General Principles of Software Validation; Final Guidance for Industry and FDA Staff

• Date: October 24, 2007
• Source: www.fda.gov

Activities and Tasks

SECTION 5. ACTIVITIES AND TASKS

Software validation is accomplished through a series of activities and tasks that are planned and executed at various stages of the software development life cycle. These tasks may be one time occurrences or may be iterated many times, depending on the life cycle model used and the scope of changes made as the software project progresses.

5.1. SOFTWARE LIFE CYCLE ACTIVITIES

This guidance does not recommend the use of any specific software life cycle model. Software developers should establish a software life cycle model that is appropriate for their product and organization. The software life cycle model that is selected should cover the software from its birth to its retirement. Activities in a typical software life cycle model include the following:

• Quality Planning
• System Requirements Definition
• Detailed Software Requirements Specification
• Software Design Specification
• Construction or Coding
• Testing
• Installation
• Operation and Support
• Maintenance
• Retirement

Verification, testing, and other tasks that support software validation occur during each of these activities. A life cycle model organizes these software development activities in various ways and provides a framework for monitoring and controlling the software development project. Several software life cycle models (e.g., waterfall, spiral, rapid prototyping, incremental development, etc.) are defined in FDA's Glossary of Computerized System and Software Development Terminology, dated August 1995. These and many other life cycle models are described in various references listed in Appendix A.

5.2. TYPICAL TASKS SUPPORTING VALIDATION

For each of the software life cycle activities, there are certain "typical" tasks that support a conclusion that the software is validated. However, the specific tasks to be performed, their order of performance, and the iteration and timing of their performance will be dictated by the specific software life cycle model that is selected and the safety risk associated with the software application. For very low risk applications, certain tasks may not be needed at all. However, the software developer should at least consider each of these tasks and should define and document which tasks are or are not appropriate for their specific application. The following discussion is generic and is not intended to prescribe any particular software life cycle model or any particular order in which tasks are to be performed.

5.2.1. Quality Planning

Design and development planning should culminate in a plan that identifies necessary tasks, procedures for anomaly reporting and resolution, necessary resources, and management review requirements, including formal design reviews. A software life cycle model and associated activities should be identified, as well as those tasks necessary for each software life cycle activity. The plan should include:

• The specific tasks for each life cycle activity;
• Enumeration of important quality factors (e.g., reliability, maintainability, and usability);
• Methods and procedures for each task;
• Task acceptance criteria;
• Criteria for defining and documenting outputs in terms that will allow evaluation of their conformance to input requirements;
• Inputs for each task;
• Outputs from each task;
• Roles, resources, and responsibilities for each task;
• Risks and assumptions; and
• Documentation of user needs.

Management must identify and provide the appropriate software development environment and resources. (See 21 CFR §820.20(b)(1) and (2).) Typically, each task requires personnel as well as physical resources. The plan should identify the personnel, the facility and equipment resources for each task, and the role that risk (hazard) management will play. A configuration management plan should be developed that will guide and control multiple parallel development activities and ensure proper communications and documentation. Controls are necessary to ensure positive and correct correspondence among all approved versions of the specifications documents, source code, object code, and test suites that comprise a software system. The controls also should ensure accurate identification of, and access to, the currently approved versions.

Procedures should be created for reporting and resolving software anomalies found through validation or other activities. Management should identify the reports and specify the contents, format, and responsible organizational elements for each report. Procedures also are necessary for the review and approval of software development results, including the responsible organizational elements for such reviews and approvals.

Typical Tasks - Quality Planning

• Risk (Hazard) Management Plan
• Configuration Management Plan
• Software Quality Assurance Plan
• - Software Verification and Validation Plan
• Verification and Validation Tasks, and Acceptance Criteria
• Schedule and Resource Allocation (for software verification and validation activities)
• Reporting Requirements

• - Formal Design Review Requirements
• - Other Technical Review Requirements

• Problem Reporting and Resolution Procedures
• Other Support Activities

5.2.2. Requirements

Requirements development includes the identification, analysis, and documentation of information about the device and its intended use. Areas of special importance include allocation of system functions to hardware/software, operating conditions, user characteristics, potential hazards, and anticipated tasks. In addition, the requirements should state clearly the intended use of the software.

