Advertisement
Articles
Advertisement

Data Analysis Helps Reduce Product Failures

Wed, 11/10/2010 - 10:06am
Andrew Lux, Ph.D.

LISTED UNDER:

One of the critical goals of a medical device manufacturing project is to eliminate the occurrence of product failure. While there are a number of methodologies to implement in order to help increase the chances of success, this article highlights one technique that relies on multi-variate data analysis to achieve the goal.

“It is not a question of how well each process works; the question is how well they all work together.” Lloyd Dobens, Thinking About Quality: Progress, Wisdom, and the Deming Philosophy

While advances in technology are fueling the development and launch of increasingly complex medical devices, post-launch device failures are becoming significantly more common. To reverse this trend and maintain high product quality even as complexity increases, manufacturers may need to consider new statistical methods as alternatives to traditional strategic process controls (SPC). Multi-variate data analysis (MVDA) is a statistical technique that analyzes data generated from multiple variables that form an interdependent system. MVDA can help reduce failure rates by identifying negative quality trends in situations that appear to be “in control” using traditional SPC methods. Newer software applications can make it easier to implement MVDA and interpret the data generated, as can working with experienced outsourcing partners who can support the successful implementation of such a methodology.

When Bad Products Happen to Good Companies
Quality issues and product failure following the launch of a medical device can be financially devastating for OEMs. Worse, they can negatively impact the care of the very patients whose lives they seek to improve. Yet, while medical device companies have been aggressively committed to maintaining quality systems and have invested heavily in SPC and CAPA software, IQ/OQ/PQ methods and procedures, documentation, training programs, and automated inspection tools, the rate of post-launch product failure continues to increase dramatically. As evidence, Class I recalls of medical devices reported to the FDA have increased 172% since 2008. So what is going wrong?

No Variable Is an Island
Part of the reason for the increase in product failure is that, in every respect, medical devices are becoming more and more complex. Consider the requirements for orthopedic implants; many newer products require post-machining surface anodization, color identification, and anti-fibrosis and bone stimulating properties. Other examples include needle manufacturing for modern dialysis applications, which must be ultra-controlled to prevent blood cell damage, and solid state devices, like ultrasound phased array transducers, which have extremely precise processing requirements to ensure image resolution. To meet these technological demands, manufacturers have had to develop more complex manufacturing processes. More complex manufacturing processes introduce more system variables, and more system variables increase the likelihood of product failure. It is in “keeping up” with the analysis of these myriad of variables where traditional SPC falls short.

One of the reasons for this shortfall is that process variables are not stable; they vary with environmental conditions, materials and lots, inspection tool drift, and many other real elements of an imperfect world. Unfortunately, SPC assumes a stable process, and the SPC charting methodology “looks” for random process variations (i.e., noise) only. If the process is not stable, the SPC methodology does not work.

Additionally, process variables are interdependent, meaning one must evaluate each variable in relation to all other variables in order to determine the risk of an overall system failure. The major limitation of SPC is that it involves measuring and charting data associated with individual variables only, and assumes that all variables in a manufacturing process are independent of one another. As a result, a number of cases have shown that a product fell within all acceptable limits of an SPC chart, yet ultimately failed final inspection<md>or worse, failed after the product shipped.

MVDA: Staying Out of Trouble
In contrast to SPC, MVDA has shown that it can account for process interdependencies and accommodate a vast number of variablesin fact, the more the better. It has been used for many years by the automotive, aerospace, chemical, and telecommunications industries to successfully identify negative quality trends that might otherwise be overlooked by SPC, leading to dramatically lower rates of product failure and higher overall product quality.

The application areas for MVDA include:

  • Process monitoring and early fault detection and classification
  • Process analytical technology
  • Quality control
  • Data mining and integration
  • Compositionproperty relationships
  • Structureactivity relationships
  • Multivariate calibration
  • Multivariate characterization
  • Image analysis

While the advantages of MVDA are compelling, it requires a solid commitment of time and resources to learn and implement. By definition, it is mathematically complex and will require an investment in a commercially available software program. Generally, a statistician will need to work closely with a manufacturing engineer to establish the modeling methodology for each specific application. Perhaps the best way to implement MVDA is by learning from a consultant or a contract manufacturer who has “done it before.”

Conclusion
Relationships are key. Whenever something complex “breaks,” whether a medical device or a deep-sea oil rig, there is rarely one single and easily identifiable cause, but rather a series of previously undetected (though not always unmonitored) weaknesses that act synergistically to cause the failure. If product quality is to keep pace with product complexity, it will be critical for medical device manufacturers to incorporate quality systems, such as MVDA, that enable them to monitor and evaluate the relationships between numerous process variables so as to predict and, ultimately, prevent device failure.

Andrew Lux, Ph.D. is senior vice president of Operations and Manufacturing at IntriMed Technologies. Driven by the notion that “good enough isn’t,” he implements continuous improvement and compliance initiatives to positively impact all phases of design and manufacturing. Dr. Lux can be reached at 805-436-8402 or andrew.lux@intrimed.com.

Advertisement

Share this Story

X
You may login with either your assigned username or your e-mail address.
The password field is case sensitive.
Loading