Cerebral Blood Flow and Glucose Metabolism in Appetite-Related Brain Regions in Type 1 Diabetic Patients After Treatment With Insulin Detemir and NPH Insulin: A randomized controlled crossover trial

From January 2009 until May 2011, patients were included in this randomized controlled crossover trial; the last follow-up visit was on 13 December 2011. Thirty-five patients with type 1 diabetes, aged 18–60 years and with a BMI of 18–35 kg/m², were included; they were recruited from the outpatient clinic of the VU University Medical Center (VUMC) and from neighboring hospitals. After giving written informed consent, all participants had a screening visit consisting of a medical history, physical examination, and fasting blood and urine analyses. Exclusion criteria were diabetes duration <1 year; A1C >8.5%; proliferative retinopathy; a history of recurrent severe hypoglycemia (defined as an episode that requires external assistance for recovery); a medical history of hypoglycemia unawareness; history of cardiovascular, renal, or liver disease or severe head trauma; any neurological or psychiatric disorder; endocrine diseases not well controlled for the last 3 months; inability to undergo magnetic resonance imaging (MRI) scanning; substance abuse; and the use of anticoagulants, oral steroids, or any centrally acting agent. Of all patients in analysis, one had microalbuminuria, four stable background retinopathy, and one peripheral neuropathy (Toronto score [18] of 9/19 and a vibration perception [19] threshold of >25 V at 5 of 12 locations). Three patients were treated with antihypertensive medication (one used an angiotensin II receptor antagonist [ARB], one an ACE inhibitor and an ARB, and one an ACE inhibitor and ARB, a diuretic, and a calcium antagonist). Three patients used cholesterol-lowering medication, and one used aspirin as well. Two patients had stable hypothyroidism treated with thyroxin, and one had stable ulcerative colitis treated with mesalazin. The study was approved by the Medical Ethics Review Committee of the VUMC and the Central Committee on Research involving Human Subjects. The study was conducted according to the Declaration of Helsinki.

The study was conducted in a randomized crossover design and was part of a larger trial (clinicaltrials.gov, clinical trial reg. no. NTC00626080). Primary outcomes were CBF and CMR_glu after a 12-week treatment period, and change in body weight after this 12-week treatment was a secondary outcome measurement. After a run-in period of at least 4 weeks, during which the current insulin therapy was optimized, patients were randomly assigned to start with either insulin detemir or NPH insulin in the evening, both in combination with insulin aspart at mealtimes. Randomization (block design) was conducted by the Trial Pharmacy of the VUMC, and the assigned treatments were concealed by envelopes; a research physician (L.W.v.G.) enrolled patients in the study and assigned them to the intervention. After assignment, no blinding was applied, since NPH insulin needs to be mixed and visually inspected before injection. Weekly seven-point self-measured blood glucose curves were made, and all fasting blood glucose levels were reported. Where appropriate, basal insulin dose was adjusted to maintain a fasting glucose level of <7 mmol/L. Regular telephone contact was available for advice on basal and prandial insulin adjustments. After 12 weeks of treatment, patients switched from basal insulin.

On the day prior to the scan session, patients refrained from food, alcohol, and coffee intake from 2200 h onward. They were carefully instructed not to forget their basal insulin injection and, if possible, not to use any insulin aspart after their dinnertime injection. Telephone calls were made both on the night before and early in the morning of the day of the PET scan, i.e., before traveling to the hospital. In addition, a similar protocol was followed at the day of MRI scanning (a week prior to the PET scan), when patients had to arrive at the hospital at the same time in a fasting state, using the same basal insulin the night before. If necessary, the insulin regimen was adjusted after the MRI scan to improve fasting glucose levels on the day of the PET scan. Patients arrived at the hospital at 0715 h in the fasting state and remained fasted during the entire imaging procedure. Upon arrival, a catheter was placed in an antecubital vein for blood collection and tracer injection. Blood glucose levels were checked and corrected if necessary (when glucose was <4 mmol/L and falling or when glucose was >15 mmol/L). To prevent further rising of glucose during the remaining duration of the test visit, a low dose of the individual's basal insulin was injected subcutaneously. No insulin aspart was used to avoid interference with the PET measurements. After we check for collateral circulation and administration of local anesthesia using intradermal 1% lidocain, a radial artery was cannulated by an experienced anesthesiologist. Both cannulas were kept patent by a 3 IE/mL 0.9% NaCl heparin solution.

