Background
Cardiovascular disease including heart failure is the leading cause for morbidity and mortality in people with diabetes mellitus making adequate therapy mandatory [
1]. Diastolic cardiac function is already impaired in the pre-diabetic phase (impaired glucose tolerance) and brings with it an increased risk of heart failure [
2,
3]. Based on the association of cardiac dysfunction with fasting and postmeal metabolic control [
4,
5], diagnosis and monitoring of sub-clinically impaired cardiac function may be valuable for monitoring therapeutic efficacy [
6]. As well in the pre-diabetic phase, the prevalence of myocardial infarction is alarming [
7]. As consequently suggested, the improvement of postprandial metabolism, that is a reduction of glucose excursions, should be made a cornerstone in metabolic control for the prevention of cardiovascular disease [
8‐
10], thereby shifting the focus from previous landmark studies related to the HbA1c, an overall mean glucose value, to cardiovascular risk. Taking this risk and that of heart failure in diabetes mellitus into account, the best risk–benefit ratio exists for metformin and insulin [
11]. In particular, the more recently-developed short-acting insulin analogues have shown superior control of postprandial glucose levels [
12] and a reduction in cardiovascular events [
13]. Likewise long-acting analogue insulin preparations with their flatter profile have shown advantages in day-to-day glucose variability compared to human insulin [
14]. Consequently, the combination of short and long-acting insulin analogues (AI) in a multiple daily injection regimen (MDI) theoretically offers cumulative effects for the reduction of glucose excursions/variability.
Whether MDI with AI can improve myocardial dysfunction in patients with type 2 diabetes has not yet been assessed. As a proof of concept, this prospective, randomized, open, long-term study tested the hypothesis that MDI with AI improves postprandial glucose better than human insulin (HI). Furthermore, a beneficial effect of AI on diastolic cardiac function was to be evaluated.
Results
There were no significant differences between the pre-study insulin regimens of the different treatment groups with regards to analogue or human insulin, basal bolus or other insulin strategies. Participants of both study groups were comparable for demographic, hemodynamic, cardiac and metabolic data including cardiovascular risk factors and concomitant medication (Table
1).
Table 1
Demographics and baseline data of patients
Age | Years | 60.4 ± 9.5 | 63.1 ± 10.6 | 0.159 |
Sex male | n (%) | 46 (75) | 31 (65) | 0.218 |
BMI | kg/m2
| 32 ± 5 | 32 ± 6 | 0.709 |
Duration of diabetes | Years | 10 {5–15} | 9.5 {5–14} | 0.580 |
Insulin dose MDI | IE/d | 56 {37–83} | 50 {36–70} | 0.485 |
Metformin | n (%) | 13 (21) | 16 (33) | 0.159 |
Hypertension | n (%) | 58 (95) | 42 (88) | 0.153 |
Smoking | n (%) | 17 (28) | 15 (31) | 0.700 |