The software requirements specification document should contain a written definition of the software functions. It is not possible to validate software without predetermined and documented software requirements. Typical software requirements specify the following:

• All software system inputs;
• All software system outputs;
• All functions that the software system will perform;
• All performance requirements that the software will meet, (e.g., data throughput, reliability, and timing);
• The definition of all external and user interfaces, as well as any internal software-to-system interfaces;
• How users will interact with the system;
• What constitutes an error and how errors should be handled;
• Required response times;
• The intended operating environment for the software, if this is a design constraint (e.g., hardware platform, operating system);
• All ranges, limits, defaults, and specific values that the software will accept; and
• All safety related requirements, specifications, features, or functions that will be implemented in software.

Software safety requirements are derived from a technical risk management process that is closely integrated with the system requirements development process. Software requirement specifications should identify clearly the potential hazards that can result from a software failure in the system as well as any safety requirements to be implemented in software. The consequences of software failure should be evaluated, along with means of mitigating such failures (e.g., hardware mitigation, defensive programming, etc.). From this analysis, it should be possible to identify the most appropriate measures necessary to prevent harm.

The Quality System regulation requires a mechanism for addressing incomplete, ambiguous, or conflicting requirements. (See 21 CFR 820.30(c).) Each requirement (e.g., hardware, software, user, operator interface, and safety) identified in the software requirements specification should be evaluated for accuracy, completeness, consistency, testability, correctness, and clarity. For example, software requirements should be evaluated to verify that:

• There are no internal inconsistencies among requirements;
• All of the performance requirements for the system have been spelled out;
• Fault tolerance, safety, and security requirements are complete and correct;
• Allocation of software functions is accurate and complete;
• Software requirements are appropriate for the system hazards; and
• All requirements are expressed in terms that are measurable or objectively verifiable.

A software requirements traceability analysis should be conducted to trace software requirements to (and from) system requirements and to risk analysis results. In addition to any other analyses and documentation used to verify software requirements, a formal design review is recommended to confirm that requirements are fully specified and appropriate before extensive software design efforts begin. Requirements can be approved and released incrementally, but care should be taken that interactions and interfaces among software (and hardware) requirements are properly reviewed, analyzed, and controlled.

Typical Tasks - Requirements

• Preliminary Risk Analysis
• Traceability Analysis
• -  Software Requirements to System Requirements (and vice versa)
• -  Software Requirements to Risk Analysis

• Description of User Characteristics
• Listing of Characteristics and Limitations of Primary and Secondary Memory
• Software Requirements Evaluation
• Software User Interface Requirements Analysis
• System Test Plan Generation
• Acceptance Test Plan Generation
• Ambiguity Review or Analysis

5.2.3. Design

In the design process, the software requirements specification is translated into a logical and physical representation of the software to be implemented. The software design specification is a description of what the software should do and how it should do it. Due to complexity of the project or to enable persons with varying levels of technical responsibilities to clearly understand design information, the design specification may contain both a high level summary of the design and detailed design information. The completed software design specification constrains the programmer/coder to stay within the intent of the agreed upon requirements and design. A complete software design specification will relieve the programmer from the need to make ad hoc design decisions.

The software design needs to address human factors. Use error caused by designs that are either overly complex or contrary to users' intuitive expectations for operation is one of the most persistent and critical problems encountered by FDA. Frequently, the design of the software is a factor in such use errors. Human factors engineering should be woven into the entire design and development process, including the device design requirements, analyses, and tests. Device safety and usability issues should be considered when developing flowcharts, state diagrams, prototyping tools, and test plans. Also, task and function analyses, risk analyses, prototype tests and reviews, and full usability tests should be performed. Participants from the user population should be included when applying these methodologies.

