Abstract
Background
Epidemiological evidence supports a positive association between circulating insulin-like growth factor-1 (IGF-1) concentrations and breast cancer risk, but both the magnitude and causality of this relationship are uncertain. We conducted observational analyses with adjustment for regression dilution bias, and Mendelian randomization (MR) analyses allowed for causal inference.
Patients and methods
We investigated the associations between circulating IGF-1 concentrations and incident breast cancer risk in 206 263 women in the UK Biobank. Multivariable hazard ratios (HRs) and 95% confidence intervals (CI) were estimated using Cox proportional hazards models. HRs were corrected for regression dilution using repeat IGF-1 measures available in a subsample of 6711 women. For the MR analyses, genetic variants associated with circulating IGF-1 and IGF-binding protein-3 (IGFBP-3) levels were identified and their association with breast cancer was examined with two-sample MR methods using genome-wide data from 122 977 cases and 105 974 controls.
Results
In the UK Biobank, after a median follow-up of 7.1 years, 4360 incident breast cancer cases occurred. In the multivariable-adjusted models corrected for regression dilution, higher IGF-1 concentrations were associated with a greater risk of breast cancer (HR per 5 nmol/l increment of IGF-1 = 1.11, 95% CI = 1.07–1.16). Similar positive associations were found by follow-up time, menopausal status, body mass index, and other risk factors. In the MR analyses, a 5 nmol/l increment in genetically-predicted IGF-1 concentration was associated with a greater breast cancer risk (odds ratio = 1.05, 95% CI = 1.01–1.10; P = 0.02), with a similar effect estimate for estrogen-positive (ER+) tumours, but no effect found for estrogen-negative (ER−) tumours. Genetically-predicted IGFBP-3 concentrations were not associated with breast cancer risk (odds ratio per 1-standard deviation increment = 1.00, 95% CI = 0.97–1.04; P = 0.98).
Conclusion
Our results support a probable causal relationship between circulating IGF-1 concentrations and breast cancer, suggesting that interventions targeting the IGF pathway may be beneficial in preventing breast tumorigenesis.
Key words: breast cancer, insulin-like growth factor-1 (IGF-1), insulin-like growth factor-binding protein-3 (IGFBP-3), Mendelian randomization, observational
Highlights
-
•
We conducted complementary observational and Mendelian randomization (MR) analyses to investigate the role of circulating insulin-like growth factor-1 (IGF-1) in breast cancer development
-
•
Consistent positive associations were found between circulating IGF-1 and breast cancer risk for both analyses
-
•
These results support a probable causal role of the IGF pathway in breast cancer development
Introduction
Insulin-like growth factor-1 (IGF-1) is a polypeptide that has mitogenic and anti-apoptotic effects.1,2 Approximately 99% of IGF-1 is bound to IGF binding proteins, with most bound to IGF-binding protein-3 (IGFBP-3).3 In experimental studies, IGFBP-3 has also been shown to not only regulate IGF-1 bioavailability, but also to have direct inhibitory effects on cell growth.4
Interest in the possible role of IGF-1 in the development of breast cancer began in the 1980s.5 An early case-control study reported higher plasma concentrations of IGF-1 in women with breast cancer than in controls,6 and in the first prospective study, circulating concentrations of IGF-1 were positively associated with breast cancer risk for premenopausal women, but not postmenopausal women.7 Most subsequent prospective studies have generally reported a positive association between IGF-1 and breast cancer risk, and a pooled individual-participant data analysis of 4790 cases from 17 prospective studies showed that women with relatively high circulating IGF-1 had a ∼30% higher risk of breast cancer than women with relatively low circulating IGF-1. There was no evidence that the association was due to reverse-causation, varied by menopausal status, or was attenuated by adjustment for other risk factors including IGFBP-3, reproductive factors, and body mass index (BMI).8 However, heterogeneity by estrogen receptor (ER) subtype was found, with the positive association present for estrogen-positive (ER+) but not estrogen-negative (ER−) tumours. In addition, this pooled analysis was based on a single IGF-1 measurement for each woman, so risk estimates could have been influenced by the combined effects of measurement error and within-person variability, leading to a likely underestimation of the true association (regression dilution).8,9
To further examine the possible causal role of IGF-1 in breast cancer risk, we conducted complementary observational and Mendelian randomization (MR) analyses. Firstly, we investigated how prediagnostic circulating concentrations of IGF-1 were related to breast cancer risk in the UK Biobank study, a large prospective cohort in which a subsample of participants has repeat IGF-1 measures enabling correction for regression dilution bias. Next, we used a two-sample MR approach to examine potential causal associations by combining genetic variants associated with circulating IGF-1 and IGFBP-3 concentrations in genome-wide association studies (GWAS), and then assessing the association of these variants with breast cancer (overall, ER+, and ER−) risk in a large consortium of 122 977 breast cancer cases and 105 974 controls.10
Methods
UK Biobank—observational analysis
Study participants
The UK Biobank is a prospective cohort of 502 536 adults aged between 40 and 69 years (229 182 men and 273 474 women) who were recruited between 2006 and 2010.11 The UK Biobank invited ∼9.2 million people to participate through postal invitation with a telephone follow-up, with a response rate of 5.7%. All participants were registered with the UK National Health Service and lived within ∼25 miles (40 km) of one of the 22 study assessment centres. The UK Biobank has approval from the North West Multi-centre Research Ethics Committee, the National Information Governance Board for Health and Social Care in England and Wales, and the Community Health Index Advisory Group in Scotland. In addition, an independent Ethics and Governance Council was formed in 2004 to oversee UK Biobank's continuous adherence to the Ethics and Governance Framework which was developed for the study (http://www.ukbiobank.ac.uk/ethics/). All participants provided written informed consent at recruitment and were to be followed up using data-linkage. This research has been conducted using the UK Biobank Resource under application numbers 3248 and 24 494.