Myocardial infarction | n (%) | 14 (23) | 8 (17) | 0.417 |
Betablocker | n (%) | 23 (38) | 18 (38) | 0.983 |
ACE-inhibitor | n (%) | 37 (60) | 33 (69) | 0.224 |
Calcium channel blocker | n (%) | 9 (15) | 10 (21) | 0.406 |
AT1 blocker | n (%) | 15 (25) | 8 (17) | 0.314 |
Diuretics | n (%) | 23 (38) | 20 (42) | 0.674 |
Statins | n (%) | 28 (46) | 25 (52) | 0.522 |
Aspirin | n (%) | 30 (49) | 23 (48) | 0.896 |
Glucose fasting | mg/dl | 163 ± 54 | 163 ± 45 | 0.962 |
Glucose pp | mg/dl | 183 ± 69 | 185 ± 50 | 0.870 |
HbA1c | % | 7.3 ± 1.7 | 7.7 ± 1.8 | 0.237 |
Intact proinsulina
| pmol/l | 5 {2.7–7.6} | 4.8 {3.2–8.2} | 0.515 |
Triglycerides | mg/dl | 137{102–177} | 117 {94–184} | 0.463 |
Cholesterol | mg/dl | 187 ± 36 | 188 ± 42 | 0.890 |
LDL | mg/dl | 114 ± 34 | 113 ± 37 | 0.941 |
HDL | mg/dl | 46 {38–53} | 48 {39–54} | 0.543 |
Creatinine | mg/dl | 0.9 {0.8–1.08} | 0.9 {0.7–1.1} | 0.493 |
ASAT | U/l | 29 {24–40} | 27 {23–39} | 0.862 |
CRP | mg/dl | 0.28 {0.13–0.5} | 0.3 {0.15–0.58} | 0.552 |
S’ | cm/s | 7.6 ± 1.1 | 7.6 ± 1 | 0.984 |
E’ | cm/s | 7.8 ± 1.4 | 8.2 ± 1.7 | 0.168 |
LV end-diastolic diameter | mm | 44 {42–48} | 44 {41–48} | 0.776 |
Septal thickness | mm | 12 {11–14} | 12 {11–13} | 0.449 |
Mitral E/A | | 0.9 {0.7–1.1} | 0.9 {0.8–1.1} | 0.651 |
E/E’ | mmHg | 8.9 {7.5–10.9} | 8.1 {7.0–9.6} | 0.174 |
Intima media thickness | mm | 0.7 ± 0.2 | 0.7 ± 0.2 | 0.714 |
Εlasticity modulus Ɛ | kPa | 140 {113–228} | 165{136–259} | 0.087 |
Pulse wave velocity | m/s | 7.3 {6.6–9.2} | 8.3 {7.0–9.9} | 0.151 |
Heart rate | bpm | 71 {63–78} | 69 {64–80} | 0.509 |
Systolic blood pressure | mmHg | 140 {129–152} | 140{123–150} | 0.712 |
Diastolic blood pressure | mmHg | 83 ± 12 | 81 ± 9 | 0.321 |
Of the 109 patients, 5 patients in group AI and 4 in group HI were lost to follow-up at the 24-month visit, and 8 patients in each group at the 36-month visit (Fig.
1). The resulting data were analyzed both using conventional statistics and the multiple imputation technique as shown in brackets [ ]. The mean treatment period was 32 ± 7 months.
At 36 months, the daily insulin dose per kg body weight had significantly increased: by 0.07 ± 0.28 U/kg with AI and by 0.18 ± 0.26 IU/kg with HI (inter-group significance p = 0.054). Weight had increased from baseline to month 36 with AI by 2.3 ± 7.8 kg (p = 0.050) and by 2.8 ± 7.6 kg (p = 0.038) with HI (inter-group significance p = 0.774).
At 36 months (Table
2), the primary endpoint, postprandial glucose, had decreased by 20 ± 62 mg/dl (p = 0.038) with AI but had a non-significant increase by 9 ± 53 mg/dl (p = 0.33) with HI thereby resulting in a significant difference between the groups (p = 0.032 [p = 0.047]). Simultaneously, the postmeal glucose levels (161 ± 39 with AI vs. 195 ± 54 mg/dl with HI) were relevantly and significantly different (p = 0.002 [p = 0.010]) as were the postmeal glucose excursions (15 ± 43 with AI vs. 57 ± 40 mg/dl with HI, p = 0.001 [p = 0.007]).
