Method validation - spike recoveries

[wchriste]

Hello,
When validating a method what would you do if linearity, accuracy and precision are acceptable but you get recoveries that are out side the desired 80-120%?
While I am asking why is the standard for spike recoveries 80-120%?

Thanks!
Wendy

[rc_12321]

Wchriste, first investigate this failure .If you cannot solve this problem, best way is to go for alternative methods.

Since you know about the potency of standard and concentration of standard solutions , standard is the preferred choice for recovery studies.

[Bruce Hamilton]

The "standard" recovery is probably 80 - 120% because that represents a historical compromise between the cumulative errors of a spiking experiment involving the nausea of extraction from complex media, and the more precise analytical procedure of a pure substance.

80 - 120% usually means that most of the target compound is recovered from the target matrix, and assures that there would be miminal danger of the API dose greatly exceeding the nomimal amount in the final product. Historically, 80 - 120% were also common limits for complex assay tehniques, such as potency testing of antibiotics.

The same applies for the wider acceptance limits for a chemical assay of a formulated product compared to the pure substance. The errors are a mix of the efficiency of the extractions, effects of the media on the compound ( degradation/binding - more common in biological monitoring ), and the effect of media on analysis.

However the acceptance limits are constrained by the need to ensure that close to the nomimal amount of API was ( biological monitoring ) or is ( pharmaceutical manufacturing ) present. Recoveries of less than 80%, or more than 120%, raise serious concerns about the quality of the data and the product.

It's not uncommon for sensitive molecules with very low therapeutic doses to have low spike recoveries, in which case the use of a closely related molecule, or a deuterated version of the molecule, is used as a standard to determine expected recovery from the matrix. Such a procedure involves a lot more validation to confirm that the recovery factors are valid for all expected situations.

For general APIs, if the spiking method can't show 80-120% recovery, you need to identify the cause, and review the reasons. You may need to modify or method, but if you want to pursue methods using low recoveries, you would have to seriously review and consider the conseqences on the monitoring experiment data and/or final product.

Bruce Hamilton

[Alfred88]

Dear wchriste:

Recovery range could (and should) be modified, based on the intended (target) concentration.

If your target is 1%, the recovery could be 93-105%, but for target 0.01%, the recovery is 85-110%. For 0.001%, you can set at 80-115%.

If the recovery is low, you need better method. If it is too high, you need better separation.

Alfred.

[Dan]

Wendy,

I have not heard of using such a wide range of recovery for an assay. Typically, the range is 100.0 +/-2.0%, 3.0% or 5.0%. Unless you have a very difficult matirx or you are analyzing low level analytes such as impurities/degradants (This is basically what Alfred88 stated).

I just re-read your statement. Something is not right!. You state that the accuracy is acceptable but that the recoivery is not! How can this be? The accuracy is determined by the recovery experiment. Can you clarify?

Regards,
Dan

[wchriste]

Wendy,

I have not heard of using such a wide range of recovery for an assay. Typically, the range is 100.0 +/-2.0%, 3.0% or 5.0%. Unless you have a very difficult matirx or you are analyzing low level analytes such as impurities/degradants (This is basically what Alfred88 stated).

I just re-read your statement. Something is not right!. You state that the accuracy is acceptable but that the recoivery is not! How can this be? The accuracy is determined by the recovery experiment. Can you clarify?

Regards,
Dan

Hi Dan,
The spike recoveries are performed in the biological matrix, while the accuracy is performed in blank matrix. I am thinking either I have a coelution issue or there is too much noise at the measured level. The only recommendations for recovery I can find are listed as "should not exceed 80-120% recovery" with no further explaination. I consider +/- 20% to be quite liberal, and I was just curious why.

I appreciate you taking the time to reply. Thanks!
Wendy

[Bruce Hamilton]

I assume that you are seeing more than 120%. As I noted above, those broad limits are used for biological monitoring, especially for analytes in low concentrations and complex matrices. Although the recoveries may be low or high, it's often possible to see trends in the data, provided the recoveries ars consistent.

If you are above 120%, you probably have several options.
You can choose a detector, or detection conditions, that avoid the interference - that's why triple quad MS is popular for clinical trial monitoring. You can change the chromatography to move the interfering compounds - could be very difficult, especially if you have a forest. You can change sample preparation to remove the interfering compounds ( eg SPE ) - probably my preferred option, especially if you don't have an LC_MS-MS lying around.

If your protocal allows you to determine the precision and accuracy of the analytical method using standard solutions, there should also be at least some checking of the method criteria, especially specificity, linearity, and precision on spiked samples or surrogates for your trial samples. Otherwise, the data will be suspect. Recoveries alone will not be sufficient.

I thought there were some standard quality guidelines on biological monitoring experiments, as that's where cGLP seems to be focussed?.
It's not my playground, so I could be wrong.

Please keep having fun

Bruce Hamilton

[Dan]

Wendy,

OK, sorry, I wasn't considering biological samples. So, a +/- 20% limit does not seem unusual for those sample types.

As to references for these limits: I can't think of any. There may not be any. The limits that I have used for drug products has been mostly historical in choice. People use +/- 2.0 or 3.0% for an assay and +/- 5 (or 5.0)% for dissolution analysis. For impurities, usually you use +/- 10% (maybe even +/- 20%).

The limit choice depends on the analyte, concentration level, analysis technique and sample matrix. People tend to use set limits that are typically used in their particular area/industry.

That makes it difficult to remember if there is any actual reference for the limits. I will post a reference if I do come across one.

Meanwhile, Bruce has a good idea to look at the results that you got for any trending. That can help you to troubleshoot. It may even help you to determine if your limits were set too stringent for your particular analysis.

Regards,
Dan