During the baseline recruitment visit, participants were asked to complete a self-administered touchscreen questionnaire, which included questions on sociodemographics (including age, sex, education, and postcode, used to assign Townsend deprivation score), health/medical history, and lifestyle exposures (including smoking habits, dietary intakes, and alcohol consumption). At the baseline visit, participants also underwent physical measurements, including body weight, height, and waist circumference. Blood samples were collected from all participants at recruitment, and repeat blood samples were collected from a subset of ∼20 000 participants who re-attended the assessment centre between 2012 and 2013. Blood samples were centrifuged, and serum stored at −80°C.
Exclusions before the onset of analyses were men (n = 229 134); women with prevalent cancer (including in situ breast cancer, but excluding non-malignant skin cancer) at recruitment (n = 18 560); participants in whom genetic sex differed from reported gender (n = 121), missing data on body size measurements (n = 1350); prevalent type-2 diabetes or unknown diabetes status at recruitment based on hospital records and self-report (because diabetes medications can affect circulating concentrations of IGF-112; n = 10 705); women who reported oral contraceptive and menopausal hormone use at recruitment (because oral estrogens alter hepatic protein production and change circulating concentrations of IGF-112; n = 20 988); and participants without an IGF-1 measurement (n = 15 415). Our analysis therefore included 206 263 women.
Blood collection and laboratory methods
As part of the UK Biobank Biomarker Project,13 serum concentrations of IGF-1 (Liaison XL, DiaSorin S.p.A., Italy), testosterone, and sex hormone binding globulin (SHBG) were determined by a chemiluminescent immunoassay (DXI 800, Beckman Coulter, London, UK). The immuno-turbidimetric method (DXI 800) was used to assay serum high sensitivity C-reactive protein (CRP) concentrations. Glycated haemoglobin (HbA1c) concentrations were determined using the high performance liquid chromatography (HPLC) Variant II Turbo 2.0 system (Bio-Rad, Hercules, CA). Full details on assay performance have been published.13 In summary, average within-laboratory (total) coefficients of variation for low, medium, and high internal quality control level samples for each biomarker ranged from 1.7% to 15.3% (for IGF-1, the coefficients of variation ranged from 5.3% to 6.2%).13 A total of 6711 women had IGF-1 concentrations measured in blood samples collected at both the recruitment and repeat assessment visit (median of 4 years apart).
Assessment of outcome
Incident cancer cases and cancer cases first recorded in death certificates within the UK Biobank cohort were identified through linkage to national cancer registries and death records. Complete follow-up was available until 31 March 2016 for England and Wales and 31 October 2015 for Scotland. Cancer incidence data were coded using the 10th Revision of the International Classification of Diseases (ICD-10). Breast cancer was defined as registration ICD-10: C50.
Statistical analysis
Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated using Cox proportional hazards models. Age was the primary time variable in all models. Time at entry was the age at recruitment. Exit time was the age at whichever of the following came first: breast cancer diagnosis (n = 4352; 99.8% of all cases), breast cancer at death without prior diagnosis (n = 8; 0.2% of all cases), death, or the last date at which follow-up was considered complete. Models were stratified by age at recruitment in 5-year categories, Townsend deprivation index fifths, and region of the recruitment assessment centre. Deviations from proportionality were assessed using an analysis of Schoenfeld residuals, with no evidence of non-proportionality being detected. IGF-1 was modelled on the continuous scale (per 5 nmol/l) and with participants grouped into sex-specific fifths of circulating concentrations.
HRs were additionally corrected for regression dilution using regression dilution ratios obtained from the subsample of 6711 women with repeated IGF-1 measurement9,14; to obtain corrected HRs, the log HRs and their standard errors were divided by the regression dilution ratio for IGF-1 (0.74) and then exponentiated.15 Possible non-linear effects were modelled using restricted cubic spline models with five knots placed at Harrell's default percentiles of circulating IGF-1 concentrations.16
The multivariable model (model 1) was adjusted for a set of breast cancer risk factors determined a priori, namely total physical activity, height, alcohol consumption frequency, smoking status and intensity, educational level, ever use of hormone replacement therapy, parity and age at first birth, and an interaction between menopausal status and BMI. We also additionally adjusted the multivariable models (model 2) for markers of inflammatory, sex hormone, and glycaemic pathways that are known to interrelate/crosstalk with the IGF system,12 namely CRP, testosterone, SHBG, and HbA1c. Statistical tests for trend were calculated using the ordinal fifths of IGF-1 entered into the model as a continuous variable.
The circulating IGF-1 and breast cancer associations were further assessed across subgroups of BMI, height, menopausal status at recruitment, ages at blood collection and diagnosis, follow-up time, smoking status, and circulating concentrations of CRP, HbA1c, testosterone, and SHBG. Interaction terms (multiplicative scale) between these variables and circulating IGF-1 concentrations were included in separate models, and the statistical significance of the cross-product terms were evaluated using likelihood ratio tests, or competing risk for follow-up time and age at diagnosis.
Statistical tests were all two-sided and a P value <0.05 was considered statistically significant. Analyses were conducted using Stata version 14 (StataCorp, College Station, TX).
Mendelian randomization
Genetic determinants of IGF-1 and IGFBP-3
Genetic markers for circulating IGF-1 and IGFBP-3 concentrations comprised SNPs identified (P < 5 × 10−8) from the largest GWAS to date.17,18 For IGF-1, this GWAS was of 194 174 women from the UK Biobank.18 The GWAS analyses of IGFBP-3 combined data on 18 995 individuals (men and women) from 13 studies.17 All participants were of European ancestry. From the genome-wide significant variants identified in these GWAS, we excluded correlated single nucleotide polymorphisms (SNPs) based on a linkage disequilibrium level of R2 < 0.01. The instruments for IGF-1 (265 SNPs) and IGFBP-3 (four SNPs) explained 5.2% (F-statistic value of 40.1) and 6.1% (F-statistic value of 308.4) of variability in circulating concentrations, respectively. Summary information on the genetic instruments, and the effect estimates for each individual SNP with IGF-1 and IGFBP-3 concentrations, are presented in supplementary Tables S1 and S2, available at Annals of Oncology online.