Table 2
Changes at 36 months from baseline
Glucose fasting (mg/dl) | −7 [−12] ± 52 [53] | 0.375 [0.095] | −17 [−19] ± 42 [44] |
0.021 [0.008]
| 0.336 [0.510] |
Glucose pp (mg/dl) | −20 [−20] ± 62 [64] |
0.038 [0.021]
| 9 [4] ± 53 [56] | 0.335 [0.676] |
0.032 [0.047]
|
HbA1c (%) | −0.4 [−0.6] ± 1.5 [1.7] | 0.063 [0.011]
| −1.3 [−1.2] ± 1.8 [1.7] |
0.000 [0.000]
|
0.024 [0.082] |
Intact proinsulina (pmol/l) | −0.9 {−2.3 to 0.8} | 0.171 | 0.3 {−0.9 to 2.7} | 0.199 |
0.048
|
Triglycerides (mg/dl) | 0 [1] {−12 [−18] to 33 [34]} | 0.496 [0.577] | 11 [8] {6 [−15] to 41 [41]} |
0.016 [0.167] | 0.153 [0.467] |
Triglycerides pp (mg/dl) | 10 [9] {−15 [−22] to 39 [45]} | 0.081 [0.198] | 24 [17] {−4 [−10] to 64 [55]} |
0.023 [0.040]
| 0.317 [0.424] |
Cholesterol (mg/dl) | 10 [10] ± 49 [50] | 0.161 [0.186] | 6 [6] ± 43 [42] | 0.437 [0.375] | 0.668 [0.699] |
LDL (mg/dl) | 5 [3] ± 40 [43] | 0.420 [0.685] | −7 [−5] ± 39 [38] | 0.262 [0.408] | 0.170 [0.367] |
HDL (mg/dl) | 1 [0] {−3 [−5] to 4 [5]} | 0.636 [0.628] | 2 [1] {−6 [−6] to 9 [9]} | 0.493 [0.487] | 0.844 [0.668] |
Creatinine (mg/dl) | 0 [0] {−0.10 [−0.10] to 0.10 [0.10]} | 0.954 [0.710] | 0.10 [0.09] {0 [−0.03] to 0.10 [0.15]} |
0.018 [0.009]
| 0.054 [0.064] |
ASAT U/l | −2 [−1] {−7 [−11] to 3 [6]} | 0.096 [0.400] | 1 [1] {−4 [−11] to 10 [11]} | 0.586 [0.734] | 0.126 [0.554] |
hsCRP (mg/dl) | 0 [−0.01] {−0.18 [−0.24] to 0.1 [0.14]} | 0.468 [0.521] | 0.05 [0.01] {−0.1 [−0.19] to 0.15 [0.18]} | 0.307 [0.669] | 0.199 [0.525] |
LV enddiast. diameter (mm) | 2 [2] {−2 [−2] to 6 [5]} |
0.041 [0.032]
| 1 [1] {0 [−1] to 3 [3]} | 0.149 [0.151] | 0.387 [0.395] |
Septal thickness (mm) | 0 [−0.1] {−2 [−1.4] to 2 [2]} | 0.948 [0.733] | 0 [0] {−2 [−1.6] to 1 [1.7]} | 0.718 [0.706] | 0.768 [0.722] |
Post. wall thickness (mm) | 1 [0.8] {−1 [−0.9] to 2 [1.4]} | 0.103 [0.082] | 1 [0.9] {0 [−0.4] to 2 [2]} |
0.010 [0.012]
| 0.388 [0.501] |
S’ (cm/s) | 0 [0] ± 0.8 [0.8] | 0.889 [0.801] | 0.1 [0.1] ± 1 [0.9] | 0.504 [0.571] | 0.516 [0.753] |
E’ (cm/s) | 0.4 [0.5] ± 1.4 [1.4] | 0.069 [0.013]
| 0.1 [0] ± 1.3 [1.3] | 0.738 [0.853] | 0.312 [0.059] |
E’ pp (cm/s) | 0.6 [0.6] ± 1.4 [1.4] |
0.009 [0.002]
| 0.2 [0.1] ± 1.4 [1.5] | 0.435 [0.430] | 0.241 [0.110] |
E/E’ (mmHg) | −0.1 [−0.01] {−1.52 [−1.54] to 1.54 [1.46]} | 0.870 [0.657] | 0.35 [0.25] {−0.9 [−1.02] to 1.99 [1.98]} | 0.136 [0.182] | 0.192 [0.198] |
IMT (mm) | 0.03 [0.05] ± 0.17 [0.18] | 0.200 [0.048]
| 0.06 [0.07] ± 0.19 [0.18] | 0.091 [0.018]
| 0.551 [0.649] |
Elasticity modulus (kPa) | 15 [8] {−36 [−40] to 49 [60]} | 0.363 [0.463] | 30 [8] {−41 [−65] to 78 [68]} | 0.313 [0.622] | 0.643 [0.740] |
PWV (m/s) | 0.2 [0.1] {−0.9 [−0.9] to 0.9 [1.1]} | 0.684 [0.494] | 0.5 [0] {−1 [−1.4] to 1.7 [1.4]} | 0.319 [0.622] | 0.448 [0.693] |
Heart rate (bpm) | −4 [−4] {−10 [−10] to 5 [4]} | 0.057 [0.026]
| −4 [−4] {−7 [−7] to 3 [3]} |
0.026 [0.079] | 0.920 [0.630] |
Systolic BP (mmHg) | 3 [7] {−9 [−9] to 20 [20]} | 0.175 [0.087] | 20 [17] {−5 [−4] to 29 [29]} |
0.008 [0.002]
| 0.065 [0.075] |
Diastolic BP (mmHg) | 5 [5] ± 12 [12] |
