Pump Data Study


Success on an insulin pump depends on determining an appropriate Total Daily Dose (TDD), and generating from this the basal infusion, carbohydrate factor, and correction factor. To derive formulas for optimal pump settings, we downloaded data from 1020 anonymous, complaint-free insulin pumps for analysis. A subset of 396 pumps were chosen because more than 85% of their glucoses (avg. 95.2%) were directly entered from an attached CoZmonitor Freestyle glucose meter. The entire group was divided into 3 tertiles based on their average glucose. The low glucose tertile (LowGT) of 132 pump wearers represented users with the best glucose control. Their average glucose was 144 mg/dl (8.0 mmol) from an 4.73 test per day averaged over 73.8 days. Optimal pump setting formulas were derived from this group and are used in the Settings Tool.


Many pump wearers were found to be using inaccurate values for their carb factor (CarbF) and correction factor (CorrF), even in the best control tertile. Rather than rely on less accurate pump CarbF settings, we calculated each user's actual CarbF by dividing the carb grams entered for each meal by the actual carb bolus taken for that meal. Actual CorrFs were derived in a similar fashion. In contrast with pump settings, actual carb and correction boluses become optimized as pump users or the pump itself adjusts doses for circumstances and for active bolus insulin on board (BOB or IOB or active insulin). Average total basal doses per day were used to determine the basal percent of the TDD.

A pump wearer's total daily dose (TDD) of insulin provides an excellent guide from which basal and carb bolus settings can be determined. The frequency of hypoglycemia and average glucose level show how well an individual's TDD is working. With an elevated average glucose, the TDD generally needs to increase; with frequent hypoglycemia the TDD needs to be decreased.

When frequent or severe hypoglycemia is present, the degree of excess in a person’s TDD cannot be measured. Here, the TDD can be lowered, such as by 5% or 10%, and this lower TDD can be used to generate new pump settings. If the lows continue, another reduction in the TDD will be needed. Taking these factors into account allows calculation of an improved TDD (iTDD).

Basal Rate

The optimal basal percentage of the TDD was found to be 48.0%. The average basal rate per hour was derived as:

Avg. basal rate (u/hr) = (TDD x 0.48) / 24hr  (or more simply as TDD x 0.02)

Carb Factor Formula

The CarbF defines how many grams of carb, after digestion and dispersal into the extracellular space (ECS), will be transported per unit of insulin into insulin sensitive cells for metabolism or energy storage. The number of grams of carb transferred across cell membranes per unit measures insulin efficiency and is directly related to an individual's insulin sensitivity, so an individual's CarbF is directly proportional to their insulin sensitivity (IS).

Carb intake per day (proportional to weight) can be divided by the carb boluses required to cover these carbs which is proportional to TDD (Wt / IS). We wanted to find the average CarbF for a large group of well-controlled individuals with Type 1 diabetes. This avg. CarbF can then be multiplied by an individual's relative insulin sensitivity to determine their personal CarbF:

Equation 7: CarbF   =   grams of carb/day   =   K* x Wt (lb)   =   K x Wt (lb) x IS   =   K x IS
carb bolus insulin/day TDD Wt (lb)

*K is a constant measured in (grams per lb)/u

Equation 8: an individual's CarbF = an avg. CarbF x the individual's relative IS (rIS)

An individual's CarbF is directly proportional to their IS. If individuals of different weights but with the same IS consume 50 grams of carb, they will use the same CarbF and take identical carb boluses to cover these carbs.

An individual's relative IS (rIS) in Type 1 diabetes can be quickly estimated as the ratio between a TDD for someone of their weight who has an average IS and compare this to their own TDD:

Equation 9: rIS   =   a TDD for that weight with an average IS   =   0.24 u/lb x Wt(lb)
The individual's TDD TDD (u)

Where 0.24 u/lb represents an average sensitivity to insulin, 3,4 and 0.24 u/lb x Wt(lb) would be the expected TDD for this person if their IS were average. Using this scale, a rIS of 100% represents an average sensitivity to insulin, less than 100% is a decreased sensitivity and greater than 100% is an increased sensitivity to insulin.

Because the CarbF is directly related to IS, once an average ActualCarbF value is known for a large group of individuals in good control with a minimum of hypoglycemia, any other individual's CarbF can be closely approximated by multiplying this average CarbF times the individual's rIS:

Equation 10: CarbF   =   Avg-ActualCarbF x rIS   =   mean-ActualCarbF   x   Wt(lb) x 0.24

Rather than rely on less accurate CarbF settings in the pump, we based CarbF determinations on the ActualCarbFs used by patients. With an average ActualCarbF of 10.8 g/u in Table 4 for the LowGT(104), Eq 11 becomes:

Equation 13: CarbF   =   10.8    x   Wt(lb) x 0.24   =   2.6 x Wt(lb)

This formula is based on carb counts that are likely undercounted by most pump wearers, so an individual who counts carbs precisely may need to raise their CarbF number calculated from Eq. 13.

Correction Factor Formula

Glucose levels become elevated in the extracellular space (ECS) when basal delivery is too low to suppress gluconeogenesis and glycogenolysis, or when carb boluses are insufficient to adequately cover carb intake. A CorrF measures how far the elevated glucose concentration in an individual will fall in the ECS per unit of insulin. The correction bolus provides backup for any deficits that occur in basal and carb bolus delivery. The fall in glucose is inversely related to a person's weight (size of their ECS), divided by IS (or ease with which glucose is transported from the ECS into insulin sensitive cells). Since “Wt / IS” equals the TDD, the CorrF is inversely related to the TDD:

Equation 11: CorrF   =   K2 x IS  =  K2

Where K2 (measured in mg/dl per unit), also called a correction factor rule number (CorrF-RN) = CorrF x TDD. It is commonly referred to as a 1500, 1700, 1800, or 2000 Rule for mg/dl, or as a 90, 100, or 110 Rule for metric users.

