The hippocampus and connected limbic brain structures are essential for the conscious retention and recollection of facts and events, that is, for the formation of declarative memory (Squire and Zola, 1996; Eichenbaum, 1999, 2004). These brain regions display a high density of insulin receptors (Baskin et al, 1994; Wickelgren, 1998), implicating that central nervous insulin signaling is involved in declarative memory processing. This assumption is supported by studies in animals as well as in humans. In rats, insulin receptor expression and phosphorylation in the hippocampus are upregulated after water maze training (Zhao et al, 1999). In humans, euglycemic intravenous infusions of insulin enhance both recall of previously learned words (Kern et al, 2001; Craft et al, 2003) and neuronal activity within the medio-temporal lobe (Rotte et al, 2005). However, owing to the strong systemic effects of intravenous insulin administration, this method does not permit assessing long-term effects of insulin on memory. Intranasal administration of bioactive compounds has been demonstrated to effectively deliver drugs to the brain without inducing systemic side effects (Born et al, 2002; Ross et al, 2004; Thorne et al, 2004; for review see Illum, 2000). Recent studies have revealed beneficial effects of acute and long-term (8 weeks) intranasal administration of regular human insulin (RH-I) on declarative memory in humans (Benedict et al, 2004; Reger et al, 2006). As RH-I molecules tend to self-associate into dimeric, tetrameric, and hexameric units, their absorption after subcutaneous administration is delayed (Kang et al, 1991). In the insulin analog insulin aspart (ASP-I), the amino-acid proline in position B28 is replaced by aspartic acid, reducing the tendency of the insulin molecule to self-associate (Brange et al, 1990; Brange and Volund, 1999), whereas the binding profile to the insulin receptor is the same as of RH-I (Kurtzhals et al, 2000). Clinical studies have shown that after subcutaneous administration, ASP-I induces a faster onset of the hypoglycemic effect than RH-I owing to its faster reabsorption from the tissue into the blood (Gammeltoft et al, 1999). As the pharmacokinetic difference between ASP-I and RH-I may affect their ability to enter the brain and to induce central nervous system effects after intranasal administration, we examined whether ASP-I may improve declarative memory function in humans better than RH-I.


36 men (age 18–35 years, body mass index, BMI<25 kg/m2) without personal or family history of diabetes were examined in a double-blind, between-subject comparison. Subjects underwent a physical examination to ensure they were healthy. Ten hours before testing they had to fast and to abstain from coffee and alcoholic beverages. The study was approved by the local Ethics Committee on Research Involving Human Subjects, and written informed consent was obtained from all subjects.

Subjects were randomly assigned to three groups (each 12 men), which were adjusted for age (RH-I: 24.92±1.63 years, ASP-I: 24.42±1.33 years, placebo: 26.25±1.66 years) and BMI (RH-I: 22.60±0.60 kg/m2, ASP-I: 22.98±0.59 kg/m2, placebo: 23.24±0.43 kg/m2). The data of the RH-I and placebo conditions were partly derived from a male subsample of eight subjects per group from a previously published study (Benedict et al, 2004). Both groups (n=12) were each supplemented by four subjects to match all groups according to age and BMI. During a 2-week baseline phase, all subjects received placebo. During the following 8-week treatment period, subjects were intranasally administered RH-I (Insulin Actrapid® HM, Novo Nordisk, Mainz, Germany), ASP-I (Insulin NovoLog® HM, Novo Nordisk, Mainz, Germany), or placebo (HOE 31 dilution buffer for H-Insulin, Aventis Pharma, Bad Soden, Germany) in the morning, around noon, in the evening, and before going to bed. Each dose consisted of either 0.4 ml ASP-I or RH-I (each containing 40 IU, respectively) or vehicle administered within four puffs of 0.1 ml (two per nostril), amounting to 1.6 ml (160 IU) insulin or vehicle per day. Based on previous experiments (Born et al, 2002), a single dose of 40 IU insulin was expected to induce temporary increases in cerebrospinal fluid concentrations of insulin distinctly above the normal level in healthy individuals. Sprays were stored in a refrigerator at 4°C and were replaced by new substance every 7 days. In order to assure compliance, subjects kept a protocol on their intake routine.

Declarative memory testing relied on the oral presentation of standardized lists of 30 nouns at a rate of 1 word/s (Fruehwald-Schultes et al, 2000; Kern et al, 2001). After a break of 3 min, subjects wrote down within 90 s all words they still remembered. In the delayed recall sessions, that is, 1 week later, subjects again had to write down all words they memorized from this list. The number of correctly recalled words and the number of words falsely remembered from previously presented lists were registered. Experimental sessions were scheduled at 0800 hours and took place (A) at the beginning of the baseline phase, (B) at the beginning of the treatment Phase, and (C) 1 week before the end of the 8-week treatment phase. Session A yielded baseline values of immediate memory recall. Session B allowed to assess the acute effects of ASP-I, RH-I, or placebo on immediate memory recall. The effects of a subchronic administration of insulin on immediate memory recall were examined in session C. Delayed memory recall was tested in separate sessions taking place 1 week after the immediate sessions (see also Figure 1). All sessions were conducted 60 min after the administration of placebo, except for session (B) when initial doses of 40 IU ASP-I, 40 IU RH-I, or placebo were given 60 min before testing in order to examine acute effects of insulin. At the end of sessions A and B and at the end of the treatment period (ie, 1 week after session C), blood samples were collected for the determination of serum insulin (Pharmacia Insulin RIA100, Pharmacia & Upjohn Inc., Uppsala, Sweden) and of plasma glucose (by the hexokinase method; Abbott; Wiesbaden, Germany).

