Biopharma · Process Characterisation
Understanding Impact Ratio in Process Characterisation
How a single metric transforms large DoE datasets into clear, defensible decisions about
critical process parameters and design space boundaries
Topic: Process Characterisation · CPP Classification
Level: Advanced
Application: Biopharma, QbD, CMC
In late-stage biopharmaceutical development, process characterisation is all about understanding how process
parameters influence product quality. One metric that proves extremely valuable at this stage is the
Impact Ratio — a powerful, quantitative tool that formalises the relative influence of each
process parameter on a critical quality attribute (CQA), transforming complex DoE datasets into clear, ranked, and
defensible CPP classifications. This methodology, originally described and applied by Roche/Genentech across
multiple monoclonal antibody programmes, has become a cornerstone of Quality by Design (QbD) regulatory
submissions worldwide
[1].
What Is the Impact Ratio?
The Impact Ratio is a normalised, quantitative metric used in process characterisation to determine
how strongly a process parameter (PP) affects a critical quality attribute (CQA) relative to the permissible
variation for that CQA at the unit operation level. It anchors the statistical effect size to real-world process
acceptability — bridging pure statistics with practical process risk management.
Impact Ratio — Formal Definition (Hakemeyer et al., 2016)
IR =
Parameter Effect (Δ CQA at limit of acceptable range)
|Process Mean − CQA Target Range limit|
Parameter Effect (numerator): The expected change in a CQA when the process parameter is
shifted from the
midpoint of its acceptable range to the
limit of its acceptable range. This
is estimated from the fitted regression model of a qualified scale-down characterisation study.
|Process Mean − CQA-TR| (denominator): The distance between the manufacturing-scale mean CQA
result (when the process runs at target) and the nearest CQA Target Range (CQA-TR) limit. The CQA-TR is
derived by narrowing the CQA Acceptance Criterion (CQA-AC) by ~5% to build in a safety margin against scale
model uncertainty
[1].
Conceptually, the numerator represents
how much a CQA may shift when a parameter reaches the
edge of its acceptable range, while the denominator represents
how much variation is
permitted before that CQA approaches its target range boundary.
Unlike a raw p-value or regression coefficient, the Impact Ratio is inherently interpretable in the
context of product quality. It tells you not just whether an effect is statistically detectable, but whether it
is practically meaningful relative to the available headroom between the current process mean and the
CQA target range.
Why CQA-TR rather than the full specification range? The CQA Target Range (CQA-TR) is used in
the denominator — not the full CQA Acceptance Criterion — because the TR is the operationally relevant limit for
a given unit operation pool. For CQAs that change across unit operations (e.g. impurities progressively removed
in downstream processing), the TR is defined to account for the capability of subsequent unit operations,
ensuring the impact ratio reflects the true risk at that stage of manufacturing
[1].
Visualising Impact Ratios : Ranking Parameters by Process Influence
When multiple parameters are evaluated simultaneously in a DoE, impact ratios allow scientists to
quickly rank which variables truly matter for each CQA. The chart below illustrates a typical parameter ranking
from a fed-batch cell culture characterisation study.
Example: Impact Ratio Ranking — Effect on Titre (CQA)
Temperature Shift
IR = 0.88
High-Impact CPP
Feed Timing
IR = 0.74
High-Impact CPP
pH Setpoint
IR = 0.61
High-Impact CPP
DO Setpoint
IR = 0.33
Low-Impact CPP
Agitation Rate
IR = 0.18
Low-Impact CPP
Inoculum Density
IR = 0.09
Non-CPP
Seed Age
IR = 0.04
Non-CPP
High-Impact CPP — IR > 0.33
Low-Impact CPP — 0.10 ≤ IR ≤ 0.33
⚠ Important: The threshold values used by Roche/Genentech — as described by Hakemeyer et al.
