Abstract
Objective: Observational studies showed that Type 2 diabetes increased the risk of breast cancer, and vice versa. However, it is uncertain whether the link is causal or just due to confounding factors. Using bidirectional Mendelian randomization analysis, we assessed the bidirectional causal relationship from a genetic level. Methods: Large genome-wide association studies yielded summary-level data for Type 2 diabetes and breast cancer. Results: Genetically predicted Type 2 diabetes presented no statistically significant association with overall breast cancer or its subtypes. Similarly, genetically predicted overall breast cancer or its subtypes had no causal effect on Type 2 diabetes. Sensitivity analyses yielded similar results. Conclusion: Our bidirectional Mendelian randomization studies revealed no causal links between Type 2 diabetes and breast cancer.
Keywords: : breast cancer, causal association, Mendelian randomization, Type 2 diabetes
Plain language summary
Summary points.
Type 2 diabetes had no causal effect on the development of overall breast cancer, estrogen receptor (ER)+ or ER- subtypes.
Overall breast cancer, ER+ and ER- subtypes had no causal effect on the development of Type 2 diabetes.
The previously observed substantial relationships between Type 2 diabetes and breast cancer in observational studies could be explained by failing to account for confounders completely.
Breast cancer is the most frequent type of cancer in females around the world [1]. There is significant heterogeneity in survival outcomes among breast cancer patients, which can be explained in part by well-established clinicopathological factors such as clinical stage, tumor grade and molecular phenotype. However, it is suggested that other factors may also be important. For example, the effect of Type 2 diabetes on breast cancer development has been investigated in multiple observational studies, which showed that Type 2 diabetes may increase the risk of breast cancer [2,3]. In addition, a small amount of evidence also indicated that the risk of future diabetes may be raised in women after a breast cancer diagnosis [4]. However, it is uncertain if the link is causal or just due to confounding factors such as obesity. Given the significant global burden of diabetes and breast cancer, if these connections are causal, they could have a significant impact on public health.
Mendelian randomization (MR) analysis is a well-established method for estimating the causal relationship between an exposure and an associated outcome by using data on inherited genetic variants (single nucleotide polymorphisms [SNPs]) that affect exposure status. Because genetic variants are assigned at random during conception and are not influenced by confounding factors, using genetic variants as instruments for modifiable exposures has the potential to avoid the limitations associated with traditional observational studies for making causal inferences [5]. For example, it has been widely used to assess the causal effect of risk factors on the development of cardiovascular diseases [6,7].
Although previous MR studies have enhanced our understanding of an intrinsic relationship underpinning Type 2 diabetes and breast cancer, the causal effect of breast cancer on Type 2 diabetes has not been investigated through an MR design. In addition, as the sample size of genome-wide association studies (GWASs) grows and data accumulate, it is vital to do a thorough reevaluation of the causal role of Type 2 diabetes in breast cancer using an MR approach. Multivariable MR adjusting potential confounders is also of great significance, yet such analysis is deficient in previous MR analyses. Therefore, the goal of this bidirectional MR study was to use germline genotype data for genetic variants known to be related to Type 2 diabetes and breast cancer to assess the causal relationship between Type 2 diabetes and breast cancer in an unbiased manner.
Materials & methods
MR analysis has to satisfy three assumptions (Figure 1): 1) genetic variants must be valid instruments that are associated with the exposure of interest; 2) genetic variants are independent of known risk factors for the outcome and 3) associations between exposure and outcome are mediated only by the exposure of interest without pleiotropy, that is, there are no other causal pathways connecting the genetic variants to the outcome.
Figure 1.
Mendelian randomization model.
Data sources
Summary results from GWAS for Type 2 diabetes were obtained from Xue et al. [8], which includes 61,714 Type 2 diabetes cases and 1178 controls. Summary-level data were available for three breast cancer phenotypes including the overall breast cancer, the estrogen receptor (ER)+ and the ER- subtypes. We obtained breast cancer data from the Breast Cancer Association Consortium (BCAC) [9], which includes 122,977 breast cancer cases, 69,501 ER+ cases, 21,468 ER- cases and 105,974 controls (Supplementary Table 1). The data included in the present analyses have been made publicly available, and ethical approval and informed consent have been obtained in the original GWAS.
