Regulatory Forum Opinion Piece: Review—Toxicological Pathology Profile and Regulatory Expectations for Nonclinical Development of Insulins and Insulin Analogues

In our own repeated dose toxicity studies in nondiabetic, normoglycemic rats, the brain was the main target organ of hypoglycemia and displayed prominent, distinct, and reproducible neuronal findings. Independent of the use of regular insulin or an insulin analogue, the hippocampus was particularly affected. Based on the subtopography (Figure 5), Cornu ammonis 1 (CA-1) to CA-3, neurons of the CA-1, and the crest of dentate gyrus have been selectively affected showing neuronal necrosis. In the hippocampus, there was a severity gradient starting from the subiculum and medial CA-1 becoming more attenuated toward the lateral CA-1. A sharp demarcation between damaged and normal areas was usually present. Affected neurons displayed eosinophilic and shrunken nuclei and cytoplasm. Neuronal necrosis was frequently accompanied by vacuolation of the neuropil (Figures 6A and B and 7A and B). Accompanying inflammatory responses including glial reaction were not observed, and lesions were not necessarily bilaterally symmetric.

In the hippocampus, these findings could be clearly distinguished from the single so-called dark neurons present also in control rats and intermingled with the majority of normal, vesicular, and faintly stained neurons (Figure 8). This was different for the cortex, were potential cortical findings occurred infrequently, and appeared to affect only individual or groups of neurons in the superficial layers, and which were difficult to discriminate from “dark neurons” that have to be considered as histological neuronal artifacts (Jortner 2006). Brain stem, basal ganglia, and cerebellum were free from lesions in our toxicity studies with insulins/insulin analogues.

The findings described above were in overall accordance with those mentioned in the literature, in particular by Auer and his colleagues (1986 and 2004) but also by Mohseni (2001). Hypoglycemia is known to induce neuronal necrosis in cerebral cortex (Agardh et al. 1980; Auer et al. 1985a), hippocampus, and caudoputamen (Kalimo et al. 1985). A distinct lesion pattern is described for the hippocampus (Auer et al. 1984; Auer et al. 1985b; Bree et al. 2009) involving CA-1 and the crest of the gyrus dentatus specifically. The lesions did not necessarily occur bilaterally or symmetrically. The usual development of hypoglycemic lesions in the cortex and hippocampus, associated with a relative resistance of the cerebellum and brain stem, facilitates the differentiation of hypoglycemic findings from those of ischemia-induced lesions (Auer 2004).

In contrast to the literature, where the CA-3 is reported to be relatively resistant to hypoglycemia, this sector was involved occasionally in our rat studies. In addition, under conditions of severe and sustained hypoglycemia, the entire hippocampus became necrotic, but with a severity gradient from CA-1 or the dentate gyrus crest to the other compartments still being present.

Similar microscopic findings have been observed in dogs; however, these were observed less frequently and consistently. Also in the literature, the hippocampus and the cerebral cortex have been described as sites of predilection for hypoglycemic lesions in dogs (Shimada et al. 2000) and cats (Okada et al. 1992).

In our repeated dose toxicity studies, cerebral lesions were exclusively seen in rats and dogs that died or were euthanized prematurely due to clinical observations suggestive (e.g., trembling, convulsions) or indicative (glucose levels) of severe hypoglycemia (Table 1). This may explain why the cerebral histopathological findings were confined to acute necrosis without any accompanying inflammatory reaction. Moreover, it suggested that the brain lesions were not compatible with the survival of the animals because findings have never been observed in animals that reached the end of the study. Subsequently, subacute or chronic stages of brain lesions could not develop. However, a disadvantage was that it was impossible to assess the reversibility of these cerebral findings. Reports in the literature on the reversibility of hypoglycemic brain findings are controversial, but the reversibility is clearly related to severity and duration of hypoglycemia. It should also be noted, however, that the potential of reversibility depends on the tissue compartment affected (Agardh et al. 1980). Nevertheless, it is evident both from the literature (Kalimo et al. 1985) and based upon our own common knowledge from general pathology that no regeneration in the CNS will be possible once necrosis has occurred.

