Key words type 1 diabetes - insulin pump therapy - insulin calculation - fat protein units
Introduction
Usually, carbohydrates consumed in the diet are regarded as the decisive factor for
an increase in postprandial glucose excursion. Consequently, with intensified
insulin therapy, the amount of carbohydrates is used to calculate the prandial
insulin dosage by multiplying the amount of carbohydrate content with an individual
carbohydrate-insulin ratio (CIR) [1 ]
[2 ]. However, besides the amount of
carbohydrates, the fat and protein contents of a meal have also been demonstrated
to
influence the postprandial glucose excursion [3 ]
[4 ]
[5 ]
[6 ]. A high fat and protein content often delays gastric emptying and the
absorption of carbohydrates from the intestine into the blood, resulting in a
delayed increase in post-prandial glucose. High protein and fat content in meals
challenge the match of prandial insulin action and post-prandial glucose. This
delayed absorption of carbohydrates can be countered therapeutically by splitting
the insulin bolus or delaying the infusion of prandial insulin as offered in modern
insulin pump therapy [1 ]
[2 ]
[7 ].
Bell et al. found an impact of a high-fat content of a meal on postprandial glucose
excursions in six out of seven reviewed studies, mainly indicating higher
post-prandial glucose excursions after fatty meals [3 ]. It is speculated that beyond the
delayed glucose absorption dietary fat might induce higher levels of free fatty
acids and this, in turn, could cause insulin resistance, requiring higher insulin
doses for optimizing postprandial glucose control [8 ].
Protein-rich meals can also have a hyperglycemic effect on post-prandial glucose
control. In particular, high amounts of proteins (> 75 g) or the
addition of a moderate amount of fat and protein have been associated with higher
postprandial glucose levels [9 ]
[10 ].
These data led to the assumption that high fat and protein content in a meal requires
additional prandial insulin. To standardize insulin dosing for fat and protein,
“fat protein units” (FPU) were introduced by Ewa Pańkowska
in 2009 [11 ]. One FPU corresponds to the
amount of fat and/or protein containing 100 kcal energy.
For fat- and/or protein-rich meals, whether and which amount of additional
prandial insulin should be administered remains debatable [3 ]
[4 ]. Pańkowska et al. [12 ]
[13 ] recommended using the
same amount of insulin for an FPU as for one carbohydrate unit, an algorithm that
has been used in many other studies [3 ]
[4 ]
[14 ], while another algorithm suggests using
only half the amount of insulin for an FPU as is used for one carbohydrate unit
[15 ]. However, Pańkowska et
al. studied only children with type 1 diabetes [12 ]
[13 ], and Lee et al. [15 ] investigated different means of bolus
administration (normal bolus vs. dual wave bolus) without randomization.
Many studies about insulin adjustment after the intake of fat- and/or
protein-rich meals have been performed in children, who have a rather short diabetes
duration and rather low insulin demand. Therefore, we conducted a randomized
controlled study and investigated the two above-mentioned algorithms in comparison
to no FPU insulin adjustment in an adult sample of type 1 diabetes patients. We
expected that the adjustment of prandial insulin doses for FPU would optimize
post-prandial glycemic control compared to that in a non-adjusted condition. We
selected the percentage of post-prandial glucose values in the range of ≥ 70
mg/dl and ≤ 180 mg/dl [16 ] as the primary outcome variable to “punish” hyper-
and hypoglycemic glucose excursions equally.
Material and Methods
Study design and participant recruitment
This study was a monocenter, open-label, parallel, randomized, controlled
crossover trial. The study design is presented in [Figure 1 ].
Fig. 1 The study flow chart
The study was carried out in an inpatient setting at the Diabetes Center
Mergentheim. Study participants were eligible if they had type 1 diabetes
mellitus. Additional inclusion criteria were treatment with insulin pump therapy
(CSII), age ≥ 18 to ≤ 70 years, and sufficient knowledge of the
German language. Exclusion criteria were severe general illness, severe
psychiatric illness, gastroenterological disease, impaired kidney function, and
heart attack, stroke/TIA, or vascular surgery within the past six months.
