APR Nov/Dec 2022 - 41

« MANUFACTURING
Figure 2. Example project charter, used from the project scoping stage onwards.
* A simple, 1-slide, project charter16,17
(Figure 2) to set common
expectations and define intended outcomes. This is written
in collaboration with all project partners after the first few
meetings. Dividing projects into work packages simplifies
project scoping and timeline setting. The charters are refined
during the project execution.
* A project portfolio table18
to manage each internship. These
simple Google Sheets™ or Excel™ lists contain project essential
information to track progress, allocate resources, and summarize
descriptive project charter information. They facilitate portfolio
analyses (see Table 1), including changes over the years, and
value demonstration to the department.
* Software solutions which enable real-time visual
development during prototyping (e.g. Google Data Studio™).
These enable a critical capability: quick, functional, mockup
generation and concept testing, regardless of the final
deployment platform.
* Comprehensive project documentation including developer
records, handover documents, best practice documents,
and user manuals. To facilitate project handover and
troubleshooting, these must be started parallel to development
and regularly revised.
Project Insights and Examples
This section starts by illustrating our approach with two hands-on
detailed examples, one scientific and one business problem. This is
followed by the insights from the analysis of the entire portfolio over
the past three years.
Scientific example: Punch sticking
Punch sticking is a phenomenon that can pose severe problems in
tablet manufacturing. It refers to strong adherence of powder onto
the tablet tooling, which can lead to significant tablet surface defects
and productivity losses.19,20
This problem is often only identified
late in development, when formulation changes carry severe time
and money penalties.20
As an early test, Roche utilizes a small-scale
removable punch sticking assay to quantify the amount of material
adhered to the tablet tooling after a small number of compactions.
Prior to our work, the data was collected in various Excel™ sheets on
different file-sharing platforms. Risk classifications were made based
only on the measurement at the highest number of compactions.
Together with our formulators, we developed an R Shiny™ application
to enable access to the consolidated data from one communal place,
and to classify formulations more robustly in terms of their sticking
risk at manufacturing scale. The Shiny application consists of one main
sidebar tab for each of these use cases. The Data Viewer tab (Figure
3A) lets the user access the previous data from one data-combining
Google Sheet™ via a Google service account and explore various
plots quantifying the sticking propensity of selected formulations. In
the Data Prediction tab, the user can enter new measurements, and
add them to the database. Furthermore, the sticking behavior of the
corresponding formulation can be predicted through data fitting
to a slightly adapted version of the function by (Paul et al, 2016).21
The mass sticking after a customizable number of compressions is
calculated together with bootstrap confidence intervals and a risk
classification (Figure 3B). Explanatory text helps the user interpret the
prediction outputs, including summary statistics on model fit. This text
warns in case of high sticking risk, or insufficient model fit, e.g., due to
large experimental noise. Finally, to facilitate reporting, a PDF report
containing all plots and summary statistics shown in the app can be
automatically generated.
Currently, the application is utilized to provide guidance in
early development. The prediction and risk classification will be
continuously optimized with future validation data as part of the
model life-cycle management.
www.americanpharmaceuticalreview.com |
| 41
»
http://www.americanpharmaceuticalreview.com
APR Nov/Dec 2022

Table of Contents for the Digital Edition of APR Nov/Dec 2022
