Page 1 of 1

help about potency andsignificant digits

Posted: Fri Nov 09, 2018 2:47 pm
by jcwagner
Hi,
i need your help about two subjects.

1. The certificate of analysis of a working std we purchased states that potency of dried substance=100,9%.
How can potency be above 100%.
I tell my manager to use 100% when calculating our samples with this standard but she denies.
What is the right thing to do?

2. A few days ago, she told us that we have to use all 10 digits regarding the areas of peaks that our LC presents , for our assay calculations.
Until now we were using 6. (ex: 14537.5 from now on 14537.58478).
This is a lot more work for the analyst and the reviewer.
I know this is more "correct" but isn't a bigger mistake statistically that we weight with 3 digits ( 50.5mg) and we report assays with 4?(100.4%)
I dont think that with 10 digits our result will be more accurate regarding the above.
I need your opinion.

Thanks

Re: help about potency andsignificant digits

Posted: Fri Nov 09, 2018 6:20 pm
by DR
You should find that using area counts 3 or 4 digits beyond your standard weight will make little or no difference in your results, assuming that they are reported to no more than 1 figure beyond your standard weight.

As to standard purity, if you are running pharmaceutical samples and have a (secondary) standard that has been characterized against a (primary) USP standard, you should use that factor, even if it is above 100% because you are supposed to be reporting your results in terms of a primary, compendial reference standard.

Re: help about potency andsignificant digits

Posted: Sat Nov 10, 2018 2:26 pm
by Consumer Products Guy
jcwagner wrote:
...manager... told us that we have to use all 10 digits regarding the areas of peaks that our LC presents , for our assay calculations.
Until now we were using 6. (ex: 14537.5 from now on 14537.58478).


Sounds like a pointy-haired boss from Dilbert !

1. I agree with DR, if you weigh stuff like standards to four decimals, you are fine using 6 digits for your calculations.

2. We had Agilent instruments, and always used the Chemstation to calculate percent/assay results for us automatically, in a Sequence Summary Report.

Re: help about potency andsignificant digits

Posted: Sat Nov 10, 2018 7:20 pm
by jcwagner
Thanks for the reply, i see your point but i disagree.

I run pharmaceutical samples and i want my result to be accurate.
If i use as potency the 100.9%, i know that i have an error of at least 1% in my final results, which i cannot afford.
100.9% happens due to human error, balance error, pipette and flask error, autosampler error etc.
Am i wrong?

Re: help about potency andsignificant digits

Posted: Sun Nov 11, 2018 3:26 pm
by jcwagner
please could you help?

could a working standard have potency above 100%?
If i use this potency for calculating my samples, my result will be inaccurate, right?
It's like i put in my flask 100mg lets say mefenamic acid and this turn to be 101mg.
When calculating your samples and you use a standard with potency 100.5% , do you use 100.5% or 100%?

Thanks

Re: help about potency andsignificant digits

Posted: Mon Nov 12, 2018 4:42 pm
by lmh
(1) The easy question: whether to use huge numbers of decimal places in areas. You must use enough decimal places such that rounding errors at intermediate stages of your calculation do not affect the least significant figure of the final published result. You shouldn't need the huge precision that your boss is requesting; most of those digits are going to be meaningless in any case.

(2) The difficult one: do you accept a potency greater than 100%? My vote: I think you should, but for rather nerdy reasons, and you're right to have a debate about it.
(i) Your final result has a defined precision (+/- X mg). This is in the realms where it's hard to differentiate between precision and accuracy. The quoted precision is influenced by a lot of things including the precision and accuracy of the standard. It may be that your method is very precise, and you can quote your results normally to +/- 0.3%. However, if the standards you obtain have an absolute error of nearly 1%, then a series of analyses using different standards would include an error of (at least) 1% derived from the variability of the standards. Your precision is limited by the standard's accuracy (or possibly the precision of measurement of the standard; we cannot tell; we have no objective way to differentiate between how much the supplier supplied, and how much they told us they supplied; either value could be inaccurate). The customer, reading your final value, should be aware of this; he wants to know how much stuff is in the sample, and if the result at his level is less precise than it should be, because you've applied a highly precise assay to an inaccurate calibration curve, he should know about it. If the supplier always supplies ">100% pure" standard, because the primary standard is actually impure, then by enforcing a cut-off at 100% you are hiding a source of error in your measurements (quite understandably, but still, I think, wrongly). So, to give your customer the most honest response, you should quote the precision of your assay, between batches of measurements, to include the variation in potency of the standard, and it follows that if you quote this, you can also use the potency as written.
(ii) errors are normally assumed to be distributed symmetrically across a central value, and if you cut off at 100%, you remove the symmetry, should that matter to anyone assessing errors. This does actually matter: imagine that someone is looking to see how accurately the manufacturing process is making the original product, over the course of many years, and using analyses derived from multiple standards. When the standard is >100% pure, the data will be low, when it's <100%, they'll be high. If the errors in the standards are random, then on average, eventually, the mean will be correct, and the conclusion will be that the manufacturing is aiming at the correct potency, but is a bit variable. If you force all values >100% to =100%, then all the low measurements get pushed upwards, while the high measurements stay where they are, giving a skewed distribution of measurements, and moving the mean upwards. The conclusion will be that the manufacturing is making, on average, a slightly over-potent product. Of course if the primary standard is impure, then the standards you buy will always be >100% pure, and your correction would be appropriate, but unless you're certain of the cause of the >100% purity, it's risky to make assumptions about the effects of its correction.
(iii) there's no point in having a measurement if you're going to ignore it. By cutting off at 100% you are, in effect, saying you don't care about the manufacturer's assessment of what they supplied, instead you're going to assume it's clean. I know you have good reason to doubt the manufacturer's assessment, but it's still a bit risky to say you decided to ignore purity data.
But it's a hard one. It's hard to report data that you know must contain a systematic error.

Re: help about potency andsignificant digits

Posted: Tue Nov 13, 2018 8:54 pm
by jcwagner
Thanks a lot for the time and effort to try to explain to me your point of view.
i really appreciate it.

Do you think though that your perspective would change a little if the potency was 102% which is within limits?

I mean, we purchased a working standard. The potency states 102.0%.
I use this standard to quantify my samples.The limits for my samples are 98-102%.
What should i do? Use this potency so multiply my results with 102%?

Re: help about potency andsignificant digits

Posted: Wed Nov 14, 2018 4:24 pm
by lmh
I would pass the buck. I would add to my report that the data have been calculated in accordance with normal good practice, taking into account the quoted purity of the standard, but since the purity of the standard is 2% higher than physically possible, it follows that the data you are providing are likely to be systematically wrong by the same amount. Our measurements are relative, and therefore cannot be more accurate than the standards we use.
It's up to whoever reads and assesses the report to decide whether that 2% is a serious issue in their decision-making.