Page 71 - Biomedical Engineering and Design Handbook Volume 2, Applications
P. 71

50  MEDICAL DEVICE DESIGN

                         Verification means confirmation by examination and provision of objective evidence that specified
                         requirements have been fulfilled.
                                                                                 Id. § 820.3(aa).


                         While validation is defined as

                         Validation means confirmation by examination and provision of objective evidence that the particular
                         requirements for a specific intended use can be consistently fulfilled.
                         (1) Process validation means establishing by objective evidence that a process consistently produces a
                            result or product meeting its predetermined specifications.
                         (2) Design validation means establishing by objective evidence that device specifications conform with
                            user needs and intended use(s).
                                                                                 Id. § 820.3(z).

                         This shows that verification is based upon testing of results, while validation is based upon sta-
                       tistical prediction of future performance. While design controls require that design output be both
                       verified and validated, process controls follow a different algorithm. A process may be either veri-
                       fied or validated. The regulation states that if a process cannot be verified, it must be validated.
                         A process can be verified if the results of that process can be consistently tested to show confor-
                       mance with specification. For example, if a process step is to paint a part blue, it can be verified on
                       the line for every part. An inspector can hold the part up to a paint chip sample, and then look for
                       chips or missed spots. Total verification can be accomplished by inspection.
                         Many processes are more complex and defy verification. For example, a weld cannot be visually
                       inspected. It is possible that all specifications for a weld cannot be satisfied without destructive testing.
                       Since you would have no product if you destructively tested each one, you need a predictive method
                       of producing quality product. These processes must be validated.
                         You validate by determining what statistical sample size is needed to predict that quality results
                       will occur within the percentage level of success you specify. For example, how many parts do I need
                       to test to predict meeting specification at the 99 percent confidence level? The FDA default for sam-
                       pling for validation is to do three runs of 30 products each. Of course, there is flexibility based upon
                       the type of product and how you validate each process. For example, if your process is the final step
                       in producing a $10,000 medical device, you do not want to do destructive testing on 90 (3 runs of 30)
                       products. That would cost near a million dollars. You would then validate previous processes and
                       argue for a smaller sample on the final test. In any case, you need to develop a statistical rationale
                       for the sampling size that can provide a “high degree of assurance” for each process to be validated.
                       Then you need to put the designated runs of product through the process and test them for meeting
                       specification. Upon this, you can project the successful future results of the process. As in all QSR
                       activities, this validation must be conducted under a controlled written procedure.
                         Of course, this predictive validation is only valid if the process runs according to specification. You
                       must establish and maintain control of your processes to make sure that they continue to meet specifi-
                       cation. Tools such as statistical process control are usually employed to assure continuing compliance.
                         As with other things established under the QSR, a change to any process must be evaluated as to
                       its effect on the specification. Revalidation my be required. Your procedure must have a robust
                       method of evaluating changes.


                       Subpart  H: Acceptance Activities.  The acceptance activities required in subpart H suffuse the
                       entire manufacturing process, from raw material to finished product. This is not just a section for
                       accepting product from outside the company. Rather it controls acceptance of product or work at
                       every step. Subpart H has taken on new significance in the era of the virtual manufacturer. If signif-
                       icant subassemblies or whole finished product is being outsourced, it is critical that these acceptance
                       activities either be performed on receipt or delegated to the supplier under the purchasing controls
                       discussed in subpart E.
   66   67   68   69   70   71   72   73   74   75   76