Interestingly, the brain and also the peripheral nervous system (PNS) has been not described as a target organ for recombinant insulins or insulin analogues by other companies (e.g., lispro, aspart, detemir, and degludec) despite the presence of clinical symptoms and unscheduled deaths due to hypoglycemia which have been reported in rodents and dogs within the approval files available on the European Medicines Agency (EMA) and Food and Drug Administration (FDA) home pages. For the animals that survived the reported clinical hypoglycemia, the duration of hypoglycemia, the reduction of the doses used, discontinuation of dosing, or therapeutic intervention to avoid premature deaths could be explanations for this. However, a question remains as to why the unscheduled deaths with preceding hypoglycemia did not show any microscopic changes in the PNS. The pathogenesis of hypoglycemic brain injury is not simply induced by energy deficiency. The presence of decreased glycolysis and subsequent lack of acetate leads to an accumulation of oxalacetate that in turn drives its transamination to aspartate which is then released into the cerebrospinal fluid. Excess of aspartate mediates excitatory and neurotoxic effects resulting in neuronal damage. In addition, intracellular alkalosis due to protein catabolism contributes to this process (Auer and Siesjö 1993; Auer 2004). The release of excitotoxic aspartate into the cerebrospinal fluid may also explain the predilection sites of hippocampus and dentate gyrus that are bathed by the cerebrospinal fluid.

Peripheral neuropathy (PN) in animals under hypoglycemic conditions has been rarely reported in the literature. After decades of development of insulins/insulin analogues, we have observed this entity only during the development program of a long-acting insulin. It was seen in both a 2-week- and a 1-month study in rats and in a 1-month study in 1 dog only and was induced by a long-acting insulin analogue later discontinued due to lack of clinical efficacy. The insulin analogue possessed an extraordinarily long pharmacokinetic/dynamic profile. Microscopically, this unique finding in sciatic nerves showed microscopic features that were compatible with axonal degeneration of the Wallerian type. At low magnification, increased vacuolation was present, and at higher magnifications, vacuoles were filled with condensed eosinophilic material (Figure 9). These eosinophilic globules are known as digestion chambers or ovoid formations (Ikegami et al. 2000; Ozaki et al. 2010) and are thought to be disintegrated and homogenized myelin sheets. Digestion chambers are considered to be a hallmark of PN of the Wallerian type. Occasionally, a subtle inflammatory response consisting of mononuclear cells was present. In most rats affected, degeneration of the sciatic nerves was associated with degenerative changes in thigh muscles considered to be secondary to paresis. These findings were not reversible after the 2-week recovery period.

According to the literature on hypoglycemia research, peripheral nerves are only affected in cases of severe (<1.5 mmol/L) and sustained (≥12 hr) hypoglycemia (Yasaki and Dyck 1990, 1991). This corresponds exactly to our unique observation of hypoglycemic PN seen after treatment with an insulin analogue that displayed supralong-acting pharmacodynamic/kinetic properties. These microscopic features induced by a supralong-acting insulin were in accordance with hypoglycemia-related PN reported in the literature for rats (Sima, Zhang, and Greene 1989; Mohseni 2000; Tabata 2000), mice (Tabata 2000), and dogs (Moore et al. 2002). Recently, PN has also been reported in rats using the long-acting insulin analogue Abasria, a biosimilar of insulin glargine (EMA 2014) but not for other insulin analogues, in particular the long-acting insulins approved to date.

The mechanism of hypoglycemic PN is different from that of the CNS because the PNS is able to use other energy sources under hypoglycemic conditions. In contrast to the brain, disturbed blood supply is thought to be involved in the pathogenesis of PN (Eaton and Tesfaye 2003). Hypoglycemia-related lesions were detected in endoneural microvessels of sciatic nerves in rats (Ohshima and Nukada, 2002). Both axons and myelin sheaths were affected. Microscopic features of hypoglycemic neuropathy resembled the Wallerian type of degeneration (Sima, Zhang, and Greene 1989; Mohseni and Hildebrand 1998; Sugimoto et al. 2003; Ozaki et al. 2010). Motor fibers are more susceptible to hypoglycemia-related lesions (Mohseni 2000), whereas sensory fibers are usually affected under hyperglycemic (diabetic) conditions (Sima, Zhang, and Greene 1989). Generally, nerve trunks appear to be more prone to lesions than nerve roots, and within the spinal roots, ventral roots seemed to be the most sensitive. Unfortunately, reversibility of PN has not been tested in the studies published so far.