(2016) — are IR > 0.33 for High-Impact CPP, 0.10–0.33 for Low-Impact CPP, and < 0.10 for Non-CPP. The
0.10 lower bound was selected because it would require at least 10 such parameters, all simultaneously at
their worst-case limits and all impacting the same CQA, to drive a failure — a scenario considered highly
unlikely in practice [1]. Thresholds must always be
pre-defined and scientifically justified before experimentation.
Interpreting Impact Ratio Thresholds
The meaning of any given impact ratio value is always relative to the CQA specification width and the
risk tolerance of the development programme. The following framework provides a practical starting point for
classification.
> 0.33
High-Impact CPP
Parameter effect consumes more than a third of the available headroom to the CQA-TR.
Robust control strategy is essential. Requires tight operational limits and explicit justification in the
design space claim.
High-Impact Critical Process Parameter
0.10 – 0.33
Low-Impact CPP
Parameter has a meaningful but not dominant effect on the CQA. Classified as CPP given
potential cumulative risk with other parameters; must be included in acceptable ranges and design space
description.
Low-Impact Critical Process Parameter
< 0.10
Non-CPP
Parameter has minimal practical influence on the CQA within the characterised range. To
drive a CQA failure, at least 10 such parameters would need to be simultaneously at their worst-case limit —
considered highly unlikely.
Non-Critical Parameter
Note on the 0.10 threshold: Roche/Genentech selected IR = 0.10 as the Non-CPP boundary to
balance identifying parameters with meaningful effects while avoiding over-classification of minor contributors.
At IR = 0.10, a parameter effect alone cannot drive a CQA failure unless ten or more such parameters act
simultaneously at worst-case — a practically remote scenario. The 0.33 boundary for High-Impact CPPs provides an
additional tier to prioritise control efforts proportionally to risk
[1].
How to Calculate the Impact Ratio — Step by Step
Impact ratio calculation follows a structured workflow from experimental design through CPP
classification, as implemented by Roche/Genentech for their monoclonal antibody programmes [1].
1
Risk Ranking & Filtering (RRF) to identify pCPPs
Before experimentation, a formal risk assessment using Risk Ranking and Filtering
(RRF) scores each process parameter on severity of impact, certainty of information, and control
capability (characterisation range vs. Normal Operating Range). Parameters scoring above defined
thresholds become potential CPPs (pCPPs) entered into characterisation studies.
2
Execute characterisation DoE in a qualified scale-down model
Statistically designed multivariate experiments (DoEs) are conducted in qualified
scale-down models (SDMs) that have been formally shown to be equivalent to manufacturing scale using a Two
One-Sided Test (TOST) with pre-defined Practically Significant Differences (PSDs). Characterisation ranges
are set at least 2–3× the Normal Operating Range.
3
Fit the regression model and extract the Parameter Effect
Fit a linear statistical model (polynomial with intercept, main effects, two-factor
interactions, and curvature terms) to each CQA response. The Parameter Effect is the predicted change in
the CQA when the parameter shifts from the midpoint to the limit of its acceptable range. Interactions and
curvature are incorporated when observed.
4
Define the CQA Target Range (CQA-TR) for the unit operation
The CQA-TR is derived by narrowing the CQA Acceptance Criterion (CQA-AC) by ~5% (2.5%
per side for two-sided TRs). For CQAs that change across the process, the TR accounts for the removal
capability of downstream steps. The denominator of the impact ratio is the distance from the
manufacturing-scale process mean to the nearest CQA-TR limit.
5
Calculate the Impact Ratio for each parameter–CQA pair
IR = Parameter Effect ÷ |Process Mean − CQA-TR limit|. Apply this calculation for
every pCPP–CQA combination. The highest IR across all CQAs for a given parameter determines its
classification. High-Impact CPP: IR > 0.33 | Low-Impact CPP: 0.10–0.33 | Non-CPP: IR < 0.10.
