Insulin-like Growth Factor-1, Insulin-like Growth Factor Binding Protein-3 and the Incidence of Malignant Neoplasms in a Nested Case–Control Study

Malignant tumors are among the most common causes of death worldwide. On the basis of estimates prepared by the World Health Organization in 2015, neoplasm is the first or second leading cause of death before the age of 70 years in 91 of 172 countries (1). Therefore, we must seek new risk factors for these diseases.

In a variety of human malignancies, signals from growth factors and their receptors are required not only for tumorigenesis, but also tumor progression (2, 3). The insulin-like growth factor (IGF) axis, including both ligands (IGF1 and IGF2) and the receptor (type 1 insulin-like growth factor receptor, IGF1R), is one of these systems (3–5). IGFs bind to IGF1R and then activate multiple downstream signal axes, the mechanism of which is regulated by multiple factors under normal homeostatic conditions (6, 7). Growth hormone, produced in the pituitary gland, stimulates the secretion of both IGFs and IGF-binding proteins (IGFBP) 1–6 in hepatocytes. Activation of IGF1R is tightly regulated by the amount of the free form of the ligands, which is modulated by IGFBPs and the nonstimulatory receptor of type 2 IGF receptor (6, 8). IGFBPs control IGF activity by reducing its bioavailability for binding to the receptor. Most of the IGF1 in the serum is in an inactive form due to binding with IGFBPs, which form a complex with IGF in a 1:1 molar ratio. IGFBP3 is the most plentiful IGFBP and accounts for almost 80% of its binding. Proteases, including matrix metalloproteinase (MMP), control the complex of IGFs and IGFBPs (9).

Because IGFs stimulate DNA synthesis and the proliferation of cells, the IGF system plays important roles not only in homeostasis, but also in premalignancy (10). In addition to cell growth, IGF shows antiapoptotic effects and thus has survival signals in several tumor cells (11–18). Matrilysin (MMP-7) can cleave all IGFBPs and can thus trigger IGF signal pathways (19). We have previously reported a positive feedback loop between matrilysin and IGF-IGF1R in the invasion and progression of gastrointestinal carcinoma (20, 21). We also previously reported that overexpression of both IGF1R and IGFs was associated with advanced pathologic parameters, higher tumor stage, recurrence, and poor prognosis (18, 21).

IGF1R could be one of the next important molecular targets in cancer therapy (3, 22). IGF1R blockade, such as with tyrosine kinase inhibitors, monoclonal antibodies, dominant negative for IGF1R, and IGFBP3, suppressed proliferation, stimulated apoptosis, and inhibited tumor dissemination (13, 18, 23, 24).

IGFBP3 is a tumor suppressor molecule. Downregulation of IGFBP3 upregulated tumor growth and suppressed apoptotic activity (25). IGFBP3 has potential as a drug, and the 16-kDa ^1-95IGFBP3 fragment could potentiate apoptosis (26–28). IGFBP3 overexpression enhanced chemotherapy-induced growth inhibition due to inhibiting NF-kB (29). In addition, promoter hypermethylation of IGFBP3 might be a diagnostic and predictive biomarker for colorectal cancer (30, 31).

Many epidemiologic studies have reported associations between diabetes mellitus and both cancer-related incidences and mortalities (32, 33). Diabetes mellitus is related to increased risk for carcinomas, especially colorectal, hepatic, and pancreatic carcinomas, in Japan (34), and colorectal, hepatic, pancreatic, breast, endometrial, and bladder cancers in the United States (35). Diabetes mellitus was correlated with an overall 20% increased risk of total tumor incidence (36). In the Japan Collaborative Cohort (JACC) study, the risk of several site-specific cancer-related mortalities was reported to be increased in subjects with diabetes mellitus (37–39).