Participants with changes in basal rate during the experimental days were also
excluded. Before inclusion, both oral and written information about the study
was provided to all participants, who then provided written informed
consent.
The study was reviewed and approved by the federal agency (Federal Institute for
Pharmaceuticals and Medical Products, BfArM), and the local ethics committee
(State Medical Association of Baden-Württemberg) and is registered in
the EU Clinical Trials Register (EudraCT Number 2017-001807-62).
Procedures
On three consecutive days, the study participants received a standardized
carbohydrate-reduced, high-fat, high-protein evening meal, the Viennese-style
chicken schnitzel breaded with potato gratin and salad (702.6 kcal, 52.7 g fat,
33.8 g protein, 23.3 g carbohydrates). The test meal contained 2.3 carbohydrate
units (CU) and 6 fat protein units (FPUs). The prandial insulin dose was
calculated using three different algorithms as defined in the clinical
investigation plan. The sequence of the three conditions was randomized.
Algorithms for Insulin Dose Adjustment
Algorithms for Insulin Dose Adjustment
Condition A (non-FPU)
Standard adjustment considering only the carbohydrate content of the meal with an
insulin dosage based on the carbohydrate-to-insulin ratio (CIR). Insulin was
delivered as a dual bolus (50% before eating, 50% delayed over
four hours).
Condition B (FPU 100%)
This algorithm took into account carbohydrates as well as fat and protein
content. The carbohydrate content was covered by the individual CIR. For each
FPU, a factor of 100% of the CIR was used. To cover carbohydrates,
insulin was delivered as a dual bolus (50% before eating; 50%
delayed over eight hours), and to cover FPU, insulin was delivered as a delayed
bolus over eight hours.
Condition C (FPU 50%)
This algorithm considered carbohydrates as well as fat and protein content. The
carbohydrate content was covered by the individual CIR. For each FPU a factor of
50% of the CIR was used. To cover carbohydrates, insulin was delivered
as a dual bolus (50% before eating; 50% delayed over eight
hours), and to cover FPU, insulin was delivered as a delayed bolus over eight
hours.
Blood Glucose Control
Before the start of the test meal (at 6:00 p.m.), blood glucose was determined based
on a routine capillary blood glucose measurement with a point-of-care (POC) blood
glucose meter (Glucometer PRO, BST Biosensor Technology GmbH, Berlin, Germany). The
prandial insulin doses were adjusted based on this POC measurement. The course of
glucose after the start of the test meals was determined using a continuous
interstitial glucose measurement system (FreeStyle Libre, Flash Glucose Monitoring
System (FGM), Abbott, Wiesbaden, Germany). Interstitial glucose values were scanned
at least before the start of the test meals, before going to bed, in the morning
after the test meals, and when routine blood glucose measurements were done. For
safety reasons four additional routine blood glucose measurements provided for
inpatient treatment were performed at 10:00 p.m., 00:00 a.m., 3:00 a.m., and 6:00
a.m. using a POC blood glucose meter. Glucose values that were too high (up to 150
mg/dl above the starting value before the test meal) were not corrected unless
ketone bodies were additionally detected in the urine (acetone test two-fold or
three-fold positive). If blood glucose was too low at the nighttime blood glucose
checks, the participants received 10–20 g of fast absorbable rescue
carbohydrates.
Follow-up
In the morning of the day following the test meal, the glucose data of the FGM
system and the insulin data of the CSII systems were readout. The occurrence of
POC confirmed level 1 hypoglycemia (<70 mg/dl) or hyperglycemia
(> 180 mg/dl), and the amount of rescue carbohydrates was recorded. The
occurrence of (severe) adverse events (AEs/SAEs) or (severe) adverse device
effects (ADEs/SADEs) were also recorded.
