LCGC North America
Here, we review the regulatory expectations of metrics for monitoring data integrity and discuss their practical implementation in a regulated chromatography laboratory.
One area of a data integrity program is the establishment of metrics for monitoring data integrity. Here, we will look at the regulatory expectations and discuss their practical implementation in a regulated chromatography laboratory.
This is the ninth article in a series on data integrity focus in the regulated chromatography laboratory (1–8), and there was also a six-part series on data integrity in 2018 co-authored by Mark Newton and myself (9–14). In two of these articles, a Data Integrity Model was presented (1,10) that consisted of a foundation layer and three levels above. One of the key items in the foundation layer is management leadership for data integrity, and this is reflected in both EU GMP Chapter 1 regulations (15), as well as data integrity guidance documents from regulatory authorities, including the Food and Drug Administration (FDA) (16), the Medicines and Healthcare Products Regulatory Agency (MHRA) (17), the World Health Organization (WHO) (18), the Pharmaceutical Inspection Convention and Pharmaceutical Inspection Co-operation Scheme (PIC/S) (19), and industry guides (20,21).
These guidance documents set a regulatory expectation that management and companies should be able to detect data errors and falsification via either second person review or quality oversight, such as data integrity audits and investigations. The alternative is a starring role in warning letters posted on the FDA's wall of shame with citations such as:
As can be seen, management held accountable for lapses in data integrity of their staff. The FDA is also interested in quality metrics, as we shall now see.
From a laboratory perspective, knowing the out-of-specification (OOS) rate is an important criterion for both quality and data integrity. One of the key questions to ask when auditing or inspecting a laboratory is to ask to see the OOS results for the past six months. However, answering, "we don't have any" can result in a regulatory surprise such as:
Since beginning manufacturing operations in 2003, your firm has initiated a total of 0 out of specification investigations for finished product (24).
Even when you have OOS results, but the overwhelming majority of them are invalidated, that can result in a winning entry for an FDA warning letter:
Observation 1: There is a failure to thoroughly review any unexplained discrepancy and the failure of a batch or any of its components to meet any of its specifications whether or not the batch has been already distributed. Specifically, from January 2016 to March 2017, your firm invalidated several OOS results as summarized in Table 1 (25).
Given that analytical procedures are subject to variation, we expect to see not only OOS, but also out-of-expectation (OOE) and out-of-trend (OOT) results. There is a specific EU GMP regulation that requires trending of laboratory results:
6.9 Some kinds of data (e.g. tests results, yields, environmental controls) should be recorded in a manner permitting trend evaluation. Any out of trend or out of specification data should be addressed and subject to investigation (26).
6.16 The results obtained should be recorded. Results of parameters identified as quality attribute or as critical should be trended and checked to make sure that they are consistent with each other.... (26).
The FDA has focused on quality metrics as a way of identifying lower risk facilities, and has issued two draft guidance for industry documents on quality metrics for comment in July 2015 and November 2016 (27). To monitor the performance of quality control (QC) laboratories, the agency has selected the invalidated out-of-specification rate (IOOSR), which is defined as the number of OOS test results for lot release and long-term stability testing invalidated as a percentage of total OOS results. Seeing the results in Table I, you can understand why the FDA has selected this metric.
What is the OOS rate in your laboratory? If you don't know, there might be some issues during the next inspection.
Tracking quality or data integrity metrics may sound like yet another task for chromatographers to perform in addition to their normal work, but many of you already are used to working with them under the name of key performance indicators (KPIs). KPIs are used by many laboratories to measure the operational and quality performance of processes.
For example, laboratory managers may need to know how many samples are being analyzed, and how long it takes to analyze them. They need to know if there are any bottlenecks, and if they have enough resources assigned to handle the work. Metrics can provide the information to measure a process or activity. If used correctly, metrics help staff understand the performance measures that a laboratory is judged against. Some common laboratory metrics include numbers of samples analyzed by unit time, turnaround time (TAT) per sample measured against targets, instrument utilization rate, and numbers of injections per chromatograph per unit time. The same approach can be taken with metrics to assess data integrity within a laboratory or organization. I can see you roll your eyes: more work, you think!
A key requirement is that metrics must be collected automatically. You may ask why. The simple reason is that if humans collect and collate the data manually, it is an error-prone process, tedious, labor intensive, and could also be subject to falsification. The latter is not the best option when generating data integrity metrics.
Automatic metric generation is the only way to create metrics that are timely, accurate, and repeatable. This is where suppliers of chromatography data systems (CDS) and other laboratory informatics applications could be more helpful by implementing applications that generate both general and data integrity metrics automatically.
Why are data integrity metrics important? It is now a regulatory expectation, as we can see from the excerpt of PIC/S guidance PI-041 shown in Table II (19). There is an expectation that data integrity metrics of processes and systems are collected for management review and that management reviews them, and acts if there are any issues.
