(FDA) issued a regulation that details criteria that must be met for the acceptance of electronic records, electronic signatures and handwritten signatures. 1 This regulation, entitled Rule 21 CFR Part 11, permits electronic records to be used in place of paper records and handwritten signatures. The rule applies to all industry segments regulated by FDA that must adhere to good laboratory practice (GLP), good clinical practice (GCP) and current good manufacturing practice (cGMP). The new regulation requires analytical laborato- ries to demonstrate the following: use of validated existing and new equipment and computer systems secure retention of electronic records to instantly reconstruct the analysis user-independent, computer-generated time-stamped audit trails system and data security, data integrity and confidentiality through limited authorized system access use of secure electronic signatures for closed and open systems use of digital signatures for open systems. Although the rule is well documented, information technology (IT) professionals and analysts in labora- tories are often unsure about its implementation. The biggest problem is to find a compromise between doing too much and satisfying minimum requirements. Frequently asked questions Exactly which records should be retained, particularly when data have to be re-evaluated several times before they can finally be accepted? How should computer-generated audit trails be implemented; what should be tracked and after To comply with the US Food and Drug Administrations 21 CFR Part 11 (and the draft guidance published on 20 February 2003), reliable, robust and validatable communication between computers and analytical instruments has become increasingly important. This article demonstrates that monitoring and feedback mechanisms for instrument control and communication can drastically simplify compliance with Part 11, because they make the generation of electronic raw data records trustworthy and reliable. It also evaluates typical communication protocols used, and lists selection criteria for instrument control and data acquisition systems. Instrument Control in Pharmaceutical Laboratories Compliance with 21 CFR Part 11 and the New Draft Guidance Wolfgang Winter is worldwide product manager, networked data systems. Tel. +49 7243 602 454 Fax +49 7243 602 501 wolfgang_winter@agilent.com Ludwig Huber is worldwide product marketing manager, HPLC. Tel. +49 7243 602 209 Fax +49 7243 602 501 ludwig_huber@agilent.com Agilent Technologies GmbH, PO Box 1280, D-76337 Waldbronn, Germany. a r t w o r k
W a r r e n
J o n e s
o r i g i n a l
i m a g e s
P h o t o d i s c which entries; and does the user of the system have to confirm the entries as being logged? How should signatures be linked with the electronic records? Is computerized instrument control in the scope of Part 11? What can be done with existing instruments that lack the appropriate functionality? How can you ensure that long- term archiving allows the possibility to replay data after 10 or more years? In 1999, FDA published compliance policy guide 7153.17 2 and issued a total of five draft guidance docu- ments relating to Part 11, which have been vividly discussed and com- mented on in the affected industries. However, on 20 February 2003, FDA published a new guidance on Scope and Applications (www.fda.gov/ OHRMS/DOCKETS/98fr/03d-0060- gdl0001.PDF and www.labcompli- ance.com/conferences/s991.htm) and withdrew all other draft guidances. The new draft guidance states FDAs intention of interpreting Part 11 more closely in line with its cGMP initiative, focussing on risk-based inspections. According to the new draft guidance, Part 11 will continue to apply to records that are required by the predicate rules, and which are maintained in electronic format in addition to paper, and are relied on to perform regulated activities. Typically, this is the case for records managed by chromatography data systems used in analytical laborato- ries in industries regulated by FDA. The authors will discuss the impact of the new draft guidance in more detail in the April 2003 issue of Pharmaceutical Technology Europe. Additionally, other associations in the industry have also acted to develop guidelines on how to imple- ment Part 11. For example, the Parenteral Drug Association (PDA) has formed a task force on Part 11 (PDAs Good Electronic Records Management [GERM]) and a special interest group of the Good Automated Manufacturing Practice (GAMP) Forum has released a draft paper. 3 FDA also plans to release more guidance documents; Part 11 itself just defines the framework now that the industry is working on concrete implementation plans to comply, many practical system and process issue arise. Ongoing updates are important and can be found at www.labcompliance.com under e-signatures (21 CFR 11). Who must comply with 21 CFR? Huber, Winter and Nickel have published a series of articles on the implementation of Part 11 in analytical laboratories. 