Outcomes
The primary study outcome was the percentage of interstitial glucose values in
the target range (≥ 70 to ≤ 180 mg/dl) within a period of 12
hours after taking the test meal.
Secondary study outcomes were the percentage of interstitial hypoglycemic glucose
values (<70 mg/dl) and hyperglycemic values (> 180 mg/dl) within
12 hours after the test meal. The percentage of POC confirmed level 1
hypoglycemia (<70 mg/dl) or hyperglycemia (> 180 mg/dl) and the
amount of rescue carbohydrates for treating POC-confirmed hypoglycemia.
Statistical Methods
Statistical analysis
In this study, a total of 26 people with type 1 diabetes were examined. The power
consideration of such a sample size statistically confirmed an effective size of
0.7 standard deviations with a power of β = 0.8 (two-sided
α = 0.025; due to alpha adjustment for multiple testing).
The full analysis dataset consisted of participants who took part in all three
test conditions. The primary and secondary outcomes were analyzed by repeated
variance analysis controlled for order effects. Specific contrasts between
single test conditions were determined if the overall test was significant. The
significance level for all tests was p <0.05.
Randomization and masking
Randomization was done centrally at the study coordinating center by staff who
were not involved in the recruitment or treatment of study participants. A
randomization sequence was generated with SYSTAT 12.0 with a 1:1:1
allocation.
The study physician received sealed envelopes with the respective allocation of
the order of prandial insulin doses. After obtaining written informed consent,
the envelope was opened. Masking of study participants and study personnel was
not feasible because of the nature of the intervention.
Results
Study participants
The study was carried out at the Diabetes Center Mergentheim from September 2017
to January 2018. A total of 35 participants who met all inclusion criteria and
who had given informed consent were recruited. One participant was excluded
again due to an unstable basal rate (exclusion criterion). Four participants
withdrew their consent to continue the study before they went through all three
test conditions. Thirty participants completed the study as per the study
requirement (participation in all three test conditions). Four of these
participants had to be excluded from analysis because of protocol violations
regarding insulin administration or because of the lack of sensor data (flash
glucose monitoring, FGM) during the outcome phase (device failure). [Figure 2 ] shows the flow diagram of
patient recruitment and data analysis.
Fig. 2 Glucose course under the three test conditions between 5:00
p.m. (pre-prandial phase) and 6:00 a.m. the next day (confidence
intervals that do not intersect mean a significant difference between
the corresponding conditions).
[Table 1 ] shows the demographic and
diabetes-specific characteristics of the sample. Study participants were
middle-aged adults with long-standing type 1 diabetes and under insulin pump
therapy for several years. With a mean HbA1c of 8.3%, glycemic control
needed improvement. [Table 2 ]
summarizes the primary and secondary outcome glucose parameters during the
12-hour follow-up period. [Figure 3 ]
shows the glucose course under the three test conditions between 5:00 p.m.
(pre-prandial phase) and 6:00 a.m. the next day.
Fig. 3 Glucose course under the three test conditions between 5:00
p.m. (pre-prandial phase) and 6:00 a.m. the next day (confidence
intervals that do not intersect mean a significant difference between
the corresponding conditions).
Table 1 Demographic and diabetes-specific characteristics
of the study sample (N=26).
Characteristic
Mean±SD resp. N (%)
Range
Age (years)
40.8±14.0
18.2–63.3
Female gender (N %)
13 (50)
BMI (kg/m2 )
27.2±5.0
19.4–42.3
HbA1c (%)
8.3±1.5
5.7–11.6
Diabetes duration (years)
22.8±12.0
7.8–47.8
CSII therapy duration (years)
7.6±7.7
0.0–29.8
Table 2 Primary and secondary outcome glucose parameters
during the 12-hour follow-up period (N=26).