Instrumental variable selection
We extracted SNPs that reached genome-wide significance (p < 5 × 10-8) from GWAS for the exposures and excluded all SNPs with linkage disequilibrium (within a 10,000 kb window, r2 >0.001) to assure statistical independence. These SNPs were then matched and harmonized with the outcome GWAS. We calculated the strength of instrumental variables to prevent weak instrument bias. If the relevant F-statistic was greater than 10, the instrument was judged suitably powerful and was included [10].
Statistical analysis
Inverse-variance weighted approach under a multiplicative random-effect model was conducted as our primary analysis. The inverse-variance weighted method presupposes that all genetic variants satisfy the MR assumptions as a whole, or that all SNPs are valid instruments. Besides this, five additional MR methods – namely weighted median, MR–Egger, multivariable MR, simple mode and weighted mode approaches – were used as sensitivity analyses to assess the robustness of the results. If more than 50% of the weight in the instrumental variable model originates from SNPs that satisfy the MR assumptions, the weighted median method can produce a valid result [11]. MR–Egger method allows for horizontal pleiotropic effects and the MR–Egger intercept estimates the average pleiotropic effect across all SNPs. If the MR–Egger intercept deviates from zero, horizontal pleiotropy is present [12]. If there is asymmetry, funnel plots can also identify horizontal pleiotropy. Multivariable MR analysis adjusting BMI, smoking and alcohol drinking was also performed. The weighted mode function can provide a reliable estimate as long as the most frequent SNP effects are contributed by valid SNP [13]. Furthermore, a leave-one-out approach was used to identify outliers that might influence the MR estimates by removing each SNP sequentially. The heterogeneity among selected SNPs was assessed using Cochrane's Q value. All analyses were performed using the ‘TwoSampleMR’ package in R software (4.0.3) [14].
Results
As shown in Figure 2, genetically predicted Type 2 diabetes presented no statistically significant association with overall breast cancer (odds ratio [OR]: 0.99; 95% CI: 0.95–1.02; p = 0.45). Such an effect remained consistent in both ER+ (OR: 0.98; 95% CI: 0.94–1.02; p = 0.29) and ER- (OR: 0.98; 95% CI: 0.94–1.04; p = 0.55) subtypes. Similarly, genetically predicted overall breast cancer (OR: 1.05; 95% CI: 0.96–1.15; p = 0.27), ER+ (OR: 1.00; 95% CI: 0.94–1.07; p = 0.89) or ER- (OR: 1.04; 95% CI: 0.82–1.32; p = 0.74) subtypes had no causal effect on Type 2 diabetes. These results were further supported by the weighted median and the MR–Egger methods (Table 1), with estimates consistent in direction except for the effect of ER+ breast cancer on Type 2 diabetes (OR: 1.06; 95% CI: 1.01–1.10; p = 0.02). After adjusting BMI, smoking and alcohol drinking, the multivariable MR analysis yielded similar results (Table 1). The simple mode method in Supplementary Table 2 also revealed similar results.
Figure 2.
Associations between Type 2 diabetes and breast cancer.
ER: Estrogen receptor; OR: Odds ratio; SNP: Single nucleotide polymorphism.
Table 1.
Associations between Type 2 diabetes and breast cancer in sensitivity analyses using the weighted median and Mendelian randomization–Egger methods.
| Outcomes | Weighted median | Mendelian randomization–Egger | Multivariable Mendelian randomization | |||
|---|---|---|---|---|---|---|
| OR (95% CI) | p-value | OR (95% CI) | p-value | OR (95% CI) | p-value | |
| Type 2 diabetes→breast cancer | 1.04 (1.00–1.07) | 0.07 | 1.06 (0.97–1.16) | 0.19 | 1.01 (0.97–1.04) | 0.71 |
| Type 2 diabetes→ER+ breast cancer | 1.01 (0.96–1.05) | 0.72 | 1.03 (0.94–1.13) | 0.54 | 0.99 (0.95–1.03) | 0.62 |
| Type 2 diabetes→ER- breast cancer | 1.08 (1.02–1.14) | 0.01 | 1.05 (0.94–1.17) | 0.40 | 1.02 (0.97–1.07) | 0.47 |
| Breast cancer→Type 2 diabetes | 1.06 (1.01–1.11) | 0.02 | 0.97 (0.79–1.19) | 0.77 | 1.06 (0.97–1.15) | 0.22 |
| ER+ breast cancer→Type 2 diabetes | 1.06 (1.01–1.10) | 0.02 | 1.09 (0.93–1.26) | 0.30 | 1.00 (0.94–1.06) | 0.91 |
| ER- breast cancer→Type 2 diabetes | 1.05 (0.98–1.12) | 0.19 | 0.85 (0.40–1.83) | 0.69 | 1.11 (1.00–1.23) | 0.051 |
ER: Estrogen receptor; OR: Odds ratio.