To determine an optimal CorrF-RN or K2, we derived CorrF-RNs from each individual's average ActualCorrF used over time multiplied by their average TDD.

In Table 5, our data shows that K2 equals 1993 and can be represented by 1960 or 2000, so:

Equation 14: CorrF   =   K2 x IS  =  1960

If the average glucose is elevated (without excess hypoglycemia) and correction boluses are routinely given, correction boluses will drive up the current TDD and begin to make up a greater than desired percentage of the TDD. Control can be improved by using the current TDD with its relatively high percentage of correction insulin to select more physiologic basal rates, CarbF and CorrF. Over a few iterations upward, or adjustment of the TDD downward for hypoglycemia, an improved TDD (iTDD) can be determined as the excess in the correction boluses is redistributed into a more appropriate basal and carb bolus balance. With an average of 4.5 glucose tests per day, the great majority of pump users are testing sufficiently to improve glucose in this way.

How We Calculate Improved TDD (iTDD) and Pump Settings

A fast way to determine an iTDD is to add to the current TDD the correction boluses that would be required to lower the elevated A1c or EAG or average meter glucose level (averaged for 14 days or more). This method is alsoof some help in correcting excess hypoglycemia. When the average glucose level is higher or lower than the LowGT's average of 144 mg/dl (8.0 mmol), a corrected or iTDD can be calculated using this formula:

Equation 15: iTDD   =   TDD  +  (AvgBG - 144 mg/dl) x 2.5
  1960 / TDD

Where 1960 is an average CorrF-RN and 2.5 is a conservative estimate of the number of times per day that the elevated average daily glucose would need to be lowered using an average correction factor.

To quickly improve an individual's overall control, more accurate basal rates, CarbF, and CorrF can be calculated from an iTDD, assuming that frequent hypoglycemia is not present or has been eliminated with appropriate dose reductions.

From this iTDD, closer estimates for each pump setting can be obtained:

Equation 16: iTDD x 0.48 = TDD +  (AvgBG - 144 mg/dl) x 2.5  x 0.48
1960 / TDD
Equation 17: iCarbF =  2.56 x Wt (lb)  =                      2.56 x Wt (lb)                    
iTDD   TDD + 
(AvgBG - 144 mg/dl) x 2.5
1960 / TDD
Equation 18: iCorrF =  1960  =                          1960                       
iTDD   TDD + 
(AvgBG - 144 mg/dl) x 2.5
1960 / TDD

In Eqs. 16, 17, and 18, the constants 2.5, 1960, 0.48, and 144 mg/dl (8.0 mmol) can all be modified for more aggressive or more gentle insulin adjustments. For example, 2.5 may be lowered to 2.0 or 1.5 for a slower increase in TDD, or raised to 3.0 or 3.5 for more aggressive corrections. The average glucose of 144 mg/dl (8.0 mmol) can be adjusted downward for pregnancy or upward for hypoglycemia unawareness as needed. The average CorrF of 1960 can be lowered for faster and more aggressive corrections or raised for safer correction boluses as needed.

When someone has erratic control with a combination of frequent highs and lows, the TDD may be lowered by 5% to 10%. We hypothesize that using this lower TDD to redistribute insulin doses using Eqs. 16-18 will provide a rapid improvement in glucose stability.

More on Where the Pump Data Came From

In 2006 and 2007, the Smiths Deltec pump company recalled their model 1700 Cozmo® insulin pumps for software enhancements. These pumps, used throughout the U.S. by a wide variety of clinicians and patients, offered a unique opportunity to analyze actual insulin pump use. We requested and received anonymous data downloads from over 1,020 complaint-free Cozmo® pumps.

Raw data was extracted and compiled from pump databases in Excel with further processing in SPSS. We selected a subset of 363 pumps that had more than 28 days of consecutive data (avg. 73.8 days), less than one glucose below 50 mg/dl (2.8 mmol) every 3 days (avg. one every 10.9 days in the best control tertile), over 85% of glucose readings entered directly from a CoZmonitor® meter (avg. 95.2% direct readings), averaged more than two glucose readings a day (avg. 4.50 tests/day), and used a TDD greater than 8.5 units.

The selected pumps were distributed into tertiles of 121 pumps based on their average glucose values. The low glucose tertile (LowGT), with a average glucose of 144 mg/dl (8.0 mmol), was used to derive dose formulas.


1  J Walsh, R Roberts, T Bailey: Guidelines for Insulin Dosing in Continuous Subcutaneous Insulin Infusion Using New Formulas from a Retrospective Study of Individuals with Optimal Glucose Levels. J Diabetes Sci Technol, 4: 1174-1181, 2010.
2 J Walsh, R Roberts, T Bailey:  Guidelines for Optimal Bolus Calculator Settings in Adults. J Diabetes Sci Technol 5(1): 1711-1717,  January 2011
3. Davidson PC, Hebblewhite HR, Steed RD, Bode BW. Analysis of guidelines for basal-bolus dosing: basal insulin, correction factor, and carbohydrate-to-insulin ratio. Endocr Pract. 2008;14(9):1095–101..
4. Adamsson U, Lins PE. Clinical views on insulin resistance in type-1 diabetes. In: Agardh CD, Berne C, Östman J. Diabetes. Stockholm: Almqvist & Wiksell; 1992, 142–50.