Data on breast cancer
Summary data for the associations of the IGF-1- and IGFBP-3-related genetic variants with breast cancer were obtained from a GWAS of 228 951 women [122 977 breast cancer (69 501 ER+, 21 468 ER−) cases and 105 974 controls] of European ancestry from the Breast Cancer Association Consortium (BCAC).10 Genotypes were imputed using the 1000 Genomes Project reference panel and the regression models adjusted for the first 10 principal components and country or study. Effect estimates for the association of each individual SNP with breast cancer are presented in supplementary Table S2, available at Annals of Oncology online.
Statistical analysis
Two-sample MR analyses using summary data and an inverse variance weighted approach were implemented. MR results correspond to an odds ratio (OR) per 5 nmol/l of genetically-predicted IGF-1 concentration and per 1-standard deviation (SD) of IGFBP-3. Heterogeneity of associations across breast cancer subtypes was assessed by calculating χ2 statistics. Cochran's Q statistics quantified heterogeneity across the individual SNPs. Sensitivity analyses were used to check and correct for the presence of horizontal pleiotropy [i.e. genetic variants influencing the outcome (breast cancer risk) via a different biological pathway from the exposures of interest (IGF-1 and IGFBP-3)]. To evaluate the extent to which directional pleiotropy (non-balanced horizontal pleiotropy in the MR risk estimates) may have affected the causal estimates for the IGF-1 and breast cancer association, we used an MR-Egger regression approach.19 We also computed OR estimates using the complementary weighted-median method that can give valid MR estimates under the presence of pleiotropy when up to 50% of the included instruments are invalid.20 The presence of pleiotropy was also assessed using the MR pleiotropy residual sum and outlier test (MR-PRESSO). In this, outlying SNPs are excluded from the instruments and the effect estimates are reassessed.21 As a visual evaluation of directional pleiotropy (asymmetry), we also examined a funnel plot of the effect estimate and standard error of each SNP within the IGF-1 instrument on breast cancer risk. For the IGFBP-3 instrument, we conducted leave-one-out analyses to assess the influence of individual variants on the observed associations. All statistical analyses were carried out using the MendelianRandomization package22 for R (R Foundation for Statistical Computing, Austria).
Results
UK Biobank—observational analysis
After a median follow-up time of 7.1 years, 4360 incident breast cancer cases were recorded. Compared with those in the lowest fifth, participants in the highest circulating IGF-1 fifth were younger, taller, had lower BMI and waist circumference, and were more likely to be never smokers, nulliparous, and never users of hormone replacement therapy (Table 1). In addition, participants in the highest circulating IGF-1 fifth had lower circulating concentrations of CRP, HbA1c, and SHBG, with higher circulating concentrations of testosterone.
Table 1.
Characteristics of UK Biobank study participants by fifth of circulating insulin-like growth factor-1 (IGF-1) concentrations (N = 206 263 women)
| Characteristic | IGF-1 concentrations |
||||
|---|---|---|---|---|---|
| 1 |
2 |
3 |
4 |
5 |
|
| (n = 41 264) | (n = 41 256) | (n = 41 239) | (n = 41 264) | (n = 41 240) | |
| IGF-1 at baseline, nmol/l | 13.8 (2.0) | 18.0 (0.9) | 20.9 (0.8) | 23.8 (0.9) | 29.4 (3.9) |
| IGF-1 at follow-up, nmol/l, n = 6711 | 14.9 (3.5) | 18.0 (3.3) | 20.2 (3.4) | 22.4 (3.6) | 26.4 (5.2) |
| Breast cancer, n | 836 | 818 | 911 | 895 | 900 |
| Age at baseline, years | 59.4 (6.9) | 57.8 (7.4) | 56.4 (7.8) | 54.9 (8.0) | 52.6 (8.2) |
| BMI, kg/m2 | 28.3 (6.0) | 27.2 (5.2) | 26.7 (4.8) | 26.4 (4.5) | 26.0 (4.2) |
| Waist circumference, cm | 87.5 (13.8) | 84.9 (12.4) | 83.7 (11.7) | 82.9 (11.2) | 82.0 (10.6) |
| Standing height, cm | 161.5 (6.3) | 162.1 (6.2) | 162.5 (6.3) | 162.9 (6.3) | 163.5 (6.3) |
| Socioeconomic status (Townsend deprivation index), n (%) | |||||
| Most deprived fifth | 9330 (22.6) | 8255 (20.0) | 7799 (18.9) | 7921 (19.2) | 7898 (19.2) |
| Qualification, n (%) | |||||
| Professional qualification, college/university degree | 21 109 (51.2) | 22 767 (55.2) | 23 763 (57.6) | 24 643 (59.7) | 25 824 (62.6) |
| Smoking, n (%) | |||||
