Moving Forward from “Fine Particle Fraction: The Good and the Bad”

Recently, a review of the fine particle fraction (FPF) metric by Newman⁽¹⁾ was reprinted in the Journal of Aerosol Medicine and Pulmonary Drug Delivery. This article was written in 2014 and was originally published in 2015 as part of the International Society for Aerosols in Medicine (ISAM) online textbook (https://isam.org). We regard this review as important because it sets out both the merits and drawbacks of fine particle fraction (FPF) as a means of characterizing the aerodynamic particle size distribution (APSD) of aerosols produced by orally inhaled drug products (OIPs). The ambiguity of FPF is correctly highlighted, but the discussion is focused on an ill-defined size limit separating fine from coarse mass fractions. Newman emphasized that FPF, by itself, does not adequately describe the bivariate nature of an APSD.

Furthermore, defining FPF with fixed size limit of 5 μm aerodynamic diameter, as required in Monograph 2.9.18 of the European Pharmacopoeia, may be a poor representative of likely deposition beyond the oropharyngeal region. For example, Sagalla and Smaldone⁽²⁾ presented an empiric relationship between in vitro testing of aerosol distributions and in vivo adult regional lung and upper airway deposition from adult patients with chronic obstructive pulmonary disease for nebulizer-generated droplets. They found that a boundary size for upper airway deposition close to 2.5 μm aerodynamic diameter was more realistic. This further emphasizes the importance of detecting changes in APSD that are not detectable by any fixed FPF value.

We wish to highlight that there is an alternative approach, termed Efficient Data Analysis (EDA).⁽³⁾ Changes in APSD can be followed unambiguously using just two independent metrics representing movements in the aerodynamic size and drug mass of unimodal APSDs typical of almost all OIP-generated aerosols. These measures are

Importantly, Tougas et al. have shown that the boundary between SPM and LPM can be set in relation to the mass median aerodynamic diameter (MMAD) between 0.3*MMAD and 3.0*MMAD without markedly affecting sensitivity.⁽⁴⁾ This flexibility will likely prove useful when comparing in vitro with in vivo data in which it becomes necessary to adjust the boundary size to improve the predictability of laboratory-generated data for likely lung and upper airway deposition. A roadmap for the implementation of EDA in the context of OIP quality control has been published in Pharmaceutical Forum.⁽⁵⁾

Goodey et al. have recently evaluated the reasons why FPF alone cannot avoid failures by examining 201 cascade impactor APSDs of the same OIP contained in the International Pharmaceutical Aerosol Consortium on Regulatory Science (IPAC-RS) database.⁽⁶⁾ They were able to identify an outlier case in which fine particle mass <5 μm (FPM_<5μm) by itself at 46% of label claim was just within the 25th and 75th percentile range for the entire data set (45%–58%) and could, therefore, likely be deemed to be within specification based only on this measure.

However, examination of the underlying APSD revealed that its MMAD was close to 3.5 μm compared with the range of MMAD for the other members of the data set that was between 1 and 2 μm. Setting the limit for FPF at 5.0 μm aerodynamic diameter (following the methodology in the European Pharmacopoeia), where there is little sensitivity to variations in FPF for individual APSDs when the expected MMAD is so far removed from this size limit, contributed to the lack of discrimination. Further assessments of APSDs from eight different OIPs identified several instances in which significant shifts in the values of MMAD of the underlying individual APSDs were not captured as corresponding shifts in fine particle dose < 5μm aerodynamic diameter (FPD_<5μm).

Summarizing, we believe that there is sufficient published evidence that FPF (and FPD) by themselves do not have the discriminating ability to detect important movements in the underlying APSD. Furthermore, we suggest that adoption of EDA can provide an effective and simple-to-implement solution to this limitation.