Figure 1
figure 1

Time schedule. Three groups of 12 males each were intranasally treated with placebo for 2 weeks. Then, one group received RH-I, the second group received insulin aspart (ASP-I), and the third group continued with placebo. Intranasal treatments were performed four times per day (160 IU insulin/day). Test session took place in the beginning of the study (session A, Se A), after the first intranasal administration of insulin (session B, Se B) and in the seventh week of the treatment period (session C, Se C). Delayed recall of words learned in each of these sessions was tested 1 week later (DeRec), respectively.

Data Reduction and Analysis

One subject of the RH-I group did not participate in the last delayed word recall testing due to illness and was excluded from analysis. For baseline adjustment, values of the baseline session were subtracted from treatment values for each individual. The differences of word list recall performances were subjected to ANOVA with the repeated measures factor time. After these global analyses had yielded significant treatment × time interaction effects, separate analyses for the acute and subchronic treatment effects on word list recall were performed with one-way ANOVA with the between-subject factor treatment condition. For hormonal parameters and plasma glucose, ANOVA with repeated measures (between-factor: treatment; within-factor: time) were calculated. Where appropriate, single time points were compared with t-tests for independent samples. A p-value <0.05 was considered significant. Degrees of freedom were adjusted using the Greenhouse–Geisser correction.


During the baseline period, immediate and delayed word list recall performances did not differ between conditions (ASP-I vs RH-I vs placebo, immediate: 11.25±0.93 vs 10.75±0.93 vs 12.17±0.93, F(2,35)=0.60, p>0.55; delayed: 5.83±1.01 vs 6.75±1.01 vs 8.33±1.01, F(2,35)=1.57, p>0.22). The word list recall performance displayed a significant time × treatment interaction (F(2,32)=13.91, p<0.001). Post hoc analyses revealed that memory performance was significantly improved by insulin in the final delayed recall of words (ie, after 8 weeks of treatment; F(2,34)=10.12, p<0.001 for ANOVA main effect; Figure 2). Although word recall after RH-I was significantly enhanced when compared to placebo (F(1,21)=1.18, p<0.03), ASP-I exerted an even stronger beneficial effect on delayed word recall (F(1,21)=0.33, p<0.05 compared to RH-I; F(1,22)=3.81, p<0.001 compared to placebo). Regarding the number of intrusions, that is, words remembered from other lists than the one presented before recall, no significant differences were found between conditions either in this session (ASP-I: 1.50±0.51 words, RH-I: 1.27±0.41 words, placebo: 1.67±0.57 words; F(2,34)=0.15, p>0.86 for main effect) or in the delayed recall following 1 week after acute administration of ASP-I, RH-I, and placebo (ASP-I: 1.42±0.45 words, RH-I: 1.25±0.43 words, placebo: 1.25±0.28 words; F(2,35)=0.06, p>0.94 for main effect). After acute insulin administration (session B), no effects on immediate (F(2,35)=0.81, p>0.92 for main effect) and delayed recall of words, F(2,34)=0.25, p>0.78) were found (Figure 2). Also, prolonged treatment with insulin compounds did not affect immediate recall of words measured in session C (F(2,34)=0.01, p>0.92 for main effect, Figure 2). Consistent with previous findings (Kern et al, 1999; Born et al, 2002) levels of plasma glucose and serum insulin were neither affected by acute nor by subchronic intranasal insulin administration (Table 1).

Figure 2
figure 2

Acute and subchronic effects of intranasal insulin on (a) immediate and (b) delayed word recall. Word recall was tested 3 min after presenting a list of 30 words (immediate recall) and 1 week later (delayed recall). Words were presented 60 min after administering 40 IU of ASP-I (black), RH-I (gray), and placebo (white, session B), and 60 min after placebo administration following 7 weeks of ASP-I, RH-I (each 4 × 40 IU/day) or placebo (session C). Note that delayed recall testing did not assess acute treatment effects but those of 1-week (acute) and 8-week (long-term) administration, respectively. Data were baseline-adjusted by subtracting values of the baseline session from treatment values. Significant differences between conditions are indicated (*p<0.05; **p<0.01).

