Multiple sclerosis and breast cancer risk: a meta-analysis of observational and Mendelian randomization studies

1. Introduction

Multiple sclerosis (MS) is a chronic autoimmune demyelinating disease (Anderson et al., 2021), and several experimental studies have found an imbalance between the inflammatory and regulatory T-cell balance in MS patients (Stephens et al., 2009; Quinn and Axtell, 2018). Because the immune system plays an initial role in MS and cancer, it is reasonable to suspect that patients with MS may have an increased incidence of cancer.

Breast cancer is the most frequent malignant tumor worldwide and is a serious threat to women’s lives and health (Sung et al., 2021). If we can identify high-risk factors for breast cancer and enhance screening, we can detect breast cancer early and thus improve the survival rate. Many researchers have explored the relationship between MS and breast cancer. Groome et al. (2022) reported a study that compared the incidence of breast cancer and colorectal cancer in MS patients through a population-based study, and Groome et al. (2022) concluded that MS patients are easily detected to have colorectal cancer but not breast cancer. Bosco-Lévy et al. (2022) found that the risk of breast cancer was higher among MS patients than among population controls (HR = 1.12, 95% CI = 1.03−1.23). Those consistencies may be caused by different study methodologies, designs, or unadjusted confounding factors (Bosco-Lévy et al., 2022). Hence, we performed a Mendelian randomization (MR) study, a new etiological investigation method, combined with previous observational studies to explore the epidemiological relationship between MS and breast cancer.

MR can identify the causal effect between risk factors and health outcomes using genetic variants as instrumental variables (Bowden and Holmes, 2019). MR can provide robust results between exposures and outcomes because mitosis allows for the random allocation of gametes to offspring, which allows for results that are not confounded by potential confounding factors and allows for a better search for the cause of the disease compared to traditional observational studies (Richmond and Davey Smith, 2022). Our study aimed to explore whether MS is a risk or protective factor for breast cancer by performing an observational meta-analysis and MR study.

2. Materials and methods 2.1. Meta-analysis 2.1.1. Search strategy

The overall design of this study is shown in Figure 1. We searched four online databases, PubMed, Embase, Web of Science, and Scopus, for articles published in English up to 23 September 2022. The search keywords were (“multiple sclerosis”) and (“breast neoplasms” or “breast cancer” or “breast tumor”).

Figure 1. Flow chart for meta-analysis and Mendelian randomization analysis.

2.1.2. Inclusion and exclusion criteria

Studies that met the following criteria were included in our meta-analysis: (i) cohort studies of patients with MS and (ii) all published articles were in English. (iii) evaluated the association or risk between MS and breast cancer. (iv) sample size not less than 300. (v) Studies provided odds ratios (ORs), relative risks (RRs), hazard ratios (HRs), or standardized incidence ratios (SIRs) with their 95% confidence intervals (CIs) of breast cancer among MS patients. The exclusion criteria were as follows: (i) systematic review, case report, letter, meeting or meta-analysis; and (ii) insufficient data to obtain effect sizes.

2.1.3. Data extraction and assessment of quality

Two researchers (Zhang and Fang) extracted the data independently. When disagreements arose, they were resolved through discussion. We extracted the following information: the name of the first author, country, year of publication, sample size, study period (including follow-up), and effect size (OR, RR, HR or SIR) with 95% CI. Two researchers (Fang and Zhang) evaluated the quality of the included studies using the nine-item Newcastle−Ottawa Quality scale. We have set a minimum follow-up period of 5 years, and no points will be awarded for less than 5 years or for not reporting the follow-up period. Concerning completeness of the follow-up, we set 5% as a cutoff level of loss during the follow-up. Those with a quality score of no more than 5 will be excluded.

2.1.4. Data synthesis

We extracted the effect size, if available, or calculated it from available data, including OR, RR, HR, or SIR. SIRs are calculated as the ratio of the number of observed cases to the number of expected cases in the exposed population. The 95% CI for log(SIR) was constructed via the term + 1.96/[square root (O)], where O was the observed case (Alder et al., 2006). Since the absolute risk of breast cancer is low, the four types of measures are expected to have similar estimates of RR. Consequently, our final summary results are presented in the form of RR (Larsson et al., 2007). The same statistical methods have been applied in other meta-analysis articles (Siristatidis et al., 2013).