49 These articles demonstrate how access to the system and to critical system functions could be limited to autho- rized personnel. They also explain how the integrity of data can be assured at the time of data analysis and evaluation and how creation, modification and deletion of records are logged in a computer-generated audit trail. They also discuss how data can be archived and accurately retrieved after several years. These articles focus mainly on compliance and management of data generated by analytical data systems. A frequently asked question is should computers that just control analytical instruments, but do not acquire data, comply with Part 11? If FDA ever looked at or asked for paper printouts of the parameters, the answer is simply, yes. Without proper documentation of the instru- ment control parameters, it becomes difficult to prove that a given result was generated according to the appropriate procedure or mono- graph. If a computer was used for this procedure, and if, at any time, the control parameters are stored on a durable storage device (typically the computers hard disk or a storage card for the instrument), Part 11 applies. In this article, we will discuss levels of instrument control and data acquisi- tion; electronic records associated with instrument control; tracking of instru- ment characteristics, events and errors in an electronic logbook; and cross- manufacturer instrument control. Levels of instrument control Analytical laboratories typically operate a diversified installed base of instruments, often from a variety of manufacturers. As most modern instruments need to be controlled by a computer system, the instrument control parameters have to be treated in the same way as the rest of the metadata. Instrument control of a device can be implemented at varying levels of sophistication and complexity (Table I). Often, instrument para- meters have to be set manually using the instruments own panel and key- board, with the signal being recorded by an analoguedigital converter (Level 1). This is often the approach chosen to integrate an instrument into a system from a different manu- facturer. In these cases, it is typically impossible to obtain a printout of the instrument settings that are used during an analysis. Analysts are forced to document instrument parameters manually. Furthermore, analoguedigital converters do not always support BCD (binary-coded decimal) or barcode input (from an autosampler) that can be used to positively correlate an injection with a given sample using the sample name or at least a vial number. Many systems implement a rudimentary level of instrument con- trol (Level 2). This is often obtained through reverse-engineering of the communication protocols for another manufacturers instrument, and supports the basic parameters of an instrument, such as solvent com- position, flow, oven temperature and detector wavelength. If the manufac- turer of the instrument did not disclose the control codes, it may be much more difficult to obtain an officially supported (that is, approved and guaranteed) solution with appropriate validation documents from the supplier. Also, additional effort should be planned to perform qualification and other validation tasks on such a system. As the manufacturer of the original instrument may be neither aware nor responsible for the implementa- tion of the communication, instru- ment firmware updates may result in non-functional communication to the data system. The implementa- tion of error handling and logging is Although the rule is well documented, information technology (IT) professionals and analysts in laboratories are often unsure about its implementation. typically weak in this category. When selecting a system that is supposed to control instrumentation from other manufacturers, it is important to verify that the control codes were obtained from the manufacturer of the instrument and not reverse-engineered. In most cases, manufacturers implement full instrument control (Level 3) for their own systems. This facilitates keeping a complete set of raw and metadata together with the proper documentation. In this category, one can also expect quite sophisticated error reporting and handling, which makes it easier to verify that an analysis was indeed completed without technical failures or to diagnose an error. Some manufacturers have implemented an additional level of instrument capabilities that can be controlled from within the data system. These functions are the basis for the execution of detailed and sophisticated instrument diagnostics as well as other service functions. This includes provisions for preventive maintenance and early maintenance feedback (EMF), a technique initially used in the aeronautics industry (to alert technical personnel to perform due maintenance jobs proactively before something fails), which is now used in the pharma- ceutical and other industries. More importantly, systems implemented at this level provide sophisticated support for tracking instruments or module serial num- bers and firmware revisions. This information is not only convenient and important for inventory tracking of validated equipment, but can also be used to verify some of the device check functions required by Part 11. In Level 4 instrument control implementations, all communication (including commands and data transfer) is performed using a hand- shake. A handshake basically means that the recipient of a data record must actively acknowledge receipt of the record by notifying the sender. For example, the controller sends a command such as START to the device, the device interprets the command and acknowledges OK, START. If, for any reason, the device is not able to execute the command, it sends a negative receipt such as NOT OK, NO START. This approach clearly avoids situations in which the controller is unaware that instructions sent to the device have not been executed. Relevance of communication protocols for data integrity Below is a discussion of widely used instrument communication protocols and their strengths and weaknesses in terms of data integrity and traceability according to 21 CFR Part 11. We will Level Parameters Impact on Part 11 Table I Levels of instrument control. Level 1 Start/stop (no digital instrument control and Metadata: instrument parameters must be Parameter set-up on the instrument, data acquisition). documented manually. synchronization using external contacts Device checks: positive ID of sample vials to start and stop an analysis, analogue may not be available (using barcodes or signal acquisition. BCD input). Level 2 Basic instrument parameter, for example, Audit trail: