Precision of the PSA test

I have already addressed this subject to a couple of fellow survivors who are worried regarding their PSA results. I will not reiterate here.

Instead, I will quantify my assertion that a slight increase in PSA is definitely in need for more data and better number crunching, while advices of the sort: "Consult a RO ASAP", _IMO_, are worthless.

So, let's start from the manufacturer's specs of this ELISA (human) PSA kit.

system.na3.netsuite.com/cor...

From par. 14.1 TABLE 3 we read the number, mean value, standard deviation and coefficient of variation for each of three control sera.

We elect TABLE 3 (Between Assay Precision) instead of TABLE 2 (Within Assay Precision) as if two tests are spaced by a couple of months, it is reasonable to assume that they were not derived from the same kit.

We also notice that 3 PSA values are specified, namely: 23.17, 3.35 and 0.98 ng/ml and that as the PSA value declines the coefficient of variation advances.

en.wikipedia.org/wiki/Coeff...

Consequently, when we are discussing values of the sort 0.0x things are less accurate for sure but we are not told by how much.

Let us now do some practical interpretations of all these figures.

Assuming normal statistical distribution this C.I. online calculator will be of help:

mathsisfun.com/data/confide...

For N=20, mean value = 1 and σ = 0.083, the 95% Confidence Interval is 1 ± 0.0364

(0.964 to 1.04). In layman's parlance this means that if you get a report of PSA=1 this can actually be somewhere from 0.964 to 1.04, 95 times out of 100.

Not bad If you asked me. BUT this is valid only if you took 20 TESTS!

Now lets try to see what are the numbers for a single test, i.e. N=1, everything else kept the same.

95% Confidence Interval: 1 ± 0.163 i.e. 0.837 to 1.16.

The +/- 3.6% initial ambiguity now has risen to 16.3%.

Take a second test (N=2) and it will drop significantly:

95% Confidence Interval: 1 ± 0.115 i.e. 0.885 to 1.11.

So, what someone can do to narrow down the ambiguity window?

What every data collecting discipline does, from GPS surveying to Forex exchange trading:

MORE DATA AGGREGATION

This is what I do and in case you asked me the advice regarding same lab is a PHD in naivety. More data anyway sourced contain much more information than less date sourced from a single source.

But, I am just a humble engineer, so don't listen to me.

I didn't charge you 400 Euros for BSing. Did I?

[Tall_Allen]

You have to address the situation. PSA is a lousy test for initially detecting prostate cancer, for many of the reasons you identify. It is only the pattern of PSA that is useful. However, it is an excellect test for detecting recurrent prostate cancer.

[GranPaSmurf]

Ahh Ha!

One cannot make diagnostic or treatment decisions based on infinitesimal variation in PSA!

I saved this post for reference later.

[Hidden in reply to GranPaSmurf]

This is the prime aim of ALL communication systems.

Get the (useful) signal out of the noise.

[dadzone43]

Live with ambiguity, maybe?

[j-o-h-n]

A newly wed woman says to her friend "I gotta be careful not to get pregnant"

Friend: Didn't your husband get a vasectomy?

Woman: Precisely.

Good Luck, Good Health and Good Humor.

j-o-h-n Tuesday 11/05/2019 9:07 PM EST