It is high time that the medical device industry adopt statistical validation methods to improve product functionality, effectiveness, and safety.
By Steve Gilder
Driven by necessity, such high-profile sectors as the automotive, aerospace, and electronics industries have significantly improved their overall statistical productivity while sharply reducing part-to-part variation and manufacturing costs. These reductions are the result of acceptance criteria that have shifted from the traditional quality control final inspection modality to more-continuous, contiguous, and effective in-line process validation. More recently, this shift has been applied successfully in the medical device manufacturing industry as well. To continue this trend, medical device suppliers must learn how to manage manufacturing acceptance criteria for both high- and low-volume medical components and devices.
Based on new technology, the shift from a final inspection to an inline process validation methodology is largely an evolutionary one. Over the past 10 to 15 years, production tools have changed dramatically, as have the statistical and measurement techniques used to qualify processes and inspection equipment. As new tooling is designed and launched, inspection equipment must follow suit. In other words, the evolution of production tooling and statistical and measurement techniques must occur in parallel. Thus, traditional reliance on final inspection sampling methods has been replaced in many sectors by statistical analysis of processes—that is, statistical process control and capability techniques, including productivity measurements. This change is continuing to unfold today.
|Capaibility analysis by molded part and cavity.|
As demonstrated outside of the medical device industry, the shift from final inspection sampling methods to statistical process analysis makes sense for several reasons. First, by performing high-level equipment qualification, operational qualification, and performance qualification (IQ/OQ/PQ) runs at every critical stage of a new product launch—including molding, assembly, functionality testing, printing, packaging, and shipping—manufacturers can ensure overall quality while reducing the amount of inspection required in both dimensional and attribute characteristics. Second, statistical process analysis methods can identify product defects earlier, reduce process variation significantly, and enhance product throughput dramatically throughout the manufacturing process—even before products have reached their full production volumes.
As in-line statistical models replace final inspection methods, manufacturers are employing less labor and more science—including automated vision inspection technology—to meet the demands of customers who now require higher statistical confidence in product reliability and reproducibility. Also, the use of initial process qualifications and in-line statistical process controls—including defect detection systems—is allowing manufacturers to rely more on their suppliers’ data in making lot-acceptance decisions and helping customers to reduce their incoming inspection levels. A game-changing shift, the turn toward process qualifications and statistical process control allows manufacturers to minimize and control product anomalies earlier in the manufacturing process, thereby increasing customer satisfaction.
The adoption of in-line control methods has been supported by manufacturers’ increasing ability to statistically analyze product qualification or validation data—an analysis that occurs at several stages during the production process. In the molding industry, for example, product, process, and quality engineers collaborate upstream to determine the critical characteristics required to achieve both functionality and manufacturability. Then, they focus on the various process parameters that affect critical dimensions, including pressure, time, temperature, and injection speed. This procedure ensures that several secondary factors—such as cure times, melting points, cross-linking, and crystallization—remain stable and optimal throughout the molding process.
During process qualifications, both the functionality and processability of each new product introduction must be considered. If a product works to design when it is assembled by hand but fails to work optimally during the feeding or assembly process, its functionality as a marketable product is compromised. A statistics-driven model minimizes this risk. As more details about a product become known early in the production process, fewer critical issues are likely to develop downstream, reducing quality concerns.
Evaluating Statistical Capability
The transition to inline statistical analysis in the medical device industry can be seen in the evolution of molding processes. A decade ago, customers of molded components often required only a First Article Inspection (FAI) report involving a limited number of samples per cavity with a minimum number of dimensions. The use of CAD models was limited, while molds were completed only according to drawing requirements and operator-influenced examinations using manual measurements predominated. Manufacturers needed a more diligent, streamlined, computerized methodology to derive benefits from this system.
|Injection molding machines equipped with end-of-arm tooling robotics can segregate parts from individual mold cavities (32 cavities in this case), allowing for the measurement of part-to-part variation to determine process capabilities.
Today, two indices are used as indicators of how well a process meets specifications: the process capability index (Cp), which measures short-term variability of part dimensions from individual mold cavities, and the process performance index (Pp), which measures long-term variability. These indices compare the part-dimension variability of an in-control and stable production process with its engineered specifications. These capability indices are generated to measure process performance versus customer requirements. A process is said to be capable if the range of the specification is wider than the actual short- and long-term production process variations.
In addition to developing key capability data, the measurement system itself must be evaluated using a statistical analysis method known as analysis of variance (ANOVA). Gauge repeatability and reproducibility (GR&R) must also be measured to determine the variability caused by the measurement system. While repeatability determines variations among the same or replicate parts processed by a single operator or instrument under the same conditions, reproducibility determines measurement variations among the same or replicate parts handled by different test operators, instruments, or test facilities. Once the measurement system has been verified using ANOVA and GR&R, there is high confidence that identical conclusions will be reached although different companies, departments, instruments, and test operators have measured the same part. Measurement studies are useless if no meaning is assigned to the data that are collected.
Throughout this process, statisticians must be cognizant that any data acquired are meaningless if they do not draw correct conclusions from them. For example, because of upgrades or changes in specification requirements, many molds must be requalified on new equipment. As the analysis model is constantly updated with the introduction of new statistical requirements, engineering teams are discovering that otherwise functional and useful products are no longer viable because of excessive mold wear. While such products may not satisfy the requirements of process engineers eager to approve process documentation, they may well be functional. Thus, keen judgment and rational thought must be applied to the data in order to contain costs resulting from iterative tooling repairs that may or may not achieve the statistical marker or to reduce losses resulting from previously viable products that are no longer considered marketable.
While the medical device industry was slow to embrace statistical validation methods, it is high time that these mathematical tools be implemented liberally, correctly, and consistently to improve product functionality, effectiveness, and safety. Implementing these methods can occur concurrently with ongoing product evaluations, saving considerable costs downstream. By employing process validations and in-line statistical analyses to improve process controls, assembly, automation, and other manufacturing steps, process variations can be identified and reduced up front.
The need to rein in costs and meet customer demands is a standard concern in the processing world, including in the medical device industry. At the same time, rising healthcare costs is prompting the medtech sphere to enhance statistical productivity. With the increasing demand for healthcare cost-effectiveness, coupled with insurers’ refusal to issue reimbursements for conditions attributable to provider faults or manufacturer defects, manufacturers must generate the technical and statistical data necessary to support their products. Suppliers that provide such data will experience improved processes; enhanced efficiency; cleaner, more consistent, and higher-quality products; and increasingly satisfied customers. As acceptance criteria move in-line, statistical productivity improves, and standardization becomes entrenched, positive results will prevail industrywide.
Steve Gilder is director of quality at Helix Medical. He has worked for several manufacturing organizations over the past 25 years, most recently in the medical device and components sphere. Gilder is a Villanova-certified Lean Six Sigma Master Black Belt and an ASQ-certified Six Sigma Black Belt. Reach him at firstname.lastname@example.org.