Before and immediately after scanning, patients completed a questionnaire, scoring their hunger (“How hungry are you right now?”), fullness (“How full are you at this moment?”), appetite (“How much do you feel like eating right now?”), prospective consumption (“How much could you eat right now?”), desire to eat (“How strong is your desire to eat right now?”), and thoughts of eating (“How much do you think about food right now?”) on a 10-point Likert scale. Furthermore, patients scored their insulin treatment satisfaction using the Diabetes Treatment Satisfaction Questionnaire, which measures satisfaction with treatment regimen, perceived frequency of hyperglycemia, and perceived frequency of hypoglycemia over the past few weeks (20).

Three-dimensional structural MRI images were acquired on a 3.0 T GE Signa HDxt scanner (General Electric, Milwaukee, WI), using a T1-weighted fast Spoiled Gradient echo sequence. PET scans were acquired with a High Resolution Research Tomograph (HRRT) (Siemens/CTI, Knoxville, TN) PET scanner. The scanning protocol consisted of a [¹⁵O]H₂O scan to measure CBF and an [¹⁸F]FDG scan to measure CMR_glu. Details on scan protocol have previously been published (21). During both scans, arterial concentrations were monitored continuously, and in addition, manual samples were taken for cross-calibration of the measured input function. Samples obtained during the [¹⁸F]FDG scan (15, 35, and 55 min postinjection) were also used to measure arterial plasma glucose levels. All scans were performed between 0930 and 1200 h to minimize diurnal variations.

List mode emission data were histogrammed into multiframe sinograms, which subsequently were normalized, and corrected for randoms, dead time, decay, scatter, and attenuation. Fully corrected sinograms were reconstructed using the standard 3D Ordinary Poisson Ordered-Subsets Expectation Maximization (OP-OSEM) reconstruction algorithm (22), resulting in 207 image planes with 256 × 256 voxels and a voxel size of 1.22 × 1.22 × 1.22 mm³ (21). The effective spatial resolution of the reconstructed images was ∼3 mm.

MRI and PET images were coregistered using the software package VINCI (23). PET images were rebinned, and PET and MRI images were cropped into a 128 × 128 × 126 matrix (21). Regions of interest (ROIs) were delineated on the MRI scan using the template defined in PVElab (24). Subsequently, all ROIs were projected onto the dynamic PET images, generating time activity curves (TACs) for the following 16 left and right regions: orbitofrontal cortex, anterior and posterior cingulate cortex, thalamus, insula, caudate nucleus, putamen, medial inferior frontal cortex, superior temporal cortex, parietal cortex, medial inferior temporal cortex, superior frontal cortex, occipital cortex, sensorimotor cortex, cerebellum, hippocampus, a single white matter region, a total gray matter region, and striatum (putamen and caudate nucleus combined). Of these ROIs, the first seven were of specific interest, as these are involved in appetite regulation and reward.

With use of standard nonlinear regression (NLR), appropriately weighted [¹⁵O]H₂O TACs were fitted to the standard one-tissue compartment model (25) to obtain regional CBF values. In addition, parametric (voxel-wise) CBF images were generated from 6-mm full-width-at-half-maximum Gaussian smoothed dynamic [¹⁵O]H₂O images using a basis function method (BFM) implementation of the same model (26).

With use of a standard NLR algorithm, appropriately weighted [¹⁸F]FDG TACs were fitted to an irreversible two-tissue compartment model with three rate constants and blood volume as fit parameters. Next, the net rate of influx K_i was calculated as K₁ ⋅ k₃/(k₂+k₃), where K₁ is the rate of transport from blood to brain, k₂ the rate of transport from brain to blood, and k₃ the rate of phosphorylation by hexokinase. Finally, K_i was multiplied with the plasma glucose concentration and divided by a lumped constant (LC) of 0.81 (27) to obtain regional CMR_glu values. In addition, parametric CMR_glu images were generated using Patlak linearization (28).

