In April of this year Abbott announced a recall of their Freestyle InsuLinx Blood Glucose Meter. As per their press release, when measuring very high glucose levels of more than 1024 mg/dL the machine reports incorrect (and much lower) values. Glucose levels that high, while rare, are dangerous.

Abbott is one of the most professional companies in the medical device industry and their response appears to have been exemplary once they discovered the issue. Fortunately nobody seems to have died as a direct result. The question is what can we learn from the event?

Curiously the Lifescan OneTouch Verio IQ glucose meter was What does the Abbot Glucose Meter recall tell us about Risk Analysis and Verification Testing? about the same time for turning off at 1024 mg/dL. This was slightly different behavior but the same fundamental failure mode. (Ed. Note: As this blog is being posted, yet another company is impacted as the FDA Warns of Massive Diabetes Test Strip Recall.)

Interestingly the number 1024 is exactly 10 bits of data. A rollover of an integer value may be related to an engineer being very economical with memory usage and allocating no more room than thought necessary (a very poor trade-off in this case). Or a 10 bit analog to digital converter may have been used without controlling for out-of-range conditions. The fact that both meters failed in similar ways at the same time may indicate that they used the same OEM module or that one device was designed to be similar to the other.

From a risk analysis standpoint it would seem that this failure mode was not considered or tested for in their design control. There would have been a few potential places to capture this problem before it got into shipping product.

1. During risk analysis (as per ISO 14971 and as mandated by IEC60601 edition 3) the possibility that an incorrect reading could take place could have been identified and specifically mitigated. Identifying and managing boundary conditions of critical components would be a good element to include at this stage.
2. During verification testing if the whole range of blood glucose levels had been tested this error would have been found. Possibly a corner was cut during testing. This work seems expensive and happens late and thus is a natural place where there is a lot of pressure to compromise.
3. During specification creation. It’s possible that the glucose range was not anticipated properly in the specs to start with.
4. During validation. It’s possible that the failure could have been caught during validation although since the very high level of blood glucose is rare it may have been missed there.

A major theme is that the drive for innovation and new product features must be balanced with an equally intense focus on thorough verification and validation. Hunting a haystack of operating conditions for an obscure failure mode needle is not sexy but it’s just as critical to the success of the venture as the innovative product concept and technology.

At risk of stating the obvious, it appears that inadequate attention to verification and risk analysis was paid in these cases. I believe that having a team of engineers that embodies both innovation and thorough execution and testing bolstered by a mature quality management system is critical for success.