Streamlining data analysis
The preceding discussion suggests that the analysis of cascade impaction data can be a somewhat complex task, and this is
true. However, there are validated computational packages that greatly reduce the burden. Some companies choose to develop
their own spreadsheet tools, but commercial offerings are arguably a more cost-effective and efficient option. Constantly
refined and updated to reflect industry requirements, these solutions promote access to best practice in the area of cascade
impaction data work-up, as that best practice evolves.
Premium software packages export data directly from HPLC systems and automatically correct stage cut-off diameters based on
the flow rate used for testing. They then process the data to produce all the necessary metrics: MMAD, typically on the basis
of two point interpolation, as discussed above, FPD and GSD. Some have the flexibility to calculate user-defined stage groupings,
a feature that can be valuable in the prevailing climate of questioning the best metrics for OINDP characterization.
Systems that alert the user to possible flaws in the data are also especially useful. For example, some will give a LOD (limit
of detection) warning for MMAD, if too much of the cumulative mass is deposited on either the highest or lowest collection
stage or there are less than a certain number of stages with a cumulative mass of greater than, 1% for example. Such warnings
alert the user to raw data that, for whatever reason, is unsuitable for computation through to a worked up set of metrics.
One further point to note is that, as with practical analysis, the higher the degree of automated data manipulation, the lower
the scope for operator error. Automating data handling to the greatest extent, with appropriate tools, saves analyst time
and increases data integrity, therefore promoting better decision-making.
On a day-to-day basis the industry continues to rely heavily on metrics such as MMAD. Understanding how to calculate these
metrics is an important element of analysis, making tools that can streamline the necessary calculations extremely valuable.
Over and above this, however, there is a broader debate going on, questioning whether new metrics or methods should be adopted
for characterizing OINDP particle size. Two key issues have combined to stimulate activity in this area:
- The need for better in vitro–in vivo correlations or relationships
- The need for greater productivity in testing.
Adopting quality by design is especially challenging for the OINDP sector. While it shares with other parts of the industry
the need to reliably predict in vivo behavior, it is difficult to reliably simulate deposition and uptake in the lung. The notoriously inconsistent use of delivery
devices, by even well-trained patients, is a further and massively complicating factor.
Productivity is now an issue in every corner of the pharmaceutical industry and cascade impaction poses a considerable challenge.
Although critical to success, cascade impaction is a time-consuming process, as well as being complex and costly to automate.
Multistage cascade impaction does not simulate the size-fractionating capabilities of the lung. Since we know this to be the
case it could be suggested that splitting the dose into so many fractions is unnecessary. In the vast majority of cases, apart
from when carrying out detailed research, cascade impaction data are actually used to answer relatively simple questions:
Has the FPD or MMAD gone down or up? Is this batch the same as the last or equivalent to that used in clinical trials to approve
the product? The focus of current debate is therefore whether fewer mass groupings and/or different metrics are the way forward
for the bulk of testing.
This idea is alluded to in the Pharmacopeia Forum Stimuli paper, which notes that the constants derived by fitting curves with either the CR or MMF models may be sensitive
metrics for differentiating APSDs (1). Elsewhere, work by the International Pharmaceutical Aerosol Consortium on Regulation
and Science (IPAC-RS) on efficient data analysis suggests that highly differentiating analysis is possible if the emitted
dose is simply considered as two mass fractions—one fine, one coarse (9). The cut-off between these two fractions can be varied
from formulation to formulation for maximum sensitivity, with results suggesting that a cut-off close to the MMAD of the formulation
is the optimum.
Physical representations of the abbreviated impactor measurement (AIM) concept are also now available, for example the Fast
Screening Impactor (MSP Corporation) and the Fast Screening Andersen (Copley Scientific). AIM systems divide a dose into just
2–3 fractions rather than the 7–8 obtained with full resolution cascade impaction. Estimated productivity gains are substantial
and solvent usage savings are also encouraging (10).
Early experimental work suggests that for metered dose inhalers and nebulizers close agreement is observed between AIM data
and corresponding metrics measured using full resolution multistage cascade impaction. With dry powder inhalers the story
is currently less clear with more research required to clarify the causes of systematic differences between data sets (10).
However, recently published work indicates good progress in understanding these differences and compensating for them (11).
Nevertheless, it is still early days for these new approaches, and the regulators remain to be convinced of their validity—an
essential step towards greater uptake.
What is clear is that inertial impaction methods will remain pivotal to the development and quality control of OINDPs. Understanding
these techniques, how to use them, and how to manipulate the resulting data are therefore core skills for those working in
David Lewis is managing director at 3DI Solutions, and Mark Copley* is sales director at Copley Scientific, Colwick Quays Business Park, Private Road No. 2, Colwick, Nottingham, NG4 2JY, UK,
tel: +44 115.961.6229, fax: +44 115.961.7637, email@example.com
*To whom all correspondence should be addressed.