Capillary blood glucose (patient monitoring) was measured using a blood glucose meter (OneTouch UltraEasy; LifeScan, Milpitas, CA). Arterial glucose samples (to determine CMR_glu) were measured using the hexokinase method (Glucoquant; Roche Diagnostics, Mannheim, Germany). A1C was measured by cation-exchange chromatography (reference values 4.3–6.1%; Menarini Diagnostics, Florence, Italy). Serum insulin concentrations were quantified using immunometric assays (Centaur; Siemens Diagnostics, Deerfield, IL); insulin detemir levels were divided by four to compensate for the difference in molar dose ratio relative to NPH insulin. Urine microalbumin was quantified using immunonephelometry (Immage 800; Beckman Coulter, Brea, CA).

Data are expressed as mean ± SD. Skewed data and ordinal values are expressed as median and interquartile (IQ) range. Differences between both insulin treatments were tested by repeated-measures analysis or the Wilcoxon signed rank test (insulin detemir vs. NPH insulin). Analyses were performed using SPSS for Windows, version 20.0 (SPSS, Chicago, IL). P < 0.05 was considered statistically significant.

Parametric images were analyzed using SPM8 software (Wellcome Trust Centre for Neuroimaging, London, U.K.). Parametric images were smoothed using a 6-mm full-width-at-half-maximum Gaussian kernel, coregistered to corresponding T1-weighted MRI images and normalized to Montreal Neurological Institute space. Paired t tests were performed (insulin detemir vs. NPH insulin).

With use of data of 18 paired H₂O PET measurements and an expected difference in total gray matter CBF of 15% (0.046 ± 0.05 mL ⋅ cm⁻³ ⋅ min⁻¹), our study had a power of 96% (α 0.05) to detect differences between treatment with insulin detemir and NPH insulin. With use of 24 paired FDG PET data and an expected difference in total gray matter CMR_glu of 7.5% (0.011 ± 0.02 μmol ⋅ cm⁻³ ⋅ min⁻¹), our study had a power of 73% to detect differences between treatments.

During the study, one patient dropped out during his first treatment period (because of NPH insulin schedule difficulties) and one in the second period (because of a hip fracture). Owing to technical problems (n = 2) and patient movement (n = 2), combined [¹⁸F]FDG and [¹⁵O]H₂O data were discarded for these four subjects. [¹⁵O]H₂O was not available for one patient on both occasions and for three patients on one occasion. After quality control of the remaining scans, paired CMR_glu data were available in 24 patients and paired CBF measurements in 18 patients.

Subject characteristics of all 28 patients included in the analyses are listed in Table 1. Of all patients included in the analyses (n = 28), 15 patients started with NPH insulin and 13 with insulin detemir. Of patients starting with NPH insulin, 5 had used insulin detemir and 10 insulin glargine, while of those starting with insulin detemir, 4 had used insulin detemir, 1 NPH insulin, and 8 insulin glargine before the trial. At the end of the treatment period, daily insulin doses and A1C did not differ between treatment (Table 2). Insulin detemir decreased body weight by 0.7 kg, whereas NPH insulin increased weight by 0.6 kg (between-treatment difference 1.3 kg, P = 0.02) (Table 2). Perceived hyperglycemia and hypoglycemia did not differ significantly between treatments (Diabetes Treatment Satisfaction Questionnaire); patient satisfaction was significantly greater when with use of insulin detemir than NPH insulin (P = 0.003). Irrespective of the treatment arm, patients scored five of six items (hunger, appetite, prospective consumption, desire to eat, and thoughts of eating) significantly higher after the scan than before the scan (P < 0.01 for each item), indicating that appetite increased during the scanning period (all were fasting). When treated with insulin detemir, patients scored higher on the sixth item, i.e., fullness, after the PET scan than patients treated with NPH insulin (mean 4.0 [IQ range 3.0–5.0] vs. 3.0 [2.0–4.0], P = 0.03 for between-group difference).