The scatter plots and forest plots of the associations between Type 2 diabetes and breast cancer are shown in Supplementary Figures 1 & 2, respectively. The leave-one-out sensitivity analysis revealed that no single SNP disproportionately affected these results (Supplementary Figure 3). The MR–Egger intercept in Table 2 revealed no evidence of horizontal pleiotropy, which was also confirmed by the funnel plots in Figure 3. There was strong evidence of heterogeneity across SNPs (Table 2), and a multiplicative random-effect model was therefore adopted as our primary analysis.
Table 2.
Analyses of horizontal pleiotropy and heterogeneity between Type 2 diabetes and breast cancer.
| Outcomes | Horizontal pleiotropy | Heterogeneity | ||
|---|---|---|---|---|
| Intercept | p-value | Q | p-value | |
| Type 2 diabetes→breast cancer | -0.0060 | 0.07 | 480 | < 0.01 |
| Type 2 diabetes→ER+ breast cancer | -0.0043 | 0.23 | 390 | < 0.01 |
| Type 2 diabetes→ER- breast cancer | -0.0053 | 0.21 | 237 | < 0.01 |
| Breast cancer→Type 2 diabetes | 0.0062 | 0.39 | 957 | < 0.01 |
| ER+ breast cancer→Type 2 diabetes | -0.0069 | 0.28 | 455 | < 0.01 |
| ER- breast cancer→Type 2 diabetes | 0.0199 | 0.59 | 799 | < 0.01 |
ER: Estrogen receptor.
Figure 3.
Funnel plot of the associations between Type 2 diabetes and breast cancer.
(A) Type 2 diabetes→breast cancer. (B) Type 2 diabetes→ER+ breast cancer. (C) Type 2 diabetes→ER- breast cancer. (D) Breast cancer→Type 2 diabetes. (E) ER+ breast cancer→Type 2 diabetes. (F) ER- breast cancer→Type 2 diabetes.
ER: Estrogen receptor; MR: Mendelian randomization.
Discussion
Using the largest available GWAS up to now, our bidirectional MR analysis suggested that Type 2 diabetes had no causal effect on breast cancer. Vice versa, breast cancer had no causal effect on Type 2 diabetes.
A possible link between Type 2 diabetes and breast cancer has long been suspected, as these diseases are diagnosed in the same individual more frequently than would be predicted by chance [3]. Actually, the effect of Type 2 diabetes on breast cancer has been thoroughly evaluated by observational studies and MR analyses. A meta-analysis based on observational studies (n = 20) suggested that Type 2 diabetes may increase the risk of breast cancer (relative risk: 1.20; 95% CI: 1.12–1.28) [15]. Another meta-analysis including 38 studies and 151,585 patients with breast cancer undergoing breast reconstruction also suggested that diabetes may increase the overall complications (11.6 vs 5.6%; p < 0.0001) and surgical complications (7.7 vs 3.3%; p < 0.0001), and these breast cancer patients with diaetes were more likely to have a prolonged hospital stay (p = 0.04) than women without diabetes [16]. In an MR analysis based on Asian individuals (24,000 participants in the Taiwan Biobank), a higher level of genetically predicted diabetes was associated with an increased risk of recurrence of and mortality from breast cancer (hazard ratio [HR]: 1.43; p < 0.001) [17]. MR analysis using data from the BCAC (European individuals) also suggested that genetic liability to Type 2 diabetes was significantly associated with worse breast cancer-specific survival (HR: 1.10; 95% CI: 1.03–1.17; p = 0.003) [18]. However, another MR analysis found no causal effect of Type 2 diabetes on over breast cancer (OR: 1.00; 95% CI: 0.98–1.02; p = 0.880), ER+ (OR: 1.00; 95% CI: 0.97–1.03; p = 0.986) or ER- (OR: 0.99; 95% CI: 0.96–1.01; p = 0.310) subtypes using BCAC data. Analysis based on UK biobank data yielded similar results (OR: 1.00; 95% CI: 0.97–1.03; p = 0.793) [19]. Therefore, the effect of Type 2 diabetes on breast cancer remains controversial, and new evidence is required to validate the causality. Using the largest available GWAS up to now, our MR analysis suggested that Type 2 diabetes had no causal effect on breast cancer.