Overall, in our toxicity studies with insulin/insulin analogues, the nervous system has been more frequently observed as a target organ in rats rather than dogs. One explanation might be that dogs were usually given lower doses based on their supposed higher sensitivity in terms of clinical hypoglycemia (see Table 1). Furthermore, based on the more evident clinical signs of hypoglycemia in dogs, it could be speculated that the decision for dose reduction, using both therapeutic intervention and premature euthanasia, may have been performed earlier. Nevertheless, a scientific explanation for the species-specific sensitivity difference has so far not been elucidated.

Islet cell findings after insulin treatment have been rarely published. We have seen islet findings in a 1-month toxicity study in rats with the extraordinarily long-acting insulin analogue that also induced the PN. In control rats, islet cells showed faintly staining homogeneous cytoplasm, and there was no specific distribution pattern visible (Figure 10A). A completely different picture was present in rats after repeated treatment for 1 month with the insulin analogue. Cells at the center of islets were eosinophilic and shrunken and showed dark, condensed nuclei, whereas the larger peripheral cells seemed sometimes to be larger with prominent vesicular and slightly basophilic nuclei, although in many cases, the nuclear density appeared also to be increased in this area when compared to controls. The changes seen in the central area were indicative of an atrophic effect, and in the periphery, the changes were considered to be suggestive of possible hypertrophic effects (Figure 10B). The distribution pattern described, corresponds exactly to the arrangement of insulin producing β cells forming the islet core, whereas glucagon producing α cells are confined to the margin (Wieczorek, Pospischil, and Perentes 1998; Germann et al. 1999) and was further confirmed by immunohistochemical insulin/glucagon double staining (Figure 10C).

With the aldehyde-thionin staining used in the past, β cell degranulation could be frequently demonstrated in both rats and dogs after insulin treatment without evidence of atrophy of islet core cells in H&E-stained slides. It was also interpreted as a kind of disuse atrophy.

After a 2-week recovery period, there was usually no evident difference between the central islet cells and those of the margin indicating probable reversibility. Both cell types displayed faintly stained cytoplasm and nuclei (Figure 10D). The more vesicular appearance of cells and the reduced nuclear density in the central compartment, seen in individual islets, may be suggestive of minimal hypertrophy of these cells.

To our knowledge, these findings in the endocrine pancreas have not been extensively described in scientific literature. We found only a single paper describing the effects of hyperglycemia on pancreatic islets in rats (Clark et al. 1982), where a decrease in the β-cell area was mentioned secondary to hypoglycemic conditions.

These findings in the pancreatic islets of Langerhans should be regarded as adaptive and/or compensatory responses rather than direct effects of insulin/insulin analogues. Exogenous, excessive insulin delivery to normoglycemic rats interferes with the endogenous insulin release of islet cells. It results in a kind of disuse atrophy of β cells which form the core of the islets (Kracht 1958). Comparable islet cell atrophy and interpretation were described for the long-acting insulin analogue Abasria, newly Abasaglar (EMA, 2014) in a repeated toxicity study in rats. It was considered to be consistent with a negative feedback mechanism in hyperinsulinemic animals and a subsequent reduction in endogenous insulin production. In our case, glucagon producing α cells at the margin of pancreatic islets became hypertrophic/hyperplastic in counteracting the exogenous insulin surcharge. After withdrawal of exogenous insulin delivery, the β cells regained their functional activity, and the α-cells are no longer stimulated. Thus, the microscopic differences seen between central and marginal cells disappeared after the treatment-free period. The hypertrophic stimulus on central cells, which is seen occasionally, may thus reflect a compensatory response.

In a 12-month toxicity study in rats with insulin glulisine and 24-month carcinogenicity studies both in rats and mice with insulin glargine, respectively, treatment-related malignant fibrous histiocytoma (MFH) was diagnosed in male and, occasionally, in female rats and mice (Greaves and Faccini 1981; Ward et al. 1981) of most groups, including the vehicle control group, at the sc administration site on a background of chronic irritation and inflammation (Stammberger et al. 2002; European Medicines Evaluation Agency [EMEA], 2005). Administration-site MFH is a well-known finding and has been reported in rodent long-term studies with drugs, nongenotoxic chemicals, plant extracts, and inert materials when administered or implanted sc (Bartholomew, Kreeger, and Morton 2014; Thomas et al. 1977). Chronic irritation and inflammation were considered to constitute the key etiologic trigger of administration-site MFH (Thomas et al. 1977; Stammberger et al. 2002; Bartholomew, Kreeger, and Morton 2014). Interestingly, administration-site MFH was seen in preclinical rodent studies with a number of other marketed drugs (Bartholomew, Kreeger, and Morton 2014). Even after decades of clinical use of insulin and insulin analogues, no reports of administration-site MFH or other injection-site neoplasia in diabetic patients were identified in the available literature. Therefore, administration-site MFH in rodent long-term studies with insulin glargine and insulin glulisine was considered not to be relevant for humans nor mechanism related.

