The following is a rudimentary calculation of an error budget for the Paradigm 722 Real Time System. Only the major sources of error are included in the calculation as the purpose of this small effort is to help users understand the data that is presented to them. All too often the statement is made that “the system is not very useful because the Sensor reading is not the same as the Meter reading”. These perceptions are the result of not understanding the bounds on the data and not understanding that it is not the absolutely values that are important, it is the TRENDS that allow better control to be achieved.

Sources of Error

Primary

• The Link – Glucose Meter
• The Sensor of the Real Time System
• A further factor, while not truly an error but must be considered is the delay time of the Sensor. This is generally user dependent.

Other

• Calibration Frequency
• Sensor Stability
• A host of other small factors that are probably not significant.

Assumptions

1. The algorithm for processing data from the strips and the sensor is stable.
2. The errors specified for the components are linear across the normal blood glucose range.
3. For the purpose of these calculations, a standard blood glucose value of 7.0 mmol/L is used.

* Please ensure that you have read the Medtronic Fact Sheet entitled “Distinctions between Standard Glucose Meters and Glucose Sensors”.

Errors

Test Strips for the Link
A review of the documentation provided with the BD Logic strips state that “a 7.36% average difference between the glucose test result on the BD Monitor compared to the glucose test result on a reference glucose instrument”. E.g. 7.37% of 7.0 mmol/L = ±0.5 mmol/L.

Medtronic Real Time System
The documentation for the Real Time component of system (Sensor Features User Guide) devotes a full chapter to the accuracy of the system. The results are presented as a percentage of time that readings fell with 20% and 30% of the laboratory readings. Therefore using the best case scenario of being within 20% of the true value the following is the possible error:

20% of 7.0 mmol/L = ±1.4 mmol/L

If we combine just these two factors, we can see errors in the range of ± 1.9 mmol/L.

Sensor Delay
While not an error, the delay factor in the sensor can appear as an error if you are not aware of the process. Dr. Bruce Perkins has talked about this delay in some of his observations. He has suggested that the delays he found were about 20 minutes. I know from some rather elementary, uncontrolled testing that I did, my delay is in the order of 35 minutes. It would appear that the delay is variable and partially depends upon the absorption of glucose in the interstitial fluid. Therefore, it depends upon the rate of change of the BG as to its effect on the displayed values of BG. So let’s look at an example.

Therefore, in this example, where the BG is rising the Meter would read 7.8 mmol/L while the Sensor would be reading 5.2 mmol/L. So we can clearly see that we must understand exactly what the current conditions are to understand what the data is telling us. This is not rocket science but may be a bit confusing until you get your head around it.

All of this to say that in my mind it doesn’t matter all that much if the Meter reading is a bit different that the Sensor reading. To me the important thing is that you understand the data and what the Sensor is telling you – the TRENDS. I believe that once we understand that, the actual numbers don’t matter much except in gross terms.

We can now proceed from dealing with something that has already happened to dealing with something that will happen in the near future. It is a very different mindset from what we have been taught about pumping to date. This is exciting and users that take on the new mind-set will likely derive significant benefits in glycemic control.

G. Ross Douglas — July 22, 2006