In addition, we provided new evidence that breast cancer had no causal effect on Type 2 diabetes. Bordeleau et al. discovered that as compared with women who did not get breast cancer, women with breast cancer were twice as likely to develop diabetes (relative risk: 2.0; 95% CI: 1.4–2.8; p = 0.0001) [4]. Lipscombe and colleagues examined whether postmenopausal women who got breast cancer had a greater incidence of diabetes than those who did not, using linked population-based health databases from Ontario, Canada [2]. Over a mean follow-up of 5.8 years, 9.7% of 24,976 breast cancer survivors and 124,880 controls developed diabetes. Diabetes risk increased 2 years after diagnosis in breast cancer survivors compared with those without breast cancer (HR: 1.07; 95% CI: 1.02–1.12) and continued to rise over time. A Korean national cohort study (National Health Insurance Service-National Sample Cohort [NHIS-NSC]) included a total of 524,089 men and women aged between 20 and 70 years without diabetes or cancer at baseline from 1 January 2003 to 31 December 2013 [20]. During 3,492,935.6 person-years of follow-up in 494,189 individuals (50.0% female), 15,130 participants developed cancer and 26,610 participants developed diabetes. Diabetes was strongly linked with the development of breast cancer (HR: 1.60; 95% CI: 1.27–2.01). However, the increased risk of diabetes in cancer patients could also be mediated by common risk factors for cancer and diabetes, such as obesity, smoking and excessive drinking, yet few studies have data on these risk factors before the development of cancer [3]. As shown in the NHIS-NSC study, individuals who developed cancer were older, more likely to drink alcohol, had a higher BMI and had more comorbidities than those who did not [20]. By applying bidirectional MR analysis, we can investigate the causal association between Type 2 diabetes and breast cancer in an unbiased way. Our results did not support a bidirectional link between Type 2 diabetes and breast cancer. Consistent with our results, another MR analysis also found no causal effect of Type 2 diabetes on breast cancer using BCAC data (OR: 0.98; 95% CI: 0.95–1.01). Similar findings were observed using data from UK Biobank (OR: 0.98; 95% CI: 0.95–1.03) [21].
The increased risk of diabetes in breast cancer patients or increased risk of breast cancer in diabetes patients in previous observational studies could be explained by the following reasons. First, chemotherapy may produce hyperglycemia and raise the risk of diabetes [22]. L-asparaginase is a well-established cause of hyperglycemia because it directly inhibits insulin release [23]. Diabetes risk may also be increased by total body irradiation [24] and immunosuppressive medications such as calcineurin inhibitors [25]. Glucocorticoids are often simultaneously used to reduce nausea associated with chemotherapy, which is linked to hyperglycemia and diabetes, a process mostly mediated by decreased insulin sensitivity [26]. The use of chemotherapy agents in conjunction with corticosteroids may have a synergistic effect on diabetes risk. Tamoxifen, an antiestrogen, may further promote diabetes [27]. However, observational studies lacked information on prescription drugs or hormone therapy. Second, patients who develop breast cancer are monitored more closely and have more opportunities for diabetes screening, which may lead to surveillance bias [28]. Misclassification of diabetes patients who did not know that they had diabetes to the reference group of participants without diabetes is also highly likely. Third, empirical evidence suggests that skewed observational connections are widespread in cancer epidemiology literature [29,30]. Typical questionable methods include selective nonpublication of ‘negative’ results and selective reporting of analyses and outcomes with great impact [31,32]. If there is a strong belief in the assumption that Type 2 diabetes increases the likelihood of breast cancer, research activities will be pushed to adhere to this notion.
Our bidirectional MR analysis represents the first attempt at discovering the bidirectional causal associations between Type 2 diabetes and breast cancer from the genetic level, and suggested no causal associations between Type 2 diabetes and breast cancer.
