadt)

07/07/2016 - 2.0

08/16/2016 - 2.2

09/29/2016 - 2.5

11/06/2016 - 2.8

12/30/2016 - 2.9 - why did it not go up more here as per previous patterns ?

02/02/2017 - 2.8 - why did it go down here from 12/30 and again, not go up as per previous patterns ?

02/23/2017 - 3.5 - back where it would have been given previous trends

03/09/2017 - 3.5 - why did it not go up here as per previous patterns/trends ?

03/23/2017 - 3.3 - why did it go down here from 03/09 test and again, not go up as per previous patterns ?

05/04/2017 - 4.0 - back where it would have been given previous trends

05/25/2017 - 4.5 - it is where it would have been or a bit higher

06/14/2017 - 4.4 - again, its not only lower but not going up as per previous trends

[Hidden]

I'm no doctor but it seems that monthly psa readings might fluctuate to the extent they have done so for you. I would worry less about bounces and focus on trend line. Never good when trend line increases so definitely would be more focused on that.

All the best

[AlanMeyer in reply to Hidden]

I agree with hankm, the trend line here is the important thing. You should research and plan for treatment. Questions to consider include: Who do you want for your medical oncologist. I don't think just any med onc will do. It should be one for whom prostate cancer is a major part of their practice and their continuing education. It should be one who is familiar with the results of the CHAARTED and STAMPEDE trials and aware of the trend toward combination therapies early in treatment.

As for the cause of the variation, I think lab error is less likely than natural variation. If there was a lab error where, for example, your blood sample was switched with or contaminated by someone else's sample, I don't think the variation would likely be as small as it is. If it is a calibration problem, I wouldn't expect it to happen so frequently.

I too have heard that PSA variation can vary by as much as 30% per day. The rad onc who told me that (this was 10+ years ago), said that no one knows what all the causes are. Your variations from the trend line are a lot less than 30%.

As you and hankm both seem to think, I too think the trend line is pretty consistent, so your work is cut out for you. Just as with primary treatment, it's important to get your next treatment as right as possible, starting with selection of a first rate medical oncologist.

Best of luck.

Alan

[eggraj8]

I am not clear about your concern about the numbers. Every measurement has some uncertainty associated with it. A single data point should not be interpreted as a point but as the center of a circle with the uncertainty of the measurement. If you knew the measurement uncertainty of each of these points then a graph of the results would show the trend very clearly. Since you do not know the uncertainty of the points then the next best thing to do is to plot each point you have on a psa vs. time graph and draw a least squares fit of the data. Calibration of equipment just tells you that the equipment should be accurate to the uncertainty in the precision associated with the equipment. You can ask the vendor about the precision of the readings. Since you are reporting to only one decimal point the equipment is not more precise than that. Add to this the uncertainty associated with the sample (daily variations in psa, etc.) and I suspect your total uncertainty maybe as high as 0.3 (probably less) for each reading. While this does not give one a warm feeling about the measurements, if you just draw a graph with circles around the points the radius of 0.3 then you would see better the trend line.

To me the data looks fine, with the trend graphically up to the right. By plotting the least fit you will get the doubling time (you need semilog paper to make this easy) which I think you have from other sources.

[PaulC2]

I see no significant variation here. As others have said, if you take each value not as a point but as the center of an interval plus-or-minus 15% wide, the linear trend of 0.2/month is regular. (As a side note, I'm slightly surprised at the linearity; I'd have expected 10%/month instead. But a straight line is the best-fit "curve" for the dataset.)

I applaud your interest in wanting to squeeze as much information out of the dataset as you can. For someone as interested in numbers and analysis as you are, I would highly recommend switching to an ultrasensitive test, if you can have it performed using the same assay each time.

Why? Well, you don't mention your life expectancy, but if you are likely to still be here in five or ten years' time, chances are fair to good that you might undergo further treatment for your PCa -- most likely ADT. If so, you'll be looking at sequence of values that resemble either this --

<0.1; <0.1; <0.1; <0.1; <0.1; <0.1; <0.1; <0.1; <0.1; 0.1; 0.1; 0.1; ...

or this --

0.06; 0.05; 0.06; 0.04; 0.05; 0.07; 0.06; 0.08; 0.09; 0.09; 0.11; 0.12; 0.14; ...

In the latter case, the dataset gives you more advance warning about the timing and extent of an impending biochemical failure.

[diller]

hankm, thanks for your comments; yes am looking at treatment situation, getting scans, etc. I realize with psa where it is now I've waited too long perhaps and much of that was my own fear and just ignoring it all, etc - I know that does not make sense but it happened and am trying to regroup.

AlanMeyer - agree with you about need to find a good and experienced dr - not easy in practice for my situation. instead of mentioning that here I'd added to your post on that overall topic - healthunlocked.com/advanced...

And to add to that post, that I forgot to put into it - I have talked with several specialist drs and was dismayed about some things I heard and some of their policies and approaches. anyway, I mention more in your post.

eggraj8 and PaulC2 - well, my head is spinning - math and statistics has not been my strong suit at all. Thanks for explaining these variances in the scientific terms that I can still understand ! - it really helps to know this, and helps me realize it was not just a lab issue. I think also that because these trends did not happen when psa was going from 0.2 up to 2.8, that that was confusing to me when these trends did start happening (from 0.2 to 2.8 it was just an increase of 0.2 or 0.3 every 5 weeks; it never went down, always up.

Paul, as to switching to ultrasensitive, its a good point - in the past I felt no need for it since had had IMRT only and the recurrence values higher than for surgery - but with ADT, and hoping indeed the psa does go down as low as your example, that the monitoring of psa rises is different than in the past re post IMRT, and thus as you mention the ultra sensitive can show trends earlier.

As to the variances, as you both explain - seems to me a person would need the psa tests more often than like every 6 months, if they are on adt and wanting to see how its going - reason being that any given test might have the potential for variance you discuss, in addition to whatever daily/weekly/time of day variations that can happen -- thus with tests not frequent enought, the results might not show the real trends of either going up or down or the same -- what the proper frequency of testing to avoid this I don't know. I

Finally,I apologize for the bad formatting of my initial post above - 2 reasons i think -

one was that i copied/pasted from an old command line kind of editor - even though in the compose window it looked ok as to paragraph separation; and two, that I did not know that in the compose window here, that if one hits return, it creates a new paragraph, with a line space in between vs just creating a new line in the same paragraph, And since there is no preview post feature, I could not see the issues until it was too late.

[Hidden]

I do not think that anyone can answer your question. Plus it seems a relevant one since clinical trials often require "progression", and if a value falls, you may be asked to "start over" regardless of "trend". It is an unsatisfactory situation to my mind. Arbitrary.