For insulin detemir, on the day of the PET scan, three patients, of whom two were excluded afterward from the CBF analyses, required several dextrose tablets to prevent or resolve a mild hypoglycemia, whereas six patients, of whom one was excluded from the CBF analyses, received ∼20 mL i.v. 20% glucose before the scan to prevent hypoglycemia. One patient received insulin detemir (12 IU s.c.) because glucose was rising upon arrival at the hospital. For NPH insulin, three patients, of whom two were excluded from the CBF analyses, required dextrose tablets because of a low or falling blood glucose level, whereas two patients, who were afterward excluded from the CBF analyses, received ∼15 mL i.v. 20% glucose before the PET scan started. Three patients, who all were included in the CBF analyses, required insulin NPH insulin (14, 10, and 5 IU s.c.) at arrival in the hospital as a result of hyperglycemia. In all patients, average arterial glucose levels were stable within 10% and >5.0 mmol/L during data acquisition. For checking whether acute glucose manipulations had affected PET measurements of CBF and CMR_glu, a separate analysis was performed in which patients who had received glucose or insulin were excluded. Results of this additional analysis, however, were similar to those of the original analysis (data not shown).

NLR analysis showed that, after treatment with insulin detemir compared with treatment with NPH insulin, CBF was higher in all regions. This was statistically significant in most appetite-related brain regions—bilateral insula, bilateral putamen and right caudate nucleus, right thalamus, and bilateral anterior and right posterior cingulate cortices—when patients received insulin detemir versus NPH insulin (Table 3). In addition, higher CBF was observed in the right medial inferior frontal cortex, bilateral parietal cortex, and bilateral sensorimotor cortex (all P < 0.05) after treatment with insulin detemir versus NPH insulin. In all other brain regions investigated, CBF was similar for both treatments. Results were similar after exclusion of patients using antihypertensive medication (n = 3) and after exclusion of the one left-handed patient. After adjustment for A1C, glucose, and insulin levels, CBF differences in appetite-related regions remained unaltered (data not shown). No significant correlation between changes in CBF and changes in glucose, insulin, and A1C levels or body weight was found. Regional analyses of parametric images showed good correlation with regional NLR analyses (slope = 0.99, R² = 0.93, for n = 5 NPH and n = 5 insulin detemir, data not shown; similar to data obtained in healthy subjects [21]). These parametric analyses (voxel level) did not provide additional findings relative to regional NLR analyses.

During the [¹⁸F]FDG scan, mean arterial plasma glucose levels did not differ between treatments; serum insulin levels were similar as well (Table 2). NLR analysis showed no significant differences in CMR_glu in appetite-related predefined (PVElab) regions (Table 3). No significant differences in transport parameters for total gray matter (K_i, K₁, k₂, and k₃), were observed (data not shown), and total gray matter CMR_glu did not differ significantly between treatments (0.15 ± 0.02 μmol ⋅ cm⁻³ ⋅ min⁻¹ after treatment with NPH insulin versus 0.16 ± 0.02 μmol ⋅ cm⁻³ ⋅ min⁻¹ after treatment with insulin detemir). Parametric analyses yielded similar results (data not shown).

The main finding of this study was that a relative loss in body weight in type 1 diabetic patients treated with insulin detemir was accompanied by an increase in CBF in insula, thalamus, anterior and posterior cingulate cortex, and striatum—regions that are involved in appetite regulation and reward. No significant differences in CMR_glu between groups were found.