Limitations
Several additional limitations and caveats should be addressed when interpreting our findings. First, our MR analysis data was derived from people primarily of European ancestry, which may not be generalizable to non-European populations. However, given that population stratification might impair MR estimations, our results are unlikely to be influenced by population stratification bias. Another restriction is the possibility of horizontal pleiotropy influencing the results. However, neither the MR–Egger analysis nor funnel plots revealed any evidence of horizontal pleiotropy. Third, there was strong evidence of heterogeneity across SNPs, and a multiplicative random-effect model was therefore adopted as our primary analysis. In addition, sensitivity analyses were performed and similar results were observed.
Conclusion
Our comprehensively bidirectional MR analysis revealed no causal links between Type 2 diabetes and breast cancer. The previously observed substantial relationships in observational studies could be explained by failing to account for confounders completely, concurrent chemotherapy, surveillance bias and selective reporting of outcomes with great impact.
Supplementary Material
Funding Statement
This study was supported by Henan Medical Science and Technology Research Project (212102310338), Henan Medical Science and Technology Project Jointly Built by the Ministry and the Province (LHGJ20190673), and Postdoctoral Research Project of Henan Province (202002077).
Financial disclosure
This study was supported by Henan Medical Science and Technology Research Project (212102310338), Henan Medical Science and Technology Project Jointly Built by the Ministry and the Province (LHGJ20190673), and Postdoctoral Research Project of Henan Province (202002077). The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.
Competing interests disclosure
The authors have no competing interests or relevant affiliations with any organization or entity with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, stock ownership or options and expert testimony.
Writing disclosure
No writing assistance was utilized in the production of this manuscript.
References
Papers of special note have been highlighted as: • of interest; •• of considerable interest
- 1.DeSantis CE, Ma J, Goding Sauer A, Newman LA, Jemal A. Breast cancer statistics, 2017, racial disparity in mortality by state. CA Cancer J. Clin. 67(6), 439–448 (2017). [DOI] [PubMed] [Google Scholar]
- 2.Lipscombe LL, Chan WW, Yun L, Austin PC, Anderson GM, Rochon PA. Incidence of diabetes among postmenopausal breast cancer survivors. Diabetologia 56(3), 476–483 (2013). [DOI] [PubMed] [Google Scholar]
- 3.Giovannucci E, Harlan DM, Archer MCet al. Diabetes and cancer: a consensus report. Diabetes Care 33(7), 1674–1685 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Bordeleau L, Lipscombe L, Lubinski Jet al. Diabetes and breast cancer among women with BRCA1 and BRCA2 mutations. Cancer 117(9), 1812–1818 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Emdin CA, Khera AV, Kathiresan S. Mendelian randomization. JAMA 318(19), 1925–1926 (2017). [DOI] [PubMed] [Google Scholar]
- 6.Hu M, Li X, Yang Y. Causal associations between cardiovascular risk factors and venous thromboembolism. Semin. Thromb. Hemost. 49(7), 679–687 (2023). [DOI] [PubMed] [Google Scholar]
- 7.Hu MJ, Tan JS, Gao XJ, Yang JG, Yang YJ. Effect of cheese intake on cardiovascular diseases and cardiovascular biomarkers. Nutrients 14(14) 1–11 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Xue A, Wu Y, Zhu Zet al. Genome-wide association analyses identify 143 risk variants and putative regulatory mechanisms for type 2 diabetes. Nat. Commun. 9(1), 2941 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Michailidou K, Lindström S, Dennis Jet al. Association analysis identifies 65 new breast cancer risk loci. Nature 551(7678), 92–94 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Burgess S, Thompson SG. Avoiding bias from weak instruments in Mendelian randomization studies. Int. J. Epidemiol. 40(3), 755–764 (2011). [DOI] [PubMed] [Google Scholar]
- 11.Burgess S, Bowden J, Fall T, Ingelsson E, Thompson SG. Sensitivity analyses for robust causal inference from Mendelian randomization analyses with multiple genetic variants. Epidemiology 28(1), 30–42 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int. J. Epidemiol. 44(2), 512–525 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Hartwig FP, Davey Smith G, Bowden J. Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption. Int. J. Epidemiol. 46(6), 1985–1998 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Hemani G, Zheng J, Elsworth Bet al. The MR-Base platform supports systematic causal inference across the human phenome. Elife 7, e34408 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Pearson-Stuttard J, Papadimitriou N, Markozannes Get al. Type 2 diabetes and cancer: an umbrella review of observational and Mendelian randomization studies. Cancer Epidemiol. Biomarkers Prev. 30(6), 1218–1228 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]; • Meta-analysis of observational studies suggested that diabetes increased the risk of breast cancer.