typically no instrument error Rudimentary digital instrument control flow rate of an HPLC (high performance information available requires additional for example, through LAN (local area liquid chromatography) pump or the inspections to determine the validity of network), RS 232, GPIB. wavelength of an HPLC detector. the measurements. Validation: could be more difficult to support if reverse-engineered. Level 3 All control parameters, including injector Audit trail and metadata: full documentation Full digital instrument control for programme, method sequencing. of instrument parameters used to example, through LAN, RS 232, GPIB. Wavelength calibration. generate a result. Error recording. Level 4 Handshake protocol between controller and Advanced error prevention and detection. Advanced functions. device (active acknowledgement of correct Validation: facilitates the execution of receipt). instrument qualification and preventive Self diagnostics and early maintenance maintenance. feedback (EMF). Qualifies for device checks required by the Automatic tracking of serial and product rule. numbers, electronic instrument logbook. Guaranteed and reproducible execution Supports advanced tagging of components, of data acquisition independent of for example column ID tags. current data system load (facilitates the Instrument performs real-time data qualification of data integrity and traceability). acquisition and synchronization independently of the computer. use GPIB (general purpose interface bus) as an example, which is widely used as the IEEE 488 standard. Contrasting examples are networking protocols, such as the well-known and ubiquitous transmis- sion control protocol/Internet pro- tocol (TCP/IP) that is used for any intranet or Internet communication. This discussion cannot, and is not intended to be, a detailed description of the technology. A wide range of publications covers these aspects accurately and in greater technical detail. 1012 Instrument communication via GPIB GPIB is a parallel communication interface that can connect up to 15 devices on a common bus. All communication, including commands and data, is performed using a hard- ware handshake for every byte. All devices connected to the bus participate in this handshake. As a consequence, every device on the bus can influence the ongoing communication and cause severe communication problems such as bus hang-ups or data corruption. These problems may be caused by a firmware error or a hardware failure in one of the participating devices (for example, the printer). But also switching a seemingly idle GPIB device on or off during ongoing com- munication may generate a problem. Even though the electrical specifica- tions of GPIB do not prohibit these scenarios, the combination of chip- set implementation, the firmware and the application software often lead to that failure. (See Table II for a summary of GPIB characteristics.) Instrument communication via LAN (TCP/IP) TCP/IP, often described as the language of the Internet, enables devices to exchange information over a network (Table III). The key idea is to break information into pieces or packets. The packets are specifi- cally structured to allow error detec- tion and error correction by using redundancy mechanisms such as checksums (this is how they differ from the majority of GPIB imple- mentations described above). A checksum is, in principle, a running total of all transmitted bits that are attached to the packet and it is used by the recipient to back-calculate and compare with the original checksum provided by the sender. If a mismatch is detected, retransmis- sion is requested. This technique guarantees an error-free data trans- port and represents an excellent basis for the implementation of device checks and system checks mandated by 21 CFR Part 11. Figure 1 illustrates a typical configuration of networked instru- ments in a distributed, client/server environment. Communication in a TCP/IP environment is, by definition, highly dynamic. Addition or removal of idle devices in the network does not affect ongoing communication between the other participants. In contrast to most GPIB implementa- tions, TCP/IP supports safety proce- dures requiring instrumentation that is not in use to be switched off without the risk of data loss. Recommendations for selecting instrument control and data acquisition systems 1. The level of control of instru- mentation used in the laboratory must be assessed. 2. If instrument communication uses proprietary interfaces, check for adherence to your internal IT support guidelines. 3. Find out what levels of control are offered by hardware manufacturers. 4. For instrumentation not directly controlled by the data system (Level 1), plan the procedures necessary to document instrument parameters. 5. For instrumentation claimed to be controlled, determine whether the Strength Weakness Recommendation Table II Characteristics of GPIB. Fast enough for high-speed data Limited number of devices on the same bus (15). N/A acquisition (for example, spectral data Restrictions on total data rate on the bus. from diode array or mass-selective Distance limitations (maximum of 2 m detectors). between devices). GPIB permits bidirectional, addressed The standard does not define power on/off Avoid powering down idle GPIB devices information to be sent (allows for the of idle devices during ongoing communication. while other devices are acquiring data, to implementation of device checks As all devices must electrically participate in minimize the risk of data loss or data according to 21 CFR Part 11). the handshake, switching a device off during corruption. communication may hang up the entire communication system. This can lead to data loss. N/A Remote monitoring of instrument functions Requires separate operational qualification requires a computer (instrument controller) of the instrument controller. that functions as an information broker next to the instrument. N/A Requires proprietary GPIB interfaces in the Subject to (re)qualification particularly instruments as well as the computers. if PC models and operating systems change. Verify compatibility with support guidelines from your IT support group. communication protocols were obtained with approval and support of the instrument manufacturer. 