Towards a consensus on validation of bioanalysis methods for large
molecules
|
A recent international conference in London considered
progress towards regulatory guidance for validation of analytical
methods for the quantitation of macromolecules in biological fluids,
for assessment of antibodies and for validation of biomarker assays.
Howard Hill reports
|
Bioval 2004 was organised by the Joint
Pharmaceutical Analysis Group (a joint venture of the Royal Pharmaceutical Society
and the Royal Society of Chemistry), and was held at the Society’s
London headquarters on 12 and 13 February. Dr Hill is chairman
of the meeting’s organising committee
|
Regulatory guidance exists for validation of bioanalytical methods
for small molecules. It includes the US Food and Drug Administration’s “Guidance
for industry bioanalytical methods validation”, published in May
2001. But agreement has yet to be reached for large molecules. In March
2000, the FDA and the American Association of Pharmaceutical Scientists
held a joint meeting on bioanalytical methods for macromolecules, and
since then the AAPS has led attempts to provide a consensus for the validation
of ligand-based assays, cell based assays, antidrug antibodies and biomarkers.
Bioval 2004 formed part of this consensus process.
Quantitation of macromolecules
The first day of Bioval 2004 was dedicated to the quantitation of macromolecules
in biological fluids and the assessment of antidrug antibodies. The
first speaker, Russ Weiner (Bristol-Myers Squibb, US) compared and
contrasted the needs for small and large molecular guidance. The differences
and degrees of increased complexity associated with ligand binding
assays include the complexity of the standard curve, the variability
and complexity of the standard itself, reagent variability, and matrix
effects further confounded by the difficulty in sample clean-up.
Because ligand binding assays are indirect assays dependent upon binding
interactions, factors such as lipaemic and haemolysed samples, binding
proteins and anticoagulants that interfere with this process will destabilise
the assay, said Dr Weiner.
These issues were further developed by subsequent speakers. Binodh DeSilva
(Amgen, US), discussing ligand binding assays for macromolecules in support
of pharmacokinetic evaluation, said that hot topics include differences
between the criteria for acceptability for a valid method and those for
accepting runs or batches when in routine use. In addition, she identified
the impact of a range of variables involved in the assay, such as reagents,
antibody specificity, analyte stability, plate coating, matrix variables,
and sample collection conditions.
Dr DeSilva also highlighted two important issues as problems associated
with choosing the correct calibration model, and the application of total
error to batch acceptability. These were then reviewed in detail by the
following two speakers.
Bruno Boulanger (Lilly, Belgium) discussed the statistics around the
total error concept, that is, the definition of accuracy being an additive
function of bias and the random error component, defined as precision.
This leads to the proposed “4:6:30” rule, where four out
of six QC samples are expected to fall within 30 per cent of nominal
for a batch to be acceptable, said Dr Boulanger, and it assumes that
the combination of random error (precision) and systematic error (bias)
for the assay is 30 per cent or less.
John Little (CentraLabS, UK) discussed the challenge to obtain the right
standard curve for ligand binding assays. It was generally accepted that
logistic regression methods are the most robust for ligand assays.
Dr Little recommended curve fitting using four-parameter logistics plots
for limited reagent immunoassays such as enzyme immunoassays and radioimmunoassays.
He considered that five-parameter logistics plots are more appropriate
for excess reagent immunoassays, such as immunoenzymic assays, enzyme-linked
immunosorbent assays (ELISA) and immuno-radiometric assays. In addition,
because the assay response error changes in a non-uniform manner, it
is essential to weight the curve to compensate for this non-uniformity.
Dr Little finished on the contentious issue of “editing” points
from the curve that fall outside a set limit, such as 15 per cent. He
showed how “choice” of editing of the same standard curve
could result in different accuracy and precision of derived data. He
posed the question, “What rational criteria can be derived for
logical unbiased editing of standard curves?” The discussion failed
to resolve the issue. Immune system
Mark Wing (CentraLabS, UK) set the scene for the afternoon session
by describing the complexity of the immune system, an important feature
of which is its ability to distinguish itself from foreign material.
In characterising the immune response to drugs, antidrug antibodies
are usually measured. Immune responses may be described as binding,
enhancing or neutralising, having no effect, increasing or decreasing
bioavailability, respectively. A further consequence of neutralising
antibodies, said Dr Wing, is the potential to see the endogenous product
as “foreign”, with potentially serious consequences. These
issues formed the basis for the rest of the afternoon’s discussions.
