Insulin-Like Growth Factors and Neoplasia

Michael N. Pollak; Eva S. Schernhammer; Susan E. Hankinson

In This Article

IGF Levels and Cancer Risk

Inter-Individual Variations in IGF Levels

Before reviewing evidence for a relation between circulating levels of IGFs and cancer risk, it is important to understand that there is considerable variation of circulating levels of IGF1, IGF2 and IGFBP concentrations between normal individuals. Although clinical endocrinologists have long measured these concentrations to aid in the diagnosis of GH deficiency and acromegaly, the traditional view has been that there is no biological or medical significance to the substantial variation that falls within the broad normal range between these pathological extremes. However, there is now evidence that this variation has implications regarding disease risk. Studies reviewed below indicate that the risk of common cancers is increased in individuals who have higher circulating levels of IGF1, compared with those who have levels at the lower end of the normal range. By contrast, among individuals with IGF1 concentrations within the normal range, risk of cardiac disease seems to vary inversely with circulating IGF1 levels.[28,29]

Genetic Factors

Genetic factors influence circulating IGF1 concentration.[30] Dozens of proteins are involved in the physiological systems that regulate IGF1 levels, and polymorphic variation of the genes encoding any of these could influence circulating concentrations. Examples include genes encoding IGF1 itself; IGFBPs and their proteases; GH and its receptor; and somatostatin, GHRH and their receptors. Only a few of these have so far been studied in the context of their ability to influence IGF1 levels within the normal range,[31,32,33,34] but some are mutated in growth disorders associated with abnormal IGF1 levels.[35,36] Some,[31,37] but not all,[38] reports indicate that polymorphic variation within these genes influences cancer risk or prognosis. Common haplotypes might account for much of the variation in circulating levels of IGFs and their binding proteins.[39]

Lifestyle Factors

Nutrition has an important influence on circulating IGF1 levels. Starvation reduces both IGF1 levels[12] and intracellular signalling distal to IGF1R, at the level of TOR.[40] These regulatory systems might have evolved to minimize the energy and protein consumption related to renewal of epithelial-cell populations at times of inadequate nutrition. The observation that the protection against carcinogenesis conferred by dietary restriction is reversible by infusing IGF1 (Ref. 41) indicates a mediating role for IGF1 in the protective effect of calorific restriction on carcinogenesis. Similarly, observations that mutations that reduce IGF signalling are sufficient to extend lifespan in model organisms[1,8] indicate a mediating role for IGF signalling on the effects of dietary restriction on ageing.

More recent studies indicate that high levels of energy or protein intake are associated with modest increases in IGF1 levels.[42,43,44] In several studies, IGF1 levels were seen to increase with increasing dairy-product intake.[42,43,45] The underlying mechanism and significance of this relationship deserves further study, particularly as prostate cancer risk has been shown to increase with both IGF1 level[23] and dairy intake.[46] Micronutrients such as retinoids also influence circulating IGFBP3 levels, and this is modified by an IGFBP3 polymorphism.[33] It is likely that ongoing research will uncover additional genetic factors that modify the influence of nutrition on IGF physiology. This work might allow the identification of certain individuals for whom specific dietary patterns have particularly important implications with respect to disease risk.

Hormones, including endogenous and exogenous steroids, have important influences on the GH-IGF1 axis. Both tamoxifen and the potent synthetic oestrogen diethylstilbestrol are effective breast cancer treatments. Both have several mechanisms of action, but it is of interest that both suppress IGF1 levels.[47,48] Oral oestrogen-replacement therapy reduces circulating levels of IGF1, probably as a consequence of its direct delivery to the liver and suppression of hepatic IGF1 gene expression.[49] This lowers circulating IGF1 levels, but might not reflect reduced IGF1 signalling in tissues. Oestrogen delivery by the transdermal route does not lower circulating IGF1 levels. In several experimental systems, oestrogens and IGFs act jointly to increase proliferation.[50,51] Polymorphisms that influence the function of IGF1R itself[34,52] or downstream signalling proteins would also be expected to add complexity to the relation between circulating IGF1 levels and IGF1R activation.

Circumstantial Evidence

The initial evidence for an association between circulating IGF1 concentration and cancer risk was circumstantial. Small but significant increases in cancer risk with increasing height have been documented in many studies.[53,54,55,56] Height per se is unlikely to be a risk factor, but the hormonal determinants of height might influence cancer risk. IGF1 levels are, at most, weakly related to adult height, but are related to height early in life.[57] So, height might be weakly related to risk in part because it is weakly related to IGF1 exposure over the first decades of life. Birth weight and size have been associated with risk of breast,[58,59] colorectal,[60] prostate[61] and childhood cancers,[62] and are positively correlated with cord-blood IGF1 levels.[63] Conversely, pre-eclampsia is associated with reduced IGF1 levels and reduced breast cancer risk.[64,65]

Mammographic density is strongly related to breast cancer risk (for a review, see Ref. 66), but the mechanisms involved are unclear. There is evidence[67,68] for a positive relation between circulating IGF1 levels (or the ratio of IGF1 to IGFBP3 levels) and mammographic density. This raises the possibility that the relation between mammographic density and breast cancer risk exists, at least in part, because variability in density functions as a surrogate for variability in IGF1 physiology.

Racial factors also influence IGF1 and IGFBP3 levels, but it is important to not over-interpret what might be a coincidence. Black men have a higher risk for prostate cancer than other groups and have been observed to have lower levels of IGFBP3 (Ref. 69), a protein that attenuates IGF signalling in experimental systems. Black women have higher IGF1 levels than other racial groups and also have a slightly higher risk of premenopausal breast cancer.[70,71]

Although there are controversies, many studies (for reviews, see Refs 72,73) provide evidence that there is an excess of neoplastic disease among acromegalics, who have IGF1 levels above the normal range. Although this increased risk represents circumstantial evidence for an association between IGF1 and neoplasia, it is surprisingly modest in magnitude, in that very high IGF1 levels are not known to be associated with extreme cancer risk. This situation might relate to the fact that IGFBP3 and IGF1 are both increased in acromegaly. Also, there is obviously little modern experience with long-term follow-up of untreated acromegaly to determine cancer incidence.

Population Studies: Methodology

To examine the possibility that levels of a circulating analyte predict risk of a future cancer diagnosis, a nested case-control study design is useful. This involves blood sampling of a large number of apparently healthy individuals before subsequent long-term clinical observation. After years of follow-up, individuals who have developed a particular cancer can be identified and assays can be performed on their stored blood samples, together with samples from an appropriate control group from the same cohort. The distribution of levels between the cases and controls is used to reach conclusions concerning the association of the analyte with subsequent risk. This method is useful because it minimizes the possibility that conclusions will be biased by effects of the disease itself on the analyte. This is an important consideration for studies of IGF1 and cancer risk, as it is known that the malnourishment that is associated with cancer cachexia lowers IGF1 levels. A non-prospective study design -- in which IGF1 levels are compared in blood taken from cases sampled after diagnosis and from a cancer-free control group -- therefore presents potential difficulties in interpretation, particularly if the 'cases' have advanced cancer.

Assay technology for measuring IGFs, IGFBPs and related analytes is evolving rapidly.[23] As epidemiological research can require assessment of small differences between groups, it demands more accurate and sensitive assays than traditionally used clinical assay applications in the diagnosis of GH deficiency and acromegaly. It is necessary to ensure that a particular assay method is appropriate for a given set of samples, as varying amounts of proteolysis might occur between the time of blood sampling and assay, and antibodies vary in their specificity for different molecular species. This issue is particularly important for IGFBP3, which is present in the circulation as a mixture of polypeptides of varying length (because of proteolytic cleavage) and varying glycosylation and phosphorylation modifications, all of which might have physiological significance. Some of the discrepancies in the literature regarding the relation of IGFBP3 levels to subsequent cancer risk might relate to technical issues concerning assay methodology.

Population Studies: Results

Earlier reviews[10,23,24,74,75] have summarized both prospective and non-prospective data. Renehan et al.[76] recently published a comprehensive overview of population studies. The authors acknowledge controversies, but conclude, on the basis of a meta-analysis, that circulating IGF1 levels are related to risk of several common cancers. Because of space constraints, we restrict our review to larger prospective data sets (which are considered the most reliable) and to cancer types that have been studied independently by several groups ( Table 1 ). Examples of recent interesting results that did not meet criteria for inclusion in  Table 1 include evidence for associations between IGF1 levels and cancers of the uterine cervix, bladder and ovary.[77,78,79]

Overall, a trend towards increasing risk with increasing levels of IGF1 is emerging. In the occasional study results are marked -- for example, within the Physicians' Health Study Cohort, the association between IGF-related analytes and prostate cancer risk was stronger than the association between cholesterol and cardiovascular disease[80,81,82] -- but overall the associations are modest. Of course, the impact of a risk factor that is common in a population and associated with modestly increased risk might exceed the disease burden attributable to strong risk factors that are encountered rarely.

It is important to point out that some studies did not detect an association of IGF1 levels with risk, particularly in the case of lung cancer.[83] This might be because IGF-related risk becomes relatively insignificant when carcinogenesis is driven by high levels of carcinogen exposure -- the lung cancer cohorts differed from the cohorts of other cancer types, as they comprised heavy smokers. Interestingly, no association was seen between prostate cancer and IGF-related analytes among heavy smokers.[84] This remains unexplained, but a mechanism involving an effect of heavy smoking on IGF1 levels can not be ruled out. In the case of breast cancer, IGF1 has been related to risk in premenopausal women only, indicating the possible importance of levels in early life or an interaction with other hormones such as oestradiol.

The data for associations between IGFBP3 and cancer risk are not consistent, which might reflect that this analyte is simply unrelated to risk. Alternatively, ongoing research might uncover technical issues related to sample storage and measurement that explain the discrepancies. It is possible, for example, that risk varies with a particular molecular species of IGFBP3 that is present in the circulation.

Laboratory Models

Several in vivo carcinogenesis models have provided data that are compatible with an effect of host IGF1 physiology on cancer risk -- these models are to be distinguished from those that study the influence of variations in IGF1 levels or IGF1 signalling on the behaviour of established cancers. It is interesting to note that most laboratory work concerning the relation of IGF1 to carcinogenesis was motivated by and followed the epidemiological research reviewed above. To the extent that these models showed trends in the same direction as those observed in the epidemiological research, the laboratory work in this area serves to increase the plausibility of the results from human populations.

Models in which circulating IGF levels are depleted by either liver-specific deletion of IGF1 or expression of a GH antagonist provide evidence that lower IGF1 concentrations are associated with a significant reduction in breast cancer following exposure to chemical carcinogens.[85,86] Prostate cancer incidence in the well-characterized tramp carcinogenesis model[87] is substantially reduced in IGF1-deficient mice, whereas organ-specific overexpression of IGF1R increased prostate neoplasia.[88] Another example concerns hepatocarcinogenesis: overexpression of IGFBP1, which can attenuate IGF1R activation by sequestering ligands, inhibited carcinogenesis following exposure to the carcinogen diethylnitrosamine.[89] It is also of interest to note that certain cancer types are more frequent in large breeds of dogs, which have higher circulating IGF1 levels.[90,91] It is important to recognize that all of these models involve variations in IGF1 levels that are large compared with the more subtle variations in IGF1 physiology that exist between healthy humans. However, it is reasonable to expect that models of large differences that operate over relatively short times are able to simulate more subtle differences that might influence human carcinogenesis over decades.

Biological Basis for a Relation Between IGF1 Signalling Level and Cancer Risk

The observed association between higher levels of IGF1 and cancer risk might arise because higher IGF1 levels are associated with acceleration of early carcinogenesis, as illustrated in Fig. 3. Carcinogenesis requires stepwise accumulation of genetic damage. This would be facilitated not only by faster rates of proliferation,[92] but also by an environment that favoured, even slightly, survival (rather than programmed cell death) of stem cells that have undergone early genetic 'hits' -- thereby increasing the pool of damaged cells available for second and subsequent hits. Higher levels of IGF1 would be expected to activate survival pathways that would make apoptotic death of damaged cells slightly less probable. When applied simultaneously to large numbers of at-risk cells over many years, even a small influence in this direction would serve to accelerate carcinogenesis. A separate mechanism might involve an effect of IGF1 on early progression of established neoplasms. This model postulates that IGF1R signalling is important in determining the time between full transformation of a single cell and the development of clinically significant disease. This is a potentially important issue, given evidence (best documented in prostate cancer[93]) that early carcinogenesis occurs very commonly by middle age and that risk of clinically detectable cancer depends on inter-individual differences in probability of progression.[94] These proposed mechanisms are not mutually exclusive.

Figure 3.

Model of the influence of insulin-like growth factor 1 signalling on the stepwise accumulation of somatic-cell genetic damage in carcinogenesis. The model of stepwise accumulation of genetic damage leading to carcinogenesis can be extended to include influences of insulin-like growth factor 1 (IGF1) signalling. These include favouring cellular proliferation over arrest and cellular survival over apoptosis. This model provides a preliminary biological framework to account for the observed association of higher levels of IGF1, or IGF1 receptor (IGF1R) activation, with cancer risk in epidemiological and laboratory studies. The model predicts that stepwise accumulation of genetic damage is facilitated in individuals with higher IGF1 levels because in these individuals there is a slightly higher rate of cell division (increasing the risk of errors) and, perhaps more importantly, because the probability of appropriate apoptosis of cells with a small number of 'hits' would be slightly reduced in a microenvironment with higher levels of IGF1R activation. The figure greatly exaggerates the magnitude of the hypothesized differences between 'high IGF' and 'low IGF' individuals in proliferation and apoptosis for purposes of illustration. Very small differences in these parameters, if applied to the very large renewing cell populations of organs such as the colon over a timespan of decades could influence the probability of emergence of a fully transformed clone. Colours indicate the following: yellow, normal cells; pale blue, cells containing one mutation or hit; dark blue, cells containing two mutations or hits; purple, apoptotic cells.