We used STATA version 14.0 (Stata Corporation, College Station, TX, USA) software for data analysis and visualization. Statistical significance is expressed as a pooled P, and a P < 0.05 was considered statistically significant. The Cochran Q-test and I2 statistics were used to measure heterogeneity; P < 0.1 and I2 value > 50% represented substantial heterogeneity (Higgins et al., 2003). The random model was used if strong heterogeneity existed, and subgroup analysis was conducted to detect potential heterogeneity. If there is no heterogeneity, the fixed-effects model will be used (DerSimonian and Laird, 2015). Sensitivity analysis was used to check data stability, and Egger’s test was used to detect publication bias.

2.2. Mendelian randomization 2.2.1. Study design

We explored the relationship between MS and breast cancer using a two-sample MR study, which applied genetic predictors of exposure to outcome (Emdin et al., 2017). Our MR study had to follow the following three assumptions: (1) There is a strong association between the instrument variants (IVs) and the risk factor. (2) There are no associations between the IVs and any confounders. (3) IVs should not influence outcomes by confounders or other ways (Davey Smith and Hemani, 2014).

2.2.2. Instrument selection

Genome-wide association studies (GWAS) of MS were derived from the International Multiple Sclerosis Genetics Consortium (IMSGC) and FinnGen datasets. In brief, the IMSGC analyzed genetic data from 15 GWASs, including 47,429 MS cases and 68,374 controls, all of whom were of European ancestry (International Multiple Sclerosis Genetics Consortium, 2019). The summary data of IMSGC were adjusted for confounding factors, age, sex, immunomodulatory drugs, batch effects, and the first 10 principal components (International Multiple Sclerosis Genetics Consortium, 2019). The diagnosis of MS was defined by the International Classification of Diseases (ICD) from the Finnish R5 release dataset, and there were 1,048 cases and 217,141 controls of Finnish ancestry. Sex, age, and 10 principal components were logistic regression covariates, and the data can also be found in MRCIEU datasets (finn-b-G6_MS). The independent single nucleotide polymorphisms (SNPs) at p < 5 × 10–8 were selected as IVs (linkage disequilibrium: R2 < 0.01; 5,000). We assessed the correlation of each IV with risk factors using the F statistic, with F < 10 representing a weak instrumental variable and exclusion (Pierce et al., 2011).

2.2.3. Data source of outcome

The summary GWAS of breast cancer was derived from the Breast Cancer Association Consortium (BCAC) with 122,977 cases (ER + breast cancer, 69,501; ER- breast cancer, 21,468) and 105,974 controls of European ancestry (Michailidou et al., 2017). Two genotyping arrays were used for genotyping: iCOGS arrays in 40,178 breast cancer cases and 35,314 controls and OncoArray in 68,242 cases and 52,367 controls. OncoArray and iCOGS were adjusted for country and study, respectively.

Ethical approval and consent for the summary statistics were obtained from the original publication.

2.2.4. Statistical analysis

The Wald ratio was used to estimate the effect of exposure on the outcome for each SNP. All effects of SNPs were meta-analyzed using the inverse-variance weighted (IVW) method. The IVW method was used as the main analysis to assess the association between MS and breast cancer in our MR study, and both fixed-effect and random-effect models were performed. Moreover, MR−Egger regression, weighted median, weighted mode, simple mode, and robust adjusted profile score (MR-RAPS) were applied to detect the robustness of our results (Bowden et al., 2017).

Several analyses were performed to check heterogeneity and pleiotropy. Cochrane’s Q-value was used to assess heterogeneity. The MR−Egger method was based on the Instrument Strength Independent of Direct Effect (InSIDE) assumption, and the value of the intercept term is far from zero, indicating horizontal pleiotropy (Bowden et al., 2015). The weighted median is more accurate when more than half of the IVs are valid (e.g., due to pleiotropy) (Bowden et al., 2016). MR pleiotropy residual sum and outlier (MR-PRESSO) was used to detect any outlier SNPs and potential horizontal pleiotropy (Verbanck et al., 2018).

2.3. Statistical analysis

All statistical analyses were performed using R software (version 4.1.1) and STATA 12.0. A P-value of <0.05 was considered statistically significant. The “TwoSampleMR,” “mr.raps,” and “MRPRESSO” packages were applied in our MR study. F analysis was used to detect weak instrument variants (F < 10). Was used R2 to estimate the ability of instruments variants present the exposure.