The software design specification should include:

• Software requirements specification, including predetermined criteria for acceptance of the software;
• Software risk analysis;
• Development procedures and coding guidelines (or other programming procedures);
• Systems documentation (e.g., a narrative or a context diagram) that describes the systems context in which the program is intended to function, including the relationship of hardware, software, and the physical environment;
• Hardware to be used;
• Parameters to be measured or recorded;
• Logical structure (including control logic) and logical processing steps (e.g., algorithms);
• Data structures and data flow diagrams;
• Definitions of variables (control and data) and description of where they are used;
• Error, alarm, and warning messages;
• Supporting software (e.g., operating systems, drivers, other application software);
• Communication links (links among internal modules of the software, links with the supporting software, links with the hardware, and links with the user);
• Security measures (both physical and logical security); and
• Any additional constraints not identified in the above elements.

The first four of the elements noted above usually are separate pre-existing documents that are included by reference in the software design specification. Software requirements specification was discussed in the preceding section, as was software risk analysis. Written development procedures serve as a guide to the organization, and written programming procedures serve as a guide to individual programmers. As software cannot be validated without knowledge of the context in which it is intended to function, systems documentation is referenced. If some of the above elements are not included in the software, it may be helpful to future reviewers and maintainers of the software if that is clearly stated (e.g., There are no error messages in this program).

The activities that occur during software design have several purposes. Software design evaluations are conducted to determine if the design is complete, correct, consistent, unambiguous, feasible, and maintainable. Appropriate consideration of software architecture (e.g., modular structure) during design can reduce the magnitude of future validation efforts when software changes are needed. Software design evaluations may include analyses of control flow, data flow, complexity, timing, sizing, memory allocation, criticality analysis, and many other aspects of the design. A traceability analysis should be conducted to verify that the software design implements all of the software requirements. As a technique for identifying where requirements are not sufficient, the traceability analysis should also verify that all aspects of the design are traceable to software requirements. An analysis of communication links should be conducted to evaluate the proposed design with respect to hardware, user, and related software requirements. The software risk analysis should be re-examined to determine whether any additional hazards have been identified and whether any new hazards have been introduced by the design.

At the end of the software design activity, a Formal Design Review should be conducted to verify that the design is correct, consistent, complete, accurate, and testable, before moving to implement the design. Portions of the design can be approved and released incrementally for implementation; but care should be taken that interactions and communication links among various elements are properly reviewed, analyzed, and controlled.

Most software development models will be iterative. This is likely to result in several versions of both the software requirement specification and the software design specification. All approved versions should be archived and controlled in accordance with established configuration management procedures.

Typical Tasks - Design

• Updated Software Risk Analysis
• Traceability Analysis - Design Specification to Software Requirements (and vice versa)
• Software Design Evaluation
• Design Communication Link Analysis
• Module Test Plan Generation
• Integration Test Plan Generation
• Test Design Generation (module, integration, system, and acceptance)

5.2.4. Construction or Coding

Software may be constructed either by coding (i.e., programming) or by assembling together previously coded software components (e.g., from code libraries, off-the-shelf software, etc.) for use in a new application. Coding is the software activity where the detailed design specification is implemented as source code. Coding is the lowest level of abstraction for the software development process. It is the last stage in decomposition of the software requirements where module specifications are translated into a programming language.

Coding usually involves the use of a high-level programming language, but may also entail the use of assembly language (or microcode) for time-critical operations. The source code may be either compiled or interpreted for use on a target hardware platform. Decisions on the selection of programming languages and software build tools (assemblers, linkers, and compilers) should include consideration of the impact on subsequent quality evaluation tasks (e.g., availability of debugging and testing tools for the chosen language). Some compilers offer optional levels and commands for error checking to assist in debugging the code. Different levels of error checking may be used throughout the coding process, and warnings or other messages from the compiler may or may not be recorded. However, at the end of the coding and debugging process, the most rigorous level of error checking is normally used to document what compilation errors still remain in the software. If the most rigorous level of error checking is not used for final translation of the source code, then justification for use of the less rigorous translation error checking should be documented. Also, for the final compilation, there should be documentation of the compilation process and its outcome, including any warnings or other messages from the compiler and their resolution, or justification for the decision to leave issues unresolved.

