• 沒有找到結果。

4.1 Key Reagents 619

4.1.1 Reference Standard 620

The reference standard should be well characterised and documented (e.g., CoA and origin). A 621

biological drug has a highly complex structure and its reactivity with binding reagents for 622

bioanalysis may be influenced by a change in the manufacturing process of the drug substance.

623

It is recommended that the manufacturing batch of the reference standard used for the 624

preparation of calibration standards and QCs is derived from the same batch of drug substance 625

as that used for dosing in the nonclinical and clinical studies whenever possible. If the reference 626

standard batch used for bioanalysis is changed, bioanalytical evaluation should be carried out 627

prior to use to ensure that the performance characteristics of the method are within the 628

acceptance criteria.

629

4.1.2 Critical Reagents 630

Critical reagents, including binding reagents (e.g., binding proteins, aptamers, antibodies or 631

conjugated antibodies) and those containing enzymatic moieties, have direct impact on the 632

results of the assay and, therefore, their quality should be assured. Critical reagents bind the 633

analyte and, upon interaction, lead to an instrument signal corresponding to the analyte 634

concentration. The critical reagents should be identified and defined in the assay method.

635

Reliable procurement of critical reagents, whether manufactured in-house or purchased 636

commercially, should be considered early in method development. The data sheet for the critical 637

reagent should include at a minimum identity, source, batch/lot number, purity (if applicable), 638

concentration (if applicable) and stability/storage conditions (Refer to Table 1). Additional 639

characteristics may be warranted.

640

A critical reagent lifecycle management procedure is necessary to ensure consistency between 641

the original and new batches of critical reagents. Reagent performance should be evaluated 642

using the bioanalytical assay. Minor changes to critical reagents would not be expected to 643

influence the assay performance, whereas major changes may significantly impact the 644

performance. If the change is minor (e.g., the source of one reagent is changed), a single 645

comparative accuracy and precision assessment is sufficient for characterisation. If the change 646

is major, then additional validation experiments are necessary. Ideally, assessment of changes 647

will compare the assay with the new reagents to the assay with the old reagents directly. Major 648

changes include, but are not limited to, change in production method of antibodies, additional 649

blood collection from animals for polyclonal antibodies and new clones or new supplier for 650

monoclonal antibody production.

651

Retest dates and validation parameters should be documented in order to support the extension 652

or replacement of the critical reagent. Stability testing of the reagents should be based upon the 653

performance in the bioanalytical assay and be based upon general guidance for reagent storage 654

conditions and can be extended beyond the expiry date from the supplier. The performance 655

parameters should be documented in order to support the extension or replacement of the critical 656

reagent.

657

4.2 Validation 658

When using LBA, study samples can be analysed using an assay format of 1 or more well(s) 659

per sample. The assay format should be specified in the protocol, study plan or SOP. If method 660

development and assay validation are performed using 1 or more well(s) per sample, then study 661

sample analysis should also be performed using 1 or more well(s) per sample, respectively. If 662

multiple wells per sample are used, the reportable sample concentration value should be 663

determined either by calculating the mean of the responses from the replicate wells or by 664

averaging the concentrations calculated from each response. Data evaluation should be 665

performed on reportable concentration values.

666

4.2.1 Specificity 667

Specificity is evaluated by spiking blank matrix samples with related molecules at the maximal 668

concentration(s) of the structurally related molecule anticipated in study samples.

669

The accuracy of the target analyte at the LLOQ and at the ULOQ should be investigated in the 670

presence of related molecules at the maximal concentration(s) anticipated in study samples. The 671

response of blank samples spiked with related molecules should be below the LLOQ. The 672

accuracy of the target analyte in presence of related molecules should be within ±25% of the 673

nominal values.

674

In the event of non-specificity, the impact on the method should be evaluated by spiking 675

increasing concentrations of interfering molecules in blank matrix and measuring the accuracy 676

of the target analyte at the LLOQ and ULOQ. It is essential to determine the minimum 677

concentration of the related molecule where interference occurs. Appropriate mitigation during 678

sample analysis should be employed, e.g., it may be necessary to adjust the LLOQ/ULOQ 679

accordingly or consider a new method.

680

During method development and early assay validation, these “related molecules” are 681

frequently not available. Additional evaluation of specificity may be conducted after the 682

original validation is completed.

