Others made similar observations, and the general thrust of the comments on the proposed change was that the presently available tools to control aseptic processing do not constitute adequate confirmation to be considered validation. The arguments against the proposed revision notwithstanding, FDA made the change to the regulation. The arguments presented above were deemed inadequate, though no clear rationale or supportive scientific evidence to the contrary was offered. In the context of this larger document, the earlier objections to FDA's CGMP revision and its relevance to this draft guidance have perhaps been made clearer, and the specific difficulties more evident.

FDA indicated that media fills, environmental monitoring, routine processing controls, and satisfactory sterility tests constituted a validation in principle if not in fact (12). That position is inconsistent with the concepts presented in the draft validation guidance, which requires substantially more robust evidence of the link between input and results. The difference may seem on one level to be semantic, yet a gulf nevertheless exists between the industry's and regulators' positions with respect to validation of aseptic processing. Regardless of how one views this conundrum, one point should cause little or no confusion: the statistical component of the guidance really doesn't work with respect to linking any process parameters directly to performance.

Manual processes. Operators' skills and proficiency play a major role in the outcome of a substantial number of important pharmaceutical processes such as manual sanitization of equipment and environments and aseptic gowning. In addition, the operator's abilities may sometimes negate other controls that are present. Processes that rely heavily on operator proficiency may not be considered adequately validated, regardless of the outcome. These processes include manual cleaning, wet granulation, sugar coating, and manual inspection. It is hard to conceive that any of these could attain the level of control associated with validation. These processes are likely best considered as verified, given their heavy dependence on the operator.

A substantial amount of clarification, and perhaps broader interpretation, is needed to reconcile the draft guidance with the elements of pharmaceutical validation that do not directly result in the preparation of an intermediate material, drug substance, or drug product. In some situations, the draft guidance certainly can be used without change, but a degree of caution is necessary. Recommendations for a life-cycle approach are certainly acceptable; many pharmaceutical firms had already instituted comparable programs. It seems that the less the particular process results in or influences a material that can be analyzed or is subject to operator variability, the less useful the statistical elements of the guidance are. Without parameters and attributes that are readily quantifiable, using statistics is certainly inappropriate. The design and development experimental evidence works quite well in some instances but appears to be a force fit in other applications, largely according to whether the process outcome is numeric. Successfully linking design and development and initial and ongoing qualification for these processes also seems to depend on the extent to which process success is readily quantifiable.

The qualification and validation model provided in the draft guidance appears fully applicable to validation of products and processes but only partially applicable in other areas. Its usefulness for sterilization and aseptic processing, two of the more important pharmaceutical processes, is highly questionable. An extensive effort to provide a common level of expectations between FDA and industry (as well as within FDA and industry) is urgently needed to clarify how implementation is to be addressed. For sterile products, where FDA theory and industry practice are most divergent, an update of FDA's 1994 Guidance Submission for Sterile Products might be the best way to accommodate the desired approach with current methods for validation (13).