Firms frequently adopt specific coding guidelines that establish quality policies and procedures related to the software coding process. Source code should be evaluated to verify its compliance with specified coding guidelines. Such guidelines should include coding conventions regarding clarity, style, complexity management, and commenting. Code comments should provide useful and descriptive information for a module, including expected inputs and outputs, variables referenced, expected data types, and operations to be performed. Source code should also be evaluated to verify its compliance with the corresponding detailed design specification. Modules ready for integration and test should have documentation of compliance with coding guidelines and any other applicable quality policies and procedures.

Source code evaluations are often implemented as code inspections and code walkthroughs. Such static analyses provide a very effective means to detect errors before execution of the code. They allow for examination of each error in isolation and can also help in focusing later dynamic testing of the software. Firms may use manual (desk) checking with appropriate controls to ensure consistency and independence. Source code evaluations should be extended to verification of internal linkages between modules and layers (horizontal and vertical interfaces), and compliance with their design specifications. Documentation of the procedures used and the results of source code evaluations should be maintained as part of design verification.

A source code traceability analysis is an important tool to verify that all code is linked to established specifications and established test procedures. A source code traceability analysis should be conducted and documented to verify that:

• Each element of the software design specification has been implemented in code;
• Modules and functions implemented in code can be traced back to an element in the software design specification and to the risk analysis;
• Tests for modules and functions can be traced back to an element in the software design specification and to the risk analysis; and
• Tests for modules and functions can be traced to source code for the same modules and functions.

Typical Tasks - Construction or Coding

• Traceability Analyses
• - Source Code to Design Specification (and vice versa)
• - Test Cases to Source Code and to Design Specification

• Source Code and Source Code Documentation Evaluation
• Source Code Interface Analysis
• Test Procedure and Test Case Generation (module, integration, system, and acceptance)

5.2.5. Testing by the Software Developer

Software testing entails running software products under known conditions with defined inputs and documented outcomes that can be compared to their predefined expectations. It is a time consuming, difficult, and imperfect activity. As such, it requires early planning in order to be effective and efficient.

Test plans and test cases should be created as early in the software development process as feasible. They should identify the schedules, environments, resources (personnel, tools, etc.), methodologies, cases (inputs, procedures, outputs, expected results), documentation, and reporting criteria. The magnitude of effort to be applied throughout the testing process can be linked to complexity, criticality, reliability, and/or safety issues (e.g., requiring functions or modules that produce critical outcomes to be challenged with intensive testing of their fault tolerance features). Descriptions of categories of software and software testing effort appear in the literature, for example:

• NIST Special Publication 500-235, Structured Testing: A Testing Methodology Using the Cyclomatic Complexity Metric;
• NUREG/CR-6293, Verification and Validation Guidelines for High Integrity Systems; and
• IEEE Computer Society Press, Handbook of Software Reliability Engineering.

Software test plans should identify the particular tasks to be conducted at each stage of development and include justification of the level of effort represented by their corresponding completion criteria.

Software testing has limitations that must be recognized and considered when planning the testing of a particular software product. Except for the simplest of programs, software cannot be exhaustively tested. Generally it is not feasible to test a software product with all possible inputs, nor is it possible to test all possible data processing paths that can occur during program execution. There is no one type of testing or testing methodology that can ensure a particular software product has been thoroughly tested. Testing of all program functionality does not mean all of the program has been tested. Testing of all of a program's code does not mean all necessary functionality is present in the program. Testing of all program functionality and all program code does not mean the program is 100% correct! Software testing that finds no errors should not be interpreted to mean that errors do not exist in the software product; it may mean the testing was superficial.

