From Gap to Success: How One Medical Device Company Built the Statistical Expertise to Hit Critical Milestones

          Paul Hakamaki, VP of Quality, Medical Devices Industry

When you’re developing a Class III medical device, there’s no room for uncertainty—especially when it comes to verification and validation. I work at a company where we have an exceptionally strong technical team with diverse backgrounds and varying levels of medical device experience. But as a relatively young company, we hadn’t yet established “our way” of handling the statistical and risk-based analysis that’s critical in medical device product development.

The Challenge

As we moved into the V&V phase, the gap became clear. We didn’t have the depth of statistical expertise needed to confidently navigate this crucial stage. Time was our enemy—we couldn’t afford to develop these capabilities organically. We needed external expertise, and we needed it fast.

Finding Kevin

As VP of Quality with Design Assurance under my responsibility, I knew finding the right consultant meant finding someone who could integrate with our team and transfer knowledge that would stick. I turned to my LinkedIn contacts, and Kevin Ashton’s name jumped out at me. I had worked with Kevin years earlier at American Medical Systems and remembered him as exceptionally knowledgeable in statistics and data analysis for medical devices. I reached out immediately, and that conversation kicked off a collaboration that proved invaluable.

Working Together

From day one, working with Kevin was a pleasure. He has a rare ability to collaborate while knowing exactly when to guide things in the right direction. But what truly set him apart was his teaching ability.

Kevin led training sessions ranging from small groups to 30+ people. He even facilitated joint sessions with our team and one of our key suppliers, ensuring everyone was aligned. His teaching style perfectly balanced theory and practical application, keeping everyone engaged. The feedback was outstanding, and many team members sought him out for one-on-one help with specific challenges.

At a strategic level, I worked with Kevin on developing SOPs for V&V, creating statistics and data analysis procedures, and establishing our training program.

The Results

Unlike a new hire who needs months to ramp up, Kevin delivered value immediately. His first opportunity came during a multi-day off-site where our R&D and Quality teams were discussing V&V plans—the perfect venue for him to begin knowledge transfer. Those sessions established him as our go-to expert for statistics and data analysis.

By the end of our engagement, Kevin had created a complete curriculum of teaching modules and materials that became an invaluable resource for our team. Combined with the SOPs we developed together, these tools gave our team the confidence and capability they needed.

The bottom line? We achieved our R&D milestone on time and to best-in-class standards. Kevin didn’t just consult—he built lasting capability within our organization.

Data Analysis Leads to Insight and Action

Garryl, a R&D Engineering Fellow, was analyzing early clinical data from a small set of patients to determine if four device performance parameters had an effect on clinical outcomes.  His engineering judgment suggested there would be two important factors and two unimportant factors.  However, the statistical analysis produced a puzzling result: one parameter, expected to be critical based on solid engineering principles, showed weak correlation with clinical success.

Was the model wrong? Or was engineering intuition misleading?

That’s when Garryl called on me for help.  Garryl remembered how my statistics classes made foundational principles easy to apply and understand, and appreciated my advice with statistical modeling on past projects.

Of course, I was eager to dive in and help.  When correlations don’t match expectations, the issue could be:

• Misapplied statistical methods
• Insufficient sample size masking true relationships
• Hidden interaction effects between variables
• Or genuinely incorrect engineering assumptions

Together, we carefully examined the model set-up and analysis method.  I was able to confirm that Garryl was applying the statistical modeling principles correctly.  The counterintuitive result was real—but why?  

Digging further with regression modeling, we tested competing hypotheses: what happens when we include vs. exclude the questionable parameter? What happens when we combine factors, treating them as linked as opposed to separate?  This revealed two critical insights:

1. The two expected important factors were strongly predictive—the small sample size had initially obscured their effect sizes
2. An interaction effect between what appeared to be a minor parameter and another key observation created a significant predictor that engineering intuition had missed 

Within 40 minutes of collaborative analysis, we transformed uncertainty into clarity. The refined model now accurately predicts clinical outcomes and actively guides physicians in case planning.

Improving patient care, one statistically sound model at a time.