Quality Designed on Day One: Built a Clinical Lab in 13 Days
Karl Andersson spent a fifth of every run on samples whose answer he already knew. That cost replaced six months of production validation.
On March 25, 2020, Karl Andersson and his team decided to build a clinical COVID testing lab in Sweden. No commercial assays existed yet. The most accessible reagents carried FDA emergency use authorization, which gave them no legal standing in Europe.
Thirteen days later, the lab met Swedish hospital regulations. The first patient sample arrived on day 14 or 15.
Karl told me this story when he joined me on “The Builders of Modern Medicine”. He started as a mathematician, moved into biology, and built the clinical laboratory behind “Stockholm3”, the prostate cancer blood test. His team took that lab from an empty room to regulatory registration in six months.
Almost every instrument available to them was labeled for research use only. Research vendors avoid clinical certification because the regulatory burden costs too much. “Stockholm3” combines around 110 values from one blood sample into a single statement, so small inaccuracies put the whole diagnostic output at risk.
Karl's answer was to build the check into the first step. His team placed known negatives and positives of different strengths into every run, around 15 to 20 percent of each plate. At the end, the software his team wrote looked for every known sample in its expected place. When they all appeared, the patient results on the same plate could be trusted.
According to Karl, it's about designing the quality test before you design the process. Meaning that operations check quality at the end, when the report is finished, or the product is built. By that point, the remaining tools are destructive testing and rework.
Karl puts the question on day one. Quality control belongs in the first design meeting, he said, because the first day is the only point where a team can build a system that reports on itself. His example was simple. Add green color at the start, and the product should still be green at the end.
“Ginolis” builds precision automation, so one distinction Karl made applies directly to my own work. Precision means a system repeats the same result again and again. Accuracy means the result matches the truth. A precise system running a noisy process will consistently repeat the wrong answer. Karl observed that sales conversations focus on precision anyway, because precision is easier to show.
His method starts with risk thinking before the line runs, then adds markers that travel with the product and data collected at several steps. A marker can be a thickness measurement tracked over time, a camera watching for falling dust, or a chemical added during test strip production. Karl was clear that this takes discipline more than mathematics. In his estimate, anyone can do more than half of it without much training.
The method has a price. Spending 20 percent of reagents and materials on validation is expensive to operate. Karl accepted that trade during COVID because a full production validation takes six months. He said himself that the model works short-term and would fail over the long term.
The controls proved their value one morning in August. A plate came back dotted with positives, as if someone had coughed on a mirror. The management meeting lasted about 35 seconds. The facility closed, the team remeasured, and the result held. People had returned from vacation and were passing the virus around. Staff came back the next day.
The “Stockholm3” lab signed off around 25,000 prostate cancer diagnostic outputs. Karl estimates that in Norway the test saved roughly one biopsy every hour or half hour, and it found more cancers than the traditional methods. He knows the lab made mistakes, and he carries them. "I know that the alternative is worse, and I decided to trust the math and the validation," Karl said.
The practical test for any new line is to ask when quality control first saw its design. If they joined at the validation stage, the line has no built-in way to report its own failures.
The “Stockholm3” lab took 180 days, and the COVID lab took 13. In both, the quality structure was in place before the first patient sample arrived.
Kauko Väinämö
Kauko Väinämö is CEO of Ginolis, a precision automation company for microfluidics and diagnostics manufacturing based in Oulu, Finland, with over 150 production lines across 37 countries.