There were other sporadic observations in laboratory animals related to insulin. Testicular findings, seen in rats treated with the fast-acting insulin analogue aspart (FDA, 2000), were also discussed as being secondary to hypoglycemia. Hypertrophy of rodent parotid and submandibular glands related to insulin treatment has been reported but solely by Wang, Purushotham, and Humphreys-Beher (1994).

These observations, however, lacked consistency and reproducibility across different products and moreover were never proven to be relevant for humans. Therefore, they were beyond the scope of this review which has focused on classical target organs of insulin treatment with potential relevance for humans.

A relatively wide spectrum of animal models for diabetes is available (see reviews from King 2012; Sasase et al. 2013). Based on the conclusion that all target organ toxicity of the insulins and insulin analogues identified so far is related to exaggerated pharmacology and hypoglycemia, why not evaluate insulin therapy in these hyperglycemic animal models of diabetes? There has been to date no published or documented experience (scientific papers or FDA/EMA approval packages) on the use of diabetic animal models in regulatory toxicity studies with insulins.

The fact that hypoglycemia can be thought of as dose limiting could speak in favor of the use diabetic animals. However, according to our experience and knowledge, there has been no difficulty in achieving favorable safety margins and exposure ratios for insulins and insulin analogues using normoglycemic animals in toxicity testing.

The position paper of Morgan et al. (2013) has outlined the benefits and limitations of animal models of human diseases in nonclinical safety assessment. Based upon the benefit limitation balance, the main recommendation of the paper was that disease models in safety testing should be reserved for the elucidation of toxicity mechanisms, in particular in the context of firstly, unexpected clinical toxicity not found in conventional preclinical studies; and secondly, differentiation of on-target toxicity from off-target toxicity. The recent investigation of unusual pathology related to a hypoglycemic compound (e.g., findings in myocardium and skeletal muscles) by using the Zucker diabetic fatty rat model (Tirmenstein et al. 2015) provides an example for the second case.

However, based on our experience and knowledge, unknown mechanisms of toxicity with unfavorable risk–benefit ratios have not been an issue in the preclinical development of insulins and insulin analogues, and therefore, no clear scientific needs have arisen to use specific animal models. The target organs are well known and characterized and exclusively related to hypoglycemia (on-target toxicity) without any differences between regular insulins and insulin analogues.

The International Conference on Harmonization (ICH) has published the ICH S6 (EMEA 1997) guideline specific to the preclinical development of biopharmaceuticals.

Overall, the ICH S6 is applicable for insulins, they being peptide hormones and also for their biosynthetic modifications. In contrast to chemical molecules, classical genotoxicity is expected to be absent for biopharmaceuticals based on the inability of direct DNA interaction and damage. Thus, chemical carcinogenicity has not been expected for peptides, and standard carcinogenicity bioassays are generally considered as inappropriate for biopharmaceuticals by the ICH S6. However, mitogenicity and the potential for tumor promotion is a considerable safety concern for growth promoting hormones (see review by Heidel and Page 2008) and for insulins in particular as already described above.

The approaches to preclinical testing of mitogenicity and carcinogenicity were initially different among companies but became more harmonized over time and have been further facilitated by upcoming regulatory guidance.

The first recombinant (biosynthetic) human insulin was already available for patients worldwide in 1982. Based upon the homology to endogenous human insulin in terms of structure and pharmacology, chronic preclinical toxicity studies, in particular those for mitogenicity and carcinogenicity evaluation were not required for the first and the subsequent recombinant human insulins.

As our understanding of the potential growth promoting effects of insulin increased, the scientific and regulatory concern for potential tumor promoting effects of recombinant insulins likewise increased. This, presumably, was initially driven by the report of Dideriksen, Jorgensen, and Drejer (1992) on the tumorigenic properties of the insulin analogue B12 Asp in rats, but more importantly on the growing molecular knowledge on the causal relationship between mitogenicity and the insulin and insulin-like growth factor (IGF) receptors. Therefore, attempts have been undertaken in the overall preclinical evaluation to study these effects more closely. Interestingly, the approaches used were diverse and have changed over time.