6. For instruments for which communication protocols were apparently developed by reverse engineering, plan additional qualification and acceptance tests to obtain a high degree of assurance that control and communication are accurate and reliable. 7. Adapt internal procedures to take advantage of the additional diagnostics, maintenance and tracking functions to validate, maintain and document the system and the measurements obtained with it. 8. Define test cases for the boundary conditions. Does the system reliably Strength Weakness Recommendation Table III Characteristics of TCP/IP. Suited to high-speed data acquisition (for example, spectral data N/A Proactively plan network infrastructure and segmentation, from diode array or mass-selective detectors). particularly in large labs with many instruments. Inherent mechanisms for error detection and correction N/A Reliable instrument communication should be verified (an important measure for data integrity and the basis for during system qualification. device checks according to 21 CFR Part 11). Turn off idle instruments not in use. Networked instruments help eliminate PCs from the lab bench. N/A Proactively plan lab layout and where computers are Remote status information is available directly from the required (and where they are superfluous!) to support instrument and does not necessarily require an extra the workflow. instrument controller (computer) as information broker. Allows the use of cheaper, widely used, standard N/A Ask for a list of recommended models from your IT interfaces (such as standard Ethernet LAN cards for the PC). support group. control. We have defined the different levels of instrument control capabili- ties that can exist in modern analytical data systems. Level 4 instrument con- trol not only supports an instruments command set, but uses an advanced control layer with handshake protocols between software and connected instruments. This level offers advanced functions that allow automatic tracking of instrument identification or configuration information, and it is a prerequisite for the implementation of EMF. Electronic records generated by an instrument are only reliable and trustworthy if the communication between instrument and the system controller is reliable and trustworthy. Level 4 instrument control is an adequate mechanism to ensure this. References 1. Code of Federal Regulations, Title 21, Food and Drugs, Part 11 Electronic Records; Electronic Signatures; Final Rule, Federal Register 62(54), 1342913466. 2. United States FDA, Compliance Policy Guide: 21 CFR Part 11; Electronic Records, Electronic Signatures (CPG 7153.17). 3. Good Automated Manufacturing Practice (GAMP) Special Interest Group, Complying with 21 CFR Part 11: Electronic Records and Signatures, Final Draft, September 2000. Available from the GAMP website (www.gamp.org). 4. L. Huber, Implementing 21 CFR in Analytical Laboratories, Part 1: Overview and Requirements, BioPharm 12(11), 2834 (1999). synchronize all devices required for an analysis? Could a contact closure problem lead to a situation where it goes unnoticed that a device such as a detector did not start? If the instrument has a local user interface, does it track parameter changes performed at the local interface while data are being acquired by the computer? Alternatively, is the local interface locked while data are being acquired? Does the system quickly detect power failures of a connected device or are data lost until a time-out situation occurs? Conclusion Part 11 enforcement practices have drastically emphasized the importance of reliable and traceable instrument Server Remote access Office PC Office PC Figure 1 Configuration of networked instruments in a distributed, client/server environment. 5. W. Winter and L. Huber, Implementing 21 CFR Part 11 in Analytical Laboratories, Part 2: Security Aspects for Systems and Applications, BioPharm 13(1), 4450 (2000). 6. W. Winter and L. Huber, Implementing 21 CFR Part 11 in Analytical Laboratories, Part 3: Ensuring Data Integrity in Electronic Signatures,BioPharm 13(3), 4549 (2000). 7. L. Huber and W. Winter, Implementing 21 CFR Part 11 in Analytical Laboratories, Part 4: Data Migration and Long-Term Archiving for Ready Retrieval, BioPharm 13(6), 5864 (2000). 8. W. Winter and L. Huber, Implementing 21 CFR Part 11 in Analytical Laboratories, Part 6: Biometric Identification: Limits and Possibilities, in Implementing 21 CFR Part 11 in Analytical Laboratories (2000) pp 4043, a supplement to BioPharm. 9. C. Nickel, W. Winter and L. Huber, Implementing 21 CFR Part 11 in Analytical Laboratories, Part 6: Making Legacy Systems Compliant, in Implementing 21 CFR Part 11 in Analytical Laboratories (2000) pp 4447, a supplement to BioPharm. 10. W. Winter, Dynamic Interprocess Communication between a Spectro- photometer and a Spreadsheet, Faculty for Physical Electronics, University of Karlsruhe (TH), Germany (1989). 11. M.F. Arnett et al., Understanding Basic Networking Concepts, in Inside TCP/IP (New Riders Publishing, Indianapolis, USA, 1994, ISBN 1-56205-354-X) pp 5154. 12. ANSI/IEEE Std 488.1-1987 revision of ANSI/IEEE Std 488-1978: An American National Standard IEEE Standard Digital Interface for Programmable Instrumentation. Article Reprinted from the 21CFR Part 11 supplement (March 2003) issue of: Article Reprint: 0548 Agilent Publication Number: 5988-9343 EN