An essential element of a software test case is the expected result. It is the key detail that permits objective evaluation of the actual test result. This necessary testing information is obtained from the corresponding, predefined definition or specification. A software specification document must identify what, when, how, why, etc., is to be achieved with an engineering (i.e., measurable or objectively verifiable) level of detail in order for it to be confirmed through testing. The real effort of effective software testing lies in the definition of what is to be tested rather than in the performance of the test.

A software testing process should be based on principles that foster effective examinations of a software product. Applicable software testing tenets include:

• The expected test outcome is predefined;
• A good test case has a high probability of exposing an error;
• A successful test is one that finds an error;
• There is independence from coding;
• Both application (user) and software (programming) expertise are employed;
• Testers use different tools from coders;
• Examining only the usual case is insufficient;
• Test documentation permits its reuse and an independent confirmation of the pass/fail status of a test outcome during subsequent review.

Once the prerequisite tasks (e.g., code inspection) have been successfully completed, software testing begins. It starts with unit level testing and concludes with system level testing. There may be a distinct integration level of testing. A software product should be challenged with test cases based on its internal structure and with test cases based on its external specification. These tests should provide a thorough and rigorous examination of the software product's compliance with its functional, performance, and interface definitions and requirements.

Code-based testing is also known as structural testing or "white-box" testing. It identifies test cases based on knowledge obtained from the source code, detailed design specification, and other development documents. These test cases challenge the control decisions made by the program; and the program's data structures including configuration tables. Structural testing can identify "dead" code that is never executed when the program is run. Structural testing is accomplished primarily with unit (module) level testing, but can be extended to other levels of software testing.

The level of structural testing can be evaluated using metrics that are designed to show what percentage of the software structure has been evaluated during structural testing. These metrics are typically referred to as "coverage" and are a measure of completeness with respect to test selection criteria. The amount of structural coverage should be commensurate with the level of risk posed by the software. Use of the term "coverage" usually means 100% coverage. For example, if a testing program has achieved "statement coverage," it means that 100% of the statements in the software have been executed at least once. Common structural coverage metrics include:

• Statement Coverage - This criteria requires sufficient test cases for each program statement to be executed at least once; however, its achievement is insufficient to provide confidence in a software product's behavior.
• Decision (Branch) Coverage - This criteria requires sufficient test cases for each program decision or branch to be executed so that each possible outcome occurs at least once. It is considered to be a minimum level of coverage for most software products, but decision coverage alone is insufficient for high-integrity applications.
• Condition Coverage - This criteria requires sufficient test cases for each condition in a program decision to take on all possible outcomes at least once. It differs from branch coverage only when multiple conditions must be evaluated to reach a decision.
• Multi-Condition Coverage - This criteria requires sufficient test cases to exercise all possible combinations of conditions in a program decision.
• Loop Coverage - This criteria requires sufficient test cases for all program loops to be executed for zero, one, two, and many iterations covering initialization, typical running and termination (boundary) conditions.
• Path Coverage - This criteria requires sufficient test cases for each feasible path, basis path, etc., from start to exit of a defined program segment, to be executed at least once. Because of the very large number of possible paths through a software program, path coverage is generally not achievable. The amount of path coverage is normally established based on the risk or criticality of the software under test.
• Data Flow Coverage - This criteria requires sufficient test cases for each feasible data flow to be executed at least once. A number of data flow testing strategies are available.

Definition-based or specification-based testing is also known as functional testing or "black-box" testing. It identifies test cases based on the definition of what the software product (whether it be a unit (module) or a complete program) is intended to do. These test cases challenge the intended use or functionality of a program, and the program's internal and external interfaces. Functional testing can be applied at all levels of software testing, from unit to system level testing.