| Never smoker | 23 657 (57.3) | 24 223 (58.7) | 24 830 (60.2) | 25 200 (61.1) | 25 806 (62.6) |
| Current smoker, ≥15 cigs/day | 1519 (3.7) | 1450 (3.5) | 1310 (3.2) | 1367 (3.3) | 1352 (3.3) |
| Alcohol intake, n (%) | |||||
| Never | 4987 (12.1) | 3712 (9.0) | 3354 (8.1) | 3239 (7.8) | 3128 (7.6) |
| Daily/almost daily | 6852 (16.6) | 7313 (17.7) | 7036 (17.1) | 6666 (16.2) | 5586 (13.5) |
| Physical activity, n (%) | |||||
| <10 METh/week | 10 481 (25.4) | 9546 (23.1) | 9314 (22.6) | 9216 (22.3) | 8921 (21.6) |
| ≥60 METh/week | 8245 (20.0) | 8544 (20.7) | 8379 (20.3) | 8383 (20.3) | 8142 (19.7) |
| HRT use, n (%) | |||||
| Never | 23 135 (56.1) | 25 505 (61.8) | 27 310 (66.2) | 28 846 (69.9) | 30 992 (75.2) |
| Past | 18 129 (43.9) | 15 751 (0.4) | 13 929 (33.8) | 12 418 (30.1) | 10 248 (24.9) |
| OCP use, n (%) | |||||
| Never | 9642 (23.4) | 8376 (20.3) | 7802 (18.9) | 6896 (16.7) | 6485 (15.7) |
| Past | 31 622 (76.6) | 32 880 (79.7) | 33 437 (81.1) | 34 368 (83.3) | 34 755 (84.3) |
| Parity, n (%) | |||||
| Nulliparous | 6650 (16.1) | 6854 (16.6) | 7551 (18.3) | 7816 (18.9) | 8709 (21.1) |
| Menopausal status, n (%) | |||||
| Pre- | 4598 (11.1) | 7382 (17.9) | 9642 (23.4) | 12 497 (30.3) | 16 675 (40.4) |
| Post- | 35 099 (85.1) | 32 049 (77.7) | 29 551 (71.7) | 26 537 (64.3) | 22 041 (53.4) |
| C-reactive protein, mg/l | 3.90 (5.6) | 2.77 (4.2) | 2.3 (3.6) | 2.0 (3.4) | 1.65 (3.1) |
| Glycated haemoglobin, mmol/mol | 35.9 (4.9) | 35.4 (4.2) | 35.1 (4.1) | 34.9 (4.0) | 34.6 (3.9) |
| Testosterone, nmol/l | 1.1 (0.6) | 1.1 (0.6) | 1.1 (0.6) | 1.2 (0.6) | 1.2 (0.7) |
| Sex hormone binding globulin, nmol/l | 64.5 (32.1) | 62.1 (28.7) | 60.9 (27.3) | 59.7 (26.3) | 57.2 (25.1) |
Mean and standard deviation unless specified.
BMI, body mass index; cig, cigarettes; METh, metabolic equivalent hours; HRT, hormone replacement therapy; OCP, oral contraceptive pill.
Association between circulating IGF-1 concentrations and breast cancer risk
Circulating IGF-1 concentrations were positively associated with breast cancer risk in the minimally adjusted model (HR for highest versus lowest fifth = 1.23, 95% CI = 1.12–1.36; Ptrend < 0.0001) (Table 2). Statistical adjustment for other breast cancer risk factors and circulating concentrations of CRP, HbA1c, testosterone, and SHBG did not materially change the association (HR for highest versus lowest fifth = 1.24, 95% CI = 1.12–1.37; Ptrend < 0.0001). In the restricted cubic spline model, no deviation from linearity for the relationship between IGF-1 and breast cancer was observed (Pnon-linear = 0.85). In the continuous multivariable model, adjusted for circulating concentrations of CRP, HbA1c, testosterone, and SHBG, a 5 nmol/l increment in IGF-1 was associated with a higher breast cancer risk (HR = 1.08, 95% CI = 1.05–1.11). Subsequent correction for regression dilution bias resulted in a larger positive relationship (HR = 1.11, 95% CI = 1.07–1.16). Similar magnitude positive associations were found according to subgroups of follow-up time, ages at blood collection and diagnosis, menopausal status at recruitment, and other breast cancer risk factors (Pheterogeneity > 0.08) (Figure 1, supplementary Table S3, available at Annals of Oncology online).
Table 2.
Risk (hazard ratios) of breast cancer associated with circulating insulin-like growth factor-1 (IGF-1) concentrations in the UK Biobank
| IGF-1 at baseline (nmol/l) | N Cases/participants | HR (95% CI) |
HR (95% CI) |
HR (95% CI) |
|---|---|---|---|---|
| Model0 | Model1 | Model2 | ||
| Fifths | ||||
| 1 (<16.4) | 836/40 428 | 1 (reference) | 1 (reference) | 1 (reference) |
| 2 (16.4 to <19.5) | 818/40 438 | 1.00 (0.91–1.10) | 1.01 (0.92–1.11) | 1.01 (0.91–1.11) |
| 3 (19.5 to <22.3) | 911/40 328 | 1.15 (1.05–1.27) | 1.16 (1.06–1.28) | 1.15 (1.05–1.27) |
| 4 (22.3 to <25.6) | 895/40 369 | 1.16 (1.06–1.28) | 1.18 (1.07–1.30) | 1.17 (1.06–1.29) |
| 5 (≥25.6) | 900/40 340 | 1.23 (1.12–1.36) | 1.25 (1.13–1.38) | 1.24 (1.12–1.37) |
| Ptrend | <0.0001 | <0.0001 | <0.0001 | |
| Per 5 nmol/l increment | 4360/206 263 | 1.08 (1.05–1.11) | 1.08 (1.05–1.11) | 1.08 (1.05–1.11) |
| Per 5 nmol/l increment (adjusted)a | 4360/206 263 | 1.11 (1.07–1.15) | 1.11 (1.07–1.15) | 1.11 (1.07–1.16) |
Model0: minimally adjusted model using age as the underlying time variable and stratified by Townsend deprivation index (fifths), region of the recruitment assessment centre, and age at recruitment (5-year categories).
