Sally's Drop Test


With a meal properly dosed, when the insulin taken for the food has finished acting, the goal is to have blood sugar level return to the same value it has when the dose for the meal was first taken. For most meals that ideal is not attained because of random meal to meal variations and/or because the dosing setup may be incorrect and/or because of the fact that food nutritional contents vary because of changes in growing conditions. In general, for most meals, the pre meal start and post meal finish BG values will differ so we restate the goal to account for that random content in the measurements made in self tests. The goal, rather than applying to every individual meal, is to have the average change in BG, from pre meal to the time when the insulin has finished, be zero when averaged over several meals.

Correction of a high starting BG is not part of the above statement of goal. After the above goal is reached, further work allows meal dosing an correction of a high or low starting BG to be included. The meal dose is adjusted up or down from what would meet the above goal so that rather then the average performance being no net change it is that the finish BG reaches a target value. For example, the goal might be to have the closing BG be 110 rather than equal to the starting BG.

An equation presented in the insulin tutorial on "reverse engineering a meal' allows useful meal dosing setup information to be extracted from a meal record when the pre meal BG and closing BG differ. The reverse engineering equation produces a dosing coefficient that has been named "K" in the reverse engineering equation, K = {gg -(finish BG -start BG)/RPG}/dose. K has the units gg/unit of dose. That is, it shows how many grams of equivalent glucose (resulting from digesting the meal) are covered by each unit of dose. In that equation, RPG is BG rise per gram of glucose load and it is best found with a bread test. Thus, looking at the equation, the BG when the meal insulin has finished acting, called finish BG in the above equation, plays a direct role in finding the dosing coefficient, K. The subject covered here is how to find out how long to wait after injecting for the meal to come to that point where the insulin has finished acting. That is, how many hours after injecting do you take the "finish BG) needed in the equation.

Reference material:

The spread sheet links on this web page, in Excel and in Quattro pro, contain Sally Hines' time to finish test data and the graphs that were used to analyze her test results and find that time to finish value. You will find it easier to follow this discussion if you first download the spread sheet file and become familiar with how the graph in it looks.

Sally drop test.xls        - Excel
Sally drop test.qpw      - Quattro Pro
Sally_Drop_test.ods    - Open Office

Humalog Time to finish test preparation:

The means used to find the insulin time to finish sounds easy enough. You take a dose of sufficient size to correct a BG in the 150 to 170 range and observe BG until you see that the action has finished.

It is necessary to start the test well after any prior fast insulin corrections or fast insulin meal doses so that the BG values that are observed during the time to finish test are not a mix of the responses to an earlier dose and to the dose employed some hours later in the test being discussed. The time to finish test needs to be spaced several hours from those prior fast insulin uses for that reason. The question, when time to finish tests have not previously been run, is how many hours between those prior and present uses are enough? The problem that first time is time to finish is not yet known.

Therefore, the action action curve data for a typical person has to be used to identify how many hours of space to leave. A data sheet with a graph of a typical insulin absorption curve is included with the fast insulin in every box where it is described as being a typical response for a group of tested people. For Novolog it is 4 hours and for Humalog it is 5 1/2 hours. Variations from one person to another make it advisable to leave additional space until a few time to finish tests have been run because a person with slower than typical absorption may be the one doing the test. If it is later seen in the data taken during the test that a shorter time can be safely used, later tests can put that to use to decrease the impact on a persons meal times and daily activities.

Therefore, six hours is advised for Novolog and 7 1/2 for Humalog when these time to finish tests are first done. Insulin is not the only influence that could distort the fast insulin time to finish test data because of persisting BG changes from prior events. Intake of any calorie bearing food or drink should end enough hours before the time to finish test to allow digestion to have finished by the time the test begins. The same time that was sufficient for the safe spacing from prior insulin, 6 hours, is marginal for safe spacing from prior calories; seven is more like it for most people, particularly when a meal was high in protein or fat. The space needed for prior insulin completion can be modified upward in a later repeat of the time to finish test if there was reason to think that the prior actions had not ended. If test results when a prior meal is skipped completely differ greatly from those when there was a prior meal, even one where the full 7 hour spacing was present, that would be a clue that even more spacing between prior meals and/or prior doses was needed.

Another test boundary condition is the BG when the test insulin is taken. BG must be under 180 because the rise in insulin resistance for higher BG values will change the drop per unit of dose, one item to be extracted from the data, to a lower value. It also changes the shape of the BG profile. BG should be under 170 to avoid that shift in insulin resistance at high BG values.

Since the test is going to produce a drop in BG, in the interest of safety the test should not be done if BG is already a low value. I think that a reading in the range of 150 to 170 is ideal for the start BG to seek when running this time to finish test. Glucose tablets finish in 45 minutes so they can be use to raise a BG that is below that range. Nothing other than glucose tabs should be used because other things act more slowly and last longer. Therefore, about an hour before the scheduled beginning of the test take a reading and use glucose tabs to get up into that desired starting BG range.

