The FDA reported that over the years 1997-2009, 13 people died due to maltose interference in glucose meters. To recall, maltose can cause large increases in a glucose reading which could lead to hypoglycemic individuals being treated with insulin.
Based on an FDA meeting, Dr. Shuren reported that there are 24 million American who have diabetes and of these, 30% (7.2 million) take insulin. Assume that each person tests 3 times per day on average. This means that over the 12 year period, there were about 94.6 billion tests.
The failure rate due to the maltose problem which caused deaths is 13/94.6 billion or 0.00000001374%. Suffice it to say that this is a small number. Were the death rate underreported by 100 fold, the percentage would still be tiny and 0.000001374%. Of course, tiny would not be any consolation if the event happened to a loved one.
So when a laboratory performs a 100 sample method comparison and finds no values in the higher regions of an error grid, it doesn’t mean that the assay has been proven to be safe. What can be done is to apply risk management studies (Failure Mode Effects Analysis, fault trees) in addition to a method comparison.