The following types of functional software testing involve generally increasing levels of effort:

• Normal Case - Testing with usual inputs is necessary. However, testing a software product only with expected, valid inputs does not thoroughly test that software product. By itself, normal case testing cannot provide sufficient confidence in the dependability of the software product.
• Output Forcing - Choosing test inputs to ensure that selected (or all) software outputs are generated by testing.
• Robustness - Software testing should demonstrate that a software product behaves correctly when given unexpected, invalid inputs. Methods for identifying a sufficient set of such test cases include Equivalence Class Partitioning, Boundary Value Analysis, and Special Case Identification (Error Guessing). While important and necessary, these techniques do not ensure that all of the most appropriate challenges to a software product have been identified for testing.
• Combinations of Inputs - The functional testing methods identified above all emphasize individual or single test inputs. Most software products operate with multiple inputs under their conditions of use. Thorough software product testing should consider the combinations of inputs a software unit or system may encounter during operation. Error guessing can be extended to identify combinations of inputs, but it is an ad hoc technique. Cause-effect graphing is one functional software testing technique that systematically identifies combinations of inputs to a software product for inclusion in test cases.

Functional and structural software test case identification techniques provide specific inputs for testing, rather than random test inputs. One weakness of these techniques is the difficulty in linking structural and functional test completion criteria to a software product's reliability. Advanced software testing methods, such as statistical testing, can be employed to provide further assurance that a software product is dependable. Statistical testing uses randomly generated test data from defined distributions based on an operational profile (e.g., expected use, hazardous use, or malicious use of the software product). Large amounts of test data are generated and can be targeted to cover particular areas or concerns, providing an increased possibility of identifying individual and multiple rare operating conditions that were not anticipated by either the software product's designers or its testers. Statistical testing also provides high structural coverage. It does require a stable software product. Thus, structural and functional testing are prerequisites for statistical testing of a software product.

Another aspect of software testing is the testing of software changes. Changes occur frequently during software development. These changes are the result of 1) debugging that finds an error and it is corrected, 2) new or changed requirements ("requirements creep"), and 3) modified designs as more effective or efficient implementations are found. Once a software product has been baselined (approved), any change to that product should have its own "mini life cycle," including testing. Testing of a changed software product requires additional effort. Not only should it demonstrate that the change was implemented correctly, testing should also demonstrate that the change did not adversely impact other parts of the software product. Regression analysis and testing are employed to provide assurance that a change has not created problems elsewhere in the software product. Regression analysis is the determination of the impact of a change based on review of the relevant documentation (e.g., software requirements specification, software design specification, source code, test plans, test cases, test scripts, etc.) in order to identify the necessary regression tests to be run. Regression testing is the rerunning of test cases that a program has previously executed correctly and comparing the current result to the previous result in order to detect unintended effects of a software change. Regression analysis and regression testing should also be employed when using integration methods to build a software product to ensure that newly integrated modules do not adversely impact the operation of previously integrated modules.

In order to provide a thorough and rigorous examination of a software product, development testing is typically organized into levels. As an example, a software product's testing can be organized into unit, integration, and system levels of testing.

1)	Unit (module or component) level testing focuses on the early examination of sub-program functionality and ensures that functionality not visible at the system level is examined by testing. Unit testing ensures that quality software units are furnished for integration into the finished software product.
2)	Integration level testing focuses on the transfer of data and control across a program's internal and external interfaces. External interfaces are those with other software (including operating system software), system hardware, and the users and can be described as communications links.
3)	System level testing demonstrates that all specified functionality exists and that the software product is trustworthy. This testing verifies the as-built program's functionality and performance with respect to the requirements for the software product as exhibited on the specified operating platform(s). System level software testing addresses functional concerns and the following elements of a device's software that are related to the intended use(s):

• Performance issues (e.g., response times, reliability measurements);
• Responses to stress conditions, e.g., behavior under maximum load, continuous use;
• Operation of internal and external security features;
• Effectiveness of recovery procedures, including disaster recovery;
• Usability;
• Compatibility with other software products;
• Behavior in each of the defined hardware configurations; and
• Accuracy of documentation.

Control measures (e.g., a traceability analysis) should be used to ensure that the intended coverage is achieved.