Glucose parameter
Condition A (non FPU)
Condition B (FPU 100%)
Condition C (FPU 50%)
overall p
p Difference Conditions A/B
p Difference Conditions A/C
Primary Outcome
Percentage of glucose values in range>70 to ≤
180 mg/dl
73.84±23.46
56.40±15.89
64.84±24.01
0.043
0.241
0.930
Secondary Outcomes
Percentage of glucose values ≤ 70 mg/dl
5.54±11.30
30.80±13.99
21.81±20.16
<0.001
<0.001
0.002
Percentage of glucose values>180 mg/dl
20.62±24.91
12.80±14.14
13.35±20.80
0.260
0.441
0.531
Mean post-prandial glucose values in mg/dl
144.47±37.51
107.91±24.49
116.42±39.05
0.003
<0.001
0.009
Mean pre-prandial glucose values (start of meal at 6:00 p.m.)
(mg/dl)
132.3±39.8
142.5±49.2
131.2±42.0
0.888
1.000
1.000
Mean amount of hypoglycaemia rescue carbohydrates (g)
4.8±9.3
39.2±22.5
16.8±17.9
<0.001
<0.001
0.001
Mean number of POC blood glucose measurements ≤ 70
mg/dl
0.1±0.3
1.2±1.0
0.6±0.8
<0.001
<0.001
0.002
Glycemic outcomes
The bolus administered according to the above-described algorithms was
3.45±1.38 IU for algorithm A (non-FPU), 10.90±4.60 IU for
algorithm B (FPU 100%), and 7.0±2.67 IU for algorithm C (FPU
50%). Mean pre-prandial glucose values at the beginning of the test
meals were comparable among groups (p =0.593).
There was a significant difference between the three test conditions considering
the primary outcome percentage of time in the range of ≥ 70
mg/dl to ≤ 180 mg/dl. However, the direction of the
difference was in contrast to the original expectation. The adjustment of
insulin with 100% CIR per FPU showed a significantly less time in the
range than with non-adjustment of prandial insulin for FPU. The adjustment with
50% CIR per FPU also showed a lower percentage of glucose values in the
range of 70 mg/dl to 180 mg/dl than with non-adjustment of
prandial insulin for FPU, but this difference was not significant. The
proportion of hypoglycemic glucose values was significantly higher if prandial
insulin was adjusted for FPU with 100% CIR and 50% CIR per FPU
(30.8% and 21.8% vs. 5.5% hypoglycemic values). The
nadir of glucose was reached at 3:00 a.m. and 4:00 a.m. in conditions B and C,
respectively.
The proportion of hyperglycemic values was not significantly affected when
considering the FPU for prandial insulin adjustment. The high prevalence of
biochemical hypoglycemia captured with ongoing FGM in conditions B and C
compared to condition A was corroborated by a significantly higher number of
nocturnal confirmatory POC blood glucose measurements<70 mg/dl.
At the same time, significantly more hypoglycemia rescue carbohydrates were
administered to stabilize the glucose level in conditions B and C than in
condition A. The only outcome which was in favor of the adjustment of prandial
insulin for FPU was the mean nocturnal post-prandial glucose level, which was
significantly lower in conditions B (107.9 mg/dl) and C (116.4
mg/dl) compared to non-adjustment for FPU (144.5 mg/dl).
Safety
No serious adverse event occurred during the study. The nocturnal monitoring of
blood glucose values ensured that no severe hypoglycemia event demanding the
injection of glucose or glucagon occurred.
Discussion
This study clearly demonstrates that additional prandial insulin for a high-fat and
-protein content of a meal does not have an advantage for a higher postprandial time
in range. In contrast, additional insulin for FPUs resulted in a higher risk of
postprandial hypoglycemia. The same was true for a reduced dose of additional
prandial insulin for FPU in terms of the time in range and hyperglycemia. Although
lower mean nocturnal glucose levels were within the range of 70–180
mg/dl when insulin was given for FPUs, one must note that these were
obtained by a very high proportion of hypoglycemic values. The results from an FGM
device were confirmed by POC blood glucose readings. Particularly, nocturnal glucose
had to be stabilized by approximately 40 g and 17 g of hypoglycemia rescue
carbohydrates, respectively, using a high-dose algorithm and a low-dose algorithm.