Arno Kromminga (Institute for Immunology, Pathology and Molecular Biology,
Hamburg, Germany) discussed the development of assays for detecting autoantibodies
in autoimmune disease. A major concern is that the assay developed is
too sensitive and may recognise autoantibodies of low affinity, with
no clinical significance.
Dr Kromminga said that administration of macromolecular therapeutic agents
leads to the production of neutralising antibodies but the route of administration
has a significant effect on the levels. For example, more antibiodies
are produced by giving interferon-beta by the subcutaneous route than
by the intramuscular route.
Geoff Hale (Oxford University and BioAnalab) described attempts to circumvent
the production of antidrug antibodies by “hiding” macromolecule
drugs from the immune system, using agents such as PEG. Another approach
is to engineer the drug — the example was the antibody Campath
(alemtuzumab) — so that the body tolerates its presence. This stealth
approach may allow for repeat, chronic administration. Professor Hale
reviewed the development of antidrug antibody assays with particular
emphasis on the problems of obtaining positive controls, determining,
and then establishing, a quantitation limit. Validation of biomarker assays
The topic for the morning session of the second day was biomarker assays.
David Perrett (Barts and The London Hospital) described how clinical
biochemists had struggled for years with problems of assaying biomarkers
as diagnostic aids in the clinical environment. Now that the pharmaceutical
industry had discovered their use in drug development, consideration
was being given to appropriate assay acceptance criteria that would
satisfy regulatory authorities, Professor Perrett said.
John Allinson (Bioanalytical Systems, UK) discussed the issues of assays
that are deemed acceptable for diagnostic purposes. Many assay kits are
validated to the US FDA 510K guidelines and are run in laboratories regulated
by the Clinical Laboratory Improvement Amendments and which have College
of American Pathologists accreditation. Proficiency tests are run under
guidelines defined by the US National Committee for Clinical Laboratory
Standards. Research-based assay kits are not used for diagnostic purposes
but can be validated to some extent for use in research biomarker assays.
Ian James (Pfizer, UK) spoke about the role of biomarkers in the early
clinical phase of drug development. He illustrated their application,
from the use of a broad range of biomarkers at the early stage in drug
development process, to a limited panel of validated biomarker assays
in the later stages.
Dr James stressed the need to ensure that the biomarkers are clinically
relevant because the cost of validating and running these assays can
add significantly to the development costs.
Jean Lee (MDS Pharma Services, US) described how the AAPS ligand binding
assay bioanalytical focus group is working towards a consensus approach
on biomarker assay validation in biological samples for drug development.
She summarised the outcomes of the focus group workshops held in Salt
Lake City in October 2004.
Currently, said Dr Lee, the seminal reference is the paper on “Validation
of bioanalytical assays for novel biomarkers: practical recommendations
for clinical investigation of new drug entities” (in Bloom JC,
Dean RA, eds. Biomarkers in clinical drug development. New York: Marcel
Dekker; 2003. pp119–48).
Some of the problems arise from a lack of “official” primary
or secondary standards, and the inability to obtain an “analyte
free” biological matrix for the preparation of standards. The assay
dynamic range is often different from the concentrations found in the
test species.
The aim of most clinical studies involving biomarkers is to compare dosed
and control populations. Relative, rather than absolute, changes in biomarker
values may be all that is required to meet the study objective. Therefore
it can be argued that assay precision and relative quantitation, rather
than absolute quantitation, would suffice. The approach used in the FDA-BMV
guidance is recommended for the validation of biomarker assays.
Some of the validation issues were discussed by Ron Bowsher (Linco Diagnostics,
US) in his talk on challenges in validating test kits for quantification
of biomarkers. Dr Bowsher described how the lack of regulatory guidance
has led to confusion among bioanalytical scientists about what procedures
are necessary and appropriate for validation of biomarker assays, especially
kit assays. He classified
assay types using analytical performance characteristics that differentiate
the assay types.
Definitive quantitative assays can be validated to the FDA-BMV guidance
and use well characterised reference standards for calibration and give
a response proportional to concentration that is continuous over the
calibration range. Relative quantitative assays are the same as definitive
quantitative assays
in all respects except that the assays use reference standards that are
uncharacterised. “Quasi-quantitative” assays have no definitive
standards and the standard curve units may be based on activity units.
