how to deal with BLQ samples?

[aniket]

Helow forum,
I just wanted to know that suppose i ma getting the analyate cocenrtration below my LOQ should i consider this value as zero or should i take the absolute value??? if this value lies in between LOD and LOQ?? how to deal with such a value? and suppose my LLOQ is 25 ng/ml, should i discard 24.8 ng/ml?? please help me on this??

[DR]

"<25ng/mL" for results >LOD and <LOQ, "ND" for <LOQ results.

[gbalock]

Aniket,
When you get a value that is below LOQ, the reliability of that result is diminished as LOQ is described as the level at which you can accurately quantittate a peak.

The ICH guidelines state that LOQ is roughly 10 times the noise in a chromatogram. Others sometimes describe LOQ as the level at which the RSD of the values at that level are not greater than X, where X is typically 10%.

What we do is if a value is below LOQ, but above LOD we report it as Below Reporting Limit (BRL). If it is below LOD, it is None Detected (ND). But, we do report values of peaks that are below LOQ for information only.

[Dan]

Aniket,

I don't want to confuse the issue, but sometimes the regulations do that.

Rounding and specifications must be considered here.

You gave an example of an LLOQ of 25 ng/ml with a result of 24.8 ng/ml. What is your specification in this instance? Is that 25 supposed to be 25.0?

If yes, then your result is 24.8 and is it below the LLOQ and you would report "<25 ng/ml"; "<LOQ" or "BRL" as the other posters stated.

But, if the value is 25 (the whole number), then your result is not 24.8 but it is the rounded value of 25. Then you don't report "<LOQ", you would have to report "25 ng/ml".

It may be confusing. It may also sound trivial. However, I can assure you that this is one of the most misunderstood area of reporting results in the phramaceutical industry.

So, consider the proper rounding with respect to the specification before you proceed with reporting a result.

[Mark Tracy]

Any analytical result has an associated confidence interval. If the result plus its upper half-interval includes LLOQ, you treat it as equal to LLOQ. (California actually write this rule into their hazardous waste laws!)

Moreover, LLOQ is by definition an imprecise number with its confidence interval. An LLOQ of "25.0" is silly because with an RSD of 10%, the confidence interval is about +/- 5.

[Dan]

I agree that the value of the LOQ is imprecise.

In my previous post, I should have made it clear that my example is from the viewpoint of reporting results in the pharma industry. We follow ICH, FDA, EMEA, etc. for setting the product specifications and the reporting of results. Results are to be reported according to the specification. The ICH gets very specific on the reporting aspect.

So, my example was for working with a specification limit. When you put into that context, then the difference between "25" and "25.0" is not trivial as I noted before.

I used an example of a specification limit that is given to one decimal. In pharma, we usually try to have an LOQ that is no more than 50% of the specification limit.

Example: If the specification is 50.0 ng/ml, then the LOQ is 25.0 (or less). That first decimal is not trivial. So, a result of 24.8 is reported as 24.8 and not as 25. That was my point.

Yes it seems a bit silly when we are dealing with the LOQ, but, in pharma, there are rules we have to live by.

Sorry, if I wasn't clear before.

Regards,
Dan

[aniket]

i just wanted to know will you take the value below LLOQ for calculationn of PK parameters, there is a lengthy discussion on PharmPK disscusssion forum. can you please elaborate on this?? what is the need of an LLOQ if it is imprecise??

Thank you all for Replies

Aniket

[bert]

Dear Aniket,

I also followed the discussion on PharmPK forum.

Generally speaking:

In my opinion, values below LOQ in the terminal elimination phase should be discarded. PK studies should be designed in such a way that the effect of discarding values <LOQ is not crucial for estimating the PK. We always want to measure elimination over at least 6 t1/2 periods and design analytical methods with an LOQ to establish this.

But remember, this is generally speaking……

Regards Bert

[Mark Tracy]

I'm sorry to hear that you have to live with rules that generate the illusion of greater precision than the measurement can possibly deliver.

[HW Mueller]

Mark, nicely put. As I mentioned before, I am getting ever more sorry for having to swallow pills.

[aniket]

dear mark can you please elaborate your point???
i know you have to live with rules?? but are not they dynamic?? are they going to change??

[Mark Tracy]

Generally, rules for regulated industries change very slowly. Once a method is established, the risk associated with changing a rule is unacceptable unless the reason is so compelling that failure to change will result in liability. Furthermore, each organization has its own understanding of the rules, and that understanding has survived auditors, so "don't mess with success" applies.

I have even heard the dictum "If you are wrong consistently enough and long enough, you are right" applied to environmental trace analysis. (This mentality is most common among lawyers for whom truth is what you say it is, and self-consistency is the only reality check. Many scientists have a hard time even thinking those thoughts.)

On the subject of LLOQ, the illusory precision is more of a philosophical offense than a practical problem in pharmaceutical testing. By the time you accumulate enough data to say something useful, and have done all the statistics, the minor flaw in reporting results washes out. The central limit theorem wins in the end.

The only time that bogus precision can really harm you is in forensic analysis where you are lucky to get a second analysis for confirmation of the first. There the statistics of small numbers are more troublesome.