Elevated serum IGF1 levels or free-IGF1 levels, which are calculated by the molar ratio of IGF1 to IGFBP3, increase the risk of developing several cancers, including breast, colon, and prostate cancers (40–42). In addition, low serum concentrations of IGFBP3 or free-IGFBP3 levels, which are estimated by the molar difference (IGFBP3 − IGF1), increase the risk of some neoplasms, such as colon and liver cancers (42, 43). However, there is insufficient information about the relationship between the incidence of whole-malignant neoplasms and serum levels of IGF1 or IGFBP3. In previous reports about the associations of IGF1 or IGFBP3 with mortalities of all-causes, cardiovascular diseases, and cancer, IGF1 and IGFBP3 were inversely associated with cancer-related death in only male subjects in one study (44), but not in either sex in another study (45). In the previous and intermediate analysis in JACC study, IGFBP3 level was inversely associated with all cancer-related mortality but IGF1 was not (46). Although several associations between IGFs and the risk of several site-specific malignancies from the JACC Study were published (43, 47, 48), the incidence of overall malignant tumors has not been reported. Thus, the aim was to investigate the relationships between these factors and malignant tumor risk in a case–control study nested in a prospective cohort study, the JACC study.

A nested case–control study within the JACC study, which evaluated cancer risk associated with lifestyle factors, was conducted. The details of the JACC study were described previously (49–52). In brief, a baseline survey was started in 1988–1990 when 110,585 apparently healthy inhabitants (40–79 years old) who had undergone a general health checkup were enrolled as a basic cohort population from 45 areas throughout Japan. Participants were asked to complete a questionnaire that included information about demographic characteristics, medical history, and lifestyle factors. Approximately 35% of the cohort participants (39,242 subjects) in 37 of 45 areas voluntarily provided peripheral blood samples, which were stored at −80°C until biochemical assays were performed.

Informed consent was obtained from each participant by having the study participants sign the cover of the questionnaire in the majority of study areas. However, it was obtained at the group level in a few areas because the concept of informed consent was not popularized during the 1980s in Japan. In that case, the municipality head gave the consent to participation representing the participants living in that area. This study was conducted in accordance with International Ethical Guidelines for Biomedical Research Involving Human Subjects. This study was approved by the human ethics review committee at Hokkaido University (Sapporo, Japan) and was performed after approval by the Institutional Review Board at Sapporo Medical University (Sapporo, Japan).

In 24 of 45 study areas, the incidence of tumors was followed (51). Subjects were followed from the baseline survey. Participants with any malignant tumor history at baseline were excluded. Individuals who moved away from the original study area were treated as study dropouts, because deaths after such moves could not be confirmed in this follow-up system. The occurrence of tumors was confirmed in population-based tumor registries or by reviewing the records of local major hospitals. Malignant neoplasms were defined as C00–C97 according to the International Statistical Classification of Diseases and Related Health Problems 10th Revision (http://www.who.int/classifications/icd/en/). Subjects diagnosed with malignant tumors by 1997 were regarded as cases in this nested case–control study. For each case, three control subjects who were matched for residential area, age, and sex were selected randomly; however, less than three controls were selected in some cases based on the selection criteria (52). This analysis involved 1,349 cases and 4,012 control subjects.

In 1999 and 2000, both serum concentrations of IGF1 and IGFBP3 were assessed at a single laboratory (SRL, Tokyo, Japan) with an immunoradiometric assay using commercially available kits (Daiichi Radioisotope Laboratories) by technicians who were blinded to case/control status. The intra-assay precision obtained using different reference sera was 2.2%–3.5% of the coefficients variation value for IGF1 and 3.2%–4.2% of those for IGFBP3. The interassay coefficient of variance was 1.1%–4.2% for IGF1 and 5.3%–8.8% for IGFBP3. Details of the assays for both serum levels of IGF1 and IGFBP3 were as described earlier (53).