The fast insulin response is not a fully isolated factor, even after taking the precaution of using that large prior event spacing. The basal insulin, taken in the morning and/or the night before, is acting during the time to finish testing hours and there may be a drift in the baseline fasting readings in response to the basal insulin during the test. It is necessary that the baseline drift and the fast insulin response be separated . If the BG continues to move well after a reasonable time to finish has elapsed, say eight hours or more, Whatever BG trend is present can be assumed to be operating during the test. If a fasting test is done on some other day with data being taken during the same hours that the time to finish data will be collected, and if the same spacing is present from prior events, the drift can be inspected directly without having to extrapolate a later trend backwards into the test hours. Once the drift is known, its opposite value can be used to adjust the data taken during the time to finish test so as to create the values that would have been seen had there been no basal BG drift.

Meters can turn up bad readings if a strip is defective or if it is not completely filled. Even a good reading can be a considerable distance from the true BG value because of random strip to strip variations seen in all brands of meters. By taking two readings at a time it becomes much easier to spot and deal with bad readings.

Therefore, to summarize up to here, spacing between the hours of the time to finish tests and prior food or insulin is needed so that actions do not overlap, the condition that BG be between 150 and 170 should be met, and a pure fasting test in the same test hours should be done to characterize baseline BG drift that the basal insulin may be causing, and dual readings should be used.

A fasting baseline tests will show more about what the basal insulin was doing during the time to finish hours and it can also help to discover if enough space was provided between the test and prior meal events.

The time to finish tests can be summarized as follows;

1) Space the beginning of the test 7 hours, from prior fast insulin or caloric intake.
2) Revise that time upward, if necessary, after seeing fasting test results, with and without a prior meal and dose.
3) Take data at 1/2 hour intervals
4) Take two readings at each reading time
5) Do not do the test unless BG is between 150 and 170 when you start, Use glucose tabs if too low and allow 45 minutes for them to finish  acting before starting the time to finish test.

Processing the data:

The data collected are a time and a pair of readings. Convert the time of day to time of a reading to elapsed minutes to facilitate graphing the data. Call the elapsed time of the first reading pair zero. If the test started at 1:00 PM and data were taken on the half hour, you would take the times of 1:00, 1:30, 2:00, and so on and describe those times as 0, 30, 60 elapsed minutes, etc.

Next, set up a spread sheet with the first column being those elapsed times in minutes and the second column being the recorded BG readings. It would not hurt to enter a text note somewhere in that spread sheet to make it a matter of record that proper separation from prior confounding events was accomplished or that a prior meal had been skipped altogether. Of course you would also include the dose used in the time to finish test as part of the record so that drop per unit of dose can be calculated. Then, at a later time, when you review that record, you will not have to wonder about validity of the test. Elsewhere in the spread sheet the data for the pure fasting tests should be included. That way you have the observed insulin profile and you have what is needed to back baseline drift out of that data all in one spreadsheet.

Once the data is entered into a spread sheet, graph it. Use the trend line feature of your spread sheet program to find a curve that describes the data points. In the example you can download from the DSM web site, EXCEL was told to derive a trend line by fitting a 6th order polynomial to the data. That trend line reduces the influence of random noise in the individual data points. That is implicit in it being what is called a least squares fit. Test strip random variations, usually around plus or minus 4 to 6% in size, depending on the meter brand, are the source of that random content in the data.

Time to finish from the spread sheet graph of Sally's humalog data:

The graph, beyond a bad data point in the first of the two tests she ran was a straight line that eventually curves over toward the time to finish. In fact it is the leveling out of the data that identifies where the insulin did finish. That happened to be between 6 1/2 and 7 hours. Thus, allowing extra leeway in the spacing from pre test events turns out to have been just the right thing to have done.

Whether horizontal, sloping up or sloping down, where the BG trend line curve becomes tangent to what the fasting basal drift curve looks like, the test is over. Leveling off is defined that way, not as the graph of the BG data becoming horizontal.

Summary of data reduction:
1) Enter the data in a spread sheet.
2) graph the data, BG vs elapsed time of the dosed test and the not dosed BG at the same times.
3) fit a polynomial trend line to the graphed data.
4) where the profile becomes tangent to the fasting baseline data determines where the insulin finished.
5) get the drop per unit of dose that the dose produced and also get the total change that was due to basal insulin during the time to finish.



The two tests are in disagreement about time to finish. Additional tests of this kind are needed to get what was originally desired, namely, time to finish. Looking at time to start, which is in better agreement between the the tests and adding her informal observation about time to start as posted on 1/26/05, time to start is in better agreement with the typical characteristics of the insulin, I feel that of the two tests the second test, with its 5 hour time to finish is more likely correct than the first one with its more than 7 hour time to finish.

In doing meal tests, to produce somewhat of a compromise between these disparage results, 6 hours will be used for time to finish unless more of these fasting profiles are performed and the time to finish firms up.

Next To DO:

The Humalog time to finish test has to be repeated several times because of test to test random variations and the lack of a reasonable consensus in these two tests. From such a group of tests, an average humalog fasting state drop per unit would be calculated and the most representative time to finish would be the one used to get finish BG when doing a meal dosing evaluation.

The fasting baseline during those same test hours has to be tested to see if baseline tilt is large enough to require adjustment of the data.