Model1: multivariable Cox regression model using age as the underlying time variable and stratified by Townsend deprivation index (fifths), region of the recruitment assessment centre, and age at recruitment (5-year categories). Models adjusted for total physical activity (<10, 10 to <20, 20 to <40, 40 to <60, ≥60 metabolic equivalent hours per week, unknown); height (per 10 cm); alcohol consumption frequency (never, special occasions only, one to three times per month, one to two times per week, three to four times per week, daily/almost daily, unknown); smoking status and intensity (never, former, current <15 per day, current ≥15 per day, current intensity unknown, unknown); educational level (CSEs/O-levels/GCSEs or equivalent, NVQ/HND/HNC/A-levels/AS-levels or equivalent, other professional qualifications, college/university degree, none of the above, unknown); ever use of hormone replacement therapy (no, yes, unknown); parity, age at first birth (nulliparous; 1–2, <25 years; 1–2, 25–30 years; 1–2, ≥30 years; 1–2, unknown; ≥3, <25 years; ≥3, 25–30 years; ≥3, ≥30 years; ≥3, unknown; unknown); and the interaction between menopausal status and body mass index (kg/m2).
Model2: Model1 plus additional adjustment for circulating concentrations (fifths, missing/unknown) of C-reactive protein (CRP; mg/l), glycated haemoglobin (HbA1c; mmol/mol), testosterone (nmol/l), and sex hormone binding globulin (SHBG; nmol/l).
CI, confidence interval; HR, hazard ratio; SD, standard deviation.
HRs per SD increment were additionally corrected for regression dilution using a regression dilution ratio (0.74) obtained from the subsample of women with repeat IGF-1 measurements.
Figure 1.
Subgroup analyses of the association between circulating insulin-like growth factor-1 (IGF-1) concentrations and breast cancer risk in the UK Biobank (per 5 nmol/l increment).
Multivariable Cox regression model using age as the underlying time variable and stratified by Townsend deprivation index (fifths), region of the recruitment assessment centre, and age at recruitment (5-year categories). Models adjusted for total physical activity (<10, 10 to <20, 20 to <40, 40 to <60, ≥60 metabolic equivalent hours per week, unknown); height (per 10 cm); alcohol consumption frequency (never, special occasions only, one to three times per month, one to two times per week, three to four times per week, daily/almost daily, unknown); smoking status and intensity (never, former, current <15 per day, current ≥15 per day, current intensity unknown, unknown); educational level (CSEs/O-levels/GCSEs or equivalent, NVQ/HND/HNC/A-levels/AS-levels or equivalent, other professional qualifications, college/university degree, none of the above, unknown); ever use of hormone replacement therapy (no, yes, unknown); parity, age at first birth (nulliparous; 1–2, <25 years; 1–2, 25–30 years; 1–2, ≥30 years; 1–2, unknown; ≥3, <25 years; ≥3, 25–30 years; ≥3, ≥30 years; ≥3, unknown; unknown); the interaction between menopausal status and body mass index (kg/m2); and circulating concentrations (fifths, missing/unknown) of C-reactive protein (CRP; mg/l), glycated haemoglobin (HbA1c; mmol/mol), testosterone (nmol/l), and sex hormone binding globulin (SHBG; nmol/l). Median values: height = 162 cm; CRP = 1.3 mg/l; HbA1c = 35.1 mmol/mol; testosterone = 1 nmol/l; SHBG = 56.3 nmol/l. HRs per 5 nmol/l increment were additionally corrected for regression dilution using a regression dilution ratio (0.74) obtained from the subsample of women with repeat IGF-1 measurements.
CI, confidence interval; HR, hazard ratio.
Mendelian randomization
Association between genetically-predicted circulating IGF-1 concentrations and breast cancer risk
In the inverse-variance weighted model, a 5 nmol/l increment in genetically-predicted IGF-1 concentrations was associated with greater breast cancer risk (OR = 1.05, 95% CI = 1.01–1.10; P = 0.02) (Table 3). IGF-1 was positively associated with ER+ (OR = 1.06, 95% CI = 1.01–1.11; P = 0.03), but not ER− (OR = 1.02, 95% CI = 0.96–1.08; P = 0.58) tumours (Pheterogeneity = 0.32). Evidence of effect heterogeneity (Cochran's Q P values < 0.0001) was found for all models. Little evidence of directional pleiotropy for the breast cancer and ER+ breast cancer models (MR-Egger intercept P values >0.2) was found, with similar magnitude effect estimates observed for the weighted median and lower powered MR-Egger models (Table 3). A similar pattern of results was found when the MR-PRESSO test detected outlier SNPs were excluded from the models (supplementary Table S4, available at Annals of Oncology online). The funnel plots for the IGF-1 instrument indicated a symmetric distribution of effect estimates (supplementary Figure S1, available at Annals of Oncology online).
Table 3.
Mendelian randomization estimates between circulating concentrations of insulin-like growth factor-1 (IGF-1) and insulin-like growth factor-binding protein-3 (IGFBP-3) and risk of breast cancer (N = 122 977 breast cancer cases and N = 105 974 controls)
| Mendelian randomization (MR) |
|||
|---|---|---|---|
| Inverse-variance-weighted (IVW) |
Weighted median |
MR-Egger |
|
| OR (95% CI) per 5 nmol/l increment | OR (95% CI) per 5 nmol/l increment | OR (95% CI) per 5 nmol/l increment | |
| IGF-1 (n = 265 SNPs) | |||
| Breast cancer | 1.05 (1.01–1.10) | 1.08 (1.03–1.13) | 1.12 (1.01–1.25) |
| Breast cancer, ER+ | 1.06 (1.01–1.11) | 1.10 (1.04–1.18) | 1.12 (1.00–1.26) |
| Breast cancer, ER− | 1.02 (0.96–1.08) | 1.03 (0.95–1.11) | 1.17 (1.02–1.34) |
| IVW |
Weighted median |
MR-Egger |
|
|---|---|---|---|
| OR (95% CI) per 1-SD increment | OR (95% CI) per 1-SD increment | OR (95% CI) per 1-SD increment | |
| IGFBP-3 (n = 4 SNPs) | |||
| Breast cancer | 1.00 (0.97–1.04) | 1.00 (0.96–1.04) | 0.95 (0.87–1.03) |
| Breast cancer, ER+ | 1.00 (0.96–1.05) | 1.00 (0.95–1.04) | 0.92 (0.83–1.02) |
| Breast cancer, ER− | 0.98 (0.92–1.04) | 0.98 (0.91–1.05) | 1.03 (0.88–1.20) |
CI, confidence interval; ER, estrogen receptor; OR, odds ratio; SD, standard deviation; SNP, single nucleotide polymorphism.