3. Results 3.1. Meta-analyses 3.1.1. Characteristics and quality assessment of the included studies

The literature search with PubMed, Embase, Web of Science, and Scopus yielded 2,876 studies. After removal of duplicates, 853 studies remained, and 62 full-text articles were reviewed after screening. Finally, 15 cohort studies were enrolled in our study, which included 173,565 female MS patients (Midgard et al., 1996; Sumelahti et al., 2004; Achiron et al., 2005; Nielsen et al., 2006; Bahmanyar et al., 2009; Hemminki et al., 2012; Kingwell et al., 2012; Sun et al., 2014; Hajiebrahimi et al., 2016; Etemadifar et al., 2017; Nørgaard et al., 2019; Grytten et al., 2020; Johnson et al., 2021; Marrie et al., 2021; Bosco-Lévy et al., 2022; Mariottini et al., 2022). The quality scores of all studies were no more than 6. The flow diagram for the included and excluded studies is shown in Figure 1, and the reasons for exclusion are listed accordingly. The characteristics of all included studies are shown in Table 1.

Table 1. Characteristics of the included studies.

3.1.2. Risk of breast cancer among MS patients

The forest plot shows the results of this meta-analysis (Figure 2), and the results indicated that there was no relationship between MS and breast cancer [relative ratio (RR) = 1.08, 95% confidence interval (CI): 0.99–1.17]. A random-effects model was performed because of significant heterogeneity among those studies (I2 = 58.2%, p = 0.002). Several subgroup analyses were conducted to detect the potential sources of significant heterogeneity.

Figure 2. Meta-analysis results of the association between MS and breast cancer.

Subgroup analysis concerning ancestries did not find an association between MS and breast cancer in Asia or Europe (Asia: RR = 1.52, 95% CI: 0.95–2.44; Europe: RR = 1.05, 95% CI = 0.98–1.13) (Figure 3A). The subgroup analysis based on the number of MS females revealed no relationship between MS and breast cancer (<3,000: RR = 1.28, 95% CI: 0.90–1.82; 3,000–7,000: RR = 1.11, 95% CI = 0.92–1.33; >7,000: RR = 1.03, 95% CI = 0.96–1.33) (Figure 3B). Subgroup analysis did not detect the source of heterogeneity.

Figure 3. (A) Association of MS and breast cancer in subgroups stratified by ancestries and (B) number of MS females.

3.1.3. Sensitivity analysis

The results did not show any evidence of publication bias (Begg’s Test 0.553; Egger’s Test = 0.297) (Supplementary Figures 1, 2). Leave-one-out sensitivity analysis indicated that the results were robust (Supplementary Figure 3).

3.2. MR analysis 3.2.1. The results based on the IMSGC dataset

The detailed SNP information is shown in Supplementary Table 1. The F analysis for all SNPs was more than 10, which indicated that no weak instrument variants were detected. The R2 values were 27.3, 27.4, and 26.7% for overall breast cancer, ER+ breast cancer and ER− breast cancer, respectively. There was no evidence that MS is related to overall breast cancer (IVW-fixed OR: 1.001, 95% CI: 0.990–1.024; IVW-random OR: 1.001, 95% CI: 0.991–1.023), ER+ breast cancer (IVW-fixed OR: 1.008, 95% CI: 0.990–1.028; IVW-random OR: 1.008, 95% CI: 0.989–1.029) and ER− breast cancer (IVW-fixed OR: 0.995, 95% CI: 0.968–1.023; IVW-random OR: 0.995, 95% CI: 0.975–1.016) based on the discovery set from the IMSGC dataset, with similar results estimated in other sensitivity analyses (Table 2). MR−Egger and Cochrane’s Q-value results indicated that there was no pleiotropy in this part, and the results are also shown in Table 3. No outliers were detected by performing the MR-PRESSO test for all the estimates.

Table 2. MR estimates for the causal effect of multiple sclerosis on breast cancer.

Table 3. Heterogeneity and horizontal pleiotropy analyses results.

