Medical Devices.

Software upgrade proves nettlesome for Nova Bio

Engineer Richard Wright of the New England district found problems with design controls for software in his inspection of Nova Biomedical of Waltham, MA, but the maker of clinical chemistry and hematology analyzers handled the six-item 483 quickly enough to evade a warning letter.

The July 1999 review, only the second since 1994, precipitated an Aug. 12 reply by the firm's director of quality assurance, Paul McDonald, who assured FDA that in addition to addressing the inspectional findings, the firm would look "beyond the observations to the full quality system at Nova for the impact" of proposed corrections. Nova apparently opted not to annotate the 483 in favor of a written response.

The 483 commenced with a citation for failure of the firm's engineering change order (ECO) procedure to "specify how design changes to be made to existing marketed products/software are to be initiated, approved, documented and controlled."

Wright wrote further that the ECO neither contained nor cross- referenced other procedures regarding verification, validation and review of design changes and that it contained no criteria for determining that validation could be bypassed in favor of design verification.

Software changes not validated?

The EIR noted changes made to software installed in the firm's Stat Profile M blood gas analyzer in 1998 and 1999, intended to improve on existing requirements and/or add new requirements.

Wright commented that while Nova had rewritten its ECO in July 1999, neither that version nor its predecessor covered the elements cited in the 483.

McDonald's August 12 response to the 483 stated that Nova had revised design control procedures to specify the requirements for initiating design changes "to existing products/software" as well as to describe how design changes would be "initiated, approved, documented and controlled."

He added that the new procedure for engineering change orders would cross-reference the design control procedure and that the design control procedure laid out steps for establishing whether verification would suffice in lieu of validation.

According to McDonald, the design control procedure had previously been known internally as a "product development procedure" and noted that the engineering change procedure now required documentation of all changes to software. The 483 cited Nova for failure to establish design input procedures for changes to the software used in the Stat Profile M and insisted that Nova had no "documented mechanisms for addressing incomplete, ambiguous or conflicting requirements."

The Checklist

Nova Biomedical,

New England District

Design controls

Software validation This finding again addressed software in the Stat Profile M, and in the EIR, Wright commented little more other than to note that there were ten input requirements for a purged version of the software. One of these was "the capability to enter BUN at the patient data screen when measured BUN is not available," and information regarding a second input was purged from the EIR.

In addressing this finding, McDonald admitted that Nova had not previously covered these design input issues, but indicated an enclosure of the new ECO that he stated "provides a mechanism" for reviewing changes for incomplete, ambiguous or conflicting requirements.

Later on the 483, Wright indicated that Nova had not established design validation procedures for the Stat Profile M software and added that "the results of the design validation activities that were reportedly conducted were not maintained/documented."

In the EIR, he commented that the company had documented ten inputs, of which six were altered going into version 2.16. The EIR stated that one verification procedure for version 2.17 involved scanning a bar code, an exit from that patient's screen, followed by scans of other bar codes.

According to the EIR, this procedure triggered a reset of the software in version 2.16 and version 2.17 included a fix for this as well as four other changes. Wright indicated no problems with the verification work other than that Nova failed to document the work.

In his letter, McDonald stated that Nova had procedures covering this type of work, but had not "fully followed" them. He stated that changes to the ECO procedure addressed this finding, then commented that the company had a validation protocol for a purged version of the software "which was prepared after the validation," at which time "the protocol was conducted and complete successfully." McDonald confessed that nobody had reviewed and approved the validation protocol.

McDonald wrote further that Nova had retrospectively run the software through a hazard analysis and design validation after the inspection, including an review of field experience data which "confirmed that there are no safety or effectiveness issues associated" with the changes and that Nova did not encounter anything suggestive of problems related to patient safety or performance. At press time, Nova had not responded to calls for comment.

Nova Biomedical, Waltham, MA, 7/20-23/99, Doc. 108989M, $7.00 plus retrieval.