System level testing also exhibits the software product's behavior in the intended operating environment. The location of such testing is dependent upon the software developer's ability to produce the target operating environment(s). Depending upon the circumstances, simulation and/or testing at (potential) customer locations may be utilized. Test plans should identify the controls needed to ensure that the intended coverage is achieved and that proper documentation is prepared when planned system level testing is conducted at sites not directly controlled by the software developer. Also, for a software product that is a medical device or a component of a medical device that is to be used on humans prior to FDA clearance, testing involving human subjects may require an Investigational Device Exemption (IDE) or Institutional Review Board (IRB) approval.

Test procedures, test data, and test results should be documented in a manner permitting objective pass/fail decisions to be reached. They should also be suitable for review and objective decision making subsequent to running the test, and they should be suitable for use in any subsequent regression testing. Errors detected during testing should be logged, classified, reviewed, and resolved prior to release of the software. Software error data that is collected and analyzed during a development life cycle may be used to determine the suitability of the software product for release for commercial distribution. Test reports should comply with the requirements of the corresponding test plans.

Software products that perform useful functions in medical devices or their production are often complex. Software testing tools are frequently used to ensure consistency, thoroughness, and efficiency in the testing of such software products and to fulfill the requirements of the planned testing activities. These tools may include supporting software built in-house to facilitate unit (module) testing and subsequent integration testing (e.g., drivers and stubs) as well as commercial software testing tools. Such tools should have a degree of quality no less than the software product they are used to develop. Appropriate documentation providing evidence of the validation of these software tools for their intended use should be maintained (see section 6 of this guidance).

Typical Tasks - Testing by the Software Developer

• Test Planning
• Structural Test Case Identification
• Functional Test Case Identification
• Traceability Analysis - Testing
• - Unit (Module) Tests to Detailed Design
• - Integration Tests to High Level Design
• - System Tests to Software Requirements

• Unit (Module) Test Execution
• Integration Test Execution
• Functional Test Execution
• System Test Execution
• Acceptance Test Execution
• Test Results Evaluation
• Error Evaluation/Resolution
• Final Test Report

5.2.6. User Site Testing

Testing at the user site is an essential part of software validation. The Quality System regulation requires installation and inspection procedures (including testing where appropriate) as well as documentation of inspection and testing to demonstrate proper installation. (See 21 CFR §820.170.) Likewise, manufacturing equipment must meet specified requirements, and automated systems must be validated for their intended use. (See 21 CFR §820.70(g) and 21 CFR §820.70(i) respectively.)

Terminology regarding user site testing can be confusing. Terms such as beta test, site validation, user acceptance test, installation verification, and installation testing have all been used to describe user site testing. For purposes of this guidance, the term "user site testing" encompasses all of these and any other testing that takes place outside of the developer's controlled environment. This testing should take place at a user's site with the actual hardware and software that will be part of the installed system configuration. The testing is accomplished through either actual or simulated use of the software being tested within the context in which it is intended to function.

Guidance contained here is general in nature and is applicable to any user site testing. However, in some areas (e.g., blood establishment systems) there may be specific site validation issues that need to be considered in the planning of user site testing. Test planners should check with the FDA Center(s) with the corresponding product jurisdiction to determine whether there are any additional regulatory requirements for user site testing.

User site testing should follow a pre-defined written plan with a formal summary of testing and a record of formal acceptance. Documented evidence of all testing procedures, test input data, and test results should be retained.

There should be evidence that hardware and software are installed and configured as specified. Measures should ensure that all system components are exercised during the testing and that the versions of these components are those specified. The testing plan should specify testing throughout the full range of operating conditions and should specify continuation for a sufficient time to allow the system to encounter a wide spectrum of conditions and events in an effort to detect any latent faults that are not apparent during more normal activities.

Some of the evaluations that have been performed earlier by the software developer at the developer's site should be repeated at the site of actual use. These may include tests for a high volume of data, heavy loads or stresses, security, fault testing (avoidance, detection, tolerance, and recovery), error messages, and implementation of safety requirements. The developer may be able to furnish the user with some of the test data sets to be used for this purpose.

In addition to an evaluation of the system's ability to properly perform its intended functions, there should be an evaluation of the ability of the users of the system to understand and correctly interface with it. Operators should be able to perform the intended functions and respond in an appropriate and timely manner to all alarms, warnings, and error messages.