3.2.2. The results based on the FinnGen dataset

Ten SNPs were selected as instrument variants (Supplementary Table 2). The F ranged from 29.6 to 259.2. In addition, MS was not a risk or protective factor for overall breast cancer (IVW-fixed OR: 1.000, 95% CI: 0.990–1.010; IVW-random OR: 1.000, 95% CI: 0.987–1.013), ER+ breast cancer (IVW-fixed OR: 1.006, 95% CI: 0.994–1.019; IVW-random OR: 1.006, 95% CI: 0.990–1.023) or ER- breast cancer (IVW-fixed OR: 0.987, 95% CI: 0.968–1.006; IVW-random OR: 0.987, 95% CI: 0.971–1.004) based on the validation set from FinnGen, with similar results assessed in other sensitivity analyses (Table 2). No heterogeneity or pleiotropy results were estimated in this MR study, and the results are also shown in Table 3. No outliers were detected by performing the MR-PRESSO test for all the estimates.

4. Discussion

The etiology of MS is unclear, but it is currently thought to be an immune-mediated demyelinating disease involving the central nervous system (Mariottini et al., 2022; Stampanoni Bassi et al., 2022). MS and breast cancer share some common features; most notably, the incidence of both diseases is much higher in women than in men (O’Malley et al., 2015). Numerous studies have explored the relationship between them, but whether MS affects the incidence of breast cancer remains controversial (Hajiebrahimi et al., 2016; Zecca et al., 2021; Groome et al., 2022). We tried to find the association between MS and breast cancer risk through meta-analysis, and finally, we included fifteen cohort studies. However, these studies were significantly different in terms of ancestry, treatment, and other factors. To avoid these confounding factors, we used Mendelian randomization as a supplement. Ultimately, Mendelian instrumental variables were obtained in 47,429 MS patients and validated in another population including 1,048 MS patients.

Fifteen cohort studies from thirteen countries were included in our meta-analysis, and five of them reported a significant relationship between MS and breast cancer risk. Our meta-analysis found that there was no relationship between MS and breast cancer risk (meta-RR = 1.08; 95% CI = 0.99–1.17; P = 0.073). Consistently, a meta-analysis of the relationship between MS and breast cancer conducted by Catalá-López and Tobías (2010), including only five cohort studies (meta-SIR = 1.02; 95% CI = 0.75–1.40). Our updated meta-analysis included 10 additional studies and therefore had increased statistical power. Moreover, Lopez’s meta-analysis study exhibited strong heterogeneity (I2 = 75.3%), and a moderate degree of heterogeneity was found in our study (I2 = 58.2%). The meta-regression did not find the source of heterogeneity. We attempted multiple subgroup analyses but also failed to find a source of heterogeneity, suggesting that the results might be affected by other potential confounders, such as environmental exposure and treatment.

To overcome the inherent limitations of traditional observational articles, such as difficult-to-detect confounders, reverse causation, and various biases (Davey Smith and Hemani, 2014; Burgess et al., 2017), we conducted a two-sample MR method by genetic variants to further investigate the association between MS and the risk of breast cancer. MR analysis is a new epidemiological method that uses genetic variation as a tool to explore the relationship between risk factors and outcomes (Davies et al., 2018). Genetic variation as an instrument variation must meet three basic conditions: (1) Genetic variations are closely associated with exposure. (2) Genetic variations are not associated with confounding factors. (3) Genetic variations do not directly influence outcomes (Davey Smith and Hemani, 2014). The first hypothesis is easier to prove, an F statistic greater than 10 proves that there is no instrumental variable, and all SNPs in our study have an F statistic greater than 10. The proof of the second and third hypotheses is to examine horizontal pleiotropy, which is the focus and difficulty of MR research (Hemani et al., 2018). Several different sensitivity analyses were used to detect and correct for any potential pleiotropic effects on outcomes in our study. The P-value of the intercept of MR−Egger proves that there is no horizontal pleiotropy, and all sensitivity analysis and IVW method results demonstrated that MS was not associated with breast cancer risk. The same conclusion was obtained in replication practice from FinnGen consortia. The lack of detected heterogeneity suggests that the results of our MR study are relatively stable.