Qualitative assays may have no accuracy or precision criteria and are
based purely on ordinal data (high, medium or low nominal data) where
results are either positive or negative. Currently, only methods having
well characterised reference standards meet the full requirements for
the FDA-BMV guidance. This can be difficult to achieve for many biomarkers. New techniques
The meeting’s final session looked to the future and considered
the development of alternative techniques. Robin Thorpe (National Institute
for Biological Sciences and Control, UK) discussed the role of ex vivo
bioassays in evaluating the biological activity of “analytes” in
a living system. This differentiates them from in vitro based immunoassay
and receptor binding assays.
Dr Thorpe described a range of bioassays, many of which are cell-based
systems, as opposed to whole animal in vivo methods. The diversity and
complexity of such assays was exemplified by more than 14 observed effects
of interleukin-1 on different target cells and tissues. The effects ranged
from fever induction at the microgram level, through the release of collagenase
from the synovium at nanogram level, to interleukin-2 receptor induction
on T-cells at the femtogram level.
Other issues ranged from standardisation of conditions to standardisation
of the “test-substance” since the purity — or activity — may
vary, depending upon the source and tests used to characterise it.
Berend Oosterhuis, (Pharma-Bioresearch, Netherlands) described a case
history of clinical application of an ex vivo stimulation assay as a
biomarker for anti-inflammatory effects. Using lipopolysaccharide stimulation
of whole blood (monocytes), he reviewed the impact of different drug
inhibitors of P38 kinase, ICE and CD14 antibody on cytokine release.
Dr Oosterhuis said that by careful optimisation of the many parameters,
it is possible to achieve coefficients of variation of 15–30 per
cent. Such characteristics mean that differences in the inhibition of
tissue necrosing factor-a can be used to differentiate different dose
levels and therefore develop useful pharmacokinetic/pharmacodynamic relationships.
Brian Shenton (University of Newcastle, UK) provided an historical overview
of the development of flow cytometry, including fluorescent activated
cell sorter (FACS), which sorts cells on the basis of physical parameter
and fluorescent tags. The development of a wide range of flow cytometers
has been accelerated by a worldwide screening programme for AIDS, the
detection of which depended upon destruction of the
T-helper lymphocyte population during viral infection.
Currently, flow cytometers are used in the clinical and research environment
as a routine tool for biomarker assays such as cytokine analysis and
lymphocyte phenotyping. It is essential to ensure comparability of results
and these depend upon well controlled quality control and quality assurance
(QA) procedures.
Dr Shenton highlighted the fact that many QC procedures using dyes and
beads are instrument-specific, making comparability of results between
laboratories difficult. Correct instrument set-up (training and qualified
operators) is essential before invoking the appropriate QC checks.
Ulrich Kunz (Boehringer Ingelheim, Germany) described the advantages
and limitations of Biacore methods for drug level and immune response
measurements of therapeutic proteins. Limitations of the technique include
matrix effects that limit the number of samples or cycles per sensor
chip, long-term stability of the chip ligand and stability of samples
in the autosampler.
Although pure “analyte” ELISA bridge assays are 100 times
more sensitive than Biacore assays, in real life applications the Biacore
system can detect the immune response as little as three weeks after
dosing whereas ELISA can only detect it after eight weeks.
Dr Kunz hypothesised that the reasons could be the access to all epitopes
by Biacore, and the ability to detect low affinity antibodies. Thus,
Biacore has the advantage of early detection, with the disadvantage of
expensive instruments and chips, and low throughput.
The final talk, by Geoff Hale, focused
on the application and validation of flow
cytometry to the measurement of the pharmacokinetics of Campath–1,
a humanised lgG1 monoclonal antibody, against the CD52 antigen of human
lymphocytes. Because the humanised drug closely resembles endogenous
human IgG (which is present at concentrations 100,000 times that of the
drug), measurement is complex.
After evaluating a wide range of ligand binding assays and bioassays
it was decided to use flow cytometry. Using this technique they were
able to validate an assay from 0.5mg/ml to 50mg/ml. When the critical
variables are identified and controlled, meaningful data can be generated.
Professor Hale illustrated the application of this assay and showed a
correlation between trough concentrations of Campath in leukaemia patients
in early therapy and their ultimate clinical response. |
The
Pharmaceutical Journal