Case Story: CCS * Control Chart Script

Report
Presentation title
Date
1
Case Story from Novo Nordisk:
CCS — Control Chart Script
Jørgen Iwersen, Principal Scientist
Torben Koustrup Sørensen, Principal Scientist
Presentation title
Date
CCS – Control Chart Script
• CCS is an GxP-critical IT-system for automated generation of control
charts and trend reports in the production area of Novo Nordisk A/S
• Current version is using of-the-shelf products such as SAS JMP®,
Citrix and Microsoft Excel
• CCS is by and large a JMP-script. Code modification is dependant of
SAS JMP–programmers
2
Presentation title
Date
CCS – Control Chart Script
• The number of users on CCS has increased from 600 users in 2011 to
over 1100 users in start 2013 and the system is in use in all
production areas of Novo Nordisk A/S
• CCS is currently handling over 5000 control charts.
This number is steadily increasing
3
Presentation title
Date
CCS – Control Chart Script
• CCS generates a control chart and report through information from
user defined Control Chart Definition Files
• The control chart and report can be saved in HTML and viewed on
the Novo Nordisk Intranet
• As a minimum a CCS-report is composed of the control chart itself
and an information block documenting the control chart parameters.
Optionally the report can contain a distribution- and capability
analysis and a table with selected raw data
4
Presentation title
Date
CCS – Control Chart Script
• Computer systems used in the pharmaceutical industry with
GxP-impact are subject to the most strict quality management
requirements, including document control and retention
• Control charts used in the GxP-environment are updated based on
controlled and approved process results stored in the GxP-data
warehouse (QDW)
5
Presentation title
CCS – Control Chart Script
Five types of control charts are available:
• I:
Individual chart
• IM:
Individual/Moving Range chart
• L:
Levey Jennings chart
• XR:
Shewhart -R chart
• XS:
Shewhart -s chart
Date
6
Presentation title
Date
CCS – Control Chart Script
• The properties of the control chart, e.g. range span/sample size etc.,
are controlled in the main parameter file
7
Presentation title
Date
CCS – Control Chart Script
Data flow
• Live data are accessed from a qualified and validated data
warehouse that contains relevant GxP-data and non-GxP data.
Data are extracted from the data warehouse to Excel and imported
into JMP
• CCS is used for both in-process data (non-GxP data) and release
data (GxP data) from production sites
8
Presentation title
Date
CCS – Control Chart Script
• GxP data
In the GxP-environment relevant QDW data can be chemical analysis
results from analytic laboratories for release test activities.
The measurements are transferred to the data warehouse on a daily
basis. CCS is updating control charts and the CCS Report ‘Review of
Submission results’
• non-GxP data
IPC-charts used in non_GxP-environment can be elaborated from data
automatically entered into excel files from local process equipment
9
Presentation title
Date
CCS – Control Chart Script
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Elaboration of Control Charts
The first step in calculating the control limits for a control chart is choosing/selecting data from a time period in which the natural variation of the process are represented. In
most cases chemists are to choose data from a relatively stable and constant period of time covering at least 20 observations sampled under normal manufacturing conditions.
Data from a time period of one year are typically used in GxP-Control Charts.
Control limits can be adjusted for controllable variation. Season variation is a typical variance component that often has to be accounted for in chemical manufacturing. Control
limits can be adjusted by season, in which change in the process level can be correlated to temperature, relative humidity, pressure etc.
Run rules can be applied. The recommended run rules are Run Rule No.1 (in JMP Test 1) “One point beyond Zone A” and Run Rule No. 5 (Test 5) “Two out of three points in a row
in Zone A or beyond”.
When elaborating a control chart for a new process the functionality of the control chart is tested in a test environment in CCS – the CCS Sandbox.
In CCS it is possible to apply the use of action limits. Action limits are typically in between control- and specification limits for the process. When the distance from control limits to
specification limits are very large action limits can be added on the control chart.