Proportions and mean values of baseline characteristics were compared between cases and controls using the t test or Chi-squared test. Results are shown as means ± SD. P values of less than 0.05 were considered significant. Serum levels were divided into quintiles based on the distribution of serum levels in all control subjects, with the first quintile used as a reference. IGF1 quintile values for quintiles 1, 2, 3, 4, and 5 were <80, 80–109, 110–130, 131–160, and >160 ng/mL, respectively. IGFBP3 quintile values for quintiles 1, 2, 3, 4, and 5 were <2.31, 2.31–2.75, 2.76–3.17, 3.18–3.66, and >3.66 μg/mL, respectively.

Because the molar ratio of IGF1/IGFBP3 is believed to correspond to the free form of IGF1, the molar ratio of IGF1/IGFBP3 was evaluated (for conversion, 1 ng/mL is 0.130 nmol/L for IGFI and 0.036 nmol/L for IGFBP3; ref. 42). Because the molar difference between IGFBP3 and IGF1 is considered to reflect the free form of IGFBP3, the molar difference of (IGFBP3 – IGF1) was assessed (43).

Using conditional logistic regression, the ORs for the incidences of malignant tumors associated with serum levels of IGF-related peptides were determined. ORs were controlled for body mass index (BMI, computed as weight in kilograms divided by the square of the height in meters: <18.5, 18.5–24.9, 25.0–29.9, or ≥30.0 kg/m², or missing), alcohol consumption (never, former, or current drinker, or missing), and cigarette smoking status (never, former, or current smoker, or missing). ORs were also adjusted for BMI, alcohol consumption, tobacco smoking status, and diabetes mellitus for subjects who were divided into three groups: never received treatment, previously or currently receiving treatment, or missing. The significance of trends across exposure quintiles was assessed by including ordinal terms for each serum concentration quintile and entering the variable as a continuous term in the model. The statistical interaction with gender was calculated by including interaction terms in this model. All P values and 95% confidence intervals (CI) presented were based on two-sided tests.

Baseline characteristics of both cases and controls are shown in Table 1. There were no differences in height, weight, or BMI between cases and controls. The percentage of current smokers was higher in the case group than in the control group, whereas the percentage of never drinkers was higher in the controls than in the cases. No history of diabetes mellitus tended to be more common in controls than in cases, but the difference was not significant. The mean serum concentration of IGF1 was not different between the two groups. The mean serum IGFBP3 level was significantly lower in cases than in controls. Table 2 shows the sites of malignant neoplasm in this study.

Concentration of total-IGF1 was not associated with the risk of malignant neoplasms in univariate or multivariate analyses (Table 3). The total-IGFBP3 level was associated inversely with the risk of malignant tumors (highest compared with lowest quintile: OR = 0.693; 95% CI, 0.558–0.861; P_trend = 0.001). After adjustment for IGF1, the result was strengthened (highest compared with lowest quintile: OR = 0.667; 95% CI, 0.518–0.859; P_trend = 0.003). After full adjustment including diabetes mellitus, the result was similar (highest compared with lowest quintile: OR = 0.707; 95% CI, 0.546–0.916; P_trend = 0.012).

A higher molar ratio of IGF1/IGFBP3, which correspond to free IGF1, was associated with an increased risk of malignant neoplasms (highest compared with lowest quintile: OR, 1.218; 95% CI, 0.957–1.549; P_trend = 0.041; Table 4). However, the trend was not significant after adjustments for other covariates.

A higher molar difference of IGFBP3 and IGF1, which represents free IGFBP3, was associated with a decreased risk of malignant neoplasms (highest compared with lowest quintile: OR, 0.698; 95% CI, 0.563–0.866; P_trend = 0.001; Table 4). After full controlling including diabetes mellitus, the result was not changed (highest compared with lowest quintile: OR, 0.737; 95% CI, 0.592–0.918; P_trend = 0.007).