0.005 [0.003]
| 10 [11] ± 12 [12] |
0.000 [0.000]
| 0.062 [0.064] |
Within the metabolic secondary endpoints, fasting glucose did not significantly decrease with AI but did significantly decrease with HI (Table
2) to comparable levels in both groups (150 ± 34 with AI vs. 144 ± 40 mg/dl with HI, p = 0.433). Intact proinsulin in the fasting state decreased insignificantly with AI and increased insignificantly with HI, resulting in a significant difference between these changes (p = 0.048). HbA1c decreased by 0.4 ± 1.5 %, p = 0.063 [p = 0.011] with AI and by 1.3 ± 1.8 %, p = 0.001 [p = 0.001] with HI. This larger decrease (p = 0.024 [p = 0.082]) resulted in comparable levels (6.7 ± 0.8 % with AI vs. 6.5 ± 0.7 % with HI, p = 0.486). Only with HI did fasting and postprandial serum triglyceride levels increase significantly as did serum creatinine with a trend to significance for this change between the groups (Table
2).
Within the cardiovascular secondary endpoints, there was no significant change of E’ with HI (Table
2) but postmeal E’ improved with AI by 0.6 ± 1.4 cm/s, p = 0.009 [p = 0.002] and fasting E’ as a trend in traditional statistics (by 0.4 ± 1.4 cm/s, p = 0.069) and significantly by multiple imputation analysis [0.5 ± 1.4 cm/s, p = 0.013]. Accordingly, diastolic dysfunction [
12] was reduced by 22 % with AI and by 12 % with HI. The traditional parameter of diastolic function, E/A, significantly improved with AI, both in the fasting state and postmeal, and as a trend with HI (Table
3). The mitral inflow E increased significantly postmeal with AI but decreased as a trend with HI. With values close to normal, E/E’ did not change significantly. None of the changes in diastolic function parameters achieved significance between the treatment groups.
Table 3
Parameters of diastolic function in the fasting state and post meal (pp)
E’ (cm/s) | Analogue Human | 7.8 ± 1.4 8.2 ± 1.7 | 8.4 ± 1.7 8.2 ± 1.7 | 0.069 0.738 |
E’ pp (cm/s) | Analogue Human | 7.6 ± 1.6 8.1 ± 1.8 | 8.5 ± 1.6 8.2 ± 1.8 |
0.009
0.435 |
E/A | Analogue Human | 0.96 ± 0.3 0.93 ± 0.25 | 1.11 ± 0.39 1.01 ± 0.30 |
0.030
0.061 |
E/A pp | Analogue Human | 0.94 ± 0.30 0.92 ± 0.26 | 1.18 ± 0.46 1.01 ± 0.32 |
0.005
0.055 |
E (cm/s) | Analogue Human | 70 ± 17 70 ± 20 | 74 ± 19 72 ± 17 | 0.296 0.097 |
E pp (cm/s) | Analogue Human | 71 ± 18 74 ± 20 | 76 ± 21 72 ± 15 |
0.001
0.068 |
A (cm/s) | Analogue Human | 76 ± 18 75 ± 21 | 70 ± 19 74 ± 23 | 0.061 0.368 |
A pp (cm/s) | Analogue Human | 71 ± 18 74 ± 20 | 68 ± 18 75 ± 22 | 0.430 0.957 |
E/E’ (mmHg) | Analogue Human | 8.9 {7.5–10.9} 8.1 {7.0–9.6} | 8.5 {7–10.3} 8.4 {7.5–10.0} | 0.870 0.136 |
E/E’ pp (mmHg) | Analogue Human | 8.3 {7.1–10.1} 8.3 {7.3–9.9} | 8.3 {7.2–10.8} 8.1 {7.4–10.8} | 0.670 0.203 |
Compared to HI, treatment with AI tended to result in smaller increases in systolic and in diastolic blood pressure (Table
2). In parallel, vascular stiffness was lower after 36 months of therapy, expressed as elasticity modulus ε (154 {105–204} vs. 182 {148–234} kPa in HI, p = 0.028 [p = 0.198].