- 16.Liu Q, Aggarwal A, Wu Met al. Impact of diabetes on outcomes in breast reconstruction: a systematic review and meta-analysis. J. Plast. Reconstr. Aesthet. Surg. 75(6), 1793–1804 (2022). [DOI] [PubMed] [Google Scholar]
- 17.Liu YS, Wu PE, Chou WCet al. Body mass index and type 2 diabetes and breast cancer survival: a Mendelian randomization study. Am. J. Cancer Res. 11(8), 3921–3934 (2021). [PMC free article] [PubMed] [Google Scholar]
- 18.Escala-Garcia M, Morra A, Canisius Set al. Breast cancer risk factors and their effects on survival: a Mendelian randomisation study. BMC Med. 18(1), 327 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]; • Systematic review of Mendelian randomization studies suggested that diabetes increased the risk of breast cancer.
- 19.Yuan S, Kar S, Carter Pet al. Is type 2 diabetes causally associated with cancer risk? Evidence from a two-sample Mendelian randomization study. Diabetes 69(7), 1588–1596 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]; • Mendelian randomization study suggested that breast cancer increased the risk of diabetes.
- 20.Hwangbo Y, Kang D, Kang Met al. Incidence of diabetes after cancer development: a Korean national cohort study. JAMA Oncol. 4(8), 1099–1105 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]; •• A large Korean cohort study suggested that cancer development increased the risk of subsequent diabetes.
- 21.Au Yeung SL, Schooling CM. Impact of glycemic traits, type 2 diabetes and metformin use on breast and prostate cancer risk: a Mendelian randomization study. BMJ Open Diabetes Res. Care 7(1), e000872 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Mohn A, Di Marzio A, Capanna R, Fioritoni G, Chiarelli F. Persistence of impaired pancreatic beta-cell function in children treated for acute lymphoblastic leukaemia. Lancet 363(9403), 127–128 (2004). [DOI] [PubMed] [Google Scholar]
- 23.Yoshida H, Imamura T, Saito AMet al. Protracted administration of L-asparaginase in maintenance phase is the risk factor for hyperglycemia in older patients with pediatric acute lymphoblastic leukemia. PLOS ONE 10(8), e0136428 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Meacham LR, Sklar CA, Li Set al. Diabetes mellitus in long-term survivors of childhood cancer. Increased risk associated with radiation therapy: a report for the childhood cancer survivor study. Arch. Intern. Med. 169(15), 1381–1388 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Davidson J, Wilkinson A, Dantal Jet al. New-onset diabetes after transplantation: 2003 international consensus guidelines. Proceedings of an international expert panel meeting. Barcelona, Spain, 19 February 2003. Transplantation 75(Suppl. 10), S3–S24 (2003). [DOI] [PubMed] [Google Scholar]
- 26.Clore JN, Thurby-Hay L. Glucocorticoid-induced hyperglycemia. Endocr. Pract. 15(5), 469–474 (2009). [DOI] [PubMed] [Google Scholar]
- 27.Lipscombe LL, Fischer HD, Yun Let al. Association between tamoxifen treatment and diabetes: a population-based study. Cancer 118(10), 2615–2622 (2012). [DOI] [PubMed] [Google Scholar]
- 28.Tsilidis KK, Kasimis JC, Lopez DS, Ntzani EE, Ioannidis JP. Type 2 diabetes and cancer: umbrella review of meta-analyses of observational studies. BMJ 350, g7607 (2015). [DOI] [PubMed] [Google Scholar]
- 29.Schoenfeld JD, Ioannidis JP. Is everything we eat associated with cancer? A systematic cookbook review. Am. J. Clin. Nutr. 97(1), 127–134 (2013). [DOI] [PubMed] [Google Scholar]
- 30.Boffetta P, McLaughlin JK, La Vecchia C, Tarone RE, Lipworth L, Blot WJ. False-positive results in cancer epidemiology: a plea for epistemological modesty. J. Natl Cancer Inst. 100(14), 988–995 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Kavvoura FK, Liberopoulos G, Ioannidis JP. Selection in reported epidemiological risks: an empirical assessment. PLOS Med. 4(3), e79 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Ioannidis JP. The importance of potential studies that have not existed and registration of observational data sets. JAMA 308(6), 575–576 (2012). [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.