Action limits are often used when the analytic results are close to e.g. LoD – Limit of Detection, and in similar situations where the variance estimates are invalidated.
Updating Control Charts
Updating the control charts are scheduled and regulated in CCS and must be undertaken as a minimum every year and after implementing process changes.
10
Presentation title
Date
CCS – Control Chart Script
Elaboration of Control Charts
• When elaborating a control chart for a new process the functionality of
the control chart is tested in a test environment in CCS – the CCS
Sandbox
• Selection of data period: data should come from a time period in which
the natural variation of the process are represented
• In most cases chemists are to choose data from a relatively stable and
constant period of time covering at least 20 observations sampled under
normal manufacturing conditions
• Data from a time period of one year are typically used in GxP-Control
Charts
11
Presentation title
Figure 1. Overview of data flow in CCS
Date
12
Presentation title
Date
CCS – Control Chart Script
Elaboration of Control Charts
• Run rules can be applied. The recommended run rules are Run Rule
No.1 (in JMP Test 1) “One point beyond Zone A” and Run Rule No. 5
(Test 5) “Two out of three points in a row in Zone A or beyond”
• Control limits can be adjusted for controllable variation. Season
variation is a typical variance component that often has to be
accounted for in chemical manufacturing. Control limits can be
adjusted by season, in which change in the process level can be
correlated to temperature, relative humidity, pressure etc.
13
Presentation title
Date
CCS – Control Chart Script
• In CCS it is possible to apply the use of action limits. Action limits
are typically in between control- and specification limits for the
process
• Action limits can be used when the analytic results are close to e.g.
LoD – Limit of Detection, and in similar situations where the variance
estimate is invalidated
14
Presentation title
Date
CCS – Control Chart Script
Updating control charts
• Updating the control charts are scheduled and regulated in
CCS and must be undertaken as a minimum every year and
after implementing process changes
15
Presentation title
Date
CCS – Control Chart Script
Example: CCS control chart for xxx_Ethanol
• Data from a period of three months has been sampled for a the GxPparameter
• The CCS report presents the following descriptive statistics from JMP
16
Presentation title
CCS – Control Chart Script
Individual Measurement (I) chart vs. Submission Creation Date
Numerical Result vs. Sample
96
UCL=94.09
94
Numerical Result
92
90
Avg=89.62
88
3
86
LCL=85.16
84
2
4
6
8
10
12
14
16
18
20
22
24
26
Sample
28
30
32
34
36
38
40
42
44
46
48
50
Date
17
Presentation title
3
.99
2
.95
.90
1
.75
.50
0
.25
-1
.10
.05
-2
.01
-3
80
84
88
92
Normal(89.6242,1.71581)
96
100
Normal Quantile Plot
Distributions
Numerical Result
Date
18
Presentation title
Moments
Quantiles
100.0%
99.5%
97.5%
90.0%
75.0%
50.0%
25.0%
10.0%
2.5%
0.5%
0.0%
Date
maximum 93.827
93.827
93.647
92.041
quartile
90.489
median
89.831
quartile
88.752
87.178
85.334
85.226
minimum 85.226
Mean
Std Dev
Std Err Mean
upper 95% Mean
lower 95% Mean
N
89.624187
1.7158077
0.2586677
90.14584
89.102533
44
Fitted Normal
Parameter Estimates
Type
Parameter
Location
μ
Dispersion σ
Estimate
89.624187
1.7158077
Lower 95%
89.102533
1.4176395
Upper 95%
90.14584
2.1739721
19
Presentation title
Date
CCS – Control Chart Script
Advancement of CCS
• On-going CCS development activities are focusing on capacity expansion
and incorporation of customer claimed enhancements to the system
• Control charts and other statistical techniques for control and
surveillance are planned to be rolled out in the framework of a new
version of CCS covering a broader spectrum than the production areas
• Future versions of CCS should account for a diverse SPC system
landscape – from control of laboratory analysis methods to surveillance
of marketed products
20

similar documents