683

4.2.2 Selectivity 684

Selectivity is the ability of the method to detect and differentiate the analyte of interest in the 685

presence of other “unrelated compounds” (non-specific interference) in the sample matrix. The 686

matrix can contain non-specific matrix component such as degrading enzymes, heterophilic 687

antibodies or rheumatoid factor which may interfere with the analyte of interest.

688

Selectivity should be evaluated at the low end of an assay where problems occur in most cases, 689

but it is recommended that selectivity is also evaluated at higher analyte concentrations.

690

Therefore, selectivity is evaluated using blank samples obtained from at least 10 individual 691

sources and by spiking the individual blank matrices at the LLOQ and at the high QC level.

692

The response of the blank samples should be below the LLOQ in at least 80% of the individual 693

sources.

694

The accuracy should be within ±25% at the LLOQ and within ±20% at the high QC level of the 695

nominal concentration in at least 80% of the individual sources evaluated.

696

Selectivity should be evaluated in lipaemic samples and haemolysed samples (Refer to Section 697

3.2.1). For lipaemic and haemolysed samples, tests can be evaluated once using a single source 698

of matrix. Selectivity should be assessed in samples from relevant patient populations. In the 699

case of relevant patient populations there should be at least five individual patients.

700

4.2.3 Calibration Curve and Range 701

The calibration curve demonstrates the relationship between the nominal analyte concentration 702

and the response of the analytical platform to the analyte. Calibration standards, prepared by 703

spiking matrix with a known quantity of analyte, span the calibration range and comprise the 704

calibration curve. Calibration standards should be prepared in the same biological matrix as the 705

study samples. The calibration range is defined by the LLOQ, which is the lowest calibration 706

standard, and the ULOQ, which is the highest calibration standard. There should be one 707

calibration curve for each analyte studied during method validation and for each analytical run.

708

A calibration curve should be generated with at least 6 concentration levels of calibration 709

standards, including LLOQ and ULOQ standards, plus a blank sample. The blank sample 710

should not be included in the calculation of calibration curve parameters. Anchor point samples 711

at concentrations below the LLOQ and above the ULOQ of the calibration curve may also be 712

used to improve curve fitting. The relationship between response and concentration for a 713

calibration curve is most often fitted by a 4- or 5-parameter logistic model if there are data 714

points near the lower and upper asymptotes, although other models may be used with suitable 715

justification.

716

A minimum of 6 independent runs should be evaluated over several days considering the factors 717

that may contribute to between-run variability.

718

The accuracy and precision of back-calculated concentrations of each calibration standard 719

should be within ±25% of the nominal concentration at the LLOQ and ULOQ, and within ±20%

720

at all other levels. At least 75% of the calibration standards excluding anchor points, and a 721

minimum of 6 concentration levels of calibration standards, including the LLOQ and ULOQ, 722

should meet the above criteria. The anchor points do not require acceptance criteria since they 723

are beyond the quantifiable range of the curve.

724

The calibration curve should preferably be prepared using freshly spiked calibration standards.

725

If freshly spiked calibration standards are not used, the frozen calibration standards can be used 726

within their defined period of stability.

727

4.2.4 Accuracy and Precision 728

4.2.4.1 Preparation of Quality Control Samples 729

The QCs are intended to mimic study samples and should be prepared by spiking matrix with 730

a known quantity of analyte, stored under the conditions anticipated for study samples and 731

analysed to assess the validity of the analytical method.

732

The dilution series for the preparation of the QCs should be completely independent from the 733

dilution series for the preparation of calibration standard samples. They may be prepared from 734

a single stock provided that its accuracy has been verified or is known. The QCs should be 735

prepared at a minimum of 5 concentration levels within the calibration curve range: The analyte 736

should be spiked at the LLOQ, within three times of the LLOQ (low QC), around the geometric 737

mean of the calibration curve range (medium QC), and at least at 75% of the ULOQ (high QC) 738

and at the ULOQ.

739

4.2.4.2 Evaluation of Accuracy and Precision 740

Accuracy and precision should be determined by analysing the QCs within each run (within-741

run) and in different runs (between-run). Accuracy and precision should be evaluated using the 742

same runs and data.

743

Accuracy and precision should be determined by analysing at least 3 replicates per run at each 744

QC concentration level (LLOQ, low, medium, high, ULOQ) in at least 6 runs over 2 or more 745

days. Reported method validation data and the determination of accuracy and precision should 746

include all results obtained, except those cases where errors are obvious and documented.