A hypothesis to explain the risk of breast cancer identified that chronic inflammation caused by MS may result in weakened activation of the immune system or immune protection against cancer becoming protumorigenic (Hofer et al., 2010). However, some people hold a different view: MS breaks myelin by upregulating immune system activity, which may have a protective effect against cancer. Immune-inflammation-associated helper T1 cells produce large amounts of antitumor factors that help prevent cancer proliferation (Fletcher et al., 2010). Therefore, speculation that it affects breast cancer risk by affecting the immune system remains controversial. A genotyping study found that mutations in the BRCA1 gene are very close to the MS gene, which may be why there are more cases of breast cancer in MS patients than in non-MS cohorts (Holzmann et al., 2013). Of course, this study was a single-family study, so the results may not be very significant. Another problem with this acceptance is that mutations in the BRCA1 gene account for only a small fraction of the causes of breast cancer (Paul and Paul, 2014). Furthermore, drug treatment, not the disease itself, may be the real cause of cancer (Bahmanyar et al., 2009; Dolladille et al., 2021; Bosco-Lévy et al., 2022). The development of disease-modifying therapies (MDTs) has revolutionized the treatment of MS (Florou et al., 2020). However, the modulation of innate immune mechanisms and suppression of the immune system induced by the use of DMT contribute to an increased risk of malignancies, such as breast cancer, lymphoma, and melanoma, in patients treated with these drugs for long periods of time (Conzett et al., 2011; Ragonese et al., 2017; Melamed and Lee, 2019). In addition, the protective role of CD20 B cells and cytotoxic T cells plays an important role in preventing the development of malignancies, and anti-CD20 therapy for MS has been found to play an important role in the development of cancer (Kelsey et al., 2021). A study found that cancer rates are three times higher among MS patients who receive immunotherapy than among those who never receive immunotherapy. This study found a particular trend toward breast cancer and cancer in the digestive tract, urinary tract, and skin (Lebrun et al., 2008). However, the 15 studies we included did not indicate whether patients with MS received the MDT or other types of treatment, which prevented us from performing subgroup analyses by treatment modality, which is a limitation of our study.

Several advantages of our study exist. First, our meta-analysis included fifteen cohort studies with a large study population (173,565 female patients with MS), providing more reliable results than previous meta-analyses (13,419 female patients with MS) (Catalá-López and Tobías, 2010). Second, this is the first MR study to estimate the causal relationship between MS and breast cancer. Our study design strictly followed the three assumptions of MR (VanderWeele et al., 2014). Third, we used two completely different cohorts of MS in our MR study, and the primary results and sensitivity analysis showed a robust conclusion, all of which suggest that the outcome was stable.

In conclusion, the results of this large MR study and meta-analysis do not support an association of MS with breast cancer risk, and our conclusion may provide insights for future studies.

Data availability statement

The original contributions presented in this study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.

Author contributions

TF and ZZ designed this study and performed the data analyses. TF conducted the analyses and drafted the manuscript. LZ directed the analytical strategy and supervised the study from conception to completion. TF, HZ, and WW revised the manuscript draft. All authors contributed to the interpretation of data and critically revised the manuscript.

Acknowledgments

We gratefully thank American Journal Experts (www.aje.com) for English language editing.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fninf.2023.1154916/full#supplementary-material

Supplementary Figure 1 | Funnel plot of Begg’s test for publication bias.

Supplementary Figure 2 | Funnel plot of Egger’s test for publication bias.

Supplementary Figure 3 | Sensitivity analysis of the leave-one-out test for the meta-analysis.

Footnotes References

Achiron, A., Barak, Y., Gail, M., Mandel, M., Pee, D., Ayyagari, R., et al. (2005). Cancer incidence in multiple sclerosis and effects of immunomodulatory treatments. Breast Cancer Res. Treat. 89, 265–270. doi: 10.1007/s10549-004-2229-4

PubMed Abstract | CrossRef Full Text | Google Scholar

Alder, N., Fenty, J., Warren, F., Sutton, A., Rushton, L., Jones, D., et al. (2006). Meta-analysis of mortality and cancer incidence among workers in the synthetic rubber-producing industry. Am. J. Epidemiol. 164, 405–420. doi: 10.1093/aje/kwj252

PubMed Abstract | CrossRef Full Text | Google Scholar

Anderson, A., Krysko, K., Rutatangwa, A., Krishnakumar, T., Chen, C., Rowles, W., et al. (2021). Clinical and radiologic disease activity in pregnancy and postpartum in MS. Neurol. Neuroimmunol. Neuroinflamm. 8:e959. doi: 10.1212/NXI.0000000000000959

PubMed Abstract | CrossRef Full Text | Google Scholar