Look at the second paragraph in section 6.5.1 in Table II that says to take care when selecting a quality metric that could result in a culture in which data integrity is lower in priority. Let us consider this statement further. Take the example of sample TAT. Imagine that the laboratory target for TAT is set for 10 days, and over the past few months the target has been missed and the laboratory manager is very peeved (this is understandable if he or she gets a bonus based on laboratory performance). The sample you are starting to analyze has been in the laboratory for nine days, and your analysis will take two days to perform. You will miss the TAT target time, unless...
This is where quality culture, management, and data integrity collide, and it is where the PIC/S guidance caution about metric comes into play. Can some metrics, intended to monitor activities, become the means of inducing behaviors that compromise data integrity? Possibly, in the case where a manager is putting pressure on staff to meet performance targets. Another factor to consider is that the areas that are monitored tend to improve because they get attention, but this can be at the expense of other activities that are not monitored.
Most data integrity reports can only identify areas for further investigation, because it is rare where a single record equates to a data integrity breach (11,28). You must carefully select what areas will be monitored by metrics, because too many can bury you in data that are intended to help you. Start small, evaluate the data, and adjust as necessary. I like a proof of concept or prototyping approach: Identify a metric, generate the data, and then assess what is happening. This is especially so when reports can be collected retrospectively so that the situation both before and after the metric was established can be evaluated.
It is virtually impossible to automatically generate and analyze metrics for manual or hybrid activities. For example, you could possibly detect missing files in a system by comparing the instrument use log (paper) against a report of sample identities from all runs stored on a CDS if a falsifier was stupid enough to complete the log honestly. However, this approach is too tedious to be done on a regular basis, and is best left for either a data integrity audit or investigation as discussed in Part VI of this series (6). In contrast, sample identities in a laboratory information management system (LIMS) or electronic laboratory notebook (ELN) could be compared against the data files on laboratory data systems using an automated script. Owing to the time involved in their preparation, manual reports should be limited to special situations, such as detecting fraud from a single suspected individual as part of an investigation.
Some data integrity issues can't be monitored or checked, such as ensuring the correct sample volumes or weights are taken during manual preparation prior to analysis. Data recorded manually outside of a computerized system are very difficult to verify but this is where data acquisition to a LIMS or ELN can improve the situation compared with the following citation:
He said that he recalibrated the balance and prepared new documentation, and subsequently discarded the original record. Furthermore, we learned that additional original calibration records of other balances had similarly been discarded (29).
Metrics could also include assessment of the performance of individuals, such as measuring the time to complete a method for several different people and looking for someone working too quickly. This is where regulatory compliance clashes with, in some countries, the works council or labor authorities who are concerned with protecting the rights of workers. The balance between ensuring data integrity of the organization producing and marketing pharmaceutical products needs to be balanced with the rights of individuals. However, the regulatory expectation is that processes must ensure the integrity of records generated.
There are some activities that are just difficult to catch with a report. For example, once a user changes the naming convention on repeat injections, you will probably not detect it on a routine report. And trying to catch everything can only be done if you have unlimited resources. So, what can we do practically?
Some suggested data integrity metrics for chromatography processes are presented below and include:
One of the best ideas for monitoring metrics often come from regulatory enforcement actions such as FDA warning letters. The key requirement is to read the cited deficiency and ask yourself, "How would we detect this situation in our own laboratory (28)?" This question will cause you to think about the data patterns that accompany the behavior, and then to formulate a query that could detect the data pattern. For example, a firm is cited for manipulating the system clock to falsify timestamps in audit trail records. If this happens, there could be a series of system audit trail entries, one for each clock adjustment. In addition, there will be some frequently written audit trails (such as inter-system messages) where the clock will appear to go backward because of the clock manipulation. So a query that checks for clock entries that do not continue to increase could flag the clock manipulation behavior. Just be careful when there are summer and winter time changes.
It is important to remember that all metrics are not created equal; some will prove more effective than others in your operation. In addition, metrics seldom identify a true issue with every reported record in a report; rather, they highlight suspicious records that require a human to investigate; this requires a time investment, and therefore becomes a limitation on reporting effectiveness. Finally, some real issues will not be detected in a report, such as reanalyzing a sample on a simple instrument (for example, a pH meter), or picking the desired outcome and forwarding it to LIMS: This is a data integrity issue that will not be detected on any report.
Tracking metrics for data integrity is a regulatory expectation. This article has provided an overview and background about quality metrics and which of those to collect for a chromatography laboratory. Metrics should be generated automatically and be used to detect trends that require human investigation. Please avoid paralysis by analysis by having too many metrics and the accompanying data: Start small and work from there.
(1) R.D. McDowall, LCGC N.Amer. 37(1), 44–51 (2019).
(2) R.D. McDowall, LCGC N.Amer.37(2), 118–123 (2019).
(3) R.D. McDowall, LCGC N.Amer.37(3), 180-184 (2019).
(4) R.D. McDowall, LCGC N.Amer.37(4), 265–268 (2019).
(5) R.D. McDowall, LCGC N.Amer. 37(5), 312–316 (2019).
