Too often, we look at the 7-day, 14-day, and 30-day trends on our blood glucose monitors, see numbers that look great (or horrid), and rather than seeing an A1c that confirms those readings, we get a number that would appear to have come completely out of left field. (Or Mars. Or the Andromeda Galaxy. It's hard to say exactly where.) We can either scratch our heads and wonder why the numbers aren't correlating, or we can take out our manual readings logs, our meter downloads, our CGM downloads, and our personal journals and try to figure "what we are doing wrong".

The answer is, we are doing nothing wrong -- and everything wrong. Most of us tend to test more frequently when we are monitoring an unexpected high, or the effect of an unusual food or eating pattern, or coming up from a low. While our numbers often change more rapidly around the peaks and dips, our regular stick-me-each-time meters don't have enough sense to determine that a group of close-by readings should not each have the same weight as readings taken further apart in time. In short, by testing more around the times of our highs and lows, as far as our glucometers are concerned, we are selectively choosing higher, or lower, readings. The process of selectively choosing which of several data points to include in, or exclude from, an analysis is called "cherry picking".

"Cherry picking" is less an issue with a CGM, which -- if worn 24/7 -- will record data at a constant rate whether your blood glucose is rising, falling, or standing still. If your computer's CGM analysis software does averaging, its output should be a closer indication of your A1c and fructosamine levels than your glucometer.

For most of us --who do not have CGMs -- the issues are when (and how often) to test, which readings to analyze, and how to analyze them. This is where a personal journal, meter (and meter software) hyper/hypo presets, and meter comments can come in handy. For example, some meter software will sort your readings by time of day (not just meal period). If you tend to go low at 11AM and test every fifteen minutes until you recover at noon, you should see not just the dip at 11AM, but a larger cluster of readings between 11AM and 12PM than between, say, noon and 4PM. If you log manually, you can average your four right-after-another-because-I-don't-believe-this readings into a single reading, and then insert that in your list of daily readings to get what might be a more accurate average. You can also look at what happens if you ignore "outlier data" -- the odd low and its surrounding readings that happened once, had no relationship to anything else, and took forever to recover from (or for that matter, the odd high that ran for two hours before you realized you had a bad set).

Sometimes, even manipulating your spot readings can't explain what's going on with your A1c. That may be because you're experiencing a drop or a rise overnight (when you are not monitoring), or because your after-meal peak happens at sixty minutes or at three hours, rather than the standard two hours. Or maybe something else that happens repeatedly during an unmonitored period of time (such as a long commute, or a gym class, or a long exercise session). For some of us, that might be reason to change our testing patterns -- be it for a day, a week, or a month -- to see what we are missing. For others, it might be part of the documentation needed to get a CGM.

Perhaps some day, someone will come up with an in-meter averaging algorithm that understands our tendency to "cherry pick" testing around our lows and highs and provide us 7-, 14-, and 30-day "averages" based on the area under a time-averaged curve, rather than a simple mathematical mean. Until then, it's a matter of understanding our testing patterns and selectively processing the data they produce to provide a meaningful assessment of our diabetes health.