Association between genetically-predicted circulating IGFBP-3 concentrations and breast cancer risk
No association was found between genetically-predicted circulating IGFBP-3 concentrations and breast cancer risk (Table 3). A similar null result was found for the weighted median, MR-Egger, and leave-one-out sensitivity analyses (Table 3, supplementary Table S5, available at Annals of Oncology online).
Discussion
In the observational analyses of UK Biobank data, we found that higher circulating concentrations of IGF-1 were associated with greater breast cancer risk. This relationship was consistent for premenopausal and postmenopausal women and by follow-up time. Consistent with this finding, in our MR analyses, we found a positive association between genetically-predicted IGF-1 concentrations and breast cancer risk, with this effect restricted to ER+ tumours. These results support a probable causal role of the IGF pathway in ER+ breast cancer development.
The positive association found between circulating IGF-1 concentrations and breast cancer in our UK Biobank observational analyses was monotonic and consistent with the result from the Endogenous Hormones and Breast Cancer Collaborative Group, a pooled analysis that included 4790 breast cancers cases and 9428 matched controls from 17 prospective studies.8 Similar to the pooled analysis, this positive association did not differ by menopausal status, follow-up time, and subgroups of other breast cancer risk factors. Our analysis, which included 4360 incident breast cancer cases, is the largest single study to examine the IGF-1 and breast cancer relationship. Uniquely, measurements of circulating IGF-1 and other biomarkers were available in the full UK Biobank cohort, and we were able to adjust our multivariable models for other serologic factors related to both circulating IGF-1 concentrations and breast cancer risk, namely testosterone, SHBG, CRP, and HbA1c.23, 24, 25, 26, 27 The risk estimates for the IGF-1 and breast cancer relationship were largely unchanged after multivariable statistical adjustment for these biomarkers and other established risk factors. A further unique aspect of our analysis was our correction for regression dilution bias using the repeat IGF-1 measurements available for 6711 women, thereby mitigating the combined effects of measurement error and within-person variability on our risk estimates.9 This correction resulted in a strengthening of the positive association, supporting the likelihood that previous epidemiological studies that relied on a single measurement of IGF-1 concentrations underestimated the strength of the positive association with breast cancer risk. For instance, the Endogenous Hormones and Breast Cancer Collaborative Group reported an OR of 1.25 (95% CI = 1.13–1.39) for an 80 percentile difference in IGF-18; while the HR at an equivalent scale in the current UK Biobank analysis rose from 1.26 (95% CI = 1.15–1.38) to 1.37 (95% CI = 1.21–1.55) after correction for regression dilution bias.
Observational analyses may be subject to residual confounding and reverse causality, making causal inference challenging. We conducted MR analyses of the associations between IGF-1, IGFBP-3, and breast cancer risk. MR uses germline genetic variants as proxies (instrumental variables) to allow causal inference between a given exposure and outcome. Unlike traditional observational analyses, MR analyses should be largely free of confounding and reverse causality due to the random assortment of alleles at meiosis and germline genetic variants being fixed at conception, and thus unaffected by the disease process. For IGF-1, the MR analysis yielded a positive effect estimate similar to our observational analysis. This positive effect was only present for ER+ and not ER− breast cancer, a result consistent with earlier observational8 and laboratory evidence,28 which suggests that crosstalk from estrogen signalling pathways may influence the IGF-1 and breast cancer relationship.
The bioavailability of IGF-1 in circulation is partly regulated by IGFBPs, with most bound to IGFBP-3. In addition to enhancing or inhibiting actions of IGF ligands, in vitro experimental models suggest that IGFBP-3 can inhibit breast cancer proliferation and induce apoptosis.4,29 Our MR analyses found no evidence of an association between genetically-predicted IGFBP-3 concentrations and breast cancer risk. This result is consistent with the Endogenous Hormones and Breast Cancer Collaborative Group analysis that reported a null association for IGFBP-3 concentrations and breast cancer risk after the multivariable models were adjusted for circulating IGF-1 concentrations.8 Taken together, these results provide little evidence of IGFBP-3 having a direct effect, independent of its role in IGF ligand binding, in breast cancer development.
A fundamental assumption of MR analyses is that the genetic instrument should not influence the outcome via a different biological pathway from the exposure of interest (horizontal pleiotropy). We conducted various sensitivity analyses to assess the possible influence of horizontal pleiotropy on our causal estimates, and our results were robust to these various tests. The possibility exists, however, that our results may have been influenced by pleiotropy from other unmeasured IGF axis components.30 To date, the only GWAS analyses that have been conducted for components of this pathway are for IGF-1 and IGFBP-3, therefore the extent of this possible pleiotropy within the IGF axis is uncertain. Genetic instruments are now required for other IGF system components to disentangle possible biological effects of specific ligands and binding proteins in breast cancer development.