6
Confirm with worst-case linkage studies
All High- and Low-Impact CPPs are included in linkage studies where multiple unit
operations are run simultaneously at their worst-case parameter settings. Linkage study results within the
CQA-TR confirm the Design Space is robust even under combined worst-case conditions across the entire
manufacturing process.
Worked example from the paper: A drug substance process with a mean HMW result of 1.0 area-%
and a CQA-TR of 3.0 area-% gives a denominator of |1.0 − 3.0| = 2.0 area-%. Any parameter shift to its
acceptable range limit that increases HMWs by ≥ 0.2 area-% (IR = 0.2/2.0 = 0.10) would be classified as a CPP
[1].
From Impact Ratio to Process Decisions
The impact ratio is not an end in itself — it is a decision-enabling metric that drives four critical
downstream outputs in late-stage bioprocess development.
Identify Critical Process Parameters (CPPs)
Impact ratio provides a statistically justified, scientifically interpretable basis
for CPP classification — replacing subjective judgment with a quantitative, reproducible criterion that
regulators can evaluate directly.
Focus Control Strategies Where They Matter Most
By ranking parameters, teams allocate analytical and engineering control resources
proportionally to risk. High-IR parameters get tight PAT monitoring and narrow operating ranges; low-IR
parameters allow operational flexibility.
Build a Statistically Justified Design Space
The design space boundaries for CPPs are set using the impact ratio and the model's
prediction uncertainty. Parameters with IR < threshold can be operated anywhere within their proven
acceptable range (PAR) without design space restriction.
Improve Process Robustness
Understanding which parameters drive quality variation enables targeted process
hardening. Scale-up decisions, raw material specifications, and equipment selection can all be informed by
the impact ratio profile — reducing risk of quality failures at commercial scale.
Regulatory Context : QbD and ICH Alignment
Alignment with ICH Q8, Q9, Q10, Q11 and QbD Principles
The Impact Ratio framework directly supports Quality by Design (QbD) principles under
ICH Q8(R2) and fulfils ICH Q9 risk management and Q11 development requirements for biological entities. As
described by Hakemeyer et al. (2016), the approach has been used in multiple BLA/MAA submissions for
monoclonal antibodies at Roche/Genentech and has received approval from all major Health Authorities. It
provides the quantitative link between process parameters and quality attributes needed to justify CPP
classification, define the design space, and build a science-based control strategy — and is increasingly
expected by FDA and EMA in biologics submissions where post-approval flexibility or RTRT is sought
[1].
Key regulatory value of Impact Ratio: It transforms the CPP classification decision from a
qualitative scientific judgement into a
quantitative, reproducible, and auditable statistical
conclusion — strengthening the scientific narrative of CTD Module 3.2.S.2.6 (Process Characterisation).
Roche/Genentech's experience demonstrates that the approach directly informs both the Attribute Testing Strategy
(ATS) and the Post-Approval Lifecycle Management (PALM) plan, enabling more efficient post-approval change
management within the approved Design Space
[1].
How Graphtal Supports Process Characterisation
Graphtal provides end-to-end statistical support for late-stage bioprocess characterisation studies —
from DoE planning through impact ratio calculation, CPP classification, design space modelling, and regulatory
report writing.
Graphtal Services
Process Characterisation & Analytics Support
DoE Planning & Design Selection
Impact Ratio Calculation & CPP Classification
Design Space Modelling & Visualisation
MVDA & Multivariate Process Understanding
Control Strategy Development
Regulatory Report Writing & Submission Support
Need help with Impact Ratio & Process Characterisation?
Graphtal's bioprocess statisticians deliver CPP classification, design space definition, and regulatory-ready
characterisation reports.
Get in Touch →
Reference
[1] Hakemeyer C, McKnight N, St. John R, Meier S, Trexler-Schmidt M, Kelley B, Zettl F,
Puskeiler R, Kleinjans A, Lim F, Wurth C. Process characterization and Design Space definition.
Biologicals. 2016;44(5):306–318. https://doi.org/10.1016/j.biologicals.2016.06.004