Several studies have investigated the effects of body weight on CBF. Some of these studies suggest that changes in CBF are causal in the development of obesity. CBF responses in appetite-related brain regions to a meal in formerly obese persons were similar to those in obese persons but different from those in lean subjects (29), indicating a predisposition to obesity that may involve areas of the brain that control complex aspects of eating behavior. This is in line with the observed increase in CBF in appetite-regulating brain regions in response to meal consumption in successful dieters (30). In minipigs, however, diet-induced obesity resulted in a decrease in CBF in several of these brain regions, suggesting that the changes in CBF were the result of weight gain (31). From the current study, it is not possible to determine whether increases in CBF in patients treated with insulin detemir are cause or consequence of the observed weight loss. Previous studies in mice and healthy humans, however, showed cortical brain activation upon acute insulin detemir versus human insulin infusion with concomitant decrease in food intake (9–11). In addition, it was shown that insulin-induced glucose lowering in type 1 diabetic patients resulted in an increase in CBF (32,33). However, whether this was caused by increasing insulin or by decreasing glucose levels could not be determined in those studies. Still, a direct effect of insulin on the brain is supported by the acute effects of insulin on cerebrovascular responses in rats (17). The present CBF findings are in contrast with a study by Hirvonen et al. (13) in eight healthy volunteers and six individuals with impaired glucose tolerance, in which no between-group CBF differences were observed and no CBF effect of insulin. In their study, acute clamp-induced hyperinsulinemic (insulin levels 5–6 times higher compared with the current study) euglycemia was imposed, which is different from the insulin effect of two chronic 12-week treatment periods. In addition, fasted, elevated (glucose level 11 mmol/L) glucose levels during PET data acquisition were higher in the current study. In addition, Hirvonen et al. investigated two different subject groups, whereas we investigated only one group of individuals with type 1 diabetes and studied the effects of a chronic treatment in a crossover study design. Finally, Hirvonen et al. may not have observed the 10% difference owing to a lack of power (although insulin levels were higher, the number of subjects was much less than in the current study) or the lower signal-to-noise ratio of the PET scanner used.

In contrast to the differential effects on CBF, the two insulin treatments did not result in significant differences in CMR_glu in any of the regions investigated. Previous studies have shown an inverse association of CMR_glu and BMI (34) and increases in CMR_glu after stimulation with food pictures (35,36). Of note, the increase in CMR_glu in appetite-related brain regions after insulin infusion was blunted in insulin-resistant men compared with insulin-sensitive men (14), and it was associated with insulin resistance and overweight.

Previously, it was shown that in type 1 diabetes changes in k₃ are observed compared with healthy volunteers (37), without significant concomitant changes in CBF. Under the assumption of absence of between-group differences in phosphorylation (which were indeed absent in the present data), the relationship between CMR_glu and CBF is nonlinear [as CMR_glu and K₁ are linearly related via CMR_glu = K_i ⋅ glucose/LC, where K_i = K₁ ⋅ k₃/(k₂ + k₃), and CMR_glu is linearly related to E ⋅ CBF, where E = 1 – exp(−PS/CBF) (38,39)], and, especially at higher flow values, an increase in CBF will induce a smaller increase in CMR_glu (37), which is what was observed in the current study, although the latter was nonsignificant.

Possible confounders that could have accounted for the differences in CBF include A1C or prevailing glucose and insulin levels. However, these parameters were not significantly different between treatments, and the insulin detemir–induced increase in CBF was similar after adjustment for A1C, glucose, and insulin levels.

Limitations of this study include its nonblinded nature owing to differences in insulin formulations. NPH insulin is a cloudy suspension that needs to be thoroughly stirred before injection, whereas insulin detemir is a clear, colorless solution that does not require stirring. Therefore, it was not possible to perform a double-blind study. Worldwide, however, NPH insulin is the standard (intermediate) long-acting human insulin and, therefore, the best active comparator. Moreover, even if patients were aware of the type of insulin treatment, it is unlikely that this will have had an effect on the present findings. It should be noted that not all patients in the study were insulin detemir naïve, i.e., five and six patients starting with NPH insulin and insulin detemir, respectively, already used insulin detemir before the start of the study. As insulin detemir–naïve patients and insulin detemir users were equally distributed between treatment groups, it is unlikely that medication prior to the study has affected the results, especially since PET scans were performed after 12 weeks of exposure to the test insulin.

Differences in CMR_glu between insulin detemir and NPH insulin were not statistically significant. Data in the current study were obtained during a resting and fasting condition. In future studies, it may be of interest to investigate responses to (visual) food stimuli in appetite regulating brain regions after both treatments. However, due to radiation exposure and practical reasons (small inner diameter of the HRRT scanner, making it difficult to present visual stimuli), this was not possible in the current study. In addition, for detection of changes in brain activation using [¹⁸F]FDG PET, two separate sessions are required to test stimulated versus nonstimulated conditions (35,36).