For the first insulin analogue, the fast-acting insulin lispro (approved 1996), the preclinical assessment of mitogenicity was focused on the in vitro evaluation of binding affinity to the insulin and IGF-1 receptors and growth effects on human mammary epithelial cells compared to regular human insulin and the use of insulin B12 Asp as a positive control. Twelve-month toxicity studies were conducted in rats and dogs (FDA, 1996; Zimmermann and Truex 1997; Llewelyn, Slieker, and Zimmermann 1998). Insulin aspart was approved 3 to 4 years later, and the preclinical mitogenicity and carcinogenicity evaluation (FDA, 2000) was comparable to that of lispro (binding affinity and mitogenicity in vitro and 12-month studies in rats and dogs). For insulin glargine, the first long-acting analogue available for patients since 2000, the approach was different. It was evaluated in classical 2-year carcinogenicity studies in rats and mice (Stammberger et al. 2002), in addition to an in vitro mitogenicity testing battery comprising comparative insulin and IGF-1 receptor affinity and thymidine incorporation assays (FDA, 2000), and there was no treatment-related carcinogenicity.

Considering the increasing scientific evidence and resulting regulatory need, the EMEA published the “Points to consider document on the non-clinical assessment of the carcinogenic potential of insulin analogues” in 2001 (pp. 1–5).

Insulin glulisine, approved in 2004, was the first analogue that was preclinically evaluated in compliance with the EMEA “points to consider” paper. Mitogenic potential was evaluated in vivo, in addition to the in vitro testing of the affinity to the insulin and IGF-1 receptor (human osteosarcoma cell line B10), activation of insulin receptor substrates (human myoblasts or rat cardiomyocytes), and DNA synthesis by ³H thymidine incorporation (human breast cell line MCF10) compared to regular human insulin and partially to insulin B12 Asp as a positive control. In vivo cell proliferation was measured in mammary glands of female rats sensitive to hyperplasia/neoplasia and Ki-67 was used as a proliferation marker. This proliferation analysis, which was incorporated into the 6- and 12-month rat toxicity studies, did not indicate any increased mitogenicity as compared to regular human insulin (Stammberger et al. 2006). This in vivo proliferation analysis was specifically requested by a national health authority.

For the long-acting insulin detemir that was approved shortly after, the in vitro evaluations were comparable to insulin glulisine but not the in vivo approach for mitogenicity and carcinogenicity (FDA, 2005). At present, the ultra-long-acting insulin analogue degludec, approved in Europe in 2012 and currently under FDA review, was the last innovative insulin analogue that went to the market. In compliance with the EMEA points to consider paper, in vitro and in vivo investigations were both conducted in equal measure. Cell proliferation was assessed in rat mammary glands from the 12-month toxicity study, using bromodeoxyuridine (BrdU) immunohistochemistry (EMA 2012).

It should be emphasized that in addition to the metabolic effects mentioned above, insulins show mitogenic activity. This potential tumor promoting activity is related to the insulin receptor and to the IGF-1 receptor linked to respective receptor kinases (see reviews from Sandow 2009; Hvid et al. 2011; Varewick and Janssen 2012) and is already inherent to endogenous regular insulin. Based on this knowledge, a hypothetical safety concern has been raised that the modified sequence and structure of insulin analogues could be associated with enhanced mitogenicity and tumor promotion. Therefore, the need for the preclinical assessment of this mitogenicity/carcinogenicity potential has been the most challenging issue in the preclinical testing of the insulin analogues.

Several approaches have been used in the history of insulin development to address this hypothetical concern as described above.

A principal strategy has been the evaluation of immunohistochemical proliferation markers in regulatory toxicity studies focused on organs with high spontaneous tumor incidences, such as the mammary gland. This approach is scientifically based on the consideration that increased mitogenic potency of modified insulins may promote initiated preformed/preneoplastic cell clones not visible in H&E-stained slides. Mammary gland tumors are among the most prevalent tumors in female rats (Sher, Jensen, and Bokelmann 1982; Brix et al. 2005; Giknis and Clifford 2013). A further basis for this strategy is interestingly a single report that the incidence of benign and malignant mammary tumors was increased in female rats at 100- to 200-fold the therapeutic dose of insulin analogues (Dideriksen, Jorgensen, and Drejer 1992), and the rat mammary gland was shown to express functional insulin and IGF-1 receptors (Hvid et al. 2011). Although the publication by Dideriksen, Jorgensen, and Drejer (1992) report has remained the only case of tumorigenic potential of newly engineered insulin analogues, the evaluation of proliferation markers in preclinical toxicity studies was recommended by a specific points to consider paper (EMEA, 2001).