After 36 months, there were two cardiovascular events with AI (myocardial infarctions) vs. five events with HI (one myocardial infarction, four strokes), equivalent to event-free percentages of 96 vs. 81 % (p = 0.142). Anti-hypertensive medication remained comparable between both groups.
Discussion
This prospective, randomized, open, controlled trial showed that long-term MDI with AI in patients with type 2 diabetes resulted in better postprandial glucose control than MDI with HI. This benefit with AI was associated with improved diastolic cardiac function whereas this function remained unchanged with HI.
The improvement of the primary endpoint postmeal glucose levels following an MDI with AI regimen could be mainly attributed to the effects of the short-acting analogues. This effect has been shown in type 1 and in type 2 diabetes [
12,
14,
23,
24]. The study design also provided a potentially beneficial effect from the basal analogue insulin which should result in a relative protection of the beta-cell from the flatter and longer action profile of the basal insulin [
14]. An augmented beta-cell function may also be related to an improvement in insulin resistance. Our data suggest an at least relative improvement of insulin resistance by the unchanged triglycerides levels with AI versus the significant rise with HI—in spite of similar weight changes and identical nutritional recommendations in both groups.
By contrast to these specifically different metabolic effects, both treatment regimens improved long-term glucose control to a similar HbA1c level after 36 months. This similarity in long-term glucose control may result from three aspects: first, a compensation of the lower postprandial glucose levels by the higher fasting glucose levels in the analogue regimen and vice versa of the higher postprandial glucose levels by the lower fasting levels with HI. Second, a numerically higher baseline HbA1c level in the HI group (7.7 vs. 7.3 %, Table
1) allowed for a more pronounced lowering; and third, a more pronounced increase in the daily insulin dose per kg body weight with HI.
Comparative trials using the same target for MDI with different insulin types allow for minimizing potential differences in glucose control so that differences in cardiovascular risk may be assessed. In the presence of comparable fasting glucose and HbA1c levels, the results can thereby be interpreted for ultimate risk–benefit assessments including diastolic cardiac function. Improvement of diastolic dysfunction, that is the antecedent to diastolic heart failure i.e. heart failure with preserved ejection fraction (HFpEF), has been a challenging topic in the treatment of patients with type 2 diabetes. Remarkably, there is no evidence-based therapy available against HFpEF [
2]. Diastolic dysfunction has high prevalence in overweight individuals with type 2 diabetes [
2,
6,
18] and results in a poor prognosis and reduced quality of life because of the associated limitation of physical activities in daily life. However, it is under-diagnosed in clinical routine and lacks evidence-based treatment strategies due to problems associated with the bashful and resigned obese patients [
6], with the measuring technique [
6,
18] and with the understanding of the complex underlying mechanisms [
25‐
27].
With regards to diagnostic issues stemming from the measuring technique, quantitative pulsed tissue Doppler imaging emerged 20 years ago and has since provided very good results for the sensitive measurement of myocardial velocity E’ during early diastole [
26], with good feasibility even in overweight individuals at bedside, sensitivity to energy restraint, correlation to exercise capacity, prognostic value and the potential to monitor therapeutic interventions [
6,
28]. Furthermore, E’ allows for the quantification of diastolic dysfunction by using the dominant influence of aging on E’ [
18], which is essential for the mathematically-correct identification of risk factors and for the diagnosis of dysfunction/HFpEF. In line with earlier reports [
2,
6] diastolic dysfunction was highly prevalent in the study cohort and was normalized in 22 % of the patients on AI treatment vs. 12 % of the patients on HI treatment.