The current study is the largest and most comprehensive investigation of the role of IGF-1 in breast cancer development. A limitation of our observational analysis is that tumour subtype data are currently unavailable in the UK Biobank; however, these data were available for our MR analyses and we found that the positive effect for IGF-1 and breast cancer was only present for ER+ tumours. For our MR analyses, we were unable to stratify the analyses by menopausal status; however, our observational analyses found no difference in the IGF-1 and breast cancer relationship between premenopausal and postmenopausal women. Our use of summary-level data for our MR analyses meant that we were unable to examine possible non-linear effects or whether the associations between IGF-1 and breast cancer differed according to subgroups of other risk factors (e.g. BMI, alcohol consumption); however, in our observational analyses, which yielded a similar association to our MR result, we found a linear association between IGF-1 and breast cancer, and detected no heterogeneity by subgroups of other risk factors. Finally, the UK Biobank participation rate was relatively low (5.7%), which may have introduced selection bias; however, this low response rate should not markedly influence etiological associations such as those reported in our observational analysis, and breast cancer incidence rates were similar to the UK national average.31,32
In conclusion, given that plausible biological mechanisms have been identified,1,2, our observational and MR results support a probable causal relationship between circulating IGF-1 concentrations and breast cancer. This result suggests that pharmacological or lifestyle interventions targeting the IGF pathway may be beneficial in preventing breast tumorigenesis.
Acknowledgements
This work has been conducted using the UK Biobank Resource under Application Numbers 3248 and 24 494, and we express our gratitude to the participants and those involved in building the resource. UK Biobank is an open access resource. Bona fide researchers can apply to use the UK Biobank dataset by registering and applying at http://ukbiobank.ac.uk/register-apply/. The breast cancer genome-wide association analyses were supported by the Government of Canada through Genome Canada and the Canadian Institutes of Health Research, the ‘Ministère de l'Économie, de la Science et de l'Innovation du Québec' through Genome Québec and grant PSR-SIIRI-701, The National Institutes of Health (U19 CA148065, X01HG007492), Cancer Research UK (C1287/A10118, C1287/A16563, C1287/A10710), and The European Union (HEALTH-F2-2009-223175 and H2020 633784 and 634935). All studies and funders are listed in Michailidou et al.10
AK is supported by the Wellcome Trust (LEAP 205212/Z/16/Z). GF, APC, RCT and TJK are supported by Cancer Research UK (C8221/A19170). RMM is supported by a Cancer Research UK program grant (C18281/A19169) and the National Institute for Health Research (NIHR) Bristol Biomedical Research Centre. The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.
Disclaimer
Where authors are identified as personnel of the International Agency for Research on Cancer/World Health Organization, the authors alone are responsible for the views expressed in this article and they do not necessarily represent the decisions, policy, or views of the International Agency for Research on Cancer/World Health Organization.
Funding
None declared.
Disclosure
The authors have declared no conflicts of interest.
Supplementary data
References
- 1.Pollak M. Insulin and insulin-like growth factor signalling in neoplasia. Nat Rev Cancer. 2008;8:915. doi: 10.1038/nrc2536. [DOI] [PubMed] [Google Scholar]
- 2.Samani A.A., Yakar S., LeRoith D. The role of the IGF system in cancer growth and metastasis: overview and recent insights. Endocr Rev. 2007;28(1):20–47. doi: 10.1210/er.2006-0001. [DOI] [PubMed] [Google Scholar]
- 3.Kelley K.M., Oh Y., Gargosky S.E. Insulin-like growth factor-binding proteins (IGFBPs) and their regulatory dynamics. Int J Biochem Cell Biol. 1996;28(6):619–637. doi: 10.1016/1357-2725(96)00005-2. [DOI] [PubMed] [Google Scholar]
- 4.Perks C.M., Holly J.M.P. IGF binding proteins (IGFBPs) and regulation of breast cancer biology. J Mammary Gland Biol Neoplasia. 2008;13(4):455–469. doi: 10.1007/s10911-008-9106-4. [DOI] [PubMed] [Google Scholar]
- 5.Furlanetto R.W., DiCarlo J.N. Somatomedin-c receptors and growth effects in human breast cells maintained in long-term tissue culture. Cancer Res. 1984;44(5):2122–2128. [PubMed] [Google Scholar]
- 6.Peyrat J.P., Louchez M.M., Lefebvre J. Plasma insulin-like growth factor-1 (IGF-1) concentrations in human breast cancer. Eur J Cancer. 1993;29(4):492–497. doi: 10.1016/s0959-8049(05)80137-6. [DOI] [PubMed] [Google Scholar]
- 7.Hankinson S.E., Willett W.C., Colditz G.A. Circulating concentrations of insulin-like growth factor I and risk of breast cancer. Lancet. 1998;351(9113):1393–1396. doi: 10.1016/S0140-6736(97)10384-1. [DOI] [PubMed] [Google Scholar]
- 8.Endogenous Hormones and Breast Cancer Collaborative Group. Key T.J., Appleby P.N., Reeves G.K., Roddam A.W. Insulin-like growth factor 1 (IGF1), IGF binding protein 3 (IGFBP3), and breast cancer risk: pooled individual data analysis of 17 prospective studies. Lancet Oncol. 2010;11(6):530–542. doi: 10.1016/S1470-2045(10)70095-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Clarke R., Shipley M., Lewington S. Underestimation of risk associations due to regression dilution in long-term follow-up of prospective studies. Am J Epidemiol. 1999;150(4):341–353. doi: 10.1093/oxfordjournals.aje.a010013. [DOI] [PubMed] [Google Scholar]