These results indicate that an adjustment of the prandial insulin dose for FPU in
the tested dosage cannot be recommended for adults with type 1 diabetes. The tested
dosage of prandial insulin was obviously too high to avoid post-prandial
hypoglycemia in the next 12 hours. The fact that the nadir of glucose was reached
at
3.00 a.m. and 4:00 am, respectively, in conditions B and C, raises the question if
the extension of the bolus over 8.00 hours is also disadvantageous besides a too
high amount of additional bolus insulin due to the FPE adjustment. A shorter
extension of the bolus may have avoided low glucose values between 3.00 am and 4:00
am in morning. The study had following strengths and limitations. The cross-over
design of this study had the advantage of allowing all participants to serve as
their own control group, as all participants took part in all three study
conditions. The carry-over effect was also limited, as can be seen from the
pre-prandial blood glucose values before the test meal. All participants were
insulin pump users. Thus, it can be expected that the effect of delayed resorption
of carbohydrates due to the relatively high-fat content could be controlled by
delayed boluses. A further limitation of the study was that the carbohydrate-related
amount of insulin was delayed over 4 hours in condition A and over 8 hours in
conditions B and C.
A cross-over design also limits the number of meals that can be tested. The study
tested only one meal, with a specific composition, in which 67.6% of the
energy content were delivered by fats, 19.2% by proteins, and 13.2%
by carbohydrates. In their review, Bell et al. [3 ] suggested that not only does the amount of fat and protein per se have
a hyperglycemic effect, but also the composition of the macronutrients that can
alter the demand of prandial insulin. Therefore, it is difficult to generalize
findings to other fat- or protein-rich meals with varying compositions of these
macronutrients.
In our study, the algorithms from Pańkowska et al. [11 ], which recommends the use of an
insulin-to-FPU ratio equal to the CIR, and from Lee et al. [15 ], which recommends the use of an
insulin-to-FPU in a ratio equal to 0.5 of the CIR, lead to an over-insulinization
with a high incidence of biochemical hypoglycemia. Outside an inpatient setting with
close surveillance of nocturnal hypoglycemia, such algorithms can be qualified as
dangerous.
Clearly, more research is needed to determine the amount of FPUs and insulin that
should be taken into account when calculating pre-prandial insulin dosage.
A smart and promising research strategy for getting more evidence on FPU adjustment
could be the use of closed-loop systems for such research questions on nutrition.
Wolpert et al. [8 ] analyzed the dosing of
prandial insulin after a low-fat and high-fat dinner while keeping the carbohydrate
content equal in both study conditions in people with type 1 diabetes using a
closed-loop system. They observed that on doubling the fat content of the dinner,
prandial insulin doses increased from 9.0 insulin units to 12.6 insulin units
(+40%). This increase in prandial insulin dose in the high-fat
condition in their closed-loop study was remarkably smaller than that observed in
our study (+100 % resp. +200%). Wolpert et al. [8 ] also observed marked interindividual
differences in the additional insulin requirements due to a high-fat content.
In summary, the results of this study could not demonstrate an advantage of
post-prandial glucose excursion when using insulin for FPUs. In fact, using insulin
for FPUs resulted in an over-insulinization with an elevated risk of hypoglycemia.
Clearly, more research with sound methodology is needed before general adjustments
of prandial insulin dosing for adults with type 1 diabetes can be established.
Authors Contributions
All authors were involved in planning the study. EH and MK collected the data, TH
and
BLG monitored the correct conduct of the study. NH and MK analyzed the data. MK and
NH wrote the manuscript. MK, NH, and TH contributed to the discussion, reviewed and
revised the manuscript. All authors have read and approved the final manuscript.