In the present study design, E’ was averaged from 6 basal myocardial regions and, therefore, taken as measure of global left ventricular function. The increase of E’ with AI was mirrored by the decrease of the traditional parameter E/A in the fasting state and postmeal. Accordingly, this study demonstrated a long-term improvement in diastolic function using MDI with AI (Table
3). However, the third parameter of diastolic function, the ratio E/E’, did not change significantly with the treatment interventions due to the parallel modifications of E and E’ (Table
3). This thereby substantiated concerns in earlier reports that this estimation of LV filling pressure may be a less sensitive measure of diastolic function, especially in metabolic disease with its predominantly mild degree of diastolic dysfunction [
18,
29].
It is of interest that the observed changes in diastolic function were more distinct in the postmeal state. As for the complex mechanisms and potential factors that may influence diastolic dysfunction [
18,
25], this study did not demonstrate any improvement in the hemodynamic factors blood pressure or vascular stiffness. Within metabolic factors, fasting glucose and HbA1c had improved to comparable levels at 36 months. In line with recent reports [
30‐
32], these data do not support an automatic translation of improved glycemic control into augmented cardiac function. Improvements in postprandial glucose excursions with AI were paralleled by functional amelioration, whereas the HI group demonstrated excessively high levels of postmeal glucose and its variability. These observations rather support the importance of controlling postprandial glucose and the underlying insulin resistance as therapeutic targets for improvement of cardiac function [
3], at least in moderately well-controlled type 2 diabetes, in line with earlier reports [
5,
18,
27,
28,
31,
33].
Pharmacological therapy with MDI or rosiglitazone and maintaining an adequate diet [
5,
28,
34] have been seen to improve diastolic cardiac function in type 2 diabetes along with improved postmeal glucose. This is in agreement with the reported link between acute glucose fluctuations and activation of oxidative stress [
35]. A recent study demonstrated an immediate and highly repetitive deterioration of cardiac function after a carbohydrate meal associated with increased oxidative stress in overweight and insulin-dependent patients with type 2 diabetes [
36]. The reverse being also true, cardiac function has been shown to improve when adhering to diets that effectively lower weight or at least postprandial glucose levels [
28,
34,
37,
38]. Small pharmacological studies have suggested positive effects on diastolic dysfunction from optimized metabolic control in patients with type 1 and type 2 diabetes [
6] in particular by improving postmeal glucose control, insulin resistance and reducing oxidative stress [
5,
37]. By using a practical combination of therapeutic strategies in the form of adapted anti-diabetic medication and adhering to effective nutrition [
28,
39] in line with the most recent dietary treatment guidelines [
40], improved postmeal glucose control appears to be a promising target for the improvement of diastolic dysfunction.
The underlying biochemical mechanisms are complex and not yet fully understood. Important metabolic mechanisms imply that a reduction of insulin resistance [
28,
31,
37], oxidative stress and augmented NO bioavailability [
28,
41,
42] enhance both cardiac efficiency i.e. intracardiocyte energy production [
43,
44] and endothelial function that translates into augmented myocardial perfusion and function. Indeed, small clinical studies in type 2 diabetes have demonstrated that improved metabolic control may enhance myocardial perfusion and function [
4] and that short-acting analogue insulin normalizes myocardial perfusion in the postmeal state by 35 % when compared to being left untreated [
45]. When looked at together, these observations and the present study suggest that analogue insulin may result in optimized postmeal regulation of myocardial perfusion and hence diastolic function in type 2 diabetes. Because of its significant clinical relevance, further studies which explore the major mechanisms of postmeal (dys-) regulation of myocardial perfusion in diabetic individuals would be highly welcome.