747

Within-run accuracy and precision data should be reported for each run. If the within-run 748

accuracy or precision criteria are not met in all runs, an overall estimate of within-run accuracy 749

and precision for each QC level should be calculated. Between-run (intermediate) precision and 750

accuracy should be calculated by combining the data from all runs.

751

The overall within-run and between-run accuracy at each concentration level should be within 752

±20% of the nominal values, except for the LLOQ and ULOQ, which should be within ±25%

753

of the nominal value. Within-run and between-run precision of the QC concentrations 754

determined at each level should not exceed 20%, except at the LLOQ and ULOQ, where it 755

should not exceed 25%.

756

Furthermore, the total error (i.e., sum of absolute value of the errors in accuracy (%) and 757

precision (%)) should be evaluated. The total error should not exceed 30% (40% at LLOQ and 758

ULOQ).

759

4.2.5 Carry-over 760

Carry-over is generally not an issue for LBA analyses. However, if the assay platform is prone 761

to carry-over, the potential of carry-over should be investigated by placing blank samples after 762

the calibration standard at the ULOQ. The response of blank samples should be below the 763

LLOQ.

764

4.2.6 Dilution Linearity and Hook Effect 765

Due to the narrow assay range in many LBAs, study samples may require dilution in order to 766

achieve analyte concentrations within the range of the assay. Dilution linearity is assessed to 767

confirm: (i) that measured concentrations are not affected by dilution within the calibration 768

range and (ii) that sample concentrations above the ULOQ of a calibration curve are not 769

impacted by hook effect (i.e., a signal suppression caused by high concentrations of the analyte), 770

whereby yielding an erroneous result.

771

The same matrix as that of the study sample should be used for preparation of the QCs for 772

dilution.

773

Dilution linearity should be demonstrated by generating a dilution QC, i.e., spiking the matrix 774

with an analyte concentration above the ULOQ, analysed undiluted (for hook effect) and 775

diluting this sample (to at least 3 different dilution factors) with blank matrix to a concentration 776

within the calibration range. For each dilution factor tested, at least 3 runs should be performed 777

using the number of replicates that will be used in sample analysis. The absence or presence of 778

response reduction (hook effect) is checked in the dilution QCs and, if observed, measures 779

should be taken to eliminate response reduction during the analysis of study samples.

780

The calculated concentration for each dilution should be within ±20% of the nominal 781

concentration after correction for dilution and the precision of the final concentrations across 782

all the dilutions should not exceed 20%.

783

The dilution factor(s) applied during study sample analysis should be within the range of 784

dilution factors evaluated during validation.

785

4.2.7 Stability 786

Stability evaluations should be carried out to ensure that every step taken during sample 787

preparation, processing and analysis as well as the storage conditions used do not affect the 788

concentration of the analyte.

789

The storage and analytical conditions applied to the stability tests, such as the sample storage 790

times and temperatures, sample matrix, anticoagulant, and container materials should reflect 791

those used for the study samples. Reference to data published in the literature is not considered 792

sufficient. Validation of storage periods should be performed on stability QCs that have been 793

stored for a time that is equal to or longer than the study sample storage periods.

794

Stability of the analyte in the studied matrix is evaluated using low and high concentration 795

stability QCs. Aliquots of the low and high stability QCs are analysed at time zero and after the 796

applied storage conditions that are to be evaluated. A minimum of three stability QCs should 797

be prepared and analysed per concentration level/storage condition/timepoint.

798

The stability QCs are analysed against a calibration curve, obtained from freshly spiked 799

calibration standards in a run with its corresponding freshly prepared QCs or QCs for which 800

stability has been proven. While the use of freshly prepared calibration standards and QCs is 801

the preferred approach, it is recognised that in some cases, for macromolecules, it may be 802

necessary to freeze them overnight. In such cases, valid justification should be provided and 803

freeze-thaw stability demonstrated. The mean concentration at each level should be within 804

±20% of the nominal concentration.

805

Since sample dilution may be required for many LBA assays due to a narrow calibration range, 806

the concentrations of the study samples may be consistently higher than the ULOQ of the 807

calibration curve. If this is the case, the concentration of the stability QCs should be adjusted, 808

considering the applied sample dilution, to represent the actual sample concentration range.

809

As mentioned in Section 3.2.8, the investigation of stability should cover bench top (short-term) 810

stability at room temperature or sample preparation temperature and freeze-thaw stability. In 811

addition, long-term stability should be studied.