During user site testing, records should be maintained of both proper system performance and any system failures that are encountered. The revision of the system to compensate for faults detected during this user site testing should follow the same procedures and controls as for any other software change.

The developers of the software may or may not be involved in the user site testing. If the developers are involved, they may seamlessly carry over to the user's site the last portions of design-level systems testing. If the developers are not involved, it is all the more important that the user have persons who understand the importance of careful test planning, the definition of expected test results, and the recording of all test outputs.

Typical Tasks - User Site Testing

• Acceptance Test Execution
• Test Results Evaluation
• Error Evaluation/Resolution
• Final Test Report

5.2.7. Maintenance and Software Changes

As applied to software, the term maintenance does not mean the same as when applied to hardware. The operational maintenance of hardware and software are different because their failure/error mechanisms are different. Hardware maintenance typically includes preventive hardware maintenance actions, component replacement, and corrective changes. Software maintenance includes corrective, perfective, and adaptive maintenance but does not include preventive maintenance actions or software component replacement.

Changes made to correct errors and faults in the software are corrective maintenance. Changes made to the software to improve the performance, maintainability, or other attributes of the software system are perfective maintenance. Software changes to make the software system usable in a changed environment are adaptive maintenance.

When changes are made to a software system, either during initial development or during post release maintenance, sufficient regression analysis and testing should be conducted to demonstrate that portions of the software not involved in the change were not adversely impacted. This is in addition to testing that evaluates the correctness of the implemented change(s).

The specific validation effort necessary for each software change is determined by the type of change, the development products affected, and the impact of those products on the operation of the software. Careful and complete documentation of the design structure and interrelationships of various modules, interfaces, etc., can limit the validation effort needed when a change is made. The level of effort needed to fully validate a change is also dependent upon the degree to which validation of the original software was documented and archived. For example, test documentation, test cases, and results of previous verification and validation testing need to be archived if they are to be available for performing subsequent regression testing. Failure to archive this information for later use can significantly increase the level of effort and expense of revalidating the software after a change is made.

In addition to software verification and validation tasks that are part of the standard software development process, the following additional maintenance tasks should be addressed:

• Software Validation Plan Revision - For software that was previously validated, the existing software validation plan should be revised to support the validation of the revised software. If no previous software validation plan exists, such a plan should be established to support the validation of the revised software.
• Anomaly Evaluation - Software organizations frequently maintain documentation, such as software problem reports that describe software anomalies discovered and the specific corrective action taken to fix each anomaly. Too often, however, mistakes are repeated because software developers do not take the next step to determine the root causes of problems and make the process and procedural changes needed to avoid recurrence of the problem. Software anomalies should be evaluated in terms of their severity and their effects on system operation and safety, but they should also be treated as symptoms of process deficiencies in the quality system. A root cause analysis of anomalies can identify specific quality system deficiencies. Where trends are identified (e.g., recurrence of similar software anomalies), appropriate corrective and preventive actions must be implemented and documented to avoid further recurrence of similar quality problems. (See 21 CFR 820.100.)
• Problem Identification and Resolution Tracking - All problems discovered during maintenance of the software should be documented. The resolution of each problem should be tracked to ensure it is fixed, for historical reference, and for trending.
• Proposed Change Assessment - All proposed modifications, enhancements, or additions should be assessed to determine the effect each change would have on the system. This information should determine the extent to which verification and/or validation tasks need to be iterated.
• Task Iteration - For approved software changes, all necessary verification and validation tasks should be performed to ensure that planned changes are implemented correctly, all documentation is complete and up to date, and no unacceptable changes have occurred in software performance.
• Documentation Updating - Documentation should be carefully reviewed to determine which documents have been impacted by a change. All approved documents (e.g., specifications, test procedures, user manuals, etc.) that have been affected should be updated in accordance with configuration management procedures. Specifications should be updated before any maintenance and software changes are made.