There are few publications describing the measurement of proliferative activity as an attempt to assess carcinogenic potential of pharmaceuticals preclinically, in particular, in the mammary gland (Stammberger et al. 2006; Christov et al. 2007), although there are a multitude of publications describing different models for the evaluation of initiation/promotion based upon the measurement of DNA synthesis, mitosis, formation of preneoplastic foci, and potential progression of preneoplastic to neoplastic lesions in different organ systems (Pitot and Sirica 1980; Ito et al. 1992; Shirai 1997; Grasl-Kraupp et al. 2000; Ittrich et al. 2003; Tsuda et al. 2003; Grassi et al. 2011; Kushida et al. 2011). Based on our own experience, the preclinical testing of the mitogenic potential is the most challenging aspect in the safety assessment of insulins/insulin analogues because all the other findings due to exaggerated pharmacodynamics are well known today, and it is not difficult to obtain a therapeutic window for the treatment of diabetic humans based upon the use of normoglycemic/nondiabetic animals.

For the assessment of putative mitogenic activity in mammary glands, several technical and scientific aspects have to be taken into account. Among the proliferation markers, Ki-67 is widely accepted and used for the evaluation of histological slides. A good correlation with BrdU has been proven for many species, including the rat (Muskhelishvili et al. 2003) and in particular for the mammary gland (Hvid et al. 2012). In addition, the in vivo injection and further shortcomings of BrdU may be avoided, and therefore, Ki-67 has been recommended as the first-choice marker in rodents (Birner et al. 2001). According to our experience, distinct and reproducible nuclear staining in paraffin sections can be achieved (Figure 11) with a defined formol fixation not longer than 48 hr and after antigen retrieval with microwave/citrate buffer using the MIB-5 antibody (Dako, Hamburg, Germany).

For the mammary gland, it is well known that the proliferative activity is dependent on the hormonal milieu (Russo and Russo 1994) and the estrus cycle stage (Schedin, Mitrenga, and Kaeck 2000; Hvid et al. 2012). Therefore, the rat mammary glands to be evaluated need to be staged for their estrus cycle, mainly assessed on the vaginal histology (Davis, Travlos, and Mc Shane 2001). It is clear that for the counting/evaluation strategy, consistency and representativeness in sampling are needed. In our approach, samples were removed from the thoracic and abdominal portions of mammary glands, and Ki-67-immunostained paraffin sections were prepared.

The evaluation of cell proliferation in vivo was initially confined to development stages of compounds close to submission (12-month studies), but the current approach is to include the in vivo evaluation of mammary gland proliferation already into the 1-month rat studies.

An image analysis system/software tool (Figure 12) was developed, tested, and validated in house for the detection/calculation of proliferation indices in the mammary gland tissue of rats. To ensure robustness and statistical power, at least 1,500 epithelial cells of alveolar and ductular structures per animal were counted in randomly chosen fields. Among the counted cells, the number of Ki-67 positive cells was recorded separately. Only cells with diffuse or stippled red pigment over the nuclei were counted as positive. The proliferative index was calculated as the ratio of stained to total mammary gland cells. Estrous cycle adjusted subgroups from insulin/insulin analogue treated rats were compared to the control group by calculating mean, standard deviation, and median. The statistical significance was tested by using the Kruskall–Wallis 2-sample test.

For all insulin analogues engineered and tested so far by Sanofi, no increased mitogenic or carcinogenic potential compared to regular insulins was demonstrated in preclinical studies, in particular for insulin glargine (Stammberger et al. 2002) and insulin glulisine (Stammberger et al. 2006) approved and on the market. Also other insulin analogues such as insulin lispro, insulin detemir, insulin aspart have not been displayed increased proliferative potential relative to regular human insulin (see review by Vahle et al. 2010).

Obviously, the regulatory environment has evolved to be in favor of investigating the mitogenic potential in vitro and in vivo instead of running classical 2-year carcinogenicity studies.