Some patients required glucose or (basal) insulin to prevent emerging hypo- or hyperglycemia, respectively. In six patients on insulin detemir versus one on NPH insulin, glucose was necessary to prevent low or falling blood glucose levels, which could have biased results, as hypoglycemia increases CBF (32,33). As three patients in the insulin detemir versus only one in the NPH insulin group required additional basal insulin to avoid hyperglycemia, one could argue that if acute injection of basal insulin would have affected CBF, this would have attenuated the difference in CBF between the groups. More importantly, the increase in CBF in the detemir versus NPH group remained unchanged after exclusion of patients who had received insulin or glucose.

Although weight gain associated with insulin treatment is relevant for type 1 diabetic patients, it is especially important for patients with type 2 diabetes. It is tempting to generalize the present findings to type 2 diabetes, but further studies in these patients are needed, especially since central insulin resistance possibly plays a role in type 2 diabetes.

The current study focused on insulin detemir action in the brain. It should be noted, however, that other mechanisms have been proposed to explain its weight-reducing effect. These include less defensive eating due to less hypoglycemia, increased energy expenditure, and higher insulin levels in the liver compared with peripheral tissue, although none of these could be firmly established (40–43). In the current study, no significant differences in perceived hypoglycemia frequency were found between treatments.

In conclusion, the present findings support the hypothesis that a differential effect on CBF, measured during a resting, fasting condition, may contribute to the consistently observed weight-sparing effect of insulin detemir treatment.

This work was supported by an investigator-initiated grant of Novo Nordisk A/S. Novo Nordisk supplied all insulin preparations. M.D. is a member of the advisory board of Abbott, Eli Lilly, Merck Sharp & Dohme (MSD), Novo Nordisk, Poxel Pharma, and Sanofi; a consultant for AstraZeneca and Bristol-Myers Squibb; and a speaker for Eli Lilly, MSD, Novo Nordisk, and Sanofi. Through M.D., the VUMC receives research grants from Amylin/Eli Lilly, MSD, Novo Nordisk, and Sanofi; M.D. receives no personal payments in connection to the above-mentioned activities—all payments are directly transferred to the Institutional Research Foundation. No other potential conflicts of interest relevant to this article were reported.

L.W.v.G. participated in the design of the study; performed the study, PET analyses, and statistical analyses; drafted the manuscript; edited the text; and made crucial revisions to the manuscript. R.G.I. clinically supervised the study, clinically commented on the manuscript, edited the text, and made crucial revisions to the manuscript. M.C.H. supervised the PET analyses, critically commented on the manuscript, edited the text, and made crucial revisions to the manuscript. J.F.H. clinically supervised the study, critically commented on the manuscript, edited the text, and made crucial revisions to the manuscript. R.P.H. was involved with patient recruitment, edited the text, and made crucial revisions to the manuscript. M.L.D. participated in the design of the study, edited the text, and made crucial revisions to the manuscript. A.A.L. participated in the design of the study, supervised PET analyses, critically commented on the manuscript, edited the text, and made crucial revisions to the manuscript. M.D. participated in the design of the study, edited the text, and made crucial revisions to the manuscript. R.G.I., M.C.H., A.A.L., and M.D. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Parts of this study were presented in abstract form (for n = 20) at BRAIN 2011, Barcelona, Spain, 24 May 2011; the 71st Scientific Sessions of the American Diabetes Association, San Diego, California, 24–28 June 2011; and the 47th Meeting of the European Association for the Study of Diabetes, Lisbon, Portugal, 12–16 September 2011.

The authors thank Arjen Binnerts (Zaans Medisch Centrum), Alex Arntzenius (Spaarne Ziekenhuis), Cees Rustemeijer (Ziekenhuis Amstelland), Jeroen de Sonnaville and Karin Daemen (Tergooi Ziekenhuizen), and Sytze van Dam and Teri Brouwer (Onze Lieve Vrouwe Gasthuis) for their help with patient recruitment; Nikie Hoetjes (VUMC) for data acquisition; the radiochemistry staff of the Department of Nuclear Medicine and PET Research (VUMC) for tracer production and blood sample analyses; Frederik Barkhof (VUMC) for MRI assessments; and Patrick Schober and Lothar Schwarte (VUMC) for arterial cannulation.