Table 1 Plasma Glucose and Serum Insulin Levels after Acute and Long-Term Intranasal Insulin Administration


The present study compared the cognitive effects of intranasal administration of the insulin analog ASP-I with those of RH-I and of placebo. As reported previously (Benedict et al, 2004), 8 weeks of RH-I administration significantly improved declarative memory performance as assessed by a delayed recall of words learned 1 week earlier. Eight weeks of ASP-I administration even exceeded this effect and yielded significantly improved delayed word recall both in comparison with the placebo condition and with RH-I. Our data indicate that the beneficial effect of intranasal insulin on memory can be enhanced by administering pharmacokinetically altered insulin analogs. On the background of increasing evidence that Alzheimer's disease (AD) is a neuroendocrine disorder with strikingly reduced CNS expression of genes encoding insulin, IGF-I and IGF-II, as well as the insulin and IGF-I receptors (Rivera et al, 2005), our finding of improved memory performance after intranasal intake of insulin and its analogs may be of significance for the treatment of memory impairments. This is supported by findings of facilitated recall of verbal memory after intravenous and intranasal treatment with insulin in memory-impaired adults (Craft et al, 1999; Reger et al, 2006).

Most recently, also acute effects of intranasal insulin on declarative memory have been reported, demonstrating that recall of previously learned words was enhanced after administration of up to 40 IU of RH-I in patients suffering from AD (Reger et al, 2006). However, corresponding with our results, healthy control subjects of this study did not benefit from acute insulin treatment. Whereas AD patients might display a higher sensitivity to the cognitive effects of central nervous insulin owing to their lower CSF-to-plasma insulin ratio (Craft et al, 1996), our results show that neither a single dose of 40 IU of RH-I nor of ASP-I is potent enough to exhibit effects on declarative memory.

Regarding the delayed recall of words, the insulin-induced improvement observed in our study occurred on the background of a generally decreasing performance (F(1,84)=11.09, p<0.02 for overall ANOVA with within-factor time). This decrease across sessions in correctly recalled words most probably was due to intrusions from previous word lists, that is, falsely recalled words learned before the presentation of the actual test list (Underwood, 1957; Postman, 1962). One or two intrusions per session registered in the present experiments show that proactive interferences occurred, but do not indicate their actual number because most of them may not have been consciously remembered by our subjects. As it is plausible that these factors had a comparable influence on both the treatment and the placebo groups, it seems justified to conclude that the superior memory performances of the treatment groups after 8 weeks of insulin indicate an improving influence of subchronic insulin on long-term memory. In this context, it is important to note that the 1 week delay between learning and recall was substantially longer than the delay of 20–30 min more frequently used in experimental memory assessments and that longer delays render memory consolidation per se more prone to interfering disturbances. Nevertheless, a 1 week interval between learning and recall is a valid means of assessing long-term memory formation (eg, Dudai, 2004).

It might be argued that differences in insulin sensitivity lead to the greater potential of ASP-I than RH-I to enhance declarative memory. Although no baseline insulin/glucose tolerance test was performed, measurements of basal homeostatic model assessment values, reflecting beta-cell function and insulin resistance (Wallace and Matthews, 2002), did not support this assumption (p>0.43). The neuronal mechanisms underlying the improvement of declarative memory after intranasal insulin administration cannot be derived from our study. Central nervous system insulin is involved in a number of neuronal mechanisms assumed to constitute memory processing (for review, see Zhao et al, 2004). Previous studies have provided clear evidence that declarative memory formation depends on intact hippocampal functioning (Kessels et al, 2001; Bayley et al, 2005). Hippocampal and cortical insulin signaling pathways have been shown to play a pivotal role in enabling long-term memory consolidation by modulating neuronal activity and triggering mechanisms that are required for establishing synaptic plasticity (Gasparini and Xu, 2003; Zhao et al, 2004; Craft and Watson, 2004; Wada et al, 2005). Thus, it is likely that after intranasal administration, insulin improves neuronal processes within these hippocampal and connected structures. Insulin may promote the expression of N-methyl-D-aspartate receptors (Skeberdis et al, 2001) and thus contribute to the formation of neuronal connections via synaptic long-term potentiation, a mechanism assumed to be essential for declarative memory formation (Castellano et al, 2001; Liu et al, 2004). The different pharmacokinetic properties of ASP-I in comparison to RH-I may add to the effects that insulin per se exerts on central nervous memory formation. In contrast to RH-I, that consists primarily of hexamers, ASP-I predominantly forms monomers so that after subcutaneous administration, it is more rapidly absorbed and lowers blood glucose more quickly than RH-I (Raslova et al, 2004; Hermansen et al, 2004). During intranasal administration, their reduced tendency to form hexamers may increase the number of ASP-I molecules transported from the nasal cavity to the brain and enhance insulin's effects on hippocampal memory processing.

In sum, our results demonstrate that after intranasal administration of identical doses, ASP-I has a distinctly greater potential than RH-I to improve memory in humans. Given that insulin and insulin-like factors not only protect brain tissue, but also favor processes such as neurogenesis and synaptogenesis (O'Kusky et al, 2000; van der Heide et al, 2005), and on the background of reduced CNS expression of genes encoding insulin and related messengers in patients suffering from AD (Steen et al, 2005), this outcome may be of considerable relevance for future clinical applications of insulin compounds in the treatment of memory disorders.