- 10.Michailidou K., Lindstrom S., Dennis J. Association analysis identifies 65 new breast cancer risk loci. Nature. 2017;551(7678):92–94. doi: 10.1038/nature24284. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.UK-Biobank . 2010. UK Biobank. Protocol for a large-scale prospective epidemiological resources.https://www.ukbiobank.ac.uk/wp-content/uploads/2011/11/UK-Biobank-Protocol.pdf Available at: [Google Scholar]
- 12.Hosgood H.D., Gunter M.J., Murphy N. The relation of obesity-related hormonal and cytokine levels with multiple myeloma and non-hodgkin lymphoma. Front Oncol. 2018;8:103. doi: 10.3389/fonc.2018.00103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.UK-Biobank . 2019. UK Biobank Biomarker Project - companion document to accompany serum biomarker data.https://nam03.safelinks.protection.outlook.com/?url=http%3A%2F%2Fbiobank.ctsu.ox.ac.uk%2Fshowcase%2Fshowcase%2Fdocs%2Fserum_biochemistry.pdf&data=02%7C01%7Ccorrections.tnq%40elsevier.com%7C7a01f4e481db4d9f468d08d7be934d60%7C9274ee3f94254109a27f9fb15c10675d%7C0%7C0%7C637187412624951252&sdata=uHrO7z1Fb%2Fc4aFPLfnspK66cQC%2FF4i6znNjsZaejh2A%3D&reserved=0 Available at: [Google Scholar]
- 14.MacMahon S., Peto R., Collins R. Blood pressure, stroke, and coronary heart disease: part 1, prolonged differences in blood pressure: prospective observational studies corrected for the regression dilution bias. Lancet. 1990;335(8692):765–774. doi: 10.1016/0140-6736(90)90878-9. [DOI] [PubMed] [Google Scholar]
- 15.Clarke R., Emberson J.R., Breeze E. Biomarkers of inflammation predict both vascular and non-vascular mortality in older men. Eur Heart J. 2008;29(6):800–809. doi: 10.1093/eurheartj/ehn049. [DOI] [PubMed] [Google Scholar]
- 16.Harrell F. Springer; New York: 2001. Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis. [Google Scholar]
- 17.Teumer A., Qi Q., Nethander M. Genomewide meta-analysis identifies loci associated with IGF-I and IGFBP-3 levels with impact on age-related traits. Aging Cell. 2016;15(5):811–824. doi: 10.1111/acel.12490. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.NealeLab UK Biobank GWAS Results. 2019. http://www.nealelab.is/uk-biobank Available at:
- 19.Bowden J., Davey Smith G., Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol. 2015;44(2):512–525. doi: 10.1093/ije/dyv080. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Bowden J., Davey Smith G., Haycock P.C. Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol. 2016;40(4):304–314. doi: 10.1002/gepi.21965. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Verbanck M., Chen C.-Y., Neale B. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet. 2018;50(5):693–698. doi: 10.1038/s41588-018-0099-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Yavorska O.O., Burgess S. MendelianRandomization: an R package for performing Mendelian randomization analyses using summarized data. Int J Epidemiol. 2017;46(6):1734–1739. doi: 10.1093/ije/dyx034. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Endogenous Hormones and Breast Cancer Collaborative Group. Key T.J., Appleby P.N. Sex hormones and risk of breast cancer in premenopausal women: a collaborative reanalysis of individual participant data from seven prospective studies. Lancet Oncol. 2013;14(10):1009–1019. doi: 10.1016/S1470-2045(13)70301-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Key T.J., Appleby P.N., Barnes I., Reeves G., Endogenous Hormones and Breast Cancer Collaborative Group Endogenous sex hormones and breast cancer in postmenopausal women: Reanalysis of nine prospective studies. J Natl Cancer Inst. 2002;94(8):606–616. doi: 10.1093/jnci/94.8.606. [DOI] [PubMed] [Google Scholar]
- 25.Kaaks R., Rinaldi S., Key T.J. Postmenopausal serum androgens, oestrogens and breast cancer risk: the European prospective investigation into cancer and nutrition. Endocr Relat Cancer. 2005;12(4):1071–1082. doi: 10.1677/erc.1.01038. [DOI] [PubMed] [Google Scholar]
- 26.Chan D.S.M., Bandera E.V., Greenwood D.C. Circulating C-reactive protein and breast cancer risk—systematic literature review and meta-analysis of prospective cohort studies. Cancer Epidemiol Biomarkers Prev. 2015;24(10):1439–1449. doi: 10.1158/1055-9965.EPI-15-0324. [DOI] [PubMed] [Google Scholar]
- 27.de Beer J.C., Liebenberg L. Does cancer risk increase with HbA1c, independent of diabetes? Br J Cancer. 2014;110(9):2361–2368. doi: 10.1038/bjc.2014.150. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Dupont J., Le Roith D. Insulin-like growth factor 1 and oestradiol promote cell proliferation of MCF-7 breast cancer cells: new insights into their synergistic effects. Mol Pathol. 2001;54(3):149–154. doi: 10.1136/mp.54.3.149. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Perks C.M., Burrows C., Holly J.M.P. Intrinsic, pro-apoptotic effects of IGFBP-3 on breast cancer cells are reversible: Involvement of PKA, Rho, and ceramide. Front Endocrinol. 2011;2:13. doi: 10.3389/fendo.2011.00013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Bonilla C., Lewis S.J., Rowlands M.-A. Assessing the role of insulin-like growth factors and binding proteins in prostate cancer using Mendelian randomization: genetic variants as instruments for circulating levels. Int J Cancer. 2016;139(7):1520–1533. doi: 10.1002/ijc.30206. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Fry A., Littlejohns T.J., Sudlow C. Comparison of sociodemographic and health-related characteristics of UK Biobank participants with those of the general population. Am J Epidemiol. 2017;186(9):1026–1034. doi: 10.1093/aje/kwx246. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Batty G.D., Gale C.R., Kivimaki M. Comparison of risk factor associations in UK Biobank against representative, general population based studies with conventional response rates: prospective cohort study and individual participant meta-analysis. BMJ. 2020:368:m131. doi: 10.1136/bmj.m131. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.