In summary, for all the insulin analogues available for patients to date and which are cited above, the preclinical safety evaluation did not show any evidence of an increased mitogenic and carcinogenic potential when compared to regular human insulin.

The findings described for target organs in the nervous system do not represent species-specific toxic effects in animals. They are relevant for humans because they are directly or indirectly related to exaggerated pharmacodynamics of the compounds resulting in hypoglycemia. Hypoglycemia is defined in man as typical symptoms characterized by plasma glucose concentration ≤3.9 mmol/L (ADA, 2005).

Observations of hypoglycemic brain damage in humans originate interestingly from psychiatry. Sakel’s therapy (1937) of schizophrenia and drug addiction was based on a desired transient hypoglycemic coma. Earlier reports are derived from patients with insulinoma. Today, cases are reported from medication errors (excess or ill-timed insulin administration) as well as from homicide (Haibach, Dix, and Shah 1987; Hood et al. 1986; Levy et al. 1985; Missliwetz 1994) and suicide attempts (Lutz et al. 1997). Severe hypoglycemia, which occurs in the setting of excess or ill-timed insulin administration, has been shown to cause brain damage.

In humans, predilection sites of hypoglycemia in the CNS are similar to animals. As in rats, hippocampus, cerebral cortex, and basal ganglia are usually affected, and the cerebellum is always spared (Kalimo and Olsen 1980; Auer 1986; Fujioka et al. 1997; Bree et al. 2009). Mechanism and pathology of hypoglycemic brain injury in humans were comprehensively investigated and described (Auer 1986; Auer et al. 1989; Auer 2004).

The few reports available for PN in humans are predominantly caused by insulinomas. Similar to rats, axonal degeneration and demyelinization were described. Despite the possible complex etiology involved, hypoglycemia is thought to be the main cause of this type of neuropathy (Eaton and Tesfaye 2003). Adaptive findings in pancreatic islets are related to the exogenous insulin supply to rats with normal endogenous insulin secretion and are, therefore, without relevance for diabetic patients.

In conclusion, it should be highlighted that the findings in the nervous system described here were induced in nondiabetic normoglycemic rats and dogs, in contrast to insulin-deficient, hyperglycemic diabetic patients. Consequently, it is usually not difficult to achieve a therapeutic window for the treatment of diabetic patients for newly developed insulins/insulin analogues.

The more challenging aspect of modern insulin development is evaluation of increased mitogenicity and potential tumor promotion and translating these nonclinical data to an effective human risk assessment. This hypothetical issue has been raised by recent, conflicting, retrospective observational studies, or meta-analyses suggesting a carcinogenic potential of insulin analogues, in particular insulin glargine (Colhoun et al. 2009; Currie, Poole, and Gale 2009; Hemkens et al. 2009; Jonasson et al. 2009; Rosenstock et al. 2009; Du et al. 2012; Ruiter et al. 2012) accompanied by an editorial (Johnson and Gale, 2010). Overall, these studies were considered as inconclusive and the European health authority (EMA, 2013) came to the conclusion that overall the data did not indicate an increased risk of cancer with insulin glargine.

Indeed, experimental evidence indicated that the proliferative effect in vitro of insulin glargine was not much different from that of regular human insulin or other insulin analogues and the rapid enzymatic conversion precluded sustained activation of the IGF-1 receptor signaling pathway (Ciaraldi and Sasaoka 2011; ter Braak et al. 2014). Moreover, pharmacokinetic data from clinical trials have shown that insulin glargine is rapidly converted into its metabolites M1 and M2 in both type 1 (Bolli et al. 2012) and 2 diabetic patients (Lucidi, Porcellati, and Rossetti 2012). The metabolic activity of M1 and M2 is lower and their mitogenic activity similar to regular insulin (Sommerfeld et al. 2010). Therefore, it has been concluded that the in vitro findings of increased mitogenicity are not translatable to the clinical setting in humans (Owens 2012; Tennagels and Werner 2013).

In addition, besides the mitogenicity hypothesis, the hyperglycemia hypothesis (providing tumor cells with more fuel) was also discussed (Johnson and Gale 2010) resulting in the conclusion that the link between insulin treatment and cancer risk might be rather driven by the high glucose levels than by any direct effect of insulin. This hypothesis was supported by the clinical trial conducted by Yang et al. (2010) indicating an increased cancer risk in diabetic patients lacking insulin treatment, whereas the insulin use was associated with reduced risk.