812

For chemical drugs, it is considered acceptable to extrapolate the stability at one temperature 813

(e.g., -20°C) to lower temperatures (e.g., -70°C).

814

For biological drugs, it is acceptable to apply a bracketing approach, e.g., in the case that the 815

stability has been demonstrated at -70°C and at -20°C, then it is not necessary to investigate the 816

stability at temperatures in between those two points at which study samples will be stored.

817

4.3 Study Sample Analysis 818

The analysis of study samples can be carried out after validation has been completed however 819

it is understood that some parameters may be completed at a later stage (e.g., long-term 820

stability). By the time the data are submitted to a regulatory authority, the bioanalytical method 821

validation should have been completed. The study samples, QCs and calibration standards 822

should be processed in accordance with the validated analytical method. Refer to Table 1 for 823

expectations regarding documentation.

824

4.3.1 Analytical Run 825

An analytical run consists of a blank sample, calibration standards at a minimum of 6 826

concentration levels, at least 3 levels of QCs (low, medium and high) applied as two sets (or at 827

least 5% of the number of study samples, whichever is higher) and the study samples to be 828

analysed. The blank sample should not be included in the calculation of calibration curve 829

parameters. The QCs should be placed in the run in such a way that the accuracy and precision 830

of the whole run is ensured taking into account that study samples should always be bracketed 831

by QCs.

832

Most often microtitre plates are used for LBAs. An analytical run may comprise of one or more 833

plate(s). Typically, each plate contains an individual set of calibration standards and QCs. If 834

each plate contains its own calibration standards and QCs then each plate should be assessed 835

on its own. However, for some platforms the sample capacity may be limited. In this case, sets 836

of calibration standards may be placed on the first and the last plate, but QCs should be placed 837

on every single plate. QCs should be placed at least at the beginning (before) and at the end 838

(after) of the study samples of each plate. The QCs on each plate and each calibration curve 839

should fulfil the acceptance criteria (Refer to Section 4.3.2). For the calculation of 840

concentrations, the calibration standards should be combined to conduct one regression analysis.

841

If the combined calibration curve does not pass the acceptance criteria the whole run fails.

842

4.3.2 Acceptance Criteria for an Analytical Run 843

Criteria for the acceptance or rejection of an analytical run should be defined in the protocol, in 844

the study plan or in an SOP. In the case that a run contains multiple batches, acceptance criteria 845

should be applied to the whole run and to the individual batches. It is possible for the run to 846

meet acceptance criteria, even if a batch within that run is rejected for failing to meet the batch 847

acceptance criteria.

848

The back-calculated concentrations of the calibration standards should be within ±20% of the 849

nominal value at each concentration level, except for the LLOQ and the ULOQ, for which it 850

should be within ±25%. At least 75% of the calibration standards, with a minimum of 6 851

concentration levels, should fulfil this criterion. This requirement does not apply to anchor 852

calibration standards. If more than 6 calibration standards are used and one of the calibration 853

standards does not meet these criteria, this calibration standard should be rejected and the 854

calibration curve without this calibration standard should be re-evaluated and a new regression 855

analysis performed.

856

If the rejected calibration standard is the LLOQ, the new lower limit for this analytical run is 857

the next lowest acceptable calibration standard of the calibration curve. If the highest calibration 858

standard is rejected, the new upper limit for this analytical run is the next acceptable highest 859

calibration standard of the calibration curve. The new lower and upper limit calibration standard 860

will retain their original acceptance criteria (i.e., ±20%). The revised calibration range should 861

cover all QCs (low, medium and high). The study samples outside of the revised assay range 862

should be reanalysed.

863

Each run should contain at least 3 levels of QCs (low, medium and high). During study sample 864

analysis, the calibration standards and QCs should mimic the analysis of the study sample with 865

regard to the number of wells used per study sample. At least 2/3 of the QCs and 50% at each 866

concentration level should be within ±20% of the nominal value at each concentration level.

867

Exceptions to these criteria should be justified and predefined in the SOP or protocol.

868

The overall mean accuracy and precision of the QCs of all accepted runs should be calculated 869

at each concentration level and reported in the analytical report. In the case that the overall 870

mean accuracy and/or precision exceeds 20%, additional investigations should be conducted to 871

determine the cause(s) of this deviation. In the case of comparative BA/BE studies it may result 872

in the rejection of the data.

873

4.3.3 Calibration Range

4.3.3 Calibration Range

相關文件