(6) R.D. McDowall, LCGC N.Amer. 37(6), 392–398 (2019).
(7) R.D. McDowall, LCGC N.Amer. 37(8), 532–537 (2019).
(8) R.D. McDowall, LCGC N.Amer. 37(9), 684–688 (2019).
(9) M.E. Newton and R.D. McDowall, LCGC N.Amer.36(5), 330–335 (2018).
(10) M.E. Newton and R.D. McDowall, LCGC N.Amer.36(1), 46–51 (2018).
(11) M.E. Newton and R.D. McDowall, LCGC N.Amer.36(4), 270-274 (2018).
(12) M.E. Newton and R.D. McDowall, LCGC N.Amer.36(7), 458–462 (2018).
(13) M.E. Newton and R.D. McDowall, LCGC N.Amer.36(8), 527–529 (2018).
(14) M.E. Newton and R.D. McDowall, LCGC N.Amer.36(9), 686–692 (2018).
(15) EudraLex - Volume 4 Good Manufacturing Practice (GMP) Guidelines, Chapter 1 Pharmaceutical Quality System. 2013, European Commission: Brussels.
(16) FDA Guidance for Industry Data Integrity and Compliance With Drug CGMP Questions and Answers (Food and Drug Administration, Silver Spring, Maryland, 2018).
(17) MHRA GXP Data Integrity Guidance and Definitions (Medicines and Healthcare Products Regulatory Agency, London, United Kingdom, 2018).
(18) WHO Technical Report Series No. 996 Annex 5 Guidance on Good Data and Records Management Practices (World Health Organization, Geneva, Switzerland, 2016)
(19) PIC/S PI-041-3 Good Practices for Data Management and Integrity in Regulated GMP/GDP Environments Draft (Pharmaceutical Inspection Cooperation Scheme, Geneva, Switzerland, 2018).
(20) GAMP Guide Records and Data integrity (International Society for Pharmaceutical Engineering, Tampa, Florida, 2018).
(21) GAMP Good Practice Guide: Data Integrity-Key Concepts (International Society for Pharmaceutical Engineering, Tampa, Florida, 2018).
(22) FDA Warning Letter Sun Pharmaceuticals (Food and Drug Administration, Rockville, Maryland, 2014).
(23) FDA Warning Letter Fresenius Kabi Oncology (WL: 320-13-20) (Food and Drug Administration, Rockville, Maryland, 2013).
(24) FDA 483 Observations, Sri Krishna Pharmaceuticals Limited (Food and Drug Administration, Silver Spring, Maryland, 2014).
(25) FDA 483 Observations: Lupin Limited (Food and Drug Administration, Rockville, Maryland, 2017).
(26) EudraLex - Volume 4 Good Manufacturing Practice (GMP) Guidelines, Chapter 6 Quality Control (European Commission, Brussels, Belgium, 2014).
(27) FDA Guidance for Industry Submission of Quality Metrics Data, Revision 1 (Food and Drug Administration, Rockville, Maryland, 2016).
(28) M.E. Newton and R.D. McDowall, LCGC Europe30(12), 679–685 (2017).
(29) FDA Warning Letter Leiner Health Laboratories (Food and Drug Administration, Rockville, Maryland, 2016).
R.D. McDowall is the director of R.D. McDowall Limited in the UK. Direct correspondence to: rdmcdowall@btconnect.com
GC–TOF-MS Finds 250 Volatile Compounds in E-Cigarette Liquids
November 1st 2024A study has used gas chromatography coupled to a time-of-flight mass spectrometer to build an electron ionization mass spectra database of more than 250 chemicals classified as either volatile or semi-volatile compounds. An additional, confirmatory layer of liquid chromatography–mass spectrometry analysis was subsequently performed.
AI and GenAI Applications to Help Optimize Purification and Yield of Antibodies From Plasma
October 31st 2024Deriving antibodies from plasma products involves several steps, typically starting from the collection of plasma and ending with the purification of the desired antibodies. These are: plasma collection; plasma pooling; fractionation; antibody purification; concentration and formulation; quality control; and packaging and storage. This process results in a purified antibody product that can be used for therapeutic purposes, diagnostic tests, or research. Each step is critical to ensure the safety, efficacy, and quality of the final product. Applications of AI/GenAI in many of these steps can significantly help in the optimization of purification and yield of the desired antibodies. Some specific use-cases are: selecting and optimizing plasma units for optimized plasma pooling; GenAI solution for enterprise search on internal knowledge portal; analysing and optimizing production batch profitability, inventory, yields; monitoring production batch key performance indicators for outlier identification; monitoring production equipment to predict maintenance events; and reducing quality control laboratory testing turnaround time.
Multivariate Design of Experiments for Gas Chromatographic Analysis
November 1st 2024Recent advances in green chemistry have made multivariate experimental design popular in sample preparation development. This approach helps reduce the number of measurements and data for evaluation and can be useful for method development in gas chromatography.