To exclude possible effects of latent malignant tumors on levels of both IGF1 and IGFBP3, the analysis was limited to subjects followed over 3 years (885 cases and 2,638 controls; Table 5). Although there was no association between total-IGF1 and the risk of malignancies, free IGF1 was related to the risk of malignancies (highest compared with lowest quintile: OR, 1.430; 95% CI, 1.053–1.942; P_trend = 0.013). After fully controlling for diabetes mellitus, free IGF1 was related to the risk of neoplasms (highest compared with lowest quintile: OR, 1.357; 95% CI, 0.994–1.853; P_trend = 0.038). This analysis strengthened the negative relationships of both total- and free-IGFBP3 with the risk of malignant tumors (highest compared with lowest quintile: OR, 0.674 and 0.669; 95% CI, 0.514–0.883 and 0.513–0.873; P_trend = 0.004 and 0.002, respectively). After full adjustment including diabetes mellitus, both total- and free-IGFBP3 were negatively associated with the risk of malignancies (highest compared with lowest quintile: OR, 0.680 and 0.715; 95% CI, 0.493–0.939 and 0.544–0.939; P_trend = 0.022 and 0.009, respectively).

Then, we evaluated the statistical interaction with gender. Although both total-IGF1 and -IGFBP3 were interacted with gender (P_interaction = 0.005 and 0.013, respectively), those interaction were not detected after adjustment each other (P_interaction = 0.109 and 0.207, respectively). Free IGFBP3 showed an interaction with gender, however free IGF1 did not (P_interaction = 0.024 and 0.645, respectively). Then, ORs were analyzed in gender subgroups. In the male population, total-IGF1 was associated inversely with the risk of tumors, but not after adjustment with IGFBP3 (highest compared with lowest quintile: OR, 0.680 and 0.979; 95% CI, 0.486–0.951 and 0.659–1.456; P_trend = 0.003 and 0.422, respectively; Table 6). Both higher total- and free-IGFBP3 were also inversely associated to the risk of malignancies in male subjects (highest compared with lowest quintile: OR, 0.599 and 0.609; 95% CI, 0.440–0.815 and 0.445–0.831; respectively, P_trend < 0.001). After several adjustments, these relationships were observed again (highest compared with lowest quintile: OR, 0.678–0.705; 95% CI, 0.492–0.935 to 0.486–1.023; P_trend = 0.002–0.028). However, there were no relationships between both total- and free-IGFBP3 and the risk of malignant tumors in the female subjects. Although total-IGF1 was not related to the incidence of malignancies, free IGF1 tended to be associated with the risk of neoplasms, but not significantly, even after several adjustments. These findings reinforced the association between a decrease in both total- and free-IGFBP3 and the risk of malignancies in males.

IGFs play several roles in carcinogenesis and tumor dissemination, although IGFBPs can inhibit those actions (3, 18, 24). In this study, although the serum level of total-IGF1 was not associated with the OR for overall malignant tumors, that of IGFBP3 was associated inversely with the OR for all neoplasms. After three adjustments, the results were observed. Although we assessed incident of malignant tumors, these results resemble the former studies in which IGFBP3 was associated cancer-related mortality, but IGF1 was not (44–46).

Although free IGF1 was associated with the risk for all cancers, the association was not significant after adjusting for BMI, alcohol intake, and smoking, with/without diabetes mellitus. IGFs form a complex with IGFBPs in a 1:1 molar ratio, and IGFBP3 is higher molar than IGF1. Thus, the molar difference (IGFBP3 − IGF1) could estimate the serum level of free IGFBP3 (43). Free IGFBP3 was related inversely to the risk of overall tumors, which was observed even by two analyses with adjustments. It might be reasonable that both total- and free-IGFBP3 were inversely related to the risk for overall malignancies, as IGFBP3 is a tumor suppressor molecule.

Limiting subjects to those followed over 3 years, high–free-IGF1 enhanced the risk of malignant neoplasms, and both high total- and high–free-IGFBP3 reduced the risk of tumors. After adjusting for BMI, alcohol intake, and smoking, with/without diabetes mellitus, these results were observed again. These results suggest that there were latent patients with cancer among the subjects in the JACC study. Moreover, it might be reasonable that tumorigenic factors of IGF1 are related to the incidence of malignant tumors.