It is conceivable that other more underlying mechanisms for diastolic dysfunction relate to insulin resistance [
27,
33], as also suggested from more recent studies [
3,
18,
32,
43]. For the present insulin-dependent patients with diabetes, however, the assessment of insulin resistance cannot be based on the HOMA-IR, but rather on surrogate parameters such as the daily insulin dose. The insulin dose per kg body weight increased by 0.07 U/kg with AI vs. 0.19 IU/kg (p = 0.054) with HI. Other surrogate parameters are the triglyceride levels which remained unchanged with AI but increased with HI therapy both fasting and postmeal. These data indicate not only the degree of impact insulin resistance has as a therapeutic target in type 2 diabetes. It also suggests selecting analogue over human MDI, possibly due to the amelioration of postmeal blood glucose excursions with the consequent reduction of acute glucose toxicity. In this respect, an insulin pump therapy using short-acting insulin analogues may be an even more interesting model for patients with type 2 diabetes: Glucose variability is further reduced compared to a MDI with short and long-acting insulin analogues [
46].
Much more knowledge is needed with regards to the effects and reversibility of insulin resistance for optimizing fat and glucose dysmetabolism—especially postmeal [
47]. Clinical studies are needed which take into account well-defined metabolic and lifestyle characteristics so that treatment strategies can evolve that make use of the inherent reversibility potential from postprandial glucose and fat dysmetabolism. This affects both the evaluation of traditional anti-diabetic medication and the most recent therapeutic developments [
48,
49].
In spite of the improvements in postprandial metabolic control and diastolic cardiac function by AI, cardiovascular events were not significantly different between both treatment groups. This is due to the low patient number (n = 109) event rate (group A: 2 events; group H: 5 events) and observation period in comparison to the Steno study (n = 180; observation period 13.3 years) or the UKPD study (n = 5102; 20 years [
10,
50]. Given the prognostic value of diastolic cardiac function, this study promotes the use of more actual measures of positive clinical outcome, such as diastolic cardiac function, to assess therapeutic efficiency in the treatment of type 2 diabetes.
Limitations
Several study limitations need to be addressed. One of these is the considerable number of patients who withdrew their study participation consent or were lost during follow up. However, the results nevertheless demonstrated a significant outcome for the primary endpoint indicating that the study had indeed remained sufficiently powered. In order to evaluate the problem of missing data, an additional statistical technique, the multiple imputation technique [
21,
22], was applied that was specifically developed for analyzing data sets with missing entries. The similarity of these results to the results from traditional statistical analysis supports the relevance of the study data and their statistical meaning in spite of the drop-out rate.
The larger number of patients in the human insulin group withdrawing their participation consent before the onset of therapy may have introduced a retrospective randomization bias. However, patients were free to withdraw participation at any time during the study and it seems a number had signed consent only in the hope to be randomized to a more modern and more expensive analogue insulin. These patients, in particular, were very motivated to apply optimal anti-diabetic care so that their absence may have contributed to the initially slightly higher HbA1c level in the human insulin group.
The study design did not allow for diagnosing and assessing insulin resistance by HOMA-IR in the insulin-dependent patients with type 2 diabetes. An assessment by clamp technique would have been beyond the scope of this study. Altogether, this diagnostic problem simply reflects an unsolved but important issue for the tailoring of optimal metabolic control in diabetology and should be further evaluated with respect to the specific effects of dietary and/or exercise intervention in support of pharmaceutical treatment in different stages of insulin resistance.
The improvement of diastolic cardiac function with AI compared to the unchanged function with HI did not result in inter-group significance in the fasting state (p = 0.312 [p = 0.059] or postmeal (p = 0.241 [p = 0.110] due to an unexpectedly high standard deviation of the change from baseline, so that the study turned out underpowered for comparative assessment of cardiac function. This limitation, however, should not detract from the clinically-relevant message that MDI with AI does improve diastolic cardiac function in moderately well-controlled patients with type 2 diabetes.
Authors’ contributions
HvB conceived of the study, participated in the design of the study and drafted the manuscript. ThS participated in the design and coordination of the study, was involved in patient recruitment and treatment, and substantially helped draft the manuscript. IK coordinated cardiological and other data from patients during their visits and made substantial contributions to the analysis and interpretation of data. MR managed the diabetological control of patients and data acquisition. TiS performed the statistical analysis and drafted the respective section of the manuscript. PS participated in design and coordination of the study and gave final approval of the version to be published. All authors read and approved the final manuscript.