Excessive insulin action associated with insulin resistance in type 2 diabetes mellitus might contribute to tumorigenicity. As both insulin and IGFs have high homology, and both insulin receptor and IGF1R keep high homology, both ligands and receptors could bind to one another (54). Thus, the effects of hyperinsulinemia on carcinogenesis might be accounted for in part by IGF1R activation. After controlling for a history of diabetes mellitus, free IGF1 and both total- and free-IGFBP3 affected the future risk of tumorigenicity in this study.

The negative relationships between total- and free-IGFBP3 and the incidence of malignant tumors in the total participants were only seen in the male participants in this study. It has been reported that there were sex differences in the effects of IGFs on the incidence of site-specific tumors (44, 55, 56). These results resemble the former report, in which both total-IGF1 and -IGFBP3 were inversely associated with cancer-related mortality in males but not in females (44). One of the reasons for the sex difference in this study might depend on the sex differences in serum distributions of both IGF1 and IGFBP3. In control participants, the serum level of IGF1 was significantly lower in female than in male participants (P < 0.001; Table 1), and that of IGFBP3 was significantly higher in female than in male participants (P < 0.001). As serum IGF1 was reported to be higher in female than in male subjects in previous reports (7), it might be a feature of this cohort that the IGF1 level in female controls was low. In the current male group, levels of both IGF1 and IGFBP3 were lower in cases than in controls (P = 0.034 and < 0.001, respectively). However, there were no differences in the serum concentrations of IGF1 and IGFBP3 between female cases and female controls in this study. Because levels of IGF1 and IGFBP3 were high in the current cases compared with cases in previous studies (7); low IGF1 levels in the male cases of this study may also be a feature of this cohort. Another reason might depend on tumor distributions. Although esophageal, gastric, liver, and lung carcinomas were more common in male than female participants, thyroid cancer was more common in female than male participants in the current cohort (Table 2).

The advantage of this study is that the samples were from a large-scale, population-based study (110,792 participants). There are, however, several limitations in this study. The first limitation is that the serum levels of IGFBP3 and IGF1 were assayed only once, at the time of the baseline inquiry. Therefore, their chronologic changes were not observed in association with the incidence of malignancies. Another limitation is that frozen serum samples were used to measure levels of both IGFBP3 and IGF1. The IGFs have been reported to remain stable in frozen serum that was stored at −80°C for 9 years (53). The third limitation is that some data about BMI, alcohol intake, smoking, and history of diabetes mellitus were missing in the JACC study because of the self-administered questionnaire (49–51). The fourth limitation is that the participants were only Japanese.

In conclusion, both low serum total-IGFBP3 and low–free-IGFBP3 (molar difference of IGFBP3 – IGF1) might be related to future risk of malignant tumors. In addition, high–free-IGF1 (molar ratio of IGF1/IGFBP3) might be associated with the future risk of malignant neoplasms.

A. Tamakoshi reports receiving grants from JSPS KAKENHI grant number JP 16H06277 and grants from Grants-in-Aid for Scientific Research on Priority Areas of Cancer; and Grants-in-Aid for Scientific Research on Priority Areas of Cancer Epidemiology from MEXT during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Adachi, M. Nojima, A. Tamakoshi

Development of methodology: Y. Adachi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Adachi, M. Mori, K. Wakai, A. Tamakoshi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Adachi, M. Nojima

Writing, review, and/or revision of the manuscript: Y. Adachi, R. Himori, T. Kubo, H. Yamano, Y. Lin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases):

Study supervision: M. Nojima, A. Tamakoshi

Other (a member of the JACC study): Y. Lin

This work was supported by grants-in-aid from the Ministry of Education, Culture, Sports, Science, and Technology, Japan, from the Ministry of Health, Labour and Welfare, Japan, and from Japan Society for